What Works, What Doesn't, What's Promising (2024)

PREVENTING CRIME:
WHAT WORKS, WHAT DOESN'T,
WHAT'S PROMISING1

A REPORT TO THE UNITED STATES CONGRESS

Prepared for the National Institute of Justice

by

Lawrence W. Sherman

Denise Gottfredson

Doris MacKenzie

John Eck

Peter Reuter

Shawn Bushway

in collaboration with members of the Graduate Program

Department of Criminology and Criminal Justice
University of Maryland

Scientific Advisers

Ronald V. Clarke
Dean and Professor
School of Criminal Justice
Rutgers University

Phillip Cook
Professor of Public Policy
Duke University

David Farrington
Professor of Psychological Criminology
Cambridge University

Carol Kumpfer
Associate Professor of Health Education
University of Utah

Joan Petersilia
Professor of Criminology, Law and Society
University of California, Irvine

Michael Tonry
Sonofsky Professor of Law
University of Minnesota

Roger Weissberg
Professor of Psychology
University of Illinois at Chicago

Charles Wellford
Professor of Criminology and Criminal Justice
University of Maryland, College Park

Partial List of Collaborating Graduate Students

Todd Armstrong, M.A.
Katherine Culotta
Laurie Alphonse, M.A.
Cynthia Lum, M.A.
Jennifer Borus
Jeffrey Bouffard, M.A.
Lynn Exum, M.A.
Veronica Puryear
John Ridgely
Stacy Skobran, M.A.
Shannon Womer
Richard Lewis, M.A.
Christine Depies
Shawn J. Anderies
Mohammed Bin Kashem, M.A.
Julie Kiernan
Aimee C. Kim
Daniel R. Lee, M.A.
Patti A. Mattson
Jennifer R. Smith
David A. Soule
Stephanie L. Weiner

NOTES

1This report was supported by National Institute of JusticeGrant Number 96MUMU0019 to the University of Maryland at College Park.Points of view or opinions stated herein are those of the authors and donot necessarily represent the official views of the United States Departmentof Justice.

Table of Contents

Overview

1. Introduction: The Congressional Mandate to Evaluate
Lawrence W. Sherman

2. Thinking About Crime Prevention
Lawrence W. Sherman

3. Communities and Crime Prevention
Lawrence W. Sherman

4. Family-Based Crime Prevention
Lawrence W. Sherman

5. School-Based Crime Prevention
Denise Gottfredson

6. Labor Markets and Crime Risk Factors
Shawn Bushway and Peter Reuter

7. Preventing Crime at Places
John Eck

8. Policing for Crime Prevention
Lawrence W. Sherman

9. Criminal Justice and Crime Prevention
Doris L. MacKenzie

10. Conclusions: The Effectiveness of Local Crime Prevention Funding
Lawrence W. Sherman

Appendix: Methodology for this Report
Lawrence W. Sherman and Denise Gottfredson

PREVENTING CRIME: AN OVERVIEW

by Lawrence W. Sherman

Mandate. In 1996 Congress required the Attorney General to providea "comprehensive evaluation of the effectiveness" of over $3Billion annually in Department of Justice grants to assist State and locallaw enforcement and communities in preventing crime. Congress requiredthat the research for the evaluation be "independent in nature,"and "employ rigorous and scientifically recognized standards and methodologies."It also called for the evaluation to give special emphasis to "factorsthat relate to juvenile crime and the effect of these programs on youthviolence," including "risk factors in the community, schools,and family environments that contribute to juvenile violence." TheAssistant Attorney General for the Office of Justice Programs asked theNational Institute of Justice to commission an independent review of therelevant scientific literature, which exceeds 500 program impact evaluations.

Primary Conclusion. This Report found that some prevention programswork, some do not, some are promising, and some have not been tested adequately.Given the evidence of promising and effective programs, the Report findsthat the effectiveness of Department of Justice funding dependsheavily on whether it is directed to the urban neighborhoods where youthviolence is highly concentrated. Substantial reductions in nationalrates of serious crime can only be achieved by prevention in areas of concentratedpoverty, where the majority of all homicides in the nation occur, and wherehomicide rates are 20 times the national average.

Primary Recommendation. Because the specific methods forpreventing crime in areas of concentrated poverty are not well-developedand tested, the Congress can make most effective use of DOJ local assistancefunding by providing better guidance about what works. A much larger partof the national crime prevention portfolio must be invested in rigoroustesting of innovative programs, in order to identify the active ingredientsof locally successful programs that can be recommended for adoption insimilar high-crime urban settings nation-wide.

SECONDARY CONCLUSIONS. The Report also reaches several secondaryconclusions:

o Institutional Settings. Most crime prevention resultsfrom informal and formal practices and programs located in seven institutionalsettings. These institutions appear to be "interdependent" atthe local level, in that events in one of these institution can affectevents in others that in turn can affect the local crime rate. These arethe seven institutions identified in Chapter Two:

* Communities

* Families

* Schools

* Labor Markets

* Places (specific premises)

* Police

* Criminal Justice

o Effective Crime Prevention in High-Violence Neighborhoods MayRequire Interventions in Many Local Institutions Simultaneously. Theinterdependency of these local institutions suggests a great need for rigoroustesting of programs that simultaneously invest in communities, families,schools, labor markets, place security, police and criminal justice. OperationWeed and Seed provides the best current example of that approach, but receivesa tiny fraction of DOJ funding.

o Crime Prevention Defined. Crime prevention is defined not byintentions or methods, but by results. There is scientific evidence, forexample, that both schools and prisons can help prevent crime. Crime preventionprograms are neither "hard" nor "soft" by definition;the central question is whether any program or institutional practice resultsin fewer criminal events than would otherwise occur. Chapter Two presentsthis analysis.

o The Effectiveness of Federal Funding Programs. The likely impactof federal funding on crime and its risk factors, especially youth violence,can only be assessed using scientifically recognized standards in the contextof what is known about each of the seven institutions. Chapter One presentsthe scientific basis for this conclusion. Each of the chapters on the seveninstitutional settings concludes with an analysis of the implications ofthe scientific findings for the likely effectiveness of the Departmentof Justice Programs.

o What Works in Each Institution. The available evidencedoes support some conclusions about what works, what doesn't, and what'spromising in each of the seven institutional settings for crime prevention.These conclusions are reported at the end of each of Chapters 3-9. In orderto reach these conclusions, however, the Report uses a relatively low thresholdof the strength of scientific evidence. This threshold is far lowerthan ideal for informing Congressional decisions about billions ofdollars in annual appropriations, and reflect the limitations of the availableevidence.

o Stronger Evaluations. The number and strength of availableevaluations is insufficient for providing adequate guidance to the nationaleffort to reduce serious crime. This knowledge gap can only be filled byCongressional restructuring of the DOJ programs to provide adequate scientificcontrols for careful testing of program effectiveness. DOJ officials currentlylack the authority and funding for strong evaluations of efforts to reduceserious violence.

o Statutory Evaluation Plan. In order to provide the Departmentof Justice with the necessary scientific tools for program evaluations,the statutory plan for evaluating crime prevention requires substantialrevision. Scientifically recognized standards for program evaluations requirestrong controls over the allocation of program funding, in close coordinationwith the collection of relevant data on the content and outcomes of theprograms. The current statutory plan does not permit the necessary levelof either scientific controls on program operations or coordination withdata collection. Funds available for data collection have also been grosslyinadequate in relation to scientific standards for measurement of programimpact.

Chapter Ten presents a statutory plan for accomplishing the Congressionalmandate to evaluate with these elements:

1. Earmark ten percent of all DOJ funding of local assistance forcrime prevention (as defined in this Report) for operational program fundsto be controlled by a central evaluation office within OJP.

2. Authorize the central evaluation office to distribute the tenpercent "evaluated program" funds on the sole criteria of producingrigorous scientific impact evaluations, the results of which can be generalizedto other locations nationwide. Allocating these funds for field testingpurposes simply adds to the total funding for which any local jurisdictionis eligible. Thus the "evaluated program" funding becomes anadditional incentive to cooperate with the scientific evaluation plan ona totally voluntary basis.

3. Set aside an additional ten percent of all DOJ funding of localassistance for crime prevention to support the conduct of scientific evaluationsby the central evaluation office. This recommendation makes clear thetrue expense of using rigorous scientific methods to evaluate program impact.Victimization interviews, offender self-reported offending, systematicobservation of high crime locations, observations of citizen-police interaction,and other methods can all cost as much or more than the program being evaluated.

DEPARTMENT OF JUSTICE FUNDING FOR LOCAL CRIME PREVENTION

Chapter One describes the basic structure and mechanisms for Departmentof Justice FY 1996 funding of State and local governments and communitiesfor assistance in crime prevention. The two major categories are $1.4 billionin funding of local police by the Office for Community-Oriented PolicingServices (COPS), and $1.8 billion in local crime prevention assistancefunding of a wide range of institutions by the Office for Justice Programs(OJP).1 This review examines both the relatively small fundingfor discretionary grants by DOJ, many of which are determined byCongressional "earmarks" to particular grantees and programs,and formula grants, which are distributed to State or local governmentsbased on statutory criteria such as population size or violent crimes.

These are the principal OJP offices administering both types of grants:the Bureau of Justice Assistance administers the $503 million Local LawEnforcement Block Grants, the $475 million Byrne Formula Grants, and the$32 Million in Byrne Discretionary Grants; the Office of Juvenile Justiceand Delinquency Prevention administers the $70 Million Juvenile JusticeFormula Grants, and the $69 Million Competitive Grants; the Violence AgainstWomen Grants Office administers the $130 Million STOP Violence AgainstWomen Formula Grants and $28 Million in Discretionary Grants To EncourageArrests; Corrections Program Office administers a $405 Million FormulaGrants for prison construction and a $27 Million Grants Program for substanceabuse treatment of prison inmates; the Drug Courts Program Office funds$15 Million (from LLEBG) to local drug courts. The Executive Office ofWeed and Seed administers the $28 Million (from Byrne) Federal componentof the Weed and Seed Program in selected high-crime inner-city areas.

SCIENTIFIC STANDARDS FOR PROGRAM EVALUATIONS

The Omnibus Crime Control and Safe Streets Act of 1968 defines an "evaluation"as "the administration and conduct of studies and analyses to determinethe impact and value of a project or program in accomplishing the statutoryobjectives of this chapter."2 By this definition, an evaluationcannot be only a description of the implementation process, or "monitoring"or "auditing" the expenditure of the funds. Such studies canbe very useful for many purposes, including learning how to implement programs.But they cannot show whether a program has succeeded in causing less crime,and if so by what magnitude. Nor can the results be easily generalized.

The scientific standards for inferring causation have been clearly establishedand have been used in other Reports to the Congress to evaluate the strengthof evidence included in each program evaluation. With some variations ineach setting, the authors of the present Report use an adapted versionof scoring system employed in the 1995 National Structured Evaluation bythe Center for Substance Abuse Prevention. The system is used to rate availableevaluations on a "scientific methods score" of 1 through 5. Thescores generally reflect the level of confidence we can place in the evaluation'sconclusions about cause and effect. Chapter Two describes the specificprocedures followed in the application of this 1-5 rating system, as wellas its limitations.

Deciding What Works

The scientific methods scores reflect only the strength of evidenceabout program effects on crime, and not the strength of the effectsthemselves. Due to the general weakness of the available evidence, theReport does not employ a standard method of rating programs according tothe magnitude of their effect size. It focuses on the prior question ofwhether there is reasonable certainty that a program has any beneficialeffect at all in preventing crime. The limitations of the available evidencefor making this classification are discussed in Chapter Two. We note theselimitations as we respond to the mandate for this Report and classify majorlocal crime prevention practices in each institutional setting as follows:

What Works. These are programs that we are reasonably certainprevent crime or reduce risk factors for crime in the kinds of social contextsin which they have been evaluated, and for which the findings should begeneralizable to similar settings in other places and times. Programs codedas "working" by this definition must have at least two level3 evaluations with statistical significance tests and the preponderanceof all available evidence showing effectiveness.

What Doesn't Work. These are programs that we are reasonablycertain fail to prevent crime or reduce risk factors for crime,using the identical scientific criteria used for deciding what works.

What's Promising. These are programs for which the level of certaintyfrom available evidence is too low to support generalizable conclusions,but for which there is some empirical basis for predicting that furtherresearch could support such conclusions. Programs are coded as "promising"if they found effective in at least one level 3 evaluation and thepreponderance of the evidence.

What's Unknown. Any program not classified in one of the threeabove categories is defined as having unknown effects.

EFFECTIVENESS OF LOCAL CRIME PREVENTION PRACTICES

The scientific evidence reviewed focuses on the local crime preventionpractices that are supported by both federal and local, public and privateresources. Conclusions about the scientifically tested effectiveness ofthese practices are organized by the seven local institutional settingsin which these practices operate.

Chapter 3: Community-Based Crime Prevention reviews evaluationsof such practices as community organizing and mobilization against crime,gang violence prevention, community-based mentoring, and after-school recreationprograms.

Chapter 4: Family-Based Crime Prevention reviews evaluationsof such practices as home visitation of families with infants, preschooleducation programs involving parents, parent training for managing troublesomechildren, and programs for preventing family violence, including batteredwomen's shelters and criminal justice programs.

Chapter 5: School-Based Prevention reviews evaluations of suchpractices as DARE, peer-group counseling, gang resistance education, anti-bullyingcampaigns, law-related education, and programs to improve school disciplineand improve social problem-solving skills.

Chapter 6: Labor Markets and Crime Risk Factors reviews evaluationsof the crime prevention effects of training and placement programs forunemployed people, including Job Corps, vocational training for prisoninmates, diversion from court to employment placements, and transportationof inner-city residents to suburban jobs.

Chapter 7: Preventing Crime At Places reviews the available evidenceon the effectiveness of practices to block opportunities for crime at specificlocations like stores, apartment buildings and parking lots, includingsuch measures as cameras, lighting, guards and alarms.

Chapter 8: Policing For Crime Prevention reviews evaluationsof such police practices as directed patrol in crime hot spots, rapid responsetime, foot patrol, neighborhood watch, drug raids, and domestic violencecrackdowns.

Chapter 9: Criminal Justice and Crime Prevention reviews theevidence on such practices as prisoner rehabilitation, mandatory drug treatmentfor convicts, boot camps, shock incarceration, intensively supervised paroleand probation, home confinement and electronic monitoring.

EFFECTIVENESS OF DEPARTMENT OF JUSTICE FUNDING PROGRAMS

DOJ funding supports a wide range of practices in all seven institutionalsettings, although much more so in some than in others. Congress has investedDOJ funding most heavily in police and prisons, with very little supportfor the other institutions. The empirical and theoretical evidence showsthat other settings for crime prevention are also important, especiallyin the small number of urban neighborhoods with high rates of youth violence.Thus the statutory allocation of investments in the crime prevention "portfolio"is lop-sided, and may be missing out on some major dividends.

The effectiveness of existing DOJ funding mechanisms is assessed atthe end of each chapter on local crime prevention practices. The followinglist of major funding programs provides an index to the Chapters in whichspecific practices funded by each of them is discussed:

Community Policing: Chapters 8 and 10.

Local Law Enforcement Block Grant Program: Chapters 3, 7, 8 and10.

Byrne Memorial Formula & Discretionary Grants Program: Chapters3, 4, 5, 6, 8, 10.

Juvenile Justice Formula and Competitive Programs: Chapters 3,4, 5, 8, 9 and 10.

Operation Weed and Seed: Chapters 3, 4, 8 and 10.

STOP Violence Against Women Grants: Chapters 3, 8, and 10.

Grants to Encourage Arrest Policies: Chapters 3, 8 and 10.

Violent Offender Prison Construction: Chapters 9 and 10.

Drug Courts Competitive Grants: Chapters 9 and 10.

CONCLUSION

The great strength of federal funding of local crime prevention is theinnovative strategies it can prompt in cities like New York, Boston, andKansas City (MO) where substantial reductions have recently occurred inhomicide and youth violence. The current limitation of that funding, however,is that it does not allow the nation to learn why some innovationswork, exactly what was done, and how they can be successfullyadapted in other cities. In short, the current statutory plan does notallow DOJ to provide effective guidance to the nation about what worksto prevent crime.

Yet despite the current limitations, DOJ has clearly demonstrated thecontribution it can make by increasing such knowledge. The Department hasalready provided far better guidance to State and local governments onthe effectiveness of all local crime prevention efforts than was availableeven a decade ago. Based on the record to date, only DOJ agencies, andnot the State and local governments, have the available resources and expertiseto produce the kind of generalizable conclusions Congress asked for inthis report. The statutory plan this report recommends would enhance thatrole, and allow DOJ to accomplish the longstanding Congressional mandateto find generally effective programs to combat serious youth violence.By focusing that effort in the concentrated poverty areas where most seriouscrime occurs, the Congress may enable DOJ to reverse the epidemic of violentcrime that has plagued the nation for three decades.

NOTES

1Total FY 1996 funding for the Office of Justice Programswas $2.7 billion, including $228 Million in collections for the Officefor Victims of Crime.

242 U.S.C. Section 3791 (10)

Chapter One

INTRODUCTION: THE CONGRESSIONAL MANDATE TO EVALUATE

by Lawrence W. Sherman

For over three decades, the federal government has provided assistancefor local crime prevention. Most of that assistance has been used to fundoperational services, such as extra police patrols. A small part of thatassistance has been used to evaluate operational services, to learnwhat works--and what doesn't--to prevent crime. Most of the operationalfunding to prevent crime, both federal and local, remains unevaluated byscientific methods (Blumstein et al 1978; Reiss and Roth, 1993).

The Congress has repeatedly stated its commitment to evaluating crimeprevention programs. In the early years of local assistance under the OmnibusCrime Control and Safe Streets Act of 1968, it was "probably the mostevaluation-conscious of all the social programs initiated in the 1960sand 1970s" (Feeley and Sarat, 1980: 130). In 1972, the Congress amendedthe Act to require evaluations of the "demonstrable results"of local assistance grants. In 1988, the Congress generally limited federalassistance under the Anti-Drug Abuse Act Byrne Grants to programs or projectsof "proven effectiveness" or a "record of success"as determined by evaluations.1 But then as now, the Congressionalmandate to evaluate remains unfulfilled, for reasons of funding structureand levels inherent in local assistance legislation for three decades.2

This report responds to the latest in the long line of Congressionalinitiatives to insure that its local assistance funding is effective atpreventing crime. It is a state-of-the-science report on what is known--andwhat is not--about the effectiveness of local crime prevention programsand practices. What is known helps to address the Congressional requestfor a scientific assessment of local programs funded by federal assistance.What is not known helps to address the underlying issue of the Congressionalmandate to evaluate crime prevention, the statutory reasons why that mandateremains unfulfilled, and the scientific basis for a statutory plan to fulfilthe mandate.

The report finds substantial advances in achieving the Congressionalmandate in recent years. The scientific strength of the best evaluationshas improved. The Department of Justice is making far greater use of evaluationresults in planning and designing programs. Within the scope of severelyconstraining statutory limitations, the level of resources the Departmentof Justice has given to evaluation has increased. The 1994 Crime Act alreadycontains piecemeal but useful precedents for a more comprehensive statutoryplan to fulfil the mandate. By asking for this report, the Congress hasopened the door for a major step forward in using the science of programevaluation better to prevent crime. That step is a clearer definition ofwhat "effectiveness" means, and a clearer plan for using impactevaluations to measure effectiveness.

THE MANDATE FOR THIS REPORT

In the 104th United States Congress, the Senate approved a major newapproach to local assistance program evaluation. The Senate bill wouldhave required the Attorney General to "reserve not less than two percent,but not more than three percent of the funds appropriated" for severallocal assistance programs to "conduct a comprehensive evaluation ofthe effectiveness of those programs." This would have been the firststatutory plan to adopt the principle of setting aside a certain percentageof DOJ's operational funds exclusively for program evaluation--aprinciple often endorsed by the same operational leaders from whose fundswould be affected,3 and one which has been adopted for otherfederal agencies.

The House version of the Justice Department's Appropriations bill didnot include the evaluation set-aside plan, so a Conference Committee ofthe two chambers reached an agreement on this point. Rather than fundingevaluations of the three specific programs named in the Senate version,the Conference Committee called for a comprehensive evaluation of the effectivenessof all Justice Department funding of local assistance for crimeprevention. The Committee also required that the review be completed withinnine months after the enactment of the legislation.

On April 27, 1996, the 104th United States Congress enacted the ConferenceReport (See Exhibit 1) requiring the Attorney General to provide an independent,comprehensive and scientific evaluation of the "diverse group of programsfunded by the Department of Justice to assist State and local law enforcementand communities in preventing crime."4 The evaluation wasrequired to focus on the effectiveness of these programs, definedin three ways:

o preventing crime, with special emphasis on youth violence

o reducing risk factors for juvenile violence, including those foundin

  • -community environments
  • -schools
  • -families

o increasing protective factors against crime and delinquency

The legislation specifically required that the evaluation employ "rigorousand scientifically recognized standards and methodologies." In orderto accomplish this task, the Assistant Attorney General for the Officeof Justice Programs directed the National Institute of Justice (NIJ), incoordination with the Bureau of Justice Assistance (BJA), the Office ofJuvenile Justice and Delinquency Prevention (OJJDP), and the ExecutiveOffice of Weed and Seed, to issue a competitive solicitation for proposals.On June 26, 1996, the National Institute of Justice released a solicitationthat began the process of building the framework for this report to achievethe mandate of the 1996 legislation.

Exhibit 1

What Works, What Doesn't, What's Promising (1)

FRAMEWORK FOR THIS REPORT

This chapter presents the broad rationale for the framework used inthis report. It begins with the scientific issues in the choice of theframework, and clarifies what the report is not. It sets the stage forthe review with a brief introduction to the scope and structure of federalfunding of local crime prevention programs. It then returns to the basicchallenge of fulfilling the mandate to evaluate as an integral part ofresponding to the Congressional request for this report. The detailed planfor the rest of the report is then presented in Chapter Two.

Scientific Issues in The Choice of Framework

The 1996 legislation featured four key factors guiding the choice ofmethods for accomplishing the evaluation mandate: its breadth, itstiming, its scientific standards, and its independence.The Justice Department programs in question cover a broad and complex arrayof activities. The short time period for producing the report ruled outany new evaluations of crime prevention effectiveness. Thus the requirementto employ scientific methods clearly implied a synthesis of already completedscientific studies.

The reliance on existing rather than new evaluations is clearly reflectedin the NIJ solicitation, which called for "an evaluation review ofthe effectiveness of broad crime prevention strategies and types of programmaticactivity..[including] family, school, and community-based strategies andapproaches, as well as law-enforcement strategies." The solicitationdefined more specifically how the evaluation was to be conducted:

    It is expected that this evaluation will not conduct new studies orengage in any detailed analysis of existing data. Rather, the evaluationreview and report should draw upon existing research and evaluation studiesand comprehensive syntheses of this work to produce a critical assessmentof the state of knowledge, including its generalizability and its potentialfor replication....Also, the review must explicitly examine the researchin light of the outcome measures specified in the Act as described above.

The Assistant Attorney General decided to award a grant to an independentresearch group to accomplish this mandate. The legislation required thatthe review's content be "independent in nature," even if provided"directly" (by federal employees) or by independent contractorsor grantees. An anonymous panel appointed by NIJ evaluated the proposalssubmitted in response to the solicitation. On the basis of the peer-reviewpanel's report, the Director of the National Institute of Justice selectedthe University of Maryland's Department of Criminology and Criminal Justicein early August, 1996 to conduct the Congressionally mandated evaluationdue on January 27, 1997.

Once the University of Maryland was selected as the independent contractor,the strategic choices for accomplishing the mandate shifted to the teamof six senior scientists who wrote this report. All decisions about theproject were left in the hands of the Maryland criminologists, who bearsole responsibility for the work. That responsibility includes the technicalchoices we made about how to employ "rigorous and scientifically recognizedstandards and methodologies" most effectively in the limited timeavailable to complete the report. The principal decision was to definethe scope of the report as follows:

    a critical assessment, based on a growing body of science, of theeffectiveness of a wide range of crime prevention strategies, operatedat the local level, with and without the support of federal funds.

This report is thus a review of scientific evaluations of categoriesof local programs and practices that are supported by broadcategories of federal funds--often by several different "programs"of funding. Using systematic procedures described in Chapter Two and theappendix, the report attempts to sort the science of local crime preventionprograms and practices supported by DOJ. It focuses primarily on the directevaluation of local program operations, and uses those findings selectivelyto support indirect and theoretical assessments of some national fundingstreams based on findings about their specific parts.

Direct Evaluations of Local Program Operations. What rigorousscience can evaluate most reliably is the effect of a specific programoperated at a local level. This report identifies over 500 studies thatattempt to do just that, with varying levels of scientific rigor. In afew areas, the science is rigorous enough, the studies are numerous enough,and the findings are consistent enough for us to draw some reasonably certainand generalizable conclusions about what works, what doesn't, andwhat is promising at the local level of operation. Such conclusions arenot yet possible for most local crime prevention strategies. That factrequires the report to address the starting point for the legislation mandatingthis report: the need for far greater investment in program evaluation.But the growing OJP support for program evaluation in recent years helpsto provide the raw material for the core of this report.

Indirect Evaluations of National Funding. In an effort to beas responsive to the Congress as possible, this report makes selectiveuse of another approach to the scientific method. That approach uses evaluationsof local programs to make indirect evaluations of federal funding streams.Those streams vary widely in their diversity, from funding streams of suchrelatively uniform programs as the hiring of the Crime Act's 100,000 policeto very diverse Local Law Enforcement Block Grants program. The extentto which it is scientifically appropriate to generalize upwards from localprogram evaluations to national funding streams varies as well. In general,the more homogeneous the federal funding stream, the more appropriate itis to evaluate the effectiveness of that funding based on local evaluations.

Theoretical Assessments of Unevaluated Programs. Where no rigorouslyscientific impact data are available on funding streams expending substantialtax dollars, the report employs theoretical analyses to provide limitedassessments of the programs. A prime example is the numerous efforts thatOJP is currently making to prevent crime in the concentrated urban ghettopoverty areas producing the majority of serious youth violence in America.These programs attempt to be comprehensive in addressing the crime riskfactors in those areas, which allows a comparison of the program contentto the available theory and data on risk factors. The need for scientificimpact assessments of these programs, however, is critical, and the theoreticalassessment should be seen merely as a stopgap approach required by thecurrent lack of measured effects.

Comprehensiveness

This report attempts to be as comprehensive as the available scienceallows. It is not, however, an annotated list of DOJ local assistance programswith a summary of scientific evidence relating to each one. Such an encyclopedicapproach would have several limitations. It would fail to identify importantissues cutting across programs. It would fail to give greater attentionto the more important crime risk factors identified in the literature.Most important, it would have nothing to say about a great proportion ofthe specific program components of DOJ local assistance programs, giventhe lack of available impact evaluations.

While the report attempts some form of scientific commentary for themajor DOJ prevention funding streams, it omits direct commentary on manyof the smaller diverse funding categories. We attempt not to omit, however,any published program impact evaluations, meeting minimal standards ofscientific rigor, that help show indirectly the effectiveness of the DOJprograms. Where such omissions have occurred, we anticipate that can becorrected in a systematic effort to keep the present findings up to datein future years.

What This Report Is Not

The Congressional mandate did not require that this report include anaudit of the use of Department Of Justice (DOJ) funds, an evaluation ofthe leadership of DOJ's Office of Justice Programs (OJP) or Community OrientedPolice Services (COPS) office, or a process or descriptive evaluation ofspecific programs at the local level supported with DOJ funds. None ofthese tasks fall within the required assessment of the scientific evidenceof the effectiveness of local assistance funds administered by DOJ in preventingcrime and risk factors.

Not an Audit of DOJ. Congress did not require the Attorney Generalto provide a detailed accounting of how DOJ local assistance funds arebeing spent. That kind of analysis requires auditing rather than scientificmethodologies; the legislation clearly indicated the use of science. Knowingexactly how much money is being spent on Drug Courts, for example, doesnot alter the conclusions that can be reached by using scientific methodsto examine the available studies of the effectiveness of drug courts. Thereport's concern with the expenditure of DOJ funds was limited to fourquestions that informed a scientific assessment:

1) Does DOJ funding support this kind of crime prevention program orpractice?

2) If not, does the scientific evidence suggest Congress should considerfunding it?

3) Are current funds allocated in relation to scientifically establishedcrime risk factors?

4) Have the funds been allocated in a way that permits scientific impactevaluation?

Not an Evaluation of DOJ Leadership. The term "evaluation"is often understood to mean something like a report card, reflecting onthe personal effectiveness of officials directing programs. There is evena substantial scientific literature in the field of industrial psychologyfor personnel or performance "evaluation" systems. The legislationclearly does not call for a performance evaluation, but for an evaluationof program effectiveness. The Congressional mandate to focus on the scienceof the programs does not require assessments, positive or negative, aboutthe performance of DOJ leadership. In order to standardize the focus onthe evidence, the report does not even employ interviews with DOJ leadership,and relies solely on analysis of legislation, written documents and publicationsabout the programs they administer.

Not A Descriptive or Process Evaluation of DOJ Programs. TheCongressional mandate clearly focuses on what scientists call "impact"evaluations, rather than "descriptive" or "process"evaluations. The distinction between the two kinds of evaluation is critical,but often misunderstood. Descriptive or process evaluations describe thenature of a program activity, usually in some detail. An impact evaluationuses scientific methods to test the theory that a program causes a givenresult or effect. Only an impact evaluation, therefore, can be used toassess the "effectiveness" of a program. Descriptive evaluationscan provide useful data for interpreting impact results based on variationsin the implementation of programs and interpretations of their effects.But they do not provide a sufficient response to the Congressional mandate.

Not a Technical "Meta-Analysis." Scientists are makingincreasing use of a statistical methodology called "meta-analysis,"in which findings from many studies are analyzed together quantitatively.This method is important because it can produce different conclusions thana summary of findings from individual studies, largely by increasing thesample size available for analysis. There are no currently published statisticalmeta-analyses comparing the effectiveness of the full array of crime preventionstrategies, from Head Start to prisons. There are several meta-analyseson specific crime prevention strategies included in the evidence used forthis report. The Congressional requirements for rapid production of thisreport, however, ruled out a formal meta-analysis of the evaluation resultsacross all crime prevention programs, however.

Evaluating Funding Mechanisms Versus Prevention Programs

The legislation did not define DOJ crime prevention "programs"as the large general funding streams. The focus on effectiveness clearlydirects the report to specific crime prevention strategies. A substantialscientific literature is available on the crime prevention effectivenessof the specific strategies. We could find no existing impact evaluation,however, of such general funding streams as the Byrne Memorial State andLocal Law Enforcement Assistance Program. This fact raises several keyissues: the definition of "programs," the science of varyingtreatments, and the barriers such variations raise to direct evaluationof internally diverse national funding streams.

Defining "Programs." A major source of confusion inpolicy analysis of federal crime prevention is the meaning of the word"program." The meanings vary on several dimensions. One dimensionis the level of government: if the federal Byrne Program funds a neighborhoodwatch program in Baltimore, which one is the DOJ "program" thisreport should evaluate for the Congress: Byrne or Baltimore's neighborhoodwatch? Or should the evaluation focus fall in between those two levelsof analysis, addressing what is known generally about neighborhood watchprograms? This report takes the latter approach.

The meanings of the term "program" also vary with respectto the required degree of internal uniformity. Neighborhood watch "programs,"for example, are fairly uniform in their content, despite some variations.A national community policing "program," in contrast, embracesa far wider range of activities and philosophies, ranging from aggressivezero tolerance enforcement campaigns "fixing broken windows"(Kelling and Coles, 1996) to outreach programs building partnerships betweenpolice and all segments of the community (Skogan, 1990).

Science and Varying Treatments. The tools of the scientific methodare only as useful as the precision of the questions they answer. Medicalscience, for example, evaluates the effectiveness of specific treatments;it is rarely able to establish the controls needed to evaluate broad categoriesof funding embracing multiple or varying treatments, such as "hospitals"or even "antibiotics." Variations in treatment place major limitationson the capacity of science to reach valid conclusions about cause and effect.The scientific study of aspirin, for example, assumes that all aspirinhas identical chemical components; violating that assumption in any givenstudy clearly weakens the science of aspirin effectiveness. The same istrue of crime prevention programs. The more a single program varies inits content, the less power science has to draw any conclusions about "the"program's content (Cohen, 1977; Weisburd, 1993).

Compare a study of the effects of a sample of 5,000 men taking aspirinto a study of the same sample taking different pills elected arbitrarilyfrom an entire pharmacy of choices. Any changes in health would be moreclearly understood with the aspirin study than with the pharmacy evaluation.Even if the whole pharmacy of pills were taken only on doctor's orders,based on a professional assessment of the most appropriate pills for eachpatient, wrapping all of the different pills' effects into the same evaluationof effectiveness would prevent an assessment of what effect each medicinehad. Science is far more effective at evaluating one kind of pill at atime than in drawing conclusions about different pills based upon a pharmacyevaluation.

Direct Evaluations of National Funding Programs. Any attemptto evaluate directly an internally diverse national funding program iscomparable to a pharmacy evaluation. Even if the right preventive treatmentsare matched to the right crime risks, a national before-and-after evaluationof a funding stream would lack vital elements of the scientific method.The lack of a control group makes it impossible to eliminate alternativetheories about why national-level crime rates changed, if at all, withthe introduction of a widely diverse national program like the Local LawEnforcement Block Grant. Federal funding of local crime prevention, forexample, increased by over five hundred percent from 1994 to 1996, andviolent crime has fallen steadily during that period. But violent crimestarted falling in 1992, for reasons that no criminologist can isolatescientifically. Isolating still further the effects of the increased fundingin 1994 is not possible to do with rigorous scientific methods. Thus wecould not have evaluated most national DOJ funding programs directly, evenif we had been allowed several years or decades.

Implications of This Approach

The choice to start with the available science on local programs ratherthan the DOJ funding mechanism programs has important implications. Onelimitation is the report's unavoidable bias towards well-researched programs.One advantage is that the report becomes a reference source for differentlegislative approaches to federal funding. The approach also becomes ademonstration of how unevenly evaluation science can proceed, and the needfor clear distinctions between science and policy analysis.

Bias Towards Well-Researched Programs. The report clearly emphasizesstrategies that have received substantial research attention, regardlessof their merits in receiving that attention. To the extent that the rigorousscience has been focused on less promising crime prevention strategies,both the report and public policymaking are at a disadvantage. The alternativemight have been to rely more on theoretical science and less on empiricalresults. The obvious danger in that course, however, is a risk of losingthe objectivity required for reliable assessments. On balance, then, thedecision to focus on the strongest scientific evidence seems to be themost useful and least problematic approach available.

A Reference for Diverse Approaches to Federal Funding. Lettingscience guide the report around local programs may help the findings tohave more lasting value. Organizing the evidence around theories and datawill provide a reference for many different possible approaches to federalfunding of local programs. While the structure of federal funding changesalmost annually, the results of program evaluations accumulate steadilyover long time periods. While the NIJ solicitation asked for special emphasisto be placed on evaluations completed in the last five years, many of themost important evaluation results are older than that. Omitting those earlierstudies from the analysis would have substantially and inappropriatelyaltered the conclusions reached. Similarly, Congressional deliberationson crime prevention policy can benefit from a reference source organizedaround the basic institutional settings for local crime prevention: communities,families, schools, labor markets, specific places, police, and criminaljustice.

The Uncertainty of Science. Guiding the report with availablefindings offers a more realistic picture of what evaluation science isable to achieve. As the U.S. Supreme Court recently concluded, hypothesesabout cause and effect cannot be "proven" conclusively like ajury verdict; they can merely be falsified using a wide array of methodsthat are more or less likely to be accurate.5 A Nobel Laureateobserves that "Scientists know that questions are not settled; rather,they are given provisional answers..."6 Science is a constantstate of double jeopardy, with repeated trials often reaching contradictoryresults. Fulfilling the mandate to evaluate will always result in an unevengrowth of evaluation results, not permanent guidance. This report directlyconfronts the problems of mixed results from methods of varying scientificrigor, and attempts to develop decision rules for applying the findingsto both research and program policy. These rules may have value not justfor this report. They may also help advance the Congressional mandate toevaluate beyond the nonscientific concept of "proven" effectivenessto the scientific concept of "likely" effectiveness.

This problem of accurately predicting the effects of a program whereverit may be implemented is an important limitation to using evaluations inpolicy analysis. Generalizing results from an evaluation in one city tothe effects of a program in another city is a very uncertain enterprise.We still lack good theories and research to predict accurately when findingscan be accurately generalized. Just as the Justice Department may funddifferent kinds of community policing programs, the same program may bevery different in different places. The nature of a "drug court"may vary enormously from one judge to the next, community policing homevisits may vary from friendly to intrusive, gang prevention programs mayhave different effects in different kinds of neighborhoods or ethnic groups.This uncertainty is best acknowledged, and addressed by ongoing evaluationsof even programs with enough evidence to be judged "likely" to"work."

Science Versus Policy Analysis. The focus on scientific resultsshould help the reader distinguish between the report's science and itspolicy analysis. The distinction is crucial. Even though scientific evaluationresults are a key part of rational policy analysis, those results cannotautomatically select the best policy. This is due not just to the scientificlimitations of generalizing results from one setting to the next. Anotherreason is that evaluations often omit key data on cost-benefit ratios;the fact that a program is "effective" may be irrelevant if thefinancial or social costs are too high. This report attempts, where possible,to distinguish summaries of science from their application to policy issuesusing judgment and other sources of information outside the evaluationresults. We expect that there will be less consensus about the policy analysisthan about the scientific findings. But we also determined after extensivedeliberation that recommendations based on policy analysis were a usefuladdition to the purely scientific summaries that form the core of the report.

The framework adopted for this report is not the only possible way tohave responded to the Congressional request. There are legitimate differencesof opinion about how best to use scientific methods for this kind of analysis.Some analysts have argued for a more "flexible" approach to programevaluation, with more emphasis on expert insight and less emphasis on whethera program "works" (Pawson and Tilley, 1994). Others call forless reliance on evaluation results that have less rigorous measurementof program context and other data needed to assess the generalizabilityof results (Ekblom and Pease, 1995). Our own preference would have beento raise the cutoff point for defining "scientific" methods muchhigher than we actually did (see Chapter Two). On balance, however, thisapproach provides an acceptable compromise between the Congressional needsfor information and the scientific strength of available evidence.

There are also multiple goals for the $4 Billion annual funding describedin this report, which may be valuable for other reasons besides its scientificallymeasurable effectiveness in preventing crime. The focus on crime preventionexcludes the very important goals of justice, fairness and equality underthe law. That limitation is not inherent in the science of program evaluation;it is merely a function of the boundaries of the specific mandate for thisreport.

LOCAL CRIME PREVENTION AND THE DEPARTMENT OF JUSTICE

The policy context for this report is the current structure of localcrime prevention assistance programs funded by the U.S. Department of Justice.This section provides a brief introduction to those programs. It beginswith a summary of the appropriated budgets for local crime prevention infiscal year 1996, the year the Congress requested this report. It thendescribes the administrative structure of the Justice Department officesadministering those funds. It concludes with a brief discussion of thetypes of funding mechanisms Congress has created for distributing the funding,and briefly details the focus and mechanisms of the largest of the fundingprograms.7

Budget

Local crime prevention offices now receive more DOJ funding than atany time in American history, a larger budget than the FBI, the DEA, orthe INS. Among all DOJ components, only the Federal Bureau of Prisons consumesa larger share of the budget. At $4 billion per year, the combined annualbudget of the $1.4 billion administered by the Director of the COPS (Community-OrientedPolicing Services) Office and the $2.6 Billion administered by the AssistantAttorney General for OJP (the Office of Justice Programs) is more thanfive times the amount the Congress allocated in the peak years of the oldLaw Enforcement Assistance Administration.

Not all of these funds can be classified as having crime preventionpurposes. The largest of these programs, the 1994 Crime Act's Title I CommunityPolicing grants, does not even specify the prevention of youth violenceas a legislative purpose of the funding, even though many observers wouldexpect youth violence prevention to result from the program. Thedefinition of crime prevention as an intention or a result is a major issueaddressed in Chapter Two, which explains this report's rationale in usinga definition focused on results. This definition thus clearly include the100,000 police. But even that broad definition does not include the State$300 State Criminal Alien Assistance Program, reimbursing states for housing38,000 illegal aliens incarcerated for felony offenses, or the $31 millionPublic Safety Officers Benefits program for families of police slain inthe line of duty. Nor does it include infrastructure programs for courtsand computerization of criminal justice records, general programs of statistics,research and evaluation, services to victims of crime, the Police Corps,or general administrative costs. As Figure 1-1 shows, the major crime preventionfunding programs within DOJ added up to about 85% of the $4 billion totalappropriations for the two local assistance offices (OJP and COPS), orabout $3.4 billion. The historical context of these appropriations levelsis indicated in Figure 1-2, which shows the three-decade trends in totalDOJ funding of its local crime prevention assistance offices (includingservices other than crime prevention).

The Department of Justice funding of local programs which may resultin crime prevention are authorized under several different Acts of Congress.The Juvenile Justice and Delinquency Prevention Act is the oldest, havingcontinued in force after the end of the Law Enforcement Assistance Administration.The 1988 Anti-Drug Abuse Act of 1988 authorized the Byrne Grants programto the states, followed by the 1994 Crime Act which took the local preventionfunding to its current historic heights. The five principal titles of the1994 Act include Public Safety and Policing (Title I), Prisons (Title II),Crime Prevention (Title III), Violence Against Women (Title IV), and DrugCourts (Title V). While this report treats all five titles as falling withina results-based scientific definition of crime prevention, it is worthnoting that the Congress has never appropriated any funds specificallylabeled as "crime prevention" under Title III. Both the1988 Anti-Drug Abuse Act and the 1996 Omnibus Appropriations Act, however,appropriated funds allowing grants to be made in a "purpose area"labeled crime prevention.

Figure 1-1

Major DOJ Crime Prevention Funding Programs

OFFICE & BUREAU FUNDING PROGRAMS FY 1996 Funding Community-Oriented Policing 100,000 Local Police $1.4 Billion Services Office of Justice Programs Bureau of Justice Local Law Enforcement Block Grant $488 Million Assistance Formula Program Byrne Memorial State and Local Law $475 Million Enforcement Assistance Formula Program Byrne Discretionary Grants Program: $32 Million (Boys and Girls Clubs Earmark) ($ 4 Million) (Nat'l. Crime Prevention Council ($ 3 Million) Earmark) (DARE Drug Abuse Prevention Earmark) ($ 2 Million) Office of Juvenile Justice Juvenile Justice Formula Grant Program $70 Million and Delinquency Prevention Competitive Grants Programs $69 Million Executive Office of Weed Operation Weed and Seed $28 Million and Seed Violence Against Women STOP (Services, Training, Officers, and $130 Million Grants Office Prosecution) Violence Against Women Formula Grant Program Rural Domestic Violence Enforcement $ 7 Million Encourage Arrest Program $ 28 Million Corrections Program Office Residential Substance Abuse Treatment $ 27 Million Violent Offender Truth in Sentencing $405 Million Prison Construction Formula Grants Drug Courts Program Office Drug Courts Competitive Grants $ 15 Million Total Major Funding $3.2 Billion 

Administrative Structure

The administration of these various programs under various Acts is organizedinto the two separate offices. One of these--the Office of Community-OrientedPolicing Services--has a single large program and a single presidentialappointee. The other--the Office of Justice Programs--has numerous programsranging widely in size, managed by an Assistant Attorney General, two DeputyAssistant Attorneys General, and five Presidentially appointed directorsor administrators of the following units the Bureau of Justice Assistance(BJA), the Bureau of Justice Statistics (BJS), the National Institute ofJustice (NIJ), the Office of Juvenile Justice and Delinquency Prevention(OJJDP), and the Office for Victims of Crime (OVC). In addition, severalother OJP offices manage funding under separate Titles of the 1994 CrimeAct: the Corrections Programs Office, the Office for Drug Courts, and theViolence Against Women Grants Office. The OJP Executive Office of Weedand Seed is supported by transfers of BJA Byrne Discretionary Grant appropriationsunder the 1988 Anti-Drug Abuse Act. Figure 1-1 summarizes the administrativeand programmatic structure of the agencies administering the major localcrime prevention programs. NIJ and BJS do not administer major local assistancegrants for crime prevention purposes, although BJS does assist states intheir implementation of the data systems requirements for compliance withthe Brady Act. The Office of Vicitms of Crime is funded by fines collectedby federal courts, and provides funding mostly for repairing the harm cuasedby crime; a few areas of potential crime prevention effects from OVC funding,such as its support for battered women's shelters, are noted in ChapterFour.

Funding Mechanisms: Formula, Discretionary, Earmarks, Competitive

The crucial point in understanding DOJ local crime prevention fundingprograms is the statutory plan for allocating the funding. The "fundingmechanisms" of this plan vary across the different authorization Acts,and use different criteria even within each funding mechanism dependingon the specific Act. Two basic types of funding mechanisms are "formula"or "block" grants versus "discretionary" grants. Manyobservers and grant recipients incorrectly assume these labels mean thatlocal units are entitled to their funding under formula grants, while DOJexecutives decide how to administer the discretionary grants. That assumptionis incorrect. There are substantial legislative requirements constrainingDOJ's allocation of "discretionary" funds, and there are alsovarious legislative requirements that grantees must satisfy in order tobecome eligible to receive their "formula" funding.

The so-called Discretionary programs are constrained by Congress inthree ways: earmarks, eligibility criteria, and competition. Earmarksare legislative directions in the Appropriations laws (as distinct fromAuthorization Acts) on how to spend certain portions of funds appropriatedwithin a larger funding program, such as the $11 million earmark for Boy'sand Girls Clubs within the 1996 appropriation for the BJA Local Law EnforcementBlock Grant Program and the $4.35 Million earmark for the same organizationunder the Byrne Discretionary grants. Earmarks are both "hard"and "soft." Hard earmarks are written into legislation, usuallywith specific amounts to be spent and the specific recipient of the fundingidentified. Soft earmarks are based upon committee hearings and conferencereports, such as the legislation for the present report, with or withoutspecified amounts.

Eligibility criteria programs are only "discretionary"in the sense that DOJ officials must decide whether the applicants areeligible to receive the funds for which they apply. The applicants do notreceive the funds unless they apply, and can demonstrate their eligibilityin the application. Congress often requires, for example, that states passcertain state laws as a condition of eligibility for receiving federalfunds under certain grant programs. The most famous example is perhapsthe limitation of maximum state speed limits to 55 miles per hour thatwas for two decades an eligibility requirement for receiving federal highwayconstruction funding. Similarly, the 1994 Crime Act makes state passageof "Truth-in-Sentencing" Legislation an eligibility requirementfor prison construction grants. Once DOJ has proof of program eligibility,however, the determination of how much funding the applicant receives mustfollow the statutory allocation plan. All those receiving funds do so onthe basis of a "formula" that may be based on population, crimerates, prison overcrowding rates or other factors. In addition, certainminimum amounts are often reserved for jurisdictions of certain size irrespectiveof the formula, such as the requirement that half of all funding for the100,000 police be allocated to applicants from cities of over 150,000 people.In that particular case, the allocation is made at least in part on a first-come,first served basis.8 Thus a more accurate label for such fundingmechanisms might be "discretionary eligibility formula grants."

Only ten percent of the total OJP appropriation is for competitivegrants, the truly discretionary programs in which applicants must competeon the merits of issues other than simple eligibility for funding. DOJofficials usually establish different criteria appropriate for each program.Examples of criteria for these grants include innovative approaches, interagencycollaboration, comprehensive targeting of crime risk factors, and potentialimpact of the program on the community. Examples of competitive local assistanceprograms include Drug Courts, Operation Weed and Seed, JUMP mentoring grantsand Encourage Arrest Grants.

Formula grant programs, in contrast to discretionary programs,have no so-called "eligibility" requirements, such as the passageof state laws. The allocation of funding is independent of such tests.Formula programs can, however, require that certain paperwork be satisfactorilycompleted. BJA Byrne grants, for example, require that an annual plan specifyhow the formula-determined allocation will be spent, and that evaluationsof all grants made with formula allocations be forwarded to BJA. Failureto satisfy these requirements presumably has the same effect as in "discretionaryeligibility" programs, which is to block the award of the funds.

These funding mechanisms offer relatively little discretion to DOJ inits choice of program areas or sites, but offers substantial directionto the state and local grant recipients. That policy choice is centralto a continuing Congressional debate. Its relevance to this report is toshow the centrality of the local programs chosen by the grant recipientsin determining the effectiveness of this funding. It is the local decisionson which prevention programs to adopt, and not the Congressionally mandatedactions by DOJ in allocating that funding, which largely determine theeffectiveness of these broad funding streams in preventing crime.

Major Funding Stream Programs

This section briefly describes the major DOJ funding stream programslisted in Figure 1-1.

COPS. This program reimburses local police agencies for up to75% of the salary and benefits of an additional police officer for threeyears, up to a maximum of $75,000 per officer. It is a discretionary-eligibility-formulagrant program in which funding is allocated on the basis of eligible applicantpopulation size, with a minimum allocation requirement that 50 percentof the funds go to police departments serving cities of over 150,000 people.In addition to this "Universal Hiring Program" to which the Congresshas restricted appropriations in 1997, the earlier years of the programoffered various competitive grant programs for domestic violence, youthfirearms, anti-gang initiatives, and other special purposes.

Byrne (BJA). The 1988 Anti-Drug Abuse Act established both formulaand discretionary grant programs in memory of New York City Police OfficerEdward Byrne, who was murdered while monitoring a crack house. The formulaprogram awards funds to states developing plans for allocating grants,originally under 21 and now under 26 purpose areas: 1) drug demand reductionprograms involving police, 2) multijurisdictional task forces against drugs,3) domestic drug factory targeting, 4) community crime prevention, 5) anti-fencingprograms, 6) white-collar and organized crime enforcement, 7) law enforcementeffectiveness techniques, 8) career criminal prosecution, 9) financialinvestigations, 10) court effectiveness, 11) correctional effectiveness,12) prison industries, 13) offender drug treatment, 14) victim-witnessassistance, 15) drug control technology, 16) innovative enforcement, 17)public housing drug markets, 18) domestic violence, 19) evaluations ofdrug control programs, 20) alternatives to incarceration, 21) urban enforcementof street drug sales, 22) DWI prosecution, 23) juvenile violence prosecution,24) gang prevention and enforcement, 25) DNA analysis, 26) death penaltylitigation. While each state is eligible to receive a minimum of 0.25 percentof total appropriations, the balance is allocated on the basis of statepopulation as a proportion of the entire U.S. All Byrne funds must be matchedby a 25% commitment of non-federal funds.

The BJA Byrne Discretionary Grants program is heavily earmarked forinitiatives such as those indicated in Figure 1-1 (e.g., Boys and GirlsClubs, DARE) as well as programs well-established with Congressional understanding,such as Weed and Seed (see below). Over 5 percent of Byrne discretionaryfunds ($3.1 million) went to program evaluation purposes in FY 1996, withanother $3.5 million allocated to program evaluation by the States fromtheir formula grants.

Local Law Enforcement Block Grants (BJA). This is a formula grantprogram that awards funds to applying local governments based on theirshare of the their state's total Part I violent offenses (homicide, rape,robbery, aggravated assault) over the previous three years. The eight purposeareas for local expenditure of the grants are 1) police hiring, 2) policeovertime, 3) police equipment and technology, 4) school security measures,5) drug courts, 6) violent offender prosecution, 7) multijurisdictionaltask forces, community crime prevention programs involving police-communitycollaboration.

STOP Violence Against Women Block Grants (VAWGO). This is a formulagrant program allocating funding to states and territories based upon population.Within each state, the grants must total at least 25% for law enforcement,prosecution, and victim services. A wide range of programs fall withineach of these categories, including both domestic and stranger violenceagainst women.

Encourage Arrest Grants (VAWGO). This is a competitive programfor which eligibility is determined by the passage of certain state lawsconcerning the arrest of suspects about whom there is probable cause tobelieve they have committed an act of domestic violence or a related offense.These grants are intended to encourage communities to adopt innovative,coordinated practices that foster collaboration among law enforcement officers,prosecutors, judges, and victim advocates to improve the response to domesticviolence.

Operation Weed and Seed (EOWS). This is a competitiveprogram funded by a transfer of BJA discretionary Byrne funding to theOJP Executive Office of Weed and Seed. The program consists of long-termfunding to a varying number of selected cities to help them create a comprehensiveprogram of reducing crime in small, high-crime areas. The DOJ funding operatesas seed money leveraging additional federal, state, local and private resources.

Juvenile Justice Formula Grants (OJJDP). This program providesannual funding to eligible states to deinstitutionalize status offenders,separate juveniles and adults in secure correctional facilities, jailsand lockups, and to reduce the number of juveniles in secure facilities.

Prison Construction Grants (Corrections Office). This programprovides funds to states to build more prison cells or to construct lessexpensive space for nonviolent offenders, to free space in secure facilitiesfor more violent offenders.

Residential Correctional Drug Abuse Treatment (CorrectionsOffice). This funding program funds state prison delivery of substanceabuse treatment to inmates.

THE STATUTORY PLAN FOR PROGRAM IMPACT EVALUATION

In theory, one of the most effective federal crime prevention programsis the evaluation of local programs. The Attorney General's Task Forceon Violent Crime called it the central role of the federal government infighting crime, the one function that could not be financed or performedas efficiently at the local level.9 With less than one percentof local criminal justice budgets supported by the federal government (notcounting the COPS program), federal funds are arguably most useful as astimulus to innovation that makes the use of local tax dollars more effective(Dunworth, et al, 1997). The three-decade old Congressional mandate toevaluate is consistent with that premise. Its implication is that a centralpurpose of federal funding of operations is to provide strong evaluations.

The Congressional mandate for this report therefore includes an evaluationof the effectiveness of DOJ-funded program evaluation itself. The centralquestion is whether those evaluations have "worked" as a federalstrategy for assisting local crime prevention. The report answers thatquestion in a different fashion from the method used to evaluate the directlocal assistance funding. Rather than directly evaluating the impact ofprogram evaluations on crime, the report indirectly examines the antecedentquestion of whether those evaluations have succeeded in producing publishedand publicly accessible scientific findings about what works to preventcrime. After presenting the scientific framework for the review in Chaptertwo, the report presents the evidence for both program and evaluation effectivenessin Chapters Three through Nine. Chapter Ten then summarizes the limitedevidence on local program effects, and returns to the underlying issueof how to accomplish the Congressional Mandate to evaluate.

This report concludes that the current statutory plan for accomplishingthat mandate is inadequate, for scientific reasons not addressed by currentlegislation. That inadequacy substantially limits the capacity to judgethe effectiveness of the federal effort to reduce serious crime and youthviolence. Part of the statutory problem is simply inadequate funding. WhileFigure 1-2 shows the steep rise in total federal support for local crimeprevention operations, Figure 1-3 shows a rough indication of the decliningproportionate support for research and evaluation: the percentage of totalOJP appropriations allocated to the National Institute of Justice.

What Works, What Doesn't, What's Promising (2)

What Works, What Doesn't, What's Promising (3)

Figure 1-3 actually overstates the amount of DOJ funding allocated toprogram evaluations. Program evaluations are also funded by OJJDP and BJA,10and actual NIJ expenditure in FY 1996 was $99 million rather than $30 (dueto inter-agency transfers).11 But Figure 1-3 reflects the totalNIJ budget for all research, technical assistance, and dissemination purposes,as a well as for program evaluation; only 27 percent ($8 million) of NIJ'sFY 1996 appropriation was allocated to evaluation. The proportionate allocationof the NIJ budget to evaluation over the past three decades has not changedsubstantially on this point. Thus while Figure 1-3 overstates the absolutedollars DOJ has been appropriated for evaluation, it is still an accurateportrayal of the absence of statutory attention to keeping evaluation fundingcommensurate with operational funding.

Evaluation funding alone, however, cannot increase the strength of scientificevidence about the effects of federally funded local programs on crime.Chapter Ten documents the need for adequate scientific controls on theexpenditures of program funds in ways that allow careful impact evaluation.A statutory plan earmarking a portion of operational funds for strong scientificprogram evaluation is the only apparent means for increasing the effectivenessof federal funding with better program evaluations. The basis for thisconclusion is central to scientific thinking about crime prevention, asthe next chapter shows.

NOTES

142 U.S.C. 3782 Sec. 801 (b) (1), (19), (20).

2U.S. Attorney General's Task Force on Violent Crime, 1981,p. 73.

3In 1988, for example, more than 30 big city police chiefsasked Congress to earmark ten percent of the Anti-Drug Abuse Act fundsfor research and evaluation. While Titles I and II of the 1994 Crime Actauthorize DOJ to spend up to 3 percent of funds for assorted purposes includingevaluation, there has never been a requirement to spend a percentage ofoperational funds exclusively on program impact evaluations demonstratingcrime prevention effectiveness.

4104th Congress, First Session, House of Representatives,Report 104-378.

5Daubert v. Merrell Dow Pharmaceuticals, Inc., 113 S. Ct.2786, 125 L. Ed. 2d 469 (1993), in which the Court adopts the scientificframework offered by Karl Popper, Conjectures and Refutations: The Growthof Scientific Knowledge, 5th Ed., 1989.

6David Baltimore, "Philosophical Differences,"THE NEW YORKER, January 27, 1997, p. 8.

7This section is based largely upon a January 17, 1997 NIJbackground memorandum from Jane Wiseman to Christy Visher prepared at theUniversity of Maryland's request.

8U.S. Department of Justice, Office of Community OrientedPolicing Services, COPS Facts: "Cops More '96." Update September18, 1996.

9Attorney General's Task Force on Violent Crime, Report,1981; James Q. Wilson, "What, if Anything, Can the Federal GovernmentDo About Crime?" Presentation in the Lecture Series on Perspectiveson Crime and Justice, sponsored by the National Institute of Justice withsupport from the Edna McConnell Clark Foundation, December, 1996.

10Total BJA expenditures on program evaluation in FY 1996were $6.6 million.

11Actual NIJ expenditures on all purposes included transfersauthorized by the Assistant Attorney General for the Office of JusticePrograms from Crime Act appropriations of $15.6 million in FY 1995 and$51.9 million in FY 1996.

REFERENCES

Blumstein, Alfred, Cohen, Jacqueline, and Daniel Nagin (eds).

1978 Deterrence and Incapacitation: Estimating The Effects of CriminalSanctions on Crime Rates. Washington, DC: National Academy of Sciences.

Cohen, J.

1977 Statistical Power Analysis for the Behavioral Sciences. N.Y.: AcademicPress.

Ekblom, Paul and Ken Pease

1995 Evaluating Crime Prevention. In Michael Tonry and David Farrington,eds., Building a Safer Society: Strategic Approaches to Crime Prevention.Crime and Justice, Vol. 19. Chicago: University of Chicago Press.

Feeley, Malcolm and Austin Sarat

1980 The Policy Dilemma: Federal Crime Policy and the Law EnforcementAssistance Administration. Minneapolis: University of Minnesota Press.

Kelling, George and Katharine Coles

1996 Fixing Broken Windows. NY: Free Press.

Pawson, R. and N. Tilley

1994 What Works in Evaluation Research. British Journal of Criminology34: 291-306.

Reiss, Albert J., Jr. and Jeffrey Roth (eds.)

1993 Understanding and Preventing Violence. Washington, D.C.: NationalAcademy of Sciences.

Skogan, Wesley

1990 Disorder and Decline. NY: Free Press.

Weisburd, David with Anthony Petrosino and Gail Mason

1993 Design Sensitivity in Criminal Justice Experiments: Reassessingthe Relationship Between Sample Size and Statistical Power. In MichaelTonry and Norval Morris, eds., Crime and Justice, Vol. 17. Chicago: Universityof Chicago Press.

Chapter Two

THINKING ABOUT CRIME PREVENTION

Lawrence W. Sherman

How effective at preventing crime are local programs with funding fromthe US Department of Justice? That question can only be answered in thecontext of a comprehensive scientific assessment of crime prevention inAmerica. That assessment shows that most crime prevention results fromthe web of institutional settings of human development and daily life.These institutions include communities, families, schools, labor marketsand places, as well as the legal institutions of policing and criminaljustice. The vast majority of resources for sustaining those institutionscomes from private initiative and local tax dollars. The resources contributedto these efforts by the federal government are almost negligible in comparison.The potential impact on local crime prevention of federally supported researchand program development, however, is enormous.

The logical starting point for assessing the current and potential impactof federal programs is the scientific evidence for the effectiveness ofcrime prevention practices in each institutional setting. This requires,in turn, great attention to the enormous variation in the strength of scientificevidence on each specific practice or program. In general, far too littleis known about the impact of crime prevention practices, regardless ofhow they are funded. But thanks largely to evaluations sponsored by theNational Institute of Justice (NIJ), the Office of Juvenile Justice andDelinquency Prevention (OJJDP) and other federal agencies, the body ofscientific evidence has grown much stronger in the past two decades. Mostimportant, it has shown a steadily increasing capacity to provide verystrong scientific evidence, even while most program evaluations remainso weak as to be scientifically useless.

The growing scientific evidence that federal support has produced allowsus to assess some programs more intensively than others. Some of the evidenceis strong enough to identify some effective and ineffective practices orprograms in most institutional settings. Some evidence is more limited,but clearly points to some promising initiatives that merit further researchand development. Reviewing this evidence in each of the seven institutionalsettings provides the strongest possible scientific basis for respondingto the Congressional mandate. By separating the question of effectivenessfrom the question of funding, we map out the entire territory of crimeprevention knowledge (including the many uncharted areas). That, in turn,provides a basis for locating both current and future Justice Departmentprograms on that map.

Chapters Three through Nine of the report each examine the evidencein one institutional setting at a time. Each chapter draws scientific conclusionsabout program effectiveness, then uses those findings to suggest policyrecommendations for both current programs and further research. ChapterTen then assembles the major findings into the Congressionally-mandatedassessment of the effectiveness of DOJ crime prevention programs. It concludesthe report with the implications of the assessment for the federal rolein generating just such evidence, and suggests a statutory plan for improvingscientific knowledge about effective crime prevention methods.

This chapter provides the four cornerstones on which the report is based.One is the crucial difference between the political and scientific definitionsof crime prevention. Making this distinction at the outset is essentialfor meeting the Congressional mandate for a scientific assessment. It alsohelps us clarify other key concepts in thinking about crime prevention.

A second cornerstone is the web of institutional settings in which crimeprevention effects are created every day all over the nation, mostly withoutany taxpayer involvement at all. From childhood moral education to employeecriminal history checks, there is tight social fabric holding most peopleback from committing crimes most of the time. Yet there are many holesand thin spots in that social fabric that crime prevention programs might,and sometimes do, address.

The third cornerstone is the logical basis for separating scientificwheat from chaff, or strong scientific evidence from weak or useless data.Not all crime prevention evaluations are created equal, but we must beclear about the rules of evidence.

The fourth and final cornerstone is the history and current status ofthe federal role in guiding and funding local crime prevention. The distinctionbetween those functions should be kept in mind in any discussion of theimplications of crime prevention research for federal policy.

KEY CONCEPTS IN CRIME PREVENTION

Crime prevention is widely misunderstood. The national debate over crimeoften treats "prevention" and "punishment" as mutuallyexclusive concepts, polar opposites on a continuum of "soft"versus "tough" responses to crime: midnight basketball versuschain gangs, for example. The science of criminology, however, containsno such dichotomy. It is as if a public debate over physics had drawn adichotomy between flame and matches. Flame is a result. Matches are onlyone tool for achieving that result. Other tools besides matches are wellknown to cause fuel to ignite into flame, from magnifying glasses to tinderboxes.

Similarly, crime prevention is a result, while punishment is only onepossible tool for achieving that result. Both midnight basketball and chaingangs may logically succeed or fail in achieving the scientific definitionof crime prevention: any policy which causes a lower number of crimes tooccur in the future than would have occurred without that policy.1Some kinds of punishment for some kinds of offenders may be preventive,while others may be "criminogenic" or crime-causing, and stillothers may have no effect at all. Exactly the same may also be true ofother programs that do not consist of legally imposed punishment, but whichare justified by a goal of preventing crime.

Crime prevention is therefore defined not by its intentions, but byits consequences. These consequences can be defined in at least two ways.One is by the number of criminal events; the other is bythe number of criminal offenders (Hirschi, 1987). Some would alsodefine it by the amount of harm prevented (Reiss and Roth, 1993:59-61) or by the number of victims harmed or harmed repeatedly (Farrell,1995). In asking the Attorney General to report on the effectiveness ofcrime prevention efforts supported by the Justice Department's Office ofJustice Programs, the U.S. Congress has embraced an even broader definitionof crime prevention: reduction of risk factors for crime (such asgang membership) and increases in protective factors (such as completinghigh school)--concepts that a National Academy of Sciences report has labeledas "primary" prevention (Reiss and Roth, 1993: 150). What allthese definitions have in common is their focus on observed effects, andnot the "hard" or "soft" content, of a program.

Which definition of crime prevention ultimately dominates public discourseis a critically important factor in Congressional and public understandingof the issues. If the crime prevention debate is framed solely in termsof the symbolic labels of punishment versus prevention, policy choicesmay be made more on the basis of emotional appeal than on solid evidenceof effectiveness. By employing the scientific definition of crime preventionas a consequence, this report responds to the Congressional mandate to"employ rigorous and scientifically recognized standards and methodologies."2This report also attempts to broaden the debate to encompass the entirerange of policies we can pursue to build a safer society. A rigorouslyempirical perspective on what works best is defined by the data from researchfindings, not from ideologically driven assumptions about human nature.

Bringing more data into the debate has already altered public understandingof several other complex issues. The prevention of disease, for example,has gained widespread public understanding of the implications of new researchfindings, especially those about lifestyle choices (like smoking, dietand exercise) that people can control themselves. The prevention of injurythrough regulation of automobile manufacturers has increasingly been debatedin terms of empirically observed consequences, rather than logically derivedtheories; the safety of passenger-side airbags, for example, has been debatednot just in terms of how they are supposed to work, but also in terms ofdata on how actual driver practices make airbags increasingly cause thedeaths of young children.3 Emotional and ideological overtonesof personal freedom and the role of government clearly affect debates aboutdisease and injury prevention, but scientific evidence appears to havegained the upper hand in those debates.

Similarly, the symbolic politics of crime prevention could eventuallygive way to empirical data in policy debates (Blumstein and Petersilia,1995). While the emotional and symbolic significance of punishment cannever be denied, it can be embedded in a broader framework of crime preventioninstitutions and programs that allows us to compare value returned formoney invested (Greenwood, et al, 1996). Even raising the question of cost-effectivenesscould help focus policy-making on empirical consequences, and their implicationsfor making choices among the extensive list of crime prevention efforts.

The value of a broad framework for analyzing crime prevention policiesis its focus on the whole forest rather than on each tree. Most debatesover crime prevention address one policy at a time. Few debates, eitherin politics or in criminology, consider the relative value of all preventionprograms competing for funding. While scientific evidence may show thattwo different programs both "work" to prevent crime, one of theprograms may be far more cost-effective than another. One may have a strongereffect, cutting criminal events by 50% while the other cuts crimesby only 20%. Or one may have a longer duration, reducing crimesamong younger people whose average remaining lifetime is 50 years, comparedto a program treating older people with an average remaining life of twentyyears. A fully informed debate about crime prevention policy choices requiresperformance measures combining duration and strength of program effect.While such accurate measures of "profitability" and "payback"periods are a standard tool in business investment decisions, they havebeen entirely lacking in crime prevention policy debates.

Yet comparative measurement is not enough. Simply comparing the returnon investment of each crime prevention policy to its alternatives can maskanother key issue: the possible interdependency between policies,or the economic and social conditions required for a specific policy tobe effective. Crime prevention policies are not delivered in a vacuum.A Head Start program may fail to prevent crime in a community where childrengrow up with daily gunfire. A chain gang may have little deterrent effectin a community with 75% unemployment. Marciniak (1994) has already shownthat arrest for domestic violence prevents crime in neighborhoods withlow unemployment and high marriage rates--but arrest increases crimein census tracts with high unemployment and low marriage rates. It maybe necessary to mount programs in several institutional settings simultaneously--suchas labor markets, families and police--in order to find programs in anyone institution to be effective.

One theory is that the effectiveness of crime prevention in eachof the seven institutional settings depends heavily on local conditionsin the other institutions. Put another way, the necessary conditionfor successful crime prevention practices in one setting is adequate supportfor the practice in related settings. Schools cannot succeed without supportivefamilies, families cannot succeed without supportive labor markets, labormarkets cannot succeed without well-policed safe streets, and police cannotsucceed without community participation in the labor market. These andother examples are an extension of the "conditional deterrence"theory in criminology (Tittle and Logan, 1973; Williams and Hawkins, 1986),which claims that legal punishment and its threat can only be effectiveat preventing crime if reinforced by the informal social controls of otherinstitutions. The conditional nature of legal deterrence may apply to othercrime prevention strategies as well. Just as exercise can only work properlyon a well-fed body, crime prevention of all kinds may only be effectivewhen the institutional context is strong enough to support it.

Over a century ago, sociologist Emile Durkheim suggested that "itis shame which doubles most punishments, and which increases with them"(Lukes and Scull, 1983, p. 62). More recently, John Braithwaite (1989)has hypothesized the institutional conditions needed to create a capacityfor shame in both communities and individuals. He concludes that shameand punishment have been de-coupled in modern society, and suggests variousapproaches to restoring their historic link. His conclusions can applyto non-criminal sanctions as well, such as school discipline, labor forceopportunities, expulsion from social groups and ostracism by neighborsand family. Conversely, it applies to rewards for compliance with the criminallaw, such as respectability, trust, and responsibility. The emotional contentof winning or losing these social assets is quite strong in settings wherecrime prevention works, but weak or counterproductive in what social scientistscall "oppositional subcultures." Any neighborhood in which goingto prison is a mark of prestige (Terry, 1993) is clearly a difficult challengefor any crime prevention practice.

The community context of crime prevention may need a critical mass ofinstitutional support for informally deterring criminal behavior. Withoutthat critical mass, neither families nor schools, labor markets nor places,police nor prisons may succeed in preventing crime. Each of these institutionsmay be able to achieve marginal success on their own. While most Americancommunities seem to offer sufficient levels of institutional support forcrime prevention, serious violence is geographically concentrated in asmall number of communities that do not. Lowering national rates of violentcrime might require programs that address several institutional settingssimultaneously, with a meaningful chance of rising to the threshold of"social capital" (Coleman, 1992) needed to make crime preventionwork.

To the extent that this theory focuses resources on the relative handfulof areas falling below that threshold, that focus can be justified by itsbenefits for the wider society. Over half of all homicides in the US occurin just 66 cities, with one-quarter of homicides in only eight cities (FBI,1994). These murders are concentrated in a small number of neighborhoodswithin those cities. The public health costs of inner-city violence, bythemselves, could provide sufficient justification for suburban investmentin inner-city crime prevention. If crime can be substantially preventedor reduced in our most desperate neighborhoods, it can probably be preventedanywhere.

By suggesting that the effectiveness of some crime prevention effortsmay depend upon their institutional contexts, we do not present a pessimisticvision of the future. While some might say that no program can work untilthe "root causes" of crime can be cured, we find no scientificbasis for that conclusion--and substantial evidence against it. What thisreport documents is the potential for something much more precise and useful,based on a more open view of the role of scientific evaluation in crimeprevention: a future in which program evaluations carefully measure, andsystematically vary, the institutional context of each program. That strategyis essential for a body of scientific knowledge to be developed about theexact connections between institutional context and program effectiveness.

We expect that greater attention to the interdependency of institutionsmay help us discover how to shape many institutional factors simultaneouslyto prevent crime--more successfully than we have been able to do so far.The apparent failure of a few efforts to do just does not mean that weshould give up our work in that direction. Such failures marked the earlystages of almost all major advances in science, from the invention of thelight bulb to the development of the polio vaccine. The fact that our reviewfinds crime prevention successes in all of seven of the institutional settingssuggests that even more trial and error could pay off handsomely. Our nationalinvestment in research and development for crime prevention to date hasbeen trivial (Reiss and Roth, 1993), especially in relation to the levelof public concern about the problem. Attacking the crime problem on manyinstitutional fronts at once should offer more, not fewer, opportunitiesfor success.

Defining crime prevention by results, rather than program intent orcontent, focuses scientific analysis on three crucial questions:

1. What is the independent effect of each program or practice on a specificmeasure of crime?

2. What is the comparative return on investment for each program orpractice, using a common metric of cost and crimes prevented?

3. What conditions in other institutional settings are required fora crime prevention program or practice to be effective, or which increaseor reduce that effectiveness?

The current state of science barely allows us to address the first question;it tells us almost nothing about the second or third. Just framing thequestions, however, reveals the potential contribution that federal supportfor crime prevention evaluations could offer. That potential may depend,in turn, on a clear understanding of the location of every crime preventionpractice or program in a broad network of social institutions.

THE INSTITUTIONAL SETTINGS OF CRIME PREVENTION

Crime prevention is a consequence of many institutional forces. Mostof them occur naturally, without government funding or intervention. Whilescholars and policymakers may disagree over the exact causes of crime,there is widespread agreement about a basic conclusion: strong parentalattachments to consistently disciplined children (Hirschi, 1995) in watchfuland supportive communities (Braithwaite, 1989) are the best vaccine againststreet crime and violence. Schools, labor markets and marriage may preventcrime, even among those who have committed crime in the past (Sampson andLaub, 1993), when they attract commitment to a conventional life patternthat would be endangered by criminality. Each person's bonds to family,community, school and work create what criminologists call "informalsocial control," the pressures to conform to the law that have littleto do with the threat of punishment. Informal controls threaten somethingthat may be far more fearsome than simply life in prison: shame and disgracein the eyes of other people you depend upon (Tittle and Logan, 1973).

The best evidence for the preventive power of informal social controlmay be the millions of unguarded opportunities to prevent crime which arepassed up each day (Cohen and Felson, 1979). Given the fact that most crimesnever result in arrest (FBI, 1996), the purely statistical odds are infavor of a rational choice to commit any given crime. The question of whyeven more people do not commit crime is therefore central to criminology,and has driven many theories (Hirschi, 1969; Cohen and Felson, 1979; Gottfredsonand Hirschi, 1990). The extent to which law enforcement can affect theperception of those odds is a matter of great debate (Blumstein, Cohenand Nagin, 1978), as is the question of whether even a low risk of punishmentis too high for most people. Yet there is widespread agreement that theinstitutions of family and community are critically important to crimeprevention.

That agreement breaks down when the institutions of family and communitythemselves appear to break down, creating a vacuum of informal social controlthat government is then invited to fill up (Black, 1976). Whether police,courts and prisons can fill the gap left by weak families and sociallymarginal communities is a question subject to debate in both politics andsocial science. But it may be the wrong question to ask, at least initially.The premise of the question is that the breakdown of the basic institutionsof crime prevention is inevitable. Yet for over a century, a wide rangeof programs has attempted to challenge that premise. Entirely new institutions,from public schools to social work to the police themselves (Lane, 1992),have been invented to provide structural support to families and communities.In recent years, the federal government has attempted a wide range of programsto assist those efforts. Rather than simply assuming their failure, itseems wiser to start by taking stock of their efforts.

Settings, Practices and Programs

Crime prevention is a result of everyday practices concentrated in seveninstitutional settings. A "setting" is a social stage for playingout various roles, such as parent, child, neighbor, employer, teacher,and church leader. There are many ways to define these settings, and theirboundaries are necessarily somewhat arbitrary. Yet much of the crime preventionliterature fits quite neatly into seven major institutional settings: 1)communities, 2) families, 3) schools, 4) labor markets, 5) places, 6) policeagencies and 7) the other agencies of criminal justice. The definitionsof these settings for crime prevention are quite broad, and sometimes theyoverlap. But as a framework for organizing research findings on crime preventioneffectiveness, we find them quite workable.

Crime prevention research examines two basic types of efforts in theseseven settings. One type is a "practice," defined as anongoing routine activity that is well established in that setting, evenif it is far from universal. Most parents make children come home at night,most schools have established starting times, most stores try to catchshoplifters, most police departments answer 911 emergency calls. Some ofthese practices have been tested for their effects on crime prevention.Most have not. Some of them (such as police patrols and school teachersalaries) are funded in part by federal programs. Most are not. Regardlessof the source of funding, we define a practice as something that may changenaturally over time, but which would continue in the absence of specificnew government policies to change or restrict them.

A "program," in contrast, is a focused effort to change,restrict or create a routine practice in a crime prevention setting. Many,but far from all, programs are federally funded. Churches may adopt programsto discourage parents from spanking children, or letting children watchviolent television shows and movies. Universities may adopt programs toescort students from the library to their cars in the hours after midnight.Shopping malls may ban juveniles unescorted by their parents on weekendevenings, and police may initiate programs to enforce long-ignored curfewor truancy laws. In time, some programs may turn into practices, with fewpeople remembering the time before the program was introduced.

Perhaps the clearest distinction between programs and practices is foundamong those programs requiring additional resources. The disciplinary practicesof parents, for example, and the hiring practices of employers are largelyindependent of tax dollars. But calling battered women to notify them oftheir assailant's imminent release from prison may be a practice that onlya federally funded program can both start and keep going. Even police enforcementof laws against drunk driving, in recent years, seems to depend almostentirely on federally funded overtime money to sustain (Ross, 1994). Whetherthese federal resources are "required" is of course a matterof local funding decisions. But in many jurisdictions, many practices begununder federal programs might die out in the absence of continued funding.

These distinctions are important to crime prevention for reasons ofevidence: newly-funded programs are more likely to be subjected to scientificevaluations than longstanding practices. The modern trend towards demandingaccountability for public expenditures has made program evaluations increasinglycommon, especially for federal programs. Paradoxically, we could know moreabout potentially marginal new ideas than we do about the mainstream practicesof the major crime prevention institutions. Police DARE programs (DrugAbuse Resistance Education), for example, have been subjected to more numerousevaluations (Lindstrom, 1996) than the far more widespread practice ofpolice patrol (Sherman and Weisburd, 1995). Similarly, neighborhood watchprograms (Hope, 1995) have been subjected to far more extensive evaluationthan the pervasive role of zoning practices in physically separating commercialand residential life in communities, reducing face-to-face contact amongthe kind of neighbors who used to see each other at the corner grocerystore.

The availability of evidence on crime prevention is itself a major issuefor the national policy debate. Where expenditures are high but evidenceis weak or non-existent, the need for evaluation research is great. Evenwhere expenditures are low, practices or programs that show good reasonto conclude that they are causing or preventing crime should merit a highpriority for research. In order to identify the key gaps in our knowledge,however, we must start not with the available evidence, but with an inventoryof crime prevention practices and programs in each institutional setting.Throughout the report, this inventory guides our review of what works,what doesn't, what's promising, and what we need to know a lot more about.

Chapter 3: Communities

We begin our review with the most broadly defined institutional setting.From small villages to large urban neighborhoods, from suburban developmentsto urban high rise public housing, both the physical and social structureof communities varies widely. So, too, does their effectiveness in preventingcrime through informal social controls. Some communities average more thantwo jobs per family; others average none. Some communities have more churchesthan taverns; others have more crack houses than grocery stores. Some havemore people on welfare than working; others have more retirees than schoolchildren.Some have more renters than homeowners; others have more adult men whoare technically homeless than those who are named on a lease or a deed.In some communities most residents recognize most other residents by nameand face; in most of the modern United States, perhaps, even face recognitionof most neighbors is extremely rare.

Communities also vary on several stark dimensions. Most of the seriousviolent juvenile crime in the US is concentrated in a relative handfulof communities (OJJDP, 1996). Some communities have homicide rates 20 timeshigher than the national average (Sherman, Shaw and Rogan, 1995). In somecommunities two-thirds of all adults are chronically unemployed (Wilson,1996: 19). In some communities 90% or more of the population is African-Americanfor miles around, a condition of "hypersegregation" unprecedentedin American history (Massey and Denton, 1993). In some communities childabuse is reported among 19% of at-risk children of white parents (Olds,et al, 1986). To a large extent, the entire rationale for the federal politicsof crime prevention is driven by the extreme criminogenic conditions ofthese relatively few communities in the US, areas of concentrated povertywhere millions of whites and an estimated 1/3 of all African-Americansreside.

Where a community winds up on these and other dimensions may not onlyaffect its crime prevention practices. There is also substantial evidencethat these factors condition the effectiveness of community-based crimeprevention programs (Hope, 1995), another excellent (but rare) exampleof interdependency. In study after study, evidence emerges that crime preventionprograms are more likely to take root, and more likely to work, in communitiesthat need them the least. Conversely, the evidence shows that communitieswith the greatest crime problems are also the hardest to reach throughinnovative program efforts.

Chapter Three reviews this evidence as pointing to the general conclusionthat such programs are too weak to make a difference in the underlyingstructural conditions causing both crime prevention and innovative programsto fail. More heavily concentrated federal efforts to address many communityfactors simultaneously have, fortunately, suggested somewhat better resultsagainst local crime risk factors. And even in the midst of great adversity,there is some evidence that "big brother" and "sister"mentoring programs can help reduce drug abuse and other risk factors forcrime--perhaps showing how much a community benefits by having strong familiesthat provide their own mentoring, also known as parenting.

Chapter 4: Families

Perhaps the most basic structural feature of any community is the conditionof its families. Basic family practices in child-rearing, marriage, andparental employment appear to matter enormously in the criminality of bothchildren and fathers (Hirschi, 1995; Sampson, 1986). The failure of manyparents to marry has been the target of many programs for preventing extramaritalpregnancy, especially among teenagers. The failure of many parents to provideconsistent affection and discipline to children has been the target ofother programs, from parent training to home visitation and consultationby nurses and other helpers. As Chapter Four shows, some of these programsare quite promising, with very encouraging evaluation results. Whetherthese programs, by themselves, can overcome the effects of surroundinga family with a high-crime community is unclear.

It is also unclear whether we have found the right programs for combattingdomestic violence, arguably a major risk factor for crime found in thefamily setting. Most of these programs are delivered to families by thecriminal justice system. These programs unfortunately fail to reach themany families whose violence goes unreported to police. For the familiesthe programs do reach, the scientific evidence is either discouraging orinadequate. Here again, the crime prevention programs seem to work bestfor the families in the strongest communities. Criminal justice programsmay be least effective in the communities where family violence is mostprevalent.

The major exception to this pattern is the use of battered women's shelters,an important emergency service at high-risk times for family violence.While shelters also lack clear evaluations showing crime prevention benefits,police data show the highest risk of such violence to lie in the immediateaftermath of the last domestic assault. Protecting women, and often theirchildren, in that short time frame may well reduce total injuries fromdomestic violence, even if shelters cannot solve the underlying familyviolence. Yet even shelters are relatively less available in the poorestcommunities, compared to communities of greater social and financial resources.

Chapter 5: Schools

The most direct link between families and communities is presently foundin schools. Measured purely by the amount of available time to reduce riskfactors for crime, schools have more opportunity to accomplish that objectivethan any other agency of government. Succeeding at their basic job of teachingchildren to read, write and compute may be the most important crime preventionpractice schools can offer. But too many schools are overwhelmed by a criminogeniccommunity context, crippled by the lack of parental support for learningand the breakdown of order in the classrooms (Toby, 1982). While some schoolssucceed at teaching basic skills despite these challenges, the odds appearto be against it.

The most intensively studied crime prevention programs in schools, however,are unrelated to academic learning. More common are the efforts to useschools to reduce non-academic crime risk factors, including drug abuseand aggression. As Chapter Five demonstrates, the extensive record of scientificallyevaluated prevention programs provides some guidance about which programsare most effective or promising. The evidence shows that school-based programsaimed at increasing resilience, for example, by teaching students "thinkingskills" necessary for social adaptation, work to reduce substanceuse and are promising for reducing delinquency. Programs that focus noton individual students, but instead on school organizations, also work.Programs that simply clarify norms about expected behavior work. As inother settings, the success of school programs and practices is largelydependent on the school's capacity to initiate and sustain innovative programs.Schools situated in crime-ridden, disorganized communities are less likelyto have the infrastructure necessary to support prevention programs, andare more likely to fail. That failure is usually more pronounced in communitieswith the weakest labor market demand for adult workers.

Chapter 6: Labor Markets

There is a long history of attempting to prevent the onset or persistenceof criminality by pulling young people into the labor market for legitimatework (Cloward and Ohlin, 1960). Theoretical and empirical support for thecrime preventive value of employment is generally quite strong in the longitudinalanalysis of individual criminal careers (Sampson and Laub, 1993; but seeShannon, 1982, and Gottfredson, 1985). It is also found in experimentalstudies of the effects of criminal sanctions, which can deter offenderswho are employed but backfire on offenders who are unemployed (Sherman,1992). Macro-level data on the short-term effects of changes in the unemploymentrate on crime are more mixed (Freeman, 1983, 1995), but the staggeringlyhigh unemployment rates in our highest-crime communities are beyond dispute(Wilson, 1996).

Programs aimed at linking labor markets more closely to high crime riskneighborhoods and individuals could have substantial crime prevention benefits.As Chapter Six shows, however, only Job Corps programs have demonstratedsuccess at enhancing the employment experience of severely unemployablepersons, and even that evidence is scientifically weak. No program hasyet shown success in tackling the unemployment rates of high crime neighborhoods.Yet of all the dimensions of neighborhood life, this one may have the mostpervasive influence on crime. Neighborhoods where work is the exceptionrather than the rule may lack the discipline necessary for conventionallife styles (Wilson, 1996). Marriage and two-parent family life deeplydeclines with the loss of labor markets for adult males, making men unnecessaryas economic partners and husbands. If inner city communities of concentratedpoverty are to be reclaimed as crime prevention institutions, revivingtheir local labor markets may be the most logical place to start. As jobsincreasingly migrate to far suburbs beyond the reach of public transit,inner city workers with no cars may depend even more on recent innovativeprograms to link them to suburban labor markets.

Inner city employment may face an even tougher problem than geography,however. As employers become increasingly sensitive to concerns about potentialtheft and violence by their employees, they have won increasing accessto measures of the criminality of prospective and current workers. Onemeasure is official records of criminal convictions, which are more readilyavailable now than at any previous time in US history (SEARCH Group, 1996).Another measure is drug testing in the workplace, which many employersrequire as a condition of employment. Both measures could either bar aworker from being hired or lead to their being fired. Extensive policecrackdowns of recent years have given millions of young men criminal recordsfor minor offenses (Blumstein, 1993; Tonry, 1995), limiting their employmentprospects and perhaps increasing their likelihood of further and more seriouscriminality.

Yet labor markets may be most powerful in preventing crime preciselybecause they respond negatively to criminal histories. While employmentmay give would-be offenders a stake in society, its crime preventive valuemay hinge on the threat of losing that stake. Maintaining that threat withoutcreating a large group of unemployable outcasts is a major crime preventionchallenge for the future of our labor market practices.

Chapter 7: Places

One of the most recently discovered "institutions" in Americanlife is the "place" (Anderson, 1978; Oldenburg, 1990). From donutshops to taverns to street corners and hotels, there is a pattern of socialorganization uniquely constructed around very small locations that areusually visible to the unaided human eye. These places vary enormouslyin their populations, core functions and activities, crime rates and criminogenicrisk factors like drugs and guns. Some places are so crime prone that theyare labeled "hot spots" of crime (Sherman, Gartin and Buerger,1989), among the 3% of addresses which produce 50% of reported crimes.

Regardless of whether these places cause crimes or merely act as "receptors"for them, the prevention of crime in places may have substantial effectsat reducing total crime in the community. Even in high crime neighborhoods,most places are crime-free for years at a time (Pierce, Spaar and Briggs,1988). The frequent recurrence of crimes in just a handful of locationsmakes the prevention of crime in such "hot spots" all the moreimportant.

Security guards, cameras, alarm systems, safes and fences have all proliferatedin the latter twentieth century, making private expenditures on crime preventionrival public spending. Whether these practices succeed in preventing crimeis generally impossible to determine from the available research, givenits limitations. Even where they do succeed at preventing crime in targetplaces, it is unclear whether the total number of criminal events in societyis reduced or merely displaced to other locations (Barr and Pease, 1990).But as the evidence reviewed in Chapter Seven shows, the control of criminogeniccommodities like alcohol, cash and firearms (Cook and Moore, 1995) canmake a great deal of difference in the rate of crime in limited accesslocations like airports and transit systems. Such strategies may even overcomethe influence of surrounding high crime communities.

Our capacity to make a limited number of places into safe havens fromcrime may also form a paradox: the safer we make places for more advantagedpeople, the less public investment there may be in making less advantagedcommunities safe (Reiss, 1989). The use of metal detectors to create ofgun-free zones has become a prized luxury, reserved for presidents andjudges, airplane passengers and (more democratically) some school children.But it may also have reduced policymakers' concern about gun crime in thestreets, especially the streets of poverty areas. People spending moremoney on private security may wish to spend less for public safety. Whilecommunities may be better off without their worst hot spots of crime, theycannot be made safe by place-based strategies alone. To the extent thatcrime prevention in places depletes efforts in other institutional settings,safe places in a dangerous community may be ultimately self-defeating.It is hard to imagine a democracy as a fortress society.

Chapter 8: Policing

The crime prevention effects of policing may pose the widest gap betweenacademic and political opinion. While public opinion polls show consensusthat police prevent crime, criminologists widely challenge that view. Citinga single, scientifically weak evaluation of police patrol presence (Kelling,et al, 1974), many criminologists generalize that variations in policepractice or numbers can make little difference in crime (Gottfredson andHirschi, 1990; Felson, 1994). This conclusion ignores a vast array of contraryevidence.

As Chapter Eight shows, there are many police practices that reducecrime, and some that even increase crime. The strength of police effectson crime is generally moderate rather than substantial, unless police presencedrops to zero when patrols go on strike--at which point all hell breaksloose. The converse of that observation could be that massive increasesof police presence focused in a small number of high crime communitieshave a major effect at preventing crime. While such concentrations havenever been attempted for sustained periods of time, it is possible thata focused crime prevention strategy could rely heavily on police presenceto regain a threshold level of public order and safety. Once beyond thisthreshold, the effectiveness of family, community, schools and the laborforce could be substantially increased.

Community policing programs offer one opportunity to increase policepresence in the highest crime communities. Like police resources generally,the 1994 Crime Act puts a large portion of its 100,000 police where thepeople are, but not where the crime is. The scientific evidence increasinglysuggests the effectiveness of much greater concentration of federal fundingin the neighborhoods which need police the most. While such policies wouldfly in the face of distributional politics (Biden, 1994), they are stronglyimplied (although not proven) by studies of police effects on crime inlow and high crime areas. The Federal funding of police overtime couldalso be more effective if available funds were channeled to the small numberof neighborhoods generating most of the handgun homicide in the nation.

Yet research also shows that police presence can backfire if it is providedin a disrespectful manner. Rude or hostile treatment of citizens, especiallyjuveniles, can provoke angry reactions that increase the risk of futureoffending (Tyler, 1991). Flooding high crime communities with aggressivepolice could backfire terribly, causing more crime than it prevents, asit has in repeated race riots over the past quarter century. The challengeis to develop programs that make policing simultaneously more focused inwhat they do to prevent crime and more polite in how theydo it.

Chapter 9: Criminal Justice

The full list of crime prevention practices and programs in criminaljustice is very long indeed. We relegate them to a single chapter in anattempt to focus more attention on how such punishment programs compareto non-punitive prevention practices. Recent reviews conclude there isvery little evidence that increased incarceration has reduced crime (Reissand Roth, 1993). Yet variations in how the criminal justice system treatsadmitted offenders can make a great deal of difference. The evidence reviewedin Chapter Nine finds encouraging support for more correctional use ofdrug treatment programs, rehabilitation programs in prison, and institutionalizationof some juvenile offenders rather than community-based supervision.

The effectiveness of any correctional treatment, however, may dependupon the community, family, and labor market context into which the offenderreturns home. In a very important sense, correctional programs competewith the same home conditions that led the offender into correctional handsin the first place. Making corrections work, at least with the offendersit treats, may require the same changes of institutional context neededto make programs and practices in other settings more effective.

Chapter 10: Justice Department Funding for Local Crime PreventionPrograms

It is important for the U.S. Congress to assess its own funding of localcrime prevention programs in the context of these seven institutional settingsfor attempting--and sometimes achieving--crime prevention results. It maybe even more important to understand the relationship among the seven settings,and the extent to which conditions in one affect conditions or resultsin another. Chapter Ten synthesizes the major findings from each institutionalsetting to draw broad conclusions about the effectiveness of DOJ localassistance programs. But many of the local programs and practices thesefunds support have never been evaluated with enough scientific rigor todraw conclusions based on direct evidence about their effects on crime.Chapter Ten therefore concludes with analysis and recommendations concerningthe structure of program evaluation for local assistance funding, suggestinghow to better achieve the longstanding Congressional mandate to evaluate.

Evaluating crime prevention is at best a delicate enterprise. Policymakersoften think, incorrectly, that an evaluation is like an "audit"or trial in which the results are usually clear cut and definitive. Eitherthe funds were spent or they weren't; either the program served its intendedbeneficiaries at a reasonable cost per client or it didn't. Such "audit"questions are much easier to answer than the "evaluation" questionsof cause and effect, often stretching out over a lifetime of the targetsof crime prevention efforts. The next section introduces some of the complicationsin drawing such conclusions scientifically. Chapter ten returns to thoseissues in terms of their implications for future evaluation policies forOJP funding. Rather than spending a little evaluation money on most programsin an "audit" model, the Congress would receive more return oninvestment by concentrating evaluation dollars on a few major examplesof key programs in a field testing model.

MEASURING CRIME PREVENTION EFFECTIVENESS

A recent review of the crime prevention evaluation literature by twoprominent English criminologists concluded the field was "dominatedby....self-serving unpublished and semi-published work that does not meeteven the most elementary criteria of evaluative probity (Ekblom and Pease,1995:585-6). What they meant by "evaluative probity" was fairlybasic to any inference of cause and effect. Measures of crime, for example,are very often missing from publicly funded crime prevention "evaluations,"which simply describe how the program worked and whether it achieved itsadministrative objectives: services provided, activities completed. Despitethe recent emphasis at reinventing government to focus on results,most crime prevention evaluations still appear to focus on efforts.

Crime Prevention and Other Worthy Goals

Many if not most government programs, of course, have multiple objectives.Even those which evaluations show ineffective at preventing crime may accomplishother worthy goals, such as justice and equality under the law. That isa very important consideration for policy analysis, one that deserves carefultreatment. This report does not explicitly examine program effects in accomplishingother goals beyond those specified in the legislation: crime, especiallyyouth violence, risk factors and (their converse) protective factors. Thatdoes not mean other goals are unimportant. Consideration of those othergoals can be entirely appropriate in other contexts, and can be examinedby scientific program evaluations. This report omits them necessarily inorder to conserve resources for answering the specific question the Congressasked.

Whether the focus of an evaluation is on crime prevention or other goals,the distinction between descriptive and impact evaluations remains crucial.Training police on domestic violence issues, for example, may not directlyreduce domestic violence. But descriptive evaluations reporting how manypolice were trained for how many hours are also unable to show whetherother goals were accomplished. Causing police to treat domestic violencevictims more politely, to provide more victim assistance, or to gatherbetter evidence at the scene could all be important objectives of policetraining. Controlled experiments could shows whether training accomplishesthose important goals. Absent a strong scientific approach to program evaluation,however, descriptive evaluations of efforts say little about results forother goals besides crime prevention.

Classifying the Strength of Scientific Evidence

Even where evaluations attempt to measure crime prevention, they oftenlack the basic scientific elements needed for inferring cause and effect.While they may report lower crime rates among people who were served bya program than those who were not, the evaluations often fail to say whichcame first, the program or the crime rates. If crime prevention programssimply attract lower crime rate people, they cannot be said to cause thoselower crime rates. Other evaluations include a temporal sequence, reportingthat crime dropped after a program was introduced, for example. But theremay be many other reasons why crime went down besides the program. Whilecomparison or "control" groups can be used to help eliminatethose other possibilities, many evaluations fail to use them. Even whenthey are used, the comparison groups chosen are often too unlike the targetgroups given the program, so that the comparison does not plausibly showwhat would have happened without the program. Only a random selection ofequally eligible program targets can conclusively eliminate alternativetheories about the effects of a crime prevention program.

Thus we must confront a body of research in which the strength of theevidence varies as much as the strength of the crime preventionprogram effects reported in the research. Making sense of this evidencerequires some scale for rating the strength of each study. While our analysisemploys more complicated classifications (see Appendix 1), there are threebasic elements we consider:

1) reliable and statistically powerful measures and correlations (includingadequate sample sizes and response rates),

2) temporal ordering of the hypothesized cause and effect--so that theprogram "cause" comes before the crime prevention "effect,"and

3) valid comparison groups or other methods to eliminate other explanations,such as "the crime rate would have dropped anyway."

The first element without the others arguably constitutes "weak"evidence, the first and second without the third comprise "moderate"evidence, and all three together define "strong" evidence. Thisstandard sets aside the question of replication of results in repeatedstudies, since it is generally so rare in federal program evaluations.Such replicated results are "very strong" evidence compared tomost program evaluations.

A SCALE OF EVIDENTIARY STRENGTH FOR CAUSE AND EFFECT

WeakModerateStrong
1. Reliable, powerful correlation testx

2. Temporal ordering of cause and effect
x
3. Elimination of Major Rival Hypotheses

x

Our analysis employs a "methodological rigor" rating basedon a scale adapted from one used in a recent national study of the effectivenessof substance abuse prevention efforts (Center for Substance Abuse Prevention,1995). Using this scientific methods scale, we rate seven differentdimensions of the methods used in each study. The overall rating is basedprimarily on these three factors:

o the study's ability to control extraneous variables (i.e.,to eliminate major rival hypotheses, accomplished through random assignmentto conditions, matching treatment and comparison groups carefully, or statisticallycontrolling for extraneous variables the minimization of measurementerror

o the statistical power to detect meaningful differences (e.g.,the power of a test to detect a true difference. The smaller the anticipatedeffects of prevention, the larger the sample size must be in order to detecta true difference.)

Other considerations contributing to the overall rating of methodologicalrigor are the response rate, attrition of cases from thestudy, and the use of appropriate statistical tests. An appendixto this report describes the methodology rating in more detail and showsthe coding sheet used to rate studies.

Using this scale, each eligible study examined for this report was givena "scientific methods score" of 1 to 5, with 5 being the strongestscientific evidence.4 While there are some minor variationsin how the authors of Chapters Three through Nine apply the basic scientificmethods criteria in making coding decisions, the criteria are standardizedwithin each chapter and highly similar across chapters. In order to reachlevel 3, a study had to employ some kind of control or comparison groupto test and refute the rival theory that crime would have had the sametrend without the crime prevention program;5 it also had toattempt to control for obvious differences between the groups, and attendto quality of measurement and to attrition issues. If that comparison wasto a more than a small number of matched or almost randomized cases, thestudy was given a score of "4".6 If the comparisonwas to a large number of comparable units selected at random to receivethe program or not, the study was scored as a "5", the highestpossible level; random assignment offers the most effective means availableof eliminating competing explanations for whatever outcome is observed.Most of the tables summarizing evaluation research in the next seven chaptersdisplay these scientific methods scores right next to the reference tothe study.

The scientific issues for inferring cause and effect vary somewhat byinstitutional setting, and the specific criteria for applying the scientificmethods scale vary accordingly. Issues such as sample "attrition,"or subjects dropping out of treatment or measurement, for example, do notapply to most evaluations of commercial security practices. But acrossall settings, our scientific methods scale does include these core criteria:

1. Correlation between a crime prevention program and a measure of crimeor crime risk factors

2. Temporal sequence between the program and the crime or risk outcomeclearly observed, or a comparison group present without demonstrated comparabilityto the treatment group

3. A comparison between two or more units of analysis, one with andone without the program 7

4. Comparison between multiple units with and without the program, controllingfor other factors, or a nonequivalent comparison group has only minor differencesevident

5. Random assignment and analysis of comparable units to program andcomparison groups

In addition, the use of statistical significance tests is employed asa key criterion in reaching program effectiveness conclusions based onthe application of the scores.

The report does not code scientific methods scores on evaluations ofevery program or practice considered. On many questions, recent literaturereviews and meta-analyses by qualified scholars were readily available.The Office of Juvenile Justice and Delinquency Prevention, in particular,was very helpful in providing the draft report of its own group of independentscholars examining the problems of serious, chronic and violent juvenileoffenders (Loeber and Farrington, forthcoming). The report uses two alternateprocedures in relying on extant secondary reviews and meta-analyses. Oneis to use data presented in the reviews to score the key original research.The other is not to use any scoring, but merely to summarize the conclusionsof the secondary review.

Risk and Protective Factors. The Congressional mandate for thisreport included risk and protective factors for crime and delinquency asoutcome measures to be considered. Different approaches to the interpretationof these terms are offered in the literature. This report defines themas inversely related: the lower the level of a risk factor, the higherthe levl of a protective factor. For example, community labor force participationis a risk factor where it is low and a protective factor where it is high.To the extent that factors such as a secure personality or strong bondingto adults may be considered protective against independent risk factors(such as neighborhood unemployment), those protective factors can alsobe treated as risk factors when they are absent.

Deciding What Works

Clear conclusions about what works and what doesn't requires a highlevel of confidence in the research results. Such claims are always suspectin science, which is an eternally provisional enterprise. New researchresults continue to fill in the gaps of our knowledge, and reanalysis ofold results in light of the new findings often produces different conclusions.The best one can ever claim to "know" is what to conclude onthe available evidence, pending the results of further research. Giventhe consequences of claim about "what works" can have major effectson crime prevention practice, it is important to use a high threshold forthe strength of scientific evidence at any point in time.

The current state of the evidence, however, creates a dilemma in respondingto the Congressional mandate. Using level 5 studies as the "gold standard"of evaluation design, the scientific methods scores for most of the availableevaluations are low. The recommendations in Chapter 10 are designed toraise the methods scores of future evaluations of DOJ programs. The dilemmathe current evidence poses is the question of how high to set the thresholdfor answering the Congressional question about program effectiveness: decidingwhat works. A very conservative approach might require at least two level5 studies showing that a program is effective (or ineffective), with thepreponderance of the evidence in favor of the same conclusion. Employinga threshold that high, however, would leave very little to based upon fromthe existing science. There is a clear tradeoff between the level of certaintyin the answers we can give to the Congress and the level of useful informationthat can be gleaned from the available science. On balance, excluding whatcan be said from moderately rigorous studies would waste a great deal ofinformation that could be useful for policymaking. The report takes themiddle road between reaching very few conclusions with great certaintyand reaching very many conclusions with very little certainty.

Based on the scientific strength and substantive findings of the availableevaluations, the report classifies all local programs into one of fourcategories: what works, what doesn't, what's promising, and what's unknown.The criteria for classification applied across all seven institutionalsettings are as follows:

What Works. These are programs that we are reasonably certainof preventing crime or reducing risk factors for crime in the kinds ofsocial contexts in which they have been evaluated, and for which the findingsshould be generalizable to similar settings in other places andtimes. Programs coded as "working" by this definition must haveat least two level 3 evaluations with statistical significance tests showingeffectiveness and the preponderance of all available evidence supportingthe same conclusion. Where the strength of the effect on crime is availablein terms of standard deviations from the mean level of crime or risk, theeffect size (Cohen, 1977) in both level 3 studies must exceed .1.

What Doesn't Work. These are programs that we are reasonablycertain fail to prevent crime or reduce risk factors for crime inthe kinds of social contexts in which they have been evaluated, and forwhich the findings should be generalizable to similar settings inother places and times. Programs coded as "not working" by thisdefinition must have at least two level 3 evaluations with statisticalsignificance tests showing ineffectiveness and the preponderanceof all available evidence supporting the same conclusion. The effect sizestandard for coding what works is also applied where available, which inthe current report is limited to the school-based prevention programs.

What's Promising. These are programs for which the level of certaintyfrom available evidence is too low to support generalizable conclusions,but for which there is some empirical basis for predicting that furtherresearch could support such conclusions. Programs are coded as "promising"if they have at least one level 3 evaluation with significance tests showingtheir effectiveness at preventing crime or reducing crime risk factors,and the preponderance of all available evidence supports the sameconclusion.

What's Unknown. Any program not coded in one of the three othercategories is defined as having unknown effects. The report lists somebut not all such programs. This category includes major variations on programcontent, social setting, and other conditions which limit the generalizabilityeven of programs coded as working or not. For example, it is unknown whetherfamily training interventions repeatedly found effective in Oregon canwork on the south side of Chicago.

The weakest aspect of this classification system is that there is nostandard means for determining exactly what variations on program contentand setting might affect generalizability. In the current state of science,that can only be accomplished by the accumulation of many tests in manysettings with all major variations on the program theme. None of the programsreviewed for this report have accumulated such a body of knowledge so far.The conclusions about what works and what doesn't should therefore be readas more certain to the extent that the conditions of the field tests canbe replicated in other settings. The greater the differences between evaluatedprograms and other programs using the same name, the less certain or generalizablethe conclusions of this report must be.

What Works and Policy Conclusions

The uses of this report for policy conclusions require two additionalcautions. One is that program evaluations alone are clearly insufficientas a basis for making policy. Other goals programs may achieve besidescrime prevention need also to be examined. So must issues of relative cost-effectivenessthat this report is unable to address. The current state of science cannotsupport detailed analyses of where crime prevention dollars can achievethe largest return on investment.

A second caution is that programs with unknown effects should not bejudged deficient. A basic tenet of science is that the absence of evidenceis not evidence of absence--of a cause and effect relationship. Merelybecause a program has not been evaluated properly does not mean that itis failing to achieve its goals. Previous reviews of crime prevention programs,especially in prison rehabilitation, have made that error, with devastatingconsequences for further funding of those efforts. In addressing the unevaluatedprograms, we must blame the lack of documented effectiveness squarely onthe evaluation process, and not on the programs themselves. Our analysismust also address programs for which there is little or weak evidence.

Given the risk of unevaluated programs being labeled ineffective, weattempt where possible to use indirect empirical evidence or theoreticalanalysis to provide some scientifically based assessment. For example,battered women's shelters have not been evaluated, but substantial epidemiologicalevidence shows that they protect women at a very high risk time for domesticviolence. Thus indirect evidence suggests they should be effective at reducingdomestic violence, even though the specific hypothesis remains untested.Such commentary beyond the scope of program evaluations seems, on balance,to be a reasonable attempt to fulfill the Congressional mandate for thisreport.

FEDERAL GUIDANCE VERSUS FEDERAL FUNDING

A recent analysis of police organizations concluded that "researchand development is the core technology of policing" (Reiss, 1992).For police officers accustomed to thinking of guns, cars or even computersas their core technology, this statement may be quite surprising. Justas R & D is the core technology of both medicine and computer softwaremanufacturing, however, so it is for crime prevention. This is no moretrue in policing than in the six other institutions. And for the federalgovernment to leverage its scarce dollars in crime prevention, ProfessorReiss's dictum may be truest of all.

The claim that R & D is a core technology for crime prevention providesa useful framework for considering the history of the federal government'srole in state and local crime. That history can been seen as a strugglebetween guiding and funding local crime prevention, betweenan emphasis on R & D and an emphasis on program funding. The two arenot necessarily exclusive, and can even be complementary to the extentthat R & D becomes the basis for more effective use of program funding.That appears to be the premise of the Congressional mandate for this report.But any consideration of federal programs for local crime prevention mustbegin by noting the two separate, and clearly unequal, responsibilitiesCongress has assigned to the U.S. Department of Justice.

Historically, crime prevention R & D preceded local funding, andpersisted during the decade in which funding was largely abolished. Thefollowing time line summarizes the two functions:

Program Funding --------- --------- Research & Development -------------------------------------------------Years 1950s 1960 1965 1969 1980 1988 1996 

Prior to World War Two, the federal role in local crime prevention waslimited to investigation and prosecution of federal crimes, such as bankrobbery. During the Eisenhower Administration, growing concern over juveniledelinquency led to research within the Department of Health, Educationand Welfare (HEW) Office of Children and Youth. These programs were expandedin the early Kennedy-Johnson administration, especially within the NationalInstitute for Mental Health, which joined the Ford Foundation as a majorsource of funding for research on youth crime. (Ford and other foundationslargely withdrew from the crime problem after the massive increases infederal funding in the 1970s). Many of the ideas emerging from that research,especially about community development, were to become key elements inthe Johnson administration's War on Poverty.

In 1965, the federal role in local crime prevention moved beyond researchinto program development, and from HEW into the Department of Justice (DOJ).In the process, the federal role evolved into a practical emphasis on providingguidance to local authorities about preventing crime. The creationof the Office of Law Enforcement Assistance within DOJ led to grants supportingnew ideas, such as the Family Crisis Intervention Unit. Developed as apartnership between the City University of New York and the New York CityPolice Department under an OLEA grant (Bard, 1970), this project becamethe first clear example of federal guidance, with these elements:

o a locally-initiated innovative idea for a crime prevention program

o federal funds to support a demonstration of the program in one location

o federal funds to support an evaluation of the program in one location

o federal funds to disseminate the results of the program nationwide

The success of the approach was dramatic. Within a few years after DOJfunded the demonstration in New York, hundreds of police agencies aroundthe country had adopted a similar approach. The capacity of the federalgovernment to help incubate a new idea and then distribute it to the nationwas clear.

What was less clear was the capacity of the federal government to insurehigh scientific standards of program evaluation (Liebman and Schwarz, 1973).Using the scale of scientific methods employed in this report, the evaluationof the New York City project would have ranked a zero. While the programsought to reduce domestic violence, the evaluation contained no measurementof that crime problem, relying only on general crime statistics. Therewas no comparison of cases that were or were not assigned to the FamilyCrisis Intervention Unit, and no basis for determining its effectiveness.Yet when both the evaluation and the DOJ pronounced the program a success,the combined authority of science and the federal government led to widespreadreplication of the program using local tax dollars.

In the past three decades, the federal capacity to produce rigorousevaluation research has increased substantially. The federal role has helpedthe entire field of criminology to grow in both the numbers and the experienceof trained evaluation scientists; the number of doctoral programs in thefield has also increased ten fold. The field itself has a much strongerbody of knowledge about scientific issues in program evaluation, notablystatistical power. The analysis presented in Chapter Ten suggests thatthe major limitations on better crime prevention evaluations today arenot technical, but statutory. There is a clear need for a statutory planspecifying both the resources and the structure of the federal role incrime prevention R & D. In the absence of such a plan, a great dealof federal funds will be spent without any opportunity to measure theireffectiveness at preventing crime.

Most of those funds will be spent on program funding for crime prevention,which have come, gone and returned to the federal role in local crime prevention.At the peak of the violent crime epidemic of the late 1960s, the idea offederal financing of local police and corrections had enormous bipartisanappeal. The Omnibus Crime Control and Safe Streets Act of 1968 was signedby President Johnson, and then implemented by President Nixon at a costof almost $1 Billion per year. The 1968 law increased the federal R &D role by creating what became the present National Institute of Justice,Bureau of Justice Statistics, and the Office of Juvenile Justice and DelinquencyPrevention as part of the new Law Enforcement Assistance Administration(LEAA) in DOJ. But most of the billion was transferred back to the states,through each Governor's Office, for spending on a wide range of unevaluatedprograms. Some of the state expenditures, like tanks for rural police agencies,became so notorious that LEAA was ultimately abolished by Congress at theend of the Carter administration.

Operational program funding slowly returned to the federal role duringthe Bush administration, as part of the national war on drugs promptedpartly by crack cocaine epidemics in several cities. Despite the urgingof almost 40 big city police chiefs that Congress set aside even 10% ofthe drug war funding for federal R & D, the return of program fundingcontained no plan for evaluating its effectiveness. Just as in the 1960sdesign of the LEAA, Congress provided no statutory plan for developingusable knowledge from state and local programs funded by federal dollars.Sound evaluations, and the costs associated with them, remained the exception,not the rule. The Crime Bill of 1994 vastly increased program funding tohistoric highs, but provided almost no statutory language for measuringthe effectiveness of the programs funded.

Discretionary reallocations of the 1994 funds by the Assistant AttorneyGeneral for Justice Programs have breathed new life into the R & Drole, putting resources for measuring effectiveness to a new high level.The National Institute of Justice, for example, was only appropriated $31million in fiscal year (FY) 1996, but actually expended $99 million. Theadditional funds came from allowable transfers of programmatic funds. Inthe short run, these reallocations seem likely to increase the scientificevidence available for assessing the effectiveness of crime preventionprograms; even a year from now, for example, a report like this one shouldhave many new findings from rigorous research. But in the long run, therole of R & D will remain marginal to the federal role without a statutoryplan for insuring its centrality.

The key issue for such a plan is the relationship between guiding andfunding crime prevention. The two can proceed on largely separate paths,much as they have in the past. The result of that approach is an enormousopportunity cost, a lost chance to learn what works, what doesn't, andwhat's promising. By tying R & D more closely to program funding, theCongress can leverage taxpayer dollars to guide local crime preventionas well as supplement its funding. The record suggests that, dollar fordollar, the small federal investment in R & D has had far more effecton local crime prevention than the large federal investment in programfunding (Blumstein and Petersilia, 1995). Program funding provides a tinyfraction of the financial capital invested in crime prevention. Researchand development, in contrast, provides a very large fraction of the intellectualcapital invested in local crime prevention. Program funding can be farmore productive if it serves to enhance R & D.

Using program funding to enhance R & D is unlikely to happen withouta Congressional mandate. No program can be properly evaluated as an afterthought.In contrast to a financial audit, a scientific evaluation requires datacollection in advance of the program startup date. It also requires anelement of control by the evaluators in how the program is delivered, inorder to provide a valid evidence about cause and effect. While not alllocations adopting a program need to be evaluated in this way, there mustbe at least a few "laboratory" locations in which controlledtesting of crime prevention effects becomes scientifically feasible. Undercurrent statutory funding arrangements, however, Congress imposes littlerequirement on funded programs to cooperate with evaluations, and littlerequirement on federal agencies to set aside program funds to support scientificallyadequate evaluations.

This historical context sets the stage for the Congressionally-mandatedreview of program effectiveness. It reveals several key points to recallin reviewing the following chapters:

1) The vast majority scientific knowledge on the effectiveness of federalprograms is itself the product of federal investment, primarily throughDOJ; such knowledge is too costly to come from state and local tax dollars

2) The short supply of available knowledge is a direct reflection offederal under- investment in crime prevention R & D.

3) Federal program funding puts the cart before the horse, then failto even harness the horse. Crime prevention programs are funded nationwidebefore they are evaluated, and then are funded in ways that make soundevaluation almost impossible to achieve.

This report is thus a scientific assessment of both federal crime preventionprograms and federal policy for evaluating those programs. Defining crimeprevention as a result rather than an intention, the report maps out thecharted and uncharted territory of crime prevention knowledge in each ofits seven institutional settings. It distinguishes between strong and weakevidence for each part of that map, most of which is unfortunately fartoo weak. It then locates federal crime prevention programs on that map,many of which fall in uncharted territory. It concludes with an assessmentof the federal role in improving that map, and a cost-effective plan forspeeding up the rate of discovery.

NOTES

1Some developmental criminologists distinguish factors andprograms that help stop people from ever becoming offenders from thosewhich help prevent further offenses after a first offense (e.g., Tremblayand Craig, 1995). Given the difficulty in detecting offenses hidden fromthe criminal justice system, however, this distinction is made primarilyfor purposes of program operation, and not for conceptual purposes.

2104th Congress, H.R. Report 104-378, December 1, 1995, Section116.

3And as the policy debate relies increasingly on data, theimportance of the scientific strength of the evidence becomes more visible.Asra Q. Nomani and Jeffrey Taylor, "Shaky Statistics Are Driving theAirbag Debate" WALL STREET JOURNAL January 22, 1997, p. B1.

4The scores are based on direct examination of studies subjectedto primary review (see Appendix). For studies summarized from secondaryreviews, the scores are inferred from descriptions of research designsprovided in the secondary reviews.

5This criterion was employed by all chapters except for ChapterSeven, in which long time series analyses absent control groups were codedas level 3.

6Chapter Five rates some studies as level four even withouta large number of units in the comparison group.

7Chapter Five also requires that differences between treatmentand control are known and partially controlled, while Chapter Seven substituteslong time series for control groups.

REFERENCES

Anderson, Elijah

1978 A Place on the Corner. Chicago: University of Chicago Press.

Bard, Morton

1970 Training Police As Specialists in Family Crisis Intervention. Washington,DC: US Department of Justice.

Barr, R. and Ken Pease

1990 Crime Placement, Displacement, and Deflection. In Michael Tonryand Norval Morris, eds. Crime and Justice: A Review of Research. Vol. 12.Chicago: University of Chicago Press.

Black, Donald

1976 The Behavior of Law. NY Academic.

Blumstein, Alfred and Joan Petersilia

1995 Investing in Criminal Justice Research. In James Q. Wilson andJoan Petersilia, Eds., Crime. San Francisco: ICS Press.

Blumstein, Alfred, Cohen, Jacqueline, and Daniel Nagin (eds).

1978 Deterrence and Incapacitation: Estimating The Effects of CriminalSanctions on Crime Rates. Washington, DC: National Academy of Sciences.

Blumstein, Alfred

1993 Making Rationality Relevant. Criminology 31: 1-16.

Braithwaite, John

1989 Crime, Shame and Reintegration. Cambridge: Cambridge UniversityPress.

Butterfield, Fox

1995 All God's Children.

Cloward, Richard and Lloyd Ohlin

1960 Delinquency and Opportunity. Glencoe, Ill: Free Press.

Cohen, J.

1977 Statistical Power for the Behavioral Sciences. NY: Academic.

Cohen, Lawrence and Marcus Felson

1979 Social Change and Crime Rate Trends: A Routine Activities Approach.American Sociological Review 44: 588-607.

Coleman, James S.

1990 Foundations of Social Theory. Cambridge: Harvard University Press.

Cook, Philip J. and Mark H. Moore

1995 Gun Control. In James Q. Wilson and Joan Petersilia, Eds., Crime.San Francisco: ICS Press.

Center for Substance Abuse Prevention

1995 The National Structured Evaluation of Alcohol and Other Drug AbusePrevention. Washington, DC: US Department of Health and Human Services.

Ekblom, Paul and Ken Pease

1995 Evaluating Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Farrell, Graham

1995 Preventing Repeat Victimization. In Michael Tonry and David P.Farrington, Eds., Building a Safer Society. Crime and Justice, Vol. 19.Chicago: University of Chicago Press.

FBI

1994 Crime in the United States, 1993. Washington, DC:

1996 Crime in the United States, 1995. Washington, DC:

Felson, Marcus

1994 Crime in Everyday Life. Thousand Oaks, CA: Pine Forge Press.

Freeman, Richard B.

1983 Crime and Unemployment. In James Q. Wilson, Ed. Crime. San Francisco:ICS Press.

1995 The Labor Market. In James Q. Wilson and Joan Petersilia, Eds.,Crime. San Francisco: ICS Press.

Gottfredson, D.C.

1985 Youth Employment, Crime, and Schooling: A Longitudinal Study ofa National Sample. DEVELOPMENTAL PSYCHOLOGY 21: 419-432.

Gottfredson, Michael and Travis Hirschi

1990 A General Theory of Crime. Stanford, CA: Stanford University Press.

Greenwood, P.W., K.E. Model, C.P. Rydell & J. Chiesa. (1996) DivertingChildren from a 1996 Life of Crime: Measuring Costs and Benefits. SantaMonica, CA: RAND.

Hirschi, Travis

1969 Causes of Delinquency. Berkeley: University of California Press.

1986 On The Compatability of Rational Choice and Social Control Theoriesof Crime. In Derek Cornish and Ronald Clarke, Eds., The Reasoning Criminal.NY: Springer-Verlag.

1995 The Family. In James Q. Wilson and Joan Petersilia, Eds., Crime.San Francisco: ICS Press.

Hope, Tim

1995 Community Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Kelling, George L., Antony M. Pate, Duane Dieckman and Charles Brown

1974 The Kansas City Preventive Patrol Experiment: Technical Report.Washington, DC: Police Foundation.

Lane, Roger

1992 Urban Police and Crime in Nineteenth Century America. In MichaelTonry and Norval Morris, eds., Modern Policing. Crime and Justice, Vol.15. Chicago: University of Chicago Press.

Liebman, Donald A. and Jeffrey Schwartz

1973 Police Programs in Domestic Crisis Intervention: A Review. In J.R. Snibbe and H.M. Snibbe, Eds., The Urban Policeman in Transition. Springfield,Ill.: Charles C. Thomas.

Lindstrom, Peter

1996 Paper Presented to the Conference on Problem-Oriented Policingas Crime Prevention, Stockholm, Swedish National Police College.

Loeber, Rolf and David P. Farrington, Editors

Forthcoming Report of the Study Group on Serious, Chronic, and ViolentJuvenile Offenders. Washington, DC: Office of Juvenile Justice and DelinquencyPrevention

Lukes, Steven and Andrew Scull, Editors

1983 Durkheim and The Law. NY: St. Martin's Press.

Marciniak, Elizabeth

1994 Community Policing of Domestic Violence: Neighborhood Differencesin the Effects of Arrest. PhD Dissertation, University of Maryland, CollegePark.

Massey, Douglas S. and Nancy A. Denton

1993 American Apartheid: Segregation and the Making of the Underclass.

Merton, Robert K.

1969 The Matthew Effect in Science. Science.

Oldenburg, Ray

1990 The Great Good Place. NY: Paragon House.

Olds, David, Charles R. Henderson, Jr., Robert Chamberlin, and RobertTatelbaum

1986 Preventing Child Abuse and Neglect: A Randomized Trial of NurseHome Visitation. Pediatrics 78: 65-78.

OJJDP (Office of Juvenile Justice and Delinquency Prevention, Officeof Justice Programs, U.S. Department of Justice)

1996 Juvenile Offenders and Victims: A National Report. Washington,DC: OJJDP.

Petersilia, Joan and Alfred Blumstein

1995 In James Q. Wilson and Joan Petersilia, Eds., Crime. San Francisco:ICS Press.

Pierce, Glenn, Susan A. Spaar, and LeBaron Briggs, IV.

1988 The Character of Police Work: Strategic and Tactical Implications.Unpublished Ms., Northeastern University, Center for Applied Social Research.

Reiss, Albert J., Jr.

1987 The Legitimacy of Intrusion Into Private Space. In Clifford D.Shearing, and Philip C. Stenning, Eds., Private Policing. Beverly Hills:Sage.

1992 In Michael Tonry and Norval Morris, Eds., Modern Policing. Crimeand Justice, Vol. 15. Chicago: University of Chicago Press.

Reiss, Albert J. Jr., and Jeffrey A. Roth, eds.

1993 Understanding and Preventing Violence. Washington, DC: NationalAcademy of Sciences.

Ross, H. Laurence

1994 personal communication

Sampson, Robert

1986 Crime In Cities: The Effects of Formal and Informal Social Control.In Albert J. Reiss, Jr. and Michael Tonry, Eds., Communities and Crime.Crime and Justice, Vol. 8. Chicago: University of Chicago Press.

Sampson, Robert and John Laub

1993 Crime in The Making. Cambridge, Mass.: Harvard University Press.

SEARCH Group

1996 Interface. Fall.

Sherman, Lawrence W.

1992 Policing Domestic Violence: Experiments and Dilemmas. NY: FreePress.

Sherman, Lawrence W., Patrick R. Gartin and Michael E. Buerger

1989 Hot Spots of Predatory Crime: Routine Activities and the Criminologyof Place. Criminology 27: 27-55.

Sherman, Lawrence W., James W. Shaw and Dennis P. Rogan

1995 The Kansas City Gun Experiment: Research in Brief. Washington,D.C.: National Institute of Justice.

Sherman, Lawrence and David A. Weisburd

1995 General Deterrent Effects of Police Patrol in Crime "Hot Spots:"A Randomized, Controlled Trial. Justice Quarterly 12: 625-648.

Tittle, Charles and Charles H. Logan

1973 Sanctions and Deviance: Evidence and Remaining Questions. Law andSociety Review 7: 371-379.

Tremblay, Richard and Wendy M. Craig

1995 Developmental Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Toby, Jackson

1982 Violence in School. In Norval Morris and Michael Tonry, Eds. Crimeand Justice: A Review of Research. Vol.4. Chicago: University of ChicagoPress.

Tonry, Michael

1995 Malign Neglect. NY: Oxford University Press.

Tyler, Tom

1991 Why People Obey the Law. New Haven: Yale University Press.

Williams, Kirk R., and Richard Hawkins

1986 Perceptual Research on General Deterrence: A Critical Review. Lawand Society Review 20: 545-572.

Wilson, William Julius

1996 When Work Disappears: The World of the New Urban Poor. NY: AlfredA. Knopf

Chapter Three:

COMMUNITIES AND CRIME PREVENTION

by Lawrence W. Sherman

..many community characteristics implicated in violence, suchas residential instability, concentration of poor, female-headed householdswith children, multiunit housing projects, and disrupted social networks,appear to stem rather directly from planned governmental policies at local,state and federal levels.

--National Academy of Sciences report, 19931

Communities are the central institution for crime prevention, the stageon which all other institutions perform. Families, schools, labor markets,retail establishments, police and corrections must all confront the consequencesof community life. Much of the success or failure of these other institutionsis affected by the community context in which they operate. Our nation'sability to prevent serious violent crime may depend heavily on our abilityto help reshape community life, at least in our most troubled communities.Our good fortune is that the number of those troubled communities is relativelysmall. Our challenge is that their problems are so profound.

Serious violent crime is not a problem for most residential communitiesin the United States. In the suburban areas where most Americans live,the homicide rate is comparable to Finland's (FBI, 1994: 191; Reiss andRoth, 1993: 52). Half of all American homicides occur in the 63 largestcities, which only house 16% of the U.S. population. Homicides in thosecities are also highly concentrated, in a handful of communities markedby concentrated poverty, hypersegregation (Massey and Denton, 1993), familydisruption and high gun density. Almost 4% of all homicides in Americainvolve gang members in Los Angeles County alone (Klein, 1995: 120). Seriousviolent crime in America is predominantly a matter of one particular kindof community, increasingly isolated and shunned by the rest of Americansociety (Wilson, 1996).

The causation of inner-city crime has received extensive diagnosis (Wilson,1987, 1996; Massey and Denton, 1993; Bursik and Grasmik, 1993; Sampsonand Lauritsen, 1993). The prevention of inner-city crime has been attemptedwith extensive programs. The connection between causes and prevention,however, has been weak at best, and often nonexistent. More than any ofthe other institutional settings, the community setting shows a strikingdivergence of causal analysis and prevention programs. The causes, or atleast the risk factors correlated with serious crime,2 are basicand interconnected, while the programs are superficial and piecemeal. Federalpolicies from urban renewal to public housing may have done more to causeinner city violence than to prevent it (Sampson and Lauritsen, 1993: 89).For most of this century, community crime prevention programs have failedto tackle the governmental policies and market forces that fuel inner-cityviolence.

A central issue in the disconnection between causes and cures is theassumptions of how these communities "got that way." As WilliamJulius Wilson has observed, "The segregated ghetto is not the resultof voluntary or positive decisions on the part of the residents... [butis] the product of systematic racial practices such as restrictive covenants,redlining by banks and insurance companies, zoning, panic peddling by realestate agents, and the creation of massive public housing projects in low-incomeareas." The result of these forces in recent years has been called"hypersegregation:" historically unprecedented levels of geographicsegregation by race and class, magnifying the effects of poverty and racialisolation (Massey and Denton, 1993). Yet community prevention programsaddress none of these causes of community composition and structure, whichin turn influence community culture and the availability of criminogenicsubstances like guns and drugs.

Ironically, a central tenet of community prevention programs has beenthe empowerment of local community leaders to design and implement theirown crime prevention strategies. This philosophy may amount to throwingpeople overboard and then letting them design their own life preserver.The scientific literature shows that the policies and market forces causingcriminogenic community structures and cultures are beyond the control ofneighborhood residents, and that "empowerment" does not includethe power to change those policies (Hope, 1995). It is one thing, for example,for tenants to manage the security guards in a public housing project.It is another thing entirely to let tenants design a new public housingpolicy and determine where in a metropolitan area households with publichousing support will live.

Even the management of modest programs with federal support are oftenbeyond the capacity of community organizations, especially where it isneeded the most. The consistent evidence of the neighborhood watch programs,for example (Skogan, 1990: chapter 6), is that the more crime and riskfactors a neighborhood suffers, the less likely it is to develop any organizedactivity to fight crime. When community organizations do get involved inadministering federal funds, there are often major problems and scandalsof financial mismanagement. "Empowering" local communities withfederal funding often turns into no applications from the worst areas andred tape nightmares for the not-so-bad areas that do get involved.

The disconnection between causation and prevention is also clear inthe official use of the term "comprehensive." To be comprehensivein addressing risk factors is very different from being comprehensive inmobilizing all available agencies of government. Recent "comprehensive"crime prevention programs merit the term more by agency participation thanby risk factors. The fit between agencies and risk factors is good in afew cases, such as home nurse visitation to address single parent childraisingpractices (see Chapter 4). But many risk factors have no obvious agencyto fix them. Even multi-agency coordination is no guarantee that the majorrisk factors, like hypersegregation and labor market isolation (see Chapter6), will be addressed.

Thus the major causes of community crime problems are like handcuffslocking a community into a high crime rate. The most frequently evaluatedcommunity-based crime prevention programs do not attempt to break thosehandcuffs. Rather, they operate inside those constraints, attempting "smallwins" within the limited range of risk factors they can manipulate.But until the handcuffs of race-based politics themselves are unlocked,many analysts expect relatively little major improvements from programsaddressing only the symptoms of those constraints.

Given the disconnection between causes and cures, it is not surprisingthat program impact evaluations provide little strong evidence of effectivecrime prevention. Setting aside programs delivered in families, schools,labor markets, places or the criminal justice system, the number of evaluationsof community-based programs is quite small and generally discouraging.While there have been some "small wins," like reduced vandalismand drug use in housing projects with recreational programs, there havebeen no scientifically documented "big wins" preventing violencein a concentrated urban poverty area. Within that context, community mobilizationefforts, gang prevention programs, gun buybacks, social worker and recreationprograms have generally failed to show much if any effect on crime.

Yet the evaluation methods for these programs have generally been quiteweak, and there is no certainty that such programs are doomed to failureeven though they sidestep the central causes reflected in the scientificliterature. Amidst generally negative results from generally weak programevaluations, there are encouraging findings from some research that maymerit further testing, even though other studies have found contradictoryresults:

o Gang violence prevention has been effective in several casestudies

o Community-Based Mentoring prevented drug abuse in one rigorousexperiment

o Afterschool Recreation programs have reduced vandalism in publichousing

These findings about community-based programs addressing "proximate"rather than "root" causes suggest a strategy for developing nationalcrime prevention policy. Both the Justice Department and the rest of thefederal government are moving towards concentration of resources on high-crimeinner-city areas, which one-third of all African-Americans reside (Masseyand Denton, 1993: 77) and where community factors generate the high homicidevictimization rate of young black males--which is twelve times higher thanthe average in the US population (Fingerhut and Kleinman, 1990). Whetherthe efforts now in planning can address the structural factors is an unansweredquestion. But a combination of programs addressing proximate causes andthe structural factors may have the best chance of success.

It is also possible that the diagnosis of community crime causationis incomplete. Even in the face of profound urban problems, it may be possibleto reduce substantially the level of serious crime. New York City homicidesand shootings dropped in half in recent years, with no documented changein concentrated urban poverty. It is not clear how or why that reductionoccurred. The leading theory is the application of the police methods foundeffective in the studies reviewed in Chapter Eight. No community-levelprevention program (or demographic change) has emerged as an alternative,competing explanation. But it remains possible to design such a program,focused more on the proximate than on the root causes of serious violence,and to test it in a randomized trial on a large multi-city sample of urbanpoverty areas. Programs currently planned by the executive branch to improveinner-city conditions can be most beneficial if they are structured toallow such a rigorous evaluation, so the nation can be very clear aboutthe precise effects of the program on crime.

This chapter compares scientific evidence about community risk factorsfor violent crime to the logic of community crime prevention programs.It briefly reviews some methodological issues in evaluating those programs.It then examines the limited impact evaluations of crime prevention programsbased in community settings outside the institutions examined in the nextsix Chapters. The chapter concludes by comparing the science of community-basedcrime prevention to major DOJ funding programs, with policy recommendationsfor both programs and research.

COMMUNITY RISK FACTORS FOR VIOLENT CRIME

The science of crime causation, while still in its infancy, offers morethan a century of research on the community characteristics associatedwith higher risks of violent crime (Quetelet, 1842). By "community,"this literature usually denotes residential areas of varying size withincities. These areas may be as small as blocks (Taylor and Gottfredson,1986) or cover several square miles (Shaw and McKay, 1942). Much of thisliterature, recently reviewed for an NIJ-funded National Academy of SciencesPanel (Sampson and Lauritsen, 1993), uses rates of homicide and other seriousviolent crimes as the major focus.

One framework for classifying community risk factors distinguishes communitycomposition, social structure, oppositional culture, legitimate opportunities,and social and physical disorder. Each of these apparent risk factors couldbe the focus of comprehensive community crime prevention programs. Mostare not. Instead, as the National Academy of Sciences report suggests,"non-crime" government policies may have done more over the pastfour decades to enhance these risk factors than to reduce them. Perhapsthe most visible example is the construction of public housing projects(Bursik, 1989), which in one study was followed by increased populationturnover and increased crime rates independent of race.

Community composition refers to the kinds of people who livein a community. Unmarried or divorced adult males, teenage males, non-workingadults, poor people, persons with criminal histories and single parentshave all been identified in the literature as the kind of people whosepresence is associated with higher rates of violent crime (Messner andTardiff, 1986; Sampson, 1986; Curry and Spergel, 1988; Bursik and Grasmik,1993). What is unclear in the literature is whether having more such peoplesimply produces a higher total of individual level risk factors, or whetherthere is a "tipping" effect associated with the concentrationsof such people (Sampson and Lauritsen, 1993). The latter theory derivesfrom substantial findings on the effects of proportions in groups and corporations(Kanter, 1977): in which behavior of entire communities changes when aproportion of one type of person goes beyond the tipping point.

Public policies contributing to the concentration of high-risk peoplein certain neighborhoods include the federally funded highway system thattook low-risk people out of urban neighborhoods to the suburbs (Skogan,1986). The suburbanization of both white middle class people through highways,and black middle class people through federal open-housing laws (Wilson,1987), helped tip the proportions of many inner city communities towardsa majority of persons or families at higher risk of crime. As long as thosehigh-risk families or persons were in a minority, their low risk neighborswere able to exercise a community protective factor against violent crime.When the high-risk families became a majority in many urban communities,a spiral of crime and the fear of crime led to further loss of middle classresidents and jobs. This in turn increased the concentration of unemployedand poor people, followed by further increases in crime (Schuerman andKobrin, 1986; Wilson, 1996.) No federal or local public policies have yetto counteract, or even challenge, these proportional imbalances.

Community Social Structure. Independently of the kinds of peoplewho live in a community, the way in which they interact may affect therisk of violent crime. Children of single parents, for example, may notbe at greater risk of crime because of their family structure. But a communitywith a high percentage of single parent households may put all its childrenat greater risk of delinquency by reducing the capacity of a communityto maintain adult networks of informal control of children. The greaterdifficulty of single parent families in supervising young males is multipliedby the association of young males with other unsupervised young males,since delinquency is well-known to be a group phenomenon (Reiss, 1988).The empirical evidence for this risk factor is particularly strong, withviolent victimization rates up to three times higher among neighborhoodsof high family disruption compared to low levels, regardless of other characteristicssuch as poverty, and the correlation between race and violent crime atthe neighborhood level disappears after controlling the percentage of female-headedhouseholds (see Sampson and Lauritsen, 1993).

Other aspects of community structure include the prevalence of unsupervisedmale teenage groups, the density (or extent of overlap) among local friendshipnetworks, and local participation in formal voluntary associations. Supportfor the inverse correlation of violent crime with voluntary associationmembership has been found at the block level in Baltimore (Taylor et al,1984). Sampson and Groves (1989) found support for dense friendship networksas a protective factor and unsupervised teen groups as a risk factor forviolence in the British Crime Survey. All of the risk factors have arguablybeen concentrated in urban neighborhoods by public policies. Skogan (1986)reviews the evidence on urban renewal's destruction of dense local friendshipnetworks, uprooting entire neighborhoods; nationwide, 20 percent of allurban housing units occupied by blacks were demolished during the 1970s(Logan and Molotch, 1987: 114, as cited in Sampson and Lauritsen, 1993:88). Wilson (1987) and Massey and Denton (1993) trace the history of publichousing policy decisions that concentrated poor, black, female-headed householdsin limited areas rather than dispersing them amidst other kinds of families(Lemann, 1991). While community mobilization programs are designed in partto build voluntary association membership and increase informal socialcontrol, the evidence to date suggests that such efforts have not succeeded(Hope, 1995).

Oppositional Culture. Observers of high crime neighborhoods havelong identified the pattern of "oppositional culture" arisingfrom a lack of participation in mainstream economic and social life: badbecomes good and good becomes bad. Given the apparent rejection of communitymembers by the larger society, the community members reject the valuesand aspirations of that society by developing an "oppositional identity"(Cohen, 1955; Clark, 1965; Braithwaite, 1989; Massey and Denton, 1993:167). This is especially notable in terms of values that oppose the protectivefactors of marriage and family, education, work and obedience to the law.As inner-city labor force participation rates have declined (Wilson, 1996)and inner-city segregation has increased over the past three decades (Masseyand Denton, 1993), the strength of the opposition has increased. Ethnographicstudies of such cultures in recent years (e.g., Anderson, 1990) show moreintense opposition than similar studies in the 1960s and 1970s (e.g., Liebow,1967; Anderson, 1978), which found more widespread acceptance of mainstreamvalues. Efforts to gain "respect" in oppositional cultures maythen rely more on violence than on other factors (Anderson, 1990). Publicpolicy has contributed to this primarily by its historical support forsegregation and its modern failure to prevent its inner-city concentration,both by race (Massey and Denton, 1993: chapter 7) and joblessness (Wilson,1996: chapter 3).

Criminogenic Commodities. Communities with very high rates ofyouth violence are places in which there are high concentrations of criminogeniccommodities (Cook and Moore, 1995). Both alcohol use (Collins, 1989) anddrug use (Goldstein, 1989) are highly correlated with violent crime atthe situational level of analysis (Miczek, et al, 1993), and gun use incrime generally causes greater risk of homicide (Cook, 1991; Reiss andRoth, 1993). Other evidence suggests that high crime communities appearto have very high concentrations of locations selling alcohol (Roncek andMaier, 1991) and drugs (Sherman and Rogan, 1995). Whether the disproportionatepresence of these substances reflects market demand arising from oppositionalculture or other reasons (including public policy) is an unresolved issuein the literature.

Social and Physical Disorder. Recent work on the "brokenwindows" (Wilson and Kelling, 1982; Kelling and Coles, 1996) theoryof community crime causation suggests some support for the theory (Skogan,1990). The theory claims that in communities where both people and buildingsappear disorderly, the visual message that the community is out of controlmay attract more serious crime (Wilson and Kelling, 1982). This may happenby a spiral of increasing fear of crime among conventional people, whouse the area less and thus provide less informal control. Communities thatdeteriorate in this respect over time are observed to suffer increasedrates of violence (Schuerman and Kobrin, 1986). Public policies contributeto such declines through nonenforcement of building code violations (Hirsch,1983) and of minor criminal conduct such as public drinking (Kelling andColes, 1996). Demolition policies to reduce the unsightly appearance ofdecayed buildings may then also reduce neighborhood density of street populations,the effect of which is not clear in the literature; lower density may eitherincrease the risk of violent crime (Wilson, 1996) or reduce it (Sampsonand Lauritsen, 1993).

All of these risk factors and more are connected to broader debatesabout race, poverty, welfare, unemployment and family life in America.These debates often ignore the extreme inner-city concentrations of theserisk factors. These concentrations are both extreme in each category andin their accumulation. Few neighborhoods in the US suffer nonemploymentrates as high as 63 to 77 percent of all adults. The ones that do are alsolikely to suffer from weak social structure, high rates of alcohol abuse,gun carrying, drug abuse, and violent youth crime. To the extent that policydebates focus on these issues outside of the inner-city areas of concentration,it may fail to attack the interdependence between these risk factors.

EVALUATING COMMUNITY CRIME PREVENTION

In order to learn whether federal policies can at least reduce violentcrime in such communities, both strong programs and strong scientific methodsshould help. In this context, "strong" programs would addressmultiple risk factors simultaneously, while "strong" scientificmethods would isolate the separate effects of different program elements.Using these definitions, the current state of the science offers no strongtests of strong community crime prevention programs.

The evaluations reviewed in this chapter generally employ weak researchdesigns to test programs focused on symptoms of community risk factors,rather than the basic risk factors themselves. This limits our abilityto draw conclusions about what effects, if any, the evaluated programsreally have. As Chapter Two explains, all evaluations are not created equal.Some of them provide far stronger evidence about cause and effect thanothers. The strong ones generally employ large samples, reliable measuresof both program operations and their intended effects, and possible rivalcauses of those effects. The weaker ones, quite common in this chapter,may measure program content and crime, but do a very poor job of measuringother factors that may affect crime besides the program.

This chapter uses the scale of scientific methods scores presented inChapter 2. On a scale of 1 to 5, each specific evaluation reviewed is rankedfor its capacity to support strong conclusions about the effect of theprogram. This strength of evidence is often unrelated to costs, or eventhe theoretical strength of the program being tested. The massive Chicagogang prevention project of the early 1960s, for example, gathered detailedrecords on thousands of interactions between the gang workers and areayouths. But because the program area was the unit of analysis, not thoseinteractions, the actual sample size was only 4 areas, and the power toinfer cause and effect was quite low. Any number of other factors couldhave caused crime in those areas to go up or down besides the presenceor absence of the intensively measured gang prevention programs.

This problem poses a serious obstacle to advancing scientific knowledgeabout community-based crime prevention. Community risk factors can onlybe addressed and measured one community at a time. The cost of measuringsome factors is very high. Multiplying that cost across a substantial sampleof communities has long been deemed prohibitive by research funding agencies.Yet the cost of inner-city violence is also very high. The cost of morerigorous program research could be well justified if it led to more effectivecommunity-based prevention programs. In the absence of such investmentto date, however, there is not a single large-sample randomized controlledtrial in which the community is the unit of analysis and the outcome measureis serious crime.

A related problem of scientific method is the simultaneous applicationof more than one program to a community at a time. These combinations oftreatments are usually premised on the rationale that the more programs,the better: comprehensively attacking many risk factors at once shouldincrease the overall chances of successful crime prevention. In the wordsof one observer, the theory is that "only everything works."The problem is that even with successful results, a combination of programsmakes it impossible as a matter of scientific method to isolate the activeingredients causing the success. It may be all of them in combination.Or it may be only one or two.

A third related issue is the choice of program elements. Many fundingprograms leave the choice of specific prevention programs up to local communities.Local assessment of specific community risk factors and local decisionsabout program content are a key part of many community-based strategies(Hawkins, et al, 1995). But from a scientific standpoint, the variabilityin these combinations across communities allows an evaluation to test theeffects of the general strategy, and not the specific program elements.Research designs in other fields have been used to systematically varythe program combinations, and determine across large samples which combinationsare most effective, holding other factors constant through random assignment.This approach, or some variant of it, can be used in evaluating communityprograms, and may be implemented soon in England (Farrington, 1997).

There is no necessary tradeoff, as some have suggested, between comprehensiveprograms and scientific evaluations. While the operational and researchproblems in multi-community designs are clearly complex, they can be addressedwith sufficient time and resources. As recent DOJ crime prevention policyhas moved in the direction of comprehensive community programs, both thenumber of treatments and the number of communities have become increasinglycritical aspects of the potential return on evaluation dollars. The scientificsolution to the methodological limitations observed so far is larger samplesizes, with varying combinations of the treatments. The best argument infavor of this "big science" solution is the evidence that follows,and the extremely limited conclusions we can draw from the $100 millionor more (in current dollars) of private and public funds that it cost overthe past three decades to conduct the studies examined below.

COMMUNITY MOBILIZATION

The most visible community-based crime prevention strategy in the latterTwentieth Century has been community mobilization. The definition of thisterm has varied widely, from the creation of formal community developmentorganizations to the mobilization of resources from outside the communityto help solve local problems like crime and unemployment. Hope's (1995)review of the evaluations of these programs finds virtually no evidencethat the programs attempted to date have achieved an impact on crime. Insome cases, as in New York City's Mobilization for Youth Project of the1960s, that is due to the lack of crime impact evaluations. In other cases,it is due to a failure to implement successfully the programs selectedby community leadership to a degree sufficient to test the theory of theprogram. Whether the approach could be successful under conditions otherthan those evaluated to date remains unknown.

The Eisenhower Foundation's support of nonprofit community organizationsin ten low-income neighborhoods in the late 1980s offers one of the bestevaluations available (Scientific Methods Score = 3; Lavrakas and Bennett,1989, as cited in Hope, 1995: 39-40). Its most encouraging finding is thateight of the ten sites actually implemented programs chosen during theplanning process. This stands in strong contrast to the police-generatedneighborhood watch programs reviewed in Chapter Eight, for which the majorproblem in low-income areas has been successfully organizing block or apartmenthouse meetings of neighborhood residents. The Eisenhower site programsthat were implemented ranged from individual-level social service provisionto attempts to change community social structure. The evaluators concludedfrom the impact evaluations that there was "little evidence that the...Program had documentable successes in achieving its major goals of crimereduction and improved quality of life."

These results may stem in part from what Hope (1995) calls the differencebetween "vertical" and "horizontal" strategies of communitycrime prevention. Horizontal strategies focus on aspects of community lifeand place accountability on community members to solve their own problems.Vertical solutions focus on the linkages between community life and decisionsmade at higher levels of power outside the community, from factory closingsto bank redlining of mortgages. Recent scholarly analyses of communitycrime causes (e.g., Wilson, 1996) focus more on vertically determined dimensionsof community life, while few prevention programs evaluated to date havedrawn heavily on a vertical approach. Uses of vertical solutions to datehave been relatively limited, such as seeking external assistance in streetclosings, assigning more police, and other city government decisions thatleave untouched most of the risk factors cited above. But even local governmentdecisions may make a difference.

In the NIJ-sponsored Hartford experiment in the early 1970s (Fowlerand Mangione, 1986), the community mobilization of a resident organizationwas successful at street closing and obtaining increased police activity.Initial reductions in crime, however, were followed by increases in thethird and fourth years of the program. This scientifically weak (ScientificMethods Score = 2) evaluation lacked a comparison area, which limits theinterpretation of the target area crime trends. But it is of interest thatin the two years after local police activity was reduced, resident mobilizationrose to its highest program levels. But despite the peak level of communitymobilization, robbery and burglary rose to their highest levels in thelife of the project.

It may be that mobilization alone cannot bear down directly on crime,and that the "horizontal" theory of community crime preventionis not likely to succeed. Further experimentation with different "vertical"tactics may be needed to find out if community mobilization or other methodsto affect decisions external to the local community can change such decisionsin ways that cause local crime prevention.

COMMUNITY PREVENTION OF GANG VIOLENCE

The disconnection between causes and cures in community crime preventionis illustrated by our nation's approach to gang violence. Five recent reviewsof this literature provide the evidence for this analysis (Klein, 1995;Spergel, 1995; Howell, 1995, forthcoming; Thornberry, forthcoming). Takentogether, this research suggests four major conclusions:

1. Most government and private programs for gang prevention have beenleft unevaluated.

2. The few evaluated programs have either failed to decrease gang violencee,or have actually increased it.

3. Gang prevention programs have ignored the most likely causes of therecent growth of gangs, the community structure of growing urban povertyghettoes.

4. Nonetheless, successful methods for preventing gang violence havebeen demonstrated in case studies, and could be subjected to controlledtesting on a larger scale.

This section reviews the connection between gang membership and seriousviolent crime, the evidence on the causes of gang membership, and the evaluationsof community-based programs for preventing gang violence. It concludesthat while most evaluations have been negative, the scientific rigor ofthe studies has been weak. The case studies demonstrating success in preventinggang violence can be tested with much greater scientific rigor as possiblenational models. The high concentration of serious juvenile violence amonggang members provides ample justification for large-scale research anddevelopment.

Gang Membership and Serious Crime

The basic question about gang prevention is whether it would have anyimpact on serious and violent crime. Success at gang prevention is onlyimportant to communities if eliminating gangs would reduce the number ofserious crimes. The answer to that question has not been clear from thescientific evidence. Fortunately, a substantial investment in researchby the U.S. Department of Justice Office of Justice Programs (OJP) hasrecently provided strong scientific evidence on the question. The Officeof Juvenile Justice and Delinquency Prevention (OJJDP) Study Group on Serious,Violent and Chronic Juvenile Offenders shared with the University of MarylandCrime Prevention Project its draft report, one chapter of which reviewsthis evidence (Thornberry, forthcoming). The chapter examines longitudinaldata on the connection between gang membership and serious crime in twobirth cohort studies. It breaks the question into two parts:

o How much serious crime is committed by gang members?

o Does gang membership make any difference in the harm caused by thepeople who join gangs, or would they have committed the same amount ofserious crime even without joining a gang? That is, do gangs facilitateserious crime, or merely recruit serious criminals?

Thornberry reports that in Rochester, NY, one-third of a panel of adolescentmales reported being a member of a gang at some point before the end ofhigh school. That same one-third committed 90 percent of the serious crimesin the entire panel, including 80% of violent crimes and 83% of drug sales.Thornberry also summarizes similar results from the NIH Center for SubstanceAbuse Prevention (CSAP)-funded study of gang members in the Seattle SocialDevelopment Project (Battin, Hill, Hawkins, Catalano, and Abbott, 1996,as cited in Thornberry, forthcoming). Gang members in Seattle comprisedonly 15% of the sample, but accounted for 85% of all robberies committedduring grades 7 to 12, and 62% of all drug selling. Thornberry reportslower gang contributions for gang crime in Denver from Esbensen and Huizinga's(1993) panel data: with 6% of respondents reporting gang membership, gangmembers reported 35% of serious offenses and 42% of drug sales.

The hypothesis that gangs cause juveniles to commit more seriouscrimes than they would commit anyway receives a rigorous test in the OJJDPRochester Youth Study. Thornberry et al (1993, as cited in Thornberry,forthcoming) report that gang members commit crimes against persons twiceas often while they are active members of gangs than before and after activemembership. Similar patterns were found for crimes in general and druguse, but not for property offenses. Thornberry (forthcoming) reports thatsimilar patterns were observed in the Seattle CSAP project, except thatinvolvement in drug sales in Seattle remained elevated even after gangmembership ended (Hill, Hawkins, Catalano, Kosterman, Abbott and Edwards,1996, as cited in Thornberry, forthcoming). More recent analyses of theRochester data also show drug sales, as well as gun carrying, persistingat elevated rates even after gang membership ends (Lizotte et al, 1996).

Large sample, multiple interview, longitudinal self-reported offendingstudies are the strongest evidence possible on these questions. The studiesreported here do not necessarily reflect the effects of gang membershipin the highest-crime areas of the very large cities where serious juvenileviolence is most concentrated. But the available evidence is clear enoughto establish gang membership as a community risk factor appropriate forpreventive programs. There is also a scientific basis for distinguishinggangs from drugs as a cause of violence, since Klein (1995) finds far moregang homicides without a drug link than with one.

Successful prevention of gang membership for substantial portions ofadolescent males might reduce their rates of serious crime. Even amonggang members, interventions to divert them from gang violence could preventmany crimes. The question then becomes how prevention or diversion canbe accomplished at the community level of intervention. As a matter ofscience, the logical starting point is to attack the causes of gang membership.

Causes of Gang Membership

At the individual level of analysis, the causes of gang membership appearlittle different from the causes of delinquency in general (Thornberry,forthcoming). While the cumulation of disadvantages in life is a risk factorfor both delinquency and gang membership, it is not clear why in the samecommunity, some boys join gangs and others do not (Spergel, 1995).

At the community level of analysis, however, the patterns are somewhatclearer. The key fact to be explained is why gangs have spread so rapidly--almostcontagiously--over the past decade, from a few big cities to virtuallyall large and middle-sized cities and many smaller cities and towns. Klein(1995: 91) reports a 345% increase in the number of cities reporting violentgangs from 1961 (54 cities) to 1992 (766 cities). The 1995 National YouthGang Survey found 2,000 jurisdictions reporting 23,000 gangs with some665,000 members (Moore, 1996, in Howell, forthcoming). Within cities inwhich gangs have been well-established for decades, gang-related homicideshave also risen dramatically, such as the 392% increase in Los AngelesCounty from 1982 to 1992 (Klein, 1995: 120). Klein (1995: 194) concludesthat while the rise of homicides is partly driven by the growth in guncarrying, the growth of gangs themselves is strongly linked to the rapidgrowth of urban "underclass" areas.

Drawing heavily on William Julius Wilson's (1987) analysis of the newurban poverty ghettoes, Klein isolates five factors: the loss of industrialjobs, out-migration of middle-class blacks, growing residential segregationof inner-city blacks, increasing failure of schools to prepare inner-citychildren for a service economy, and the consequent strains on family lifeof the declining ratio of "marriageable" (that is, employed)males to females of child-bearing years. Hagedorn (1988) applies this theoryto the case study of Milwaukee, and finds a good fit with the facts: gangmembership and violence rose as the Wilson model of concentrated urbanpoverty developed in that city. Huff's (1989) comparison of gangs in Columbusand Cleveland found much more rapid growth in Cleveland, where the Wilsonmodel had rapidly accelerated, than in Columbus, where community factorshad remained fairly static. Jackson (1991) found across a large sampleof cities that two factors predicted whether they developed gangs, jobopportunities and the proportion of the population ages 15 to 24.

Klein's own work with Fagan (reported at Klein, 1995: 204) finds that1970 Census data on community characteristics at the city level predictgang emergence in the 1980s. Specifically, racial segregation and a lowproportion of persons in the labor force in 1970, although not concentrationof poverty in 1970, predicts the 1980s emergence of gangs. So does an interactionof the loss of manufacturing jobs and unemployment rates. Different patternsare evident, however, for blacks and Hispanics, with strong effects forthe former but not the latter. Curry and Spergel (1992) also report black-Hispanicdifferences in causes of gang growth, with more emphasis on cultural factorsfor Hispanics and structural factors for blacks. These findings lead Klein(1995: 205) to this conclusion about the design of gang prevention programs:"at least some portion of the gang proliferation problem is reflectiveof larger social ills. Merely addressing gang problems through gang intervention,be it street work or suppression, won't have much effect."

Evaluations of Gang Prevention Programs

The impact evaluation literature is largely consistent with Klein'sconclusion. Howell's (1995, forthcoming) review of these data for OJJDPincludes nine studies, from which "nothing has been demonstrated throughrigorous evaluation to be effective in preventing or reducing serious andviolent gang delinquency, [although] a number of promising strategies areavailable" (Howell, forthcoming, p. 21). Spergel's (1995: 256) independentreview of the same evidence reaches the same conclusion: "traditionalsocial intervention programs, whether agency-based, outreach or streetwork, or crisis intervention, have shown little effect or may even haveworsened the youth gang problem."

Gang Membership Prevention. Three studies test a gang membershipprevention program on a population of potential gang members. The firstevaluation dates to the 1930s, when University of Chicago gang scholarFrederic Thrasher (1936, as cited in Howell, forthcoming) directed a four-yearstudy of the "character-building" and recreation programs ofa New York City Boys' Club. His conclusion sounds much like Klein's a half-centurylater: the program was unable to prevent gang membership due to family,school and poverty problems. "These influences for the most part werebeyond the power of the Boys' Club to neutralize" (p. 78). The secondstudy is a description of a grass-roots residential and nonresidential"sanctuary" from street life in Philadelphia (Woodson, 1981),without a comparison group. The House of Umoja also initiated "gangsummits," so it is difficult to credit the city-wide drop from thirty-ninegang homicides in 1973 to one in 1977 to prevention alone.

The third prevention program (Thompson and Jason, 1988, as cited inHowell, forthcoming) consists of a gang prevention curriculum and afterschoolrecreational activities offered to eighth grade students suggests. Theevaluation's conclusion that the program was successful is based on a differenceof three more students who became gang members in the comparison group(4 out of 43) than in the experimental group (1 out of 74). The evaluationdesign also suffered substantial attrition between exposure to treatmentand followup interview, as well as the common problem of school-based evaluations(see Chapter Five): the treatment was assigned at the level of the school,but evaluated at the level of the student. The design featured three pairsof schools, with one in each pair assigned to receive the program. Theoutcome data are not reported at the school level, but the base rate ofgang membership in the short followup period renders most other aspectsof the design less important. In sum, there is little empirical basis forpromise in the Thompson and Jason (1988) evaluation of the gang preventioncurriculum and afterschool program.

Gang Intervention. The programs for intervening with alreadyactive gangs and gang members are somewhat more rigorously evaluated. Whilethe oldest and most influential of all gang intervention and preventionprojects, the Chicago Area Project, has never been evaluated, its primarycomponent has been evaluated several times. That component is the "detachedworker," a trained youth counselor who spends most working hours onthe streets with gang members. The role and function of these workers variessomewhat across projects, largely on a dimension of how much formal programmingthey organize, such as club meetings or outings to major league baseballgames. Some detached workers also try to organize adults into voluntaryassociations, and to develop community-level capacity for leadership andproblem-solving. The workers vary in the extent to which they focused ongangs as groups or on gang members as individuals. The common core of theirrole is an attempt to redirect gang energy towards legitimate activity,including school and work, as well as to discourage crime.

Despite these variations on the theme, none of the evaluations of detachedworker programs found any evidence of reduced crime. Klein (1971), in fact,found just the opposite in an African-American area of Los Angeles: thedetached workers increased the level of crime, which declined after theprogram was terminated. His explanation for that result is that the detachedworkers enhanced group cohesion, which in turn increased the "productivity"of the gang with its major product, crime. The theoretical significanceof that conclusion is enormous, given the implications for other gang programsthat may also increase cohesion. Durkheim's basic principle that groupsolidarity is increased by external attack would apply, for example, topolice efforts to lock up a gang. Such a struggle with authorities canprovide glory and meaning to otherwise barren lives, and simply encouragemore violence.

In a followup study, Klein (1995: 146)) applied the group cohesion theoryin an explicit attempt to minimize it. The Ladino Hills program testeda strategy of working only with 100 Hispanic gang members as individuals,not with the gangs as a group. Detached workers in this evaluation encouragedgang members to drop out of the gang, which some of them did as long asthe workers were around; gang arrests declined 35% during that period.Gang cohesion also remained low for a six month followup period after theprogram ended. Several years after the program ended, Klein reports, gangcohesion and crime returned to its baseline levels. He concludes (1995:147) that gangs "cannot long be controlled by attacks on symptomsalone; community structure and capacity must also be targeted."

Limited evidence against the cohesion hypothesis, however, comes froma California Youth Authority program in Los Angeles in the mid-1970s (Torres,1981, cited in Klein, 1995: 149). Over four years, cohesion-building effortswith seven Hispanic gangs, including sports activities, served as a basisfor truce meetings and feud mediation. Homicides and intergang violencedeclined among the targeted gangs, but not between targeted gangs and othergroups. Klein (1995: 149) is skeptical about the reliability of the policedata on "gang" crimes, but concludes that "further researchattention to such intensive efforts as took place in this CYA project certainlyseem warranted."

Table 3-1

Findings from Gang Prevention and Intervention Evaluations

(Secondary Sources: Howell 1995, forthcoming; Klein, 1995)

Primary Evaluation Scientific Rigor Program Content Program Effects Score Gang Membership Prevention Thrasher 1936 ? NYC Boy's Club No preventive effect Woodson 1981 2 House of Umoja, Gang Murders declined Philadelphia Thompson & Jason 2 12 Gang Prevention Major attrition, 1988 Classes; some small N joined gangs; afterschool options 1 of 74 Experimentals, 4 of 43 Comparison Gang Member Intervention Miller 1962 3 Goal: turn gangs No effect on into clubs, 7 delinquency measures detached workers, of targets 205 boys Gold & Mattick 1974 3 Detached Workers No effect on area cited in Spergel focused on gangs; crime or gang crime; 1995: 249 community slight effect on organization educational goals Bibb 1967 ? NYC Detached Workers No effect on gang with gangs crime Klein 1969 2 LA Group Guidance Project increased 5 detached workers delinquency; more 5 gangs, weekly program, more crime; meetings, program crime reduced when program ended Klein 1968, 2 100 Ladino Hills 35% reduction in gang 1995:145-147 gang members arrests from less encouraged to leave gang cohesion; effect gangs, 18 months lost after 2 yrs Torres 1981 2 Older gang leaders Homicides and hired as intergang violence consultants, truces declined among target and feud mediation gangs, not other gangs Spergel 1986 3 Crisis intervention Less serious crime & mediation by for juveniles, more detached workers for adults, in target than control Spergel 1995 3 Conflict mediation, 50% less serious job and school violence for target referrals, police gangs and social workers Goldstein, Glick and ? Anger Replacement Reductions in gang Carthan (1994) training for gang arrests members 

Most other evaluated gang programs had far less success than the CYAor Ladino Hills projects, even with the symptoms of community structure.It was not for lack of effort. The intensity of gang worker efforts isdescribed in one summary of the six years of work of the Chicago YouthDevelopment Project (CYDP), a privately-sponsored program combining detachedgang workers with community organization (Carney, Mattick and Callaway,1969: 15, as quoted in Klein, 1995:144):

Staff succeeded in finding 750 jobs for 490 young people; similarly,950 school dropouts were returned to school 1,400 times. CYDP outreachworkers made 1,250 appearances at police stations and courts on behalfof 800 youngster.. Finally CYDP workers made 2,700 follow-up visits tothe homes of 2,000 juveniles who were arrested during the last thirty monthsof the project, in an effort to get them involved in one aspect or anotherof the project's programs. Despite this effort, the careful evaluationfound that the youth unemployment rate remained unchanged, the school dropoutrate increased somewhat, and the arrest rates of juveniles in CYDP areasincreased over time.

A different and more recent strategy for using gang workers is crisisintervention and conflict mediation. A test of this approach by detachedworkers in a Puerto Rican area of Chicago had more encouraging, if, complexresults (Spergel, 1986, as cited in Spergel, 1995: 255). While the programarea had a slower rate of increase in serious gang crimes by juvenilesthan the comparison area, the program area also had a faster rate of increasein serious crimes by adults. Attempts to organize the target communitywere less successful than efforts to mediate juvenile gang conflicts toprevent violence. More recently, Spergel has found some evidence that acoordinated police-probation-detached worker program to monitor gang offenderson community supervision has slowed their rate of committing serious violence(Spergel and Grossman, 1995, as cited in Howell, forthcoming). Encouragingresults from another conflict-oriented program have been reported for NewYork (Goldstein, Glick and Carthan, 1994, as cited in Howell, forthcoming).Using a cognitive skills approach called "Anger Replacement Training,"the evaluators report decreases in arrests of gang members.

Perhaps the most encouraging findings about gangs come from Boston,where they have nothing to do with traditional gang prevention. Preliminaryresults of a gang-related project to reduce juvenile firearms crime areextremely encouraging (Kennedy, Piehl and Braga, 1996). An effort to detergang-related gun violence by massive police response to any shootings issupported by probation officers who have the statutory authority to searchprobationers at will. The probation officers work with police to send outthe word that any shootings will get anyone even tangentially involvedinto a lot of trouble. This approach has apparently given some gang membersa convenient excuse to opt out of planned conflicts, much as the policecrackdown on drunk driving in Australia has given barroom drinkers an excuseto refuse extra drinks (Homel, 1994). If the final results of this projectconfirm preliminary findings, it will be another example of substantiallyreduced gun crime without any structural changes in community conditions.

The Future of Gang Violence Prevention

While the results of available evaluations are generally negative, thenumber of careful field tests remains quite small. The average level ofscientific rigor in the available evaluations is quite low. Taken together,the studies show weak evidence of no effect. None of the programsaddress the underlying community risk factors associated with the recentexplosive growth in gang activity. Yet new models of gang violence preventionnow under development at Harvard and the University of Chicago might wellsucceed in reducing gang violence without solving the structural problemsof the inner-city. Combinations of police, probation officers and civilianswho keep gangs under close surveillance may be successful at heading offplanned conflicts leading to gun violence. Unplanned encounters of rivalgangs leading to shootouts may be harder to prevent, but reduced gun carryingcould accomplish that as well. Police-civilian teams checking known andconvicted gang members for guns, with appropriate legal authority, couldin theory reduce gun carrying and spontaneous shootings.

The enormous concentration of serious violence among gang members suggeststhe value of further research and development efforts to find effectiveprevention methods for gang violence. But the state of the scientific evidencesuggests the risks of funding gang programs without careful evaluations,whether through block grants or discretionary programs. University of SouthernCalifornia gang violence scholar Malcolm Klein (1995: 138) states the caseclearly:

Consider California, more affected by street gangs than any other stateis, by far...the state has 196 cities with street gangs, 60 in Los AngelesCounty alone. The state's Office of Criminal Justice Planning in fiscalyear 1990-91 poured almost $6 million into sixty projects under its GangViolence Suppression Program. Included were school programs, street workprograms, community mobilization, diversion alternatives, and a wide varietyof criminal justice enforcement projects. Yet not a dollar went to an independentevaluation of the effectiveness of these projects. Sixty wasted opportunitiesto assess our efforts seems to be an inexcusable exercise in public irresponsibility.

The fact that Klein's own work demonstrated that a gang "prevention"program actually increased crime rather than reducing it lends specialforce to his conclusion. The theoretical implications of Klein's work ongang cohesion suggest that much of what police are doing--often supportedby federal funds--to suppress gang violence may also be increasing ratherthan preventing that violence. The seriousness of gang violence provideseven more reason, not less, for a high standard of scientific rigor inevaluating gang prevention. What evidence we have clearly shows that goodintentions are not enough.

Both old and new strategies could be subjected to more rigorous evaluations.Despite the strength of Klein's findings, for example, they are based onquasi-experimental pre-post designs generally lacking control groups. Alarge scale test of gang worker strategies across a sample of 100 gangs,with 50 gangs randomly assigned to intervention, might well produce differentresults. The Ladino Hills project Klein (1995:146) reports is actuallyquite encouraging; the program was a clear success at diverting gang membersfrom gangs as long as the gang workers stayed on the job. Klein's emphasison the project's failure to end gang activity in the area for up to twoyears after the gang workers were withdrawn seems to set an unrealisticallyhigh standard. Just because a maintenance therapy did not rise to the levelof a permanent vaccine does not make it worthless. Rather, the evidencesuggests that Klein has found a way to reduce gang membership. This isa promising finding that merits replication with a more rigorous researchdesign.

New strategies for gang prevention should also be tested at much higherlevels of scientific rigor. OJJDP is currently supporting the developmentand testing of comprehensive community gang prevention efforts, coordinatingmultiple local agencies and attempting to mobilize community involvement.NIJ is currently supporting firearms crime reduction efforts. Neither approachis currently undergoing a randomized controlled test (level 5) using communities,or gangs, as the unit of analysis. Indeed, it may well be premature tobe doing so at this stage until the strategies are sufficiently well-developed.But a clear plan to develop a strategy that can be subjected to more rigoroustesting could help move the nation more quickly to discovering effectivemethods for reducing gang violence.

One objection to this approach is that every city has a unique gangsituation, and must design its own program (Klein, 1995: 154). The responseto that objection is that most cities lack sufficient data to conduct rigorousevaluations: enough neighborhoods, enough gangs, enough gang violence tocontrol for all the chance factors that can affect results. Limiting evaluationsto one gang program or one city at a time would do little to increase availableevidence about how to prevent gang crime. It is only by seeking out thecommonalities of successful gang prevention programs across areas and typesof gangs that the scientific basis for effective prevention can be advanced.

COMMUNITY-BASED MENTORING PROGRAMS

Community-based mentoring programs take a much broader focus on riskfactors than gang prevention programs. Both the empirical evidence andtheoretical linkages to community risk factors gives solid reason to supportmuch more research and development on this strategy. While it does nothave the gang programs' efficiency of focusing on the limited number ofjuveniles committing the most serious violence, mentoring offers the promiseof effectiveness across a much broader population. Some members of thatpopulation could well become gang members or serious violent criminals.Mentoring could be a way to prevent that.

Theoretical Rationale for Mentoring

Why should mentoring of a larger at-risk population of pre- and earlyadolescents be any more effective than detached social workers focusedon gangs? Gang social workers, after all, are in effect mentors to gangmembers. But the general failure of detached workers may be due to theirfocus on older youths who are already active offenders. Many developmentaltheorists argue that ages 10 to 14 provide a more promising focus for interventionand prevention (Carnegie Council on Adolescent Development, 1995). Thepower of peer groups may not be as great in that age-range, and an intensiverelationship with a conventional adult could be a powerful influence foryouths on the cusp of delinquency.

A more powerful reason for the failure of detached workers with gangsmay be insufficient dosage. Given their workloads, they may not have beenable to spend enough time with their individual clients, irrespective ofage, in order to become a strong role model. A more intense relationship,with "quantity time" of "quality time," between a "mainstream"male adult and a preadolescent or early adolescent boy may directly addressseveral community risk factors for crime:

o fatherless boys; 17 million children now in single parent homes, 25%of all youth and 50% of minority youth (Tierney, et al, 1995: 49)

o lack of legitimate role models

o insufficient "intergenerational closure" with adult influencescounteracting peers (Wilson, 1996: 62)

Mentoring provides the highest dosage of adult-child interaction ofany formal community-based program. Compared to street workers and recreationprogram supervisors, mentors can develop much stronger bonds with juvenilesat risk. In theory, they can gain the power of "legitimacy" (Tyler,1990) based on a pattern of respect and support the mentor establisheswith the juvenile, so that the mentor's approval and attention becomesa valued resource. That resource then gives the juvenile a "stakein conformity" (Toby, 1957), something to lose if the juvenile getsinto trouble with the law.

Mentoring programs described in available evaluations feature threeto four meetings a month or more between mentor and child, with each meetinglasting at least for several hours. Community-based mentors see juvenilesin a wide range of settings, including home, movies, professional sports,plays and concerts. They may talk frequently on the telephone, with menteescalling mentors as well as vice versa. In contrast to school-based mentoringprograms (reviewed in Chapter 5) which generally operate with a heavieremphasis on academic issues and truancy, community-based mentors tend tobe involved in more domains of the child's life. They may also providemore resources in the form of entertainment outings. Mentors may be paidor unpaid, college students or adults. All of them receive some sort oftraining, although the infrastructure supporting mentoring relationshipsvaries. Adult volunteers in the oldest formal mentoring program, the 90-plusyear-old Big Brothers and Big Sisters of America (BB/BSA), for example,are subjected to extensive background examination to screen out potentialchild molesters.

Results of Community Mentoring Evaluations

Careful examination of community-based mentoring evaluations supportsa conclusion that they are a promising approach to preventing crime riskfactors, notably drug use. While most of the evaluations show no effect,the most rigorous modern evaluation shows a strong effect at reducing druguse, and clear effects at reducing alcohol use and "hitting"among at-risk children. The short-term measurement of those beneficialeffects, however, must stand in the shadow of much less encouraging resultsfrom a thirty-year followup of an equally rigorous Depression-era mentoringtest, the privately-funded Cambridge-Somerville experiment.

Controlled Experiments. The first controlled test of mentoringbegan in 1937, when recent college graduates were hired and trained toprovide an average of two visits a month to the experimental half of asample of 650 at-risk boys under age 12 at the program's outset.3The paid social worker mentors met with their clients at home, in the street,or at project headquarters. They provided academic tutoring, trips to concertsand sports events, and general emotional support for the boys. The programalso provided the boys' families with help for medical and employment problems,and sent the treatment group boys to summer camp. By 1942, 253 of the original325 treatment group boys were still in the program, when it was ended sothe counselors could join the armed forces.

The results of this intensive mentoring showed no difference betweentreatment and control groups in criminal records, either in 1942 (Powersand Witmer, 1972) or in 1975-76 (McCord, 1978). The longterm followup,however, did show significantly higher levels of diagnosed alcoholism,serious mental illness, and stress-related physical health problems. Ahigher level of unfavorable life outcomes, although not specifically greatercrime, among the treatment group seems clear. What is less clear is themeaning of the results for the value of mentoring programs today.

Three theories compete to explain these results. One is that mentoringsimply backfires, somehow creating an artificial source of support thatmakes it harder for mentored boys to adjust as adults. A more plausibletheory is that the abrupt departure of these long-term counselors fromthe boys' lives was as damaging emotionally to the boys as a divorce orother loss of parental involvement, compounded in many cases by the boys'previous loss of their own natural fathers' support. A third theory isthat the difference in diagnosed mental health problems is only an artifactof the treatment group's greater exposure to professional and medical servicesas part of the treatment content. Under this theory, the treatment boyshad no greater rate of personal problems, but when they had problems theywere simply more likely to seek professional help of the kind the programhad taught them to seek.

The fundamental principle of science here is that one experiment alone,no matter how rigorous, cannot provide a "definitive" test ofany hypothesis. Social experiments in particular require replication todetermine their generalizability to other times and places. A three-decadefollowup is an excellent basis for drawing conclusions about the lifetimeeffects of a treatment, but it has a substantial drawback for policy analysis:by the time the results are in, the world has changed so much that theresults may no longer be valid. The modern social conditions of inner-citypoverty and segregation are so different from the context of the Cambridge-Somervilleexperiment that it is not clear that the identical program would producesimilar results.

If three decades are too long, one year is probably too short. Unfortunately,that is all we have in our modern controlled experiment in community-basedmentoring for pre- and early adolescents (Tierney and Grossman with Resch,1995). The virtues of this experiment, however, are many, including thesubstantial risk factors in the sample. The 959 eligible applicants forthe Big Brothers/Big Sisters program in eight cities came from homes inwhich 40% of the parents were divorced or separated, 15% had suffered adeath of a parent, 40% had a family history of substance abuse, and 28%had a history of domestic violence. The children themselves, of whom 60%were minorities, 40% girls, and all aged 10-14, included 27% who had beenabused as children. As Chapter Four reports, child abuse substantiallyincreases the risk of criminality in later life.

How much the Big Brothers/Big Sisters program reduces criminality laterin life is not clear. What is clear from this tightly randomized experimentis that there were substantial benefits in one year (average) treatment.After spending around 12 hours monthly with their volunteer adult mentors,the treatment group children had 45% less reported onset of drug abusethan the control group children, who had been put on the waiting list.4They also had 27% less onset of alcohol use, and 32% less frequency ofhitting someone. The program also reduced truancy: treatment group childrenskipped 52% fewer days of school and 37% fewer classes on days they werein school.

These results were achieved at a very modest cost. Since the mentorsvolunteer their time, the only cost is the infrastructure needed to recruit,screen, train and properly "match" the mentors to children forsuccessful long-term relationships. The cost is estimated at about $1,000per match (Tierney and Grossman, with Resch, 1995: 52). While the fullcrime prevention benefits of that cost cannot be specified without a longer-termfollowup study, the short-term benefits alone might justify federal supportof this apparently underfunded program. At a price of $1,000 per year ofdrug abuse prevented, the taxpayer would be well ahead spending money onthis program instead.

Table 3-2

Community-Based Mentoring Evaluations

Primary Source Scientific Methods Program Content Program Effects (secondary) Score McCord 1978, 1992 5 2 visits monthly by No effect on criminal Powers and Witmer paid male counselors record; treatment 1972 for 5.5 years with group did worse on 253 At-risk Boys diagnosed mental under 12 in 1937-42; health WW2 end Tierney et al 1995 5 Big Brothers & 46% reduction in drug Sisters, 1 year for use onset, 32% 10-14 yr.-olds, 60% reduction in hitting minority & 27% people, relative to abused; 3 hrs wkly controls Green 1980 4 Big Brothers for No effects on (Howell 1995) fatherless white disruptive class boys behavior; no measures 1/2 day weekly for 6 of drug use months Goodman 1972 2 College Student high control group (Howell 1995) Mentors of 10-11 attrition; program yr-old boys 6 hrs effects on crime wkly over 2 years unknown Dicken, Bryson and 3 College Student no difference in Kass 1977 mentors for 6-13 teacher-rated (Howell 1995) yr.-olds, 6 hrs behavior of mentees wkly, 4 months Fo and O'Donnell 5 12 weeks of paid Truancy reduced 1974 community mentors significantly under (Howell 1995) with at-risk 11 to some conditions 17 year olds; N = 26 Fo and O'Donnell 5 1 year of paid Lower recidivism for 1975 community mentors treatment groups with (Howell 1995) meeting weekly with priors, higher at-risk 10-17 without yr-olds 

Two other randomized experiments in paid "Buddy System" mentoringconducted in Hawaii were published in the early 1970s. The ages of theat-risk youth ranged from 11 to 17, while the ages of the paid mentorsranged from 17 to 65. The first experiment (Fo and O'Donnel, 1974, as citedin Howell, 1995: 91) lasted only 12 weeks, during which it randomly assigned26 subjects into four treatment groups ( an average of 6 per group). Thissmall experiment used an elaborate theoretical model, in which treatmentgroups varied on several dimensions. The dimensions included the conditionsof mentor approval for the mentees, dichotomized as contingent, or not,on appropriate behavior by the mentees. A third treatment group was paid$10 a month on the same contingent basis. The results showed that truancydeclined for the subjects receiving contingent approval, but not for thosereceiving unconditional approval.

A larger experiment by the same authors abandoned the theoretical distinctions,comparing crime rates between randomly assigned 10-17 year olds receivingmentoring or not (Fo and O'Donnell, 1975, as cited in Howell, 1995: 92).The one-year experiment found that treatment backfired among those withno prior record; those in the experimental group had more offenses duringtreatment than control group youths who also had no prior record in thebaseline period. Among youth who had prior records at the outset of theexperiment, however, the results were the opposite: mentees had less recidivismthan the control group. The possible reasons for this difference were notreported.

Non-Randomized Evaluations. The other community-based mentoringstudies offer little scientific evidence for policy purposes. The Green(1980, as cited in Howell, 1995: 92) evaluation of a Big Brothers' programin Nassau County, for example, lacks any outcome measure of drug abuse,violence or crime. Green does find no difference in disruptive classroombehavior, but so did the Tierney and Grossman with Resch (1995) experiment.The non-randomized design and 6 month followup period also limit its value.

None of the remaining tests are strong enough to contradict the positiveeffects found in the recent test of Big Brothers/Big Sisters. The Goodman(1972, as cited in Howell, 1995: 90) two-year test of paid mentors in Berkeley(CA) showed some evidence of worse school behavior among mentored at-riskboys than among the controls. Substantial attrition in the control grouponly, however, made the comparison difficult to interpret. A nonrandomtest of a similar approach using unpaid college students for a semesterfound no differences in teacher ratings of behavior (Dicken, Bryson andKass, 1977, as cited in Howell, 1995: 91). All of these negative resultsfrom what were essentially "start-up" programs may be due tofactors that are not present in the standardized, long-practiced methodsof the national Big Brothers/Big Sisters program.

The Future of Community-Based Mentoring

The major question about mentoring remains the meaning of the Cambridge-Somervilleexperiment for contemporary public policy. The answer to that questionis unlikely to come from further analysis of that experiment, but fromits replication under modern conditions. The Big Brothers/Big Sisters experiment(Tierney and Grossman with Resch, 1995) is an excellent start in that direction,and would be even more valuable if followed by many years of followup datacollection. Its promising results, however, suggest the value of a largertest, one that incorporates the diagnosis of community risk factors, assuggested in the conclusions of this chapter.

Based solely on the research available at present, there seems to besufficient basis to reach somewhat different conclusions than those reachedby one OJP publication prepared prior to the publication of the Tierneyand Grossman with Resch (1995) experiment, which substantially alters theweight of the evidence. The OJJDP Guide for Implementing the ComprehensiveStrategy for Serious, Violent and Chronic Offenders (Howell, 1995:128) suggests that "mentoring relationships that are noncontingentand uncritically supportive" are "not effective," but that"mentoring relationships that include behavior management techniques"are "potentially promising." The Big Brothers/Big Sisters programreports no contingency policy for mentor approval of mentees. Its successat reducing drug use onset would thus seem to falsify the "contingentapproval" hypothesis. The small sample size (N =26) of the one findingconsistent with that hypothesis makes the much larger recent study morecompelling evidence (Fo and O'Donnell, 1974).

The most important conclusion from this research restates the conclusionof the gang prevention evaluations. Even with the encouraging findingsfrom the most recent controlled test of community mentoring, there is toolittle information for adequate policymaking. The priority is for moreresearch, not more unevaluated programs. The danger of doing harm is fartoo great to promote and fund mentoring on a broad scale without carefullycontrolled evaluations. No such evaluations, to our knowledge, are presentlyon the drawing boards. They could readily be included, however, as partof a broader test of a comprehensive interventions package in high-crimeareas. While the community context of mentoring experiments under thoseconditions would be unique, the addition of other programs addressing communityrisk factors could well enhance the potential for crime prevention willadding to scientific knowledge.

COMMUNITY-BASED RECREATION PROGRAMS

The hypothesis that recreation can prevent crime has become one of themost acrimonious in the history of crime policy. More than any other issue,the debate reflects the inappropriate definition of prevention discussedin Chapter 2. What is most revealing about the debate, however, is thevirtual indifference it has displayed to empirical evidence. Rather thanarguing on theoretical grounds alone, it would seem more valuable to testthe hypothesis scientifically. Chapter Five presents evidence that school-basedprograms have been tested an found ineffective at preventing crime anddelinquency. This section presents more limited evidence on community-basedrecreation centers, where the evidence is thinner but marginally more promising.

An OJJDP publication (Howell, 1995: 95) provides a clear statement ofthe recreation hypothesis:

    Afterschool recreation programs can address the risk factors of alienationand association with delinquent and violent peers. Protective factors mayinclude opportunities for involvement with prosocial youth and adults,skills for leisure activities, and bonding to prosocial others.

An equally plausible negative hypothesis can be suggested on theoreticalgrounds. In a neighborhood plagued by inter-gang rivalries and everydayanger (Bernard, 1990), after-school recreation creates opportunities forvictims and offenders to intersect in time and space (Cohen and Felson,1979), creating conflicts and potential for violence. One Philadelphianightclub shooting in the early 1980s, for example, was generated by afight that began on a recreation center basketball court. A middle groundhypothesis is that the effects of after-school recreation may vary substantiallyby neighborhood context and how the recreation center is run.

Results of Recreation Evaluations

The scientific evidence on these hypotheses is currently quite limited.What evidence there is all positive, supporting the proponents of recreationprograms. While the scientific rigor of the three available evaluationsis modest, it shows fairly strong effects, two on crime and one on drugs.Two are based on Boys' and Girls' Clubs (BGC), and two are in public housing.

Table 3-3

After-School Recreation Programs

Primary Source Scientific Methods Program Content Program Effects (Secondary source) Score Jones and Offord 3 Canadian Public 75% reduction in 1989 Housing Project juvenile arrests for (Howell 1995: 95) children 5 to 15 experimental, 67% offered intensive increase for control recreation, 3 years location Schinke, Orlandi and 4 3 groups of 5 public Recreation centers Cole 1992 housing projects with drug prevention each, 1 group had lowest drug use; Boys/Girls Club vandalized housing (BGC), 1 BCG plus units down 25% in drug prevention, 1 drug prevention sites control no BGC Brown and Dodson 3 Boys' Club area Program area 1959 compared to 2 delinquency declined (Howell 1995: 95) comparison areas, 9 after two years, years comparison rose 

The test in a Canadian public housing project offers the strongest evidence.Over 32 months, the low-income children ages 5 to 15 were provided an intensiveafter-school program in sports, music dancing, and scouting. A comparisonpublic housing project had only minimal city services. The majority ofage-eligible children in the test site participated in the recreation program.Compared to a baseline period of two years prior to the program, arrestsof juveniles in the program site declined 75 percent. In the same timeperiod, arrests of juvenile in the comparison site rose 67%. Sixteen monthsafter the program ended the effect had worn off, providing further evidenceof a program effect (Jones and Offord 1989, as cited in Howell, 1995:95).

The American public housing test covered three groups of five housingprojects each. One group already had a traditional BGC program operatingin the community center. A second group received newly established BGCprograms, supplemented by the SMART Moves (Self-Management and ResistanceTraining) substance abuse prevention program aimed at parents as well aschildren. A third group of three projects had no BGC and remained thatway as a control group. Observational and police data indicated a declinein drug use in the new BGC/SMART Moves sites. Archival records showed thatvandalized housing units dropped from 8% to 6% of total units in the newBGC sites, while rising from 8% to 9% in the controls and remaining unchangedin the existing BGC sites (Schinke, Orlandi and Cole, 1989).

A nine-year, 1950s study examined juvenile delinquency in a LouisvilleKentucky area served by a Boys' Club (Brown and Dodson, 1959). The clubincluded both traditional activities at the building and a summer campprogram. The study found declining juvenile delinquency relative to twocomparison areas without a Club. The first two years after the Club beganoperation, however, showed similar trends in delinquency in the programand comparison areas. While the prevention effect could plausibly havetaken several years to become evidence, the lack of significance testsand other checks on validity limit the value of this study.

The Past and Future of Recreation Programs

Recreation programs merit further research and development for theirpotential crime prevention benefits, if only because they continue to drawCongressional support (e.g., Washington POST, January 16, 1997, p. A4).This conclusion is based not just on the three available impact evaluations,but on the long history of such programs in mainstream American life. Thewidespread availability of such programs in low-crime areas is anotherstructural difference between suburban and inner-city communities, onethat may contribute to the latter's higher crime rates.

The danger of violent conflicts being generated by club activities isjust as open a question as the potential benefits of the programs. Carefulresearch is needed to assess the net frequency of such conflicts with andwithout recreation, since shootouts can start off the basketball courtsas well as on them. The potential prevention benefits from such programsmay well exceed the benefits of prison, perhaps at much lower cost. Butwe will never know unless we invest in careful evaluation research. Morefunding of operations alone will leave the policy decision vulnerable toideological and symbolic politics, rather than a rational decision on themerits of reliable evidence.

REMOVING CRIMINOGENIC COMMODITIES

Perhaps the most immediate proximate contributing cause to many criminalevents is a "criminogenic substance" (Cook and Moore, 1995).Guns, drugs, alcohol and cash, in the right circumstances, can all providethe additional, if not sufficient, cause which helps make a crime happen.That does not mean, however, that these substances will always be in theright circumstances, even when they are available in the community. Guns,for example, may not do much harm if they are kept locked in a safe, eventhough the potential for theft of the guns may make them a potential causeof a shooting on the street. Similarly, the context and use of alcoholvaries widely, and is only criminogenic in some settings.

One approach to community crime prevention is to limit access to criminogenicsubstances. Community groups often lobby against the renewal of tavernliquor licenses, for example, on the grounds that the alcohol access increasesthe rates of robbery and assault in the community. Many cities are increasinglyconcerned about 24-hour bank cash-dispensing machines, with increasingregulatory control of their locations and security measures (Sherman, 1995).Low-income communities have possibly had fewer robberies and thefts sincedirect bank deposit of welfare and Social Security checks became commona decade ago.

These ideas are generally theoretically sound, given the prevailingtheory of criminal events (Felson, 1994). Few of them have been evaluated.One specific approach that has been evaluated, gun buyback programs, suggeststhat there can be a major gap between theory and practice.

Gun buyback programs are based on two hypotheses. One is that the moreguns in a community, the more gun violence there is. There is substantialevidence to support that claim (Reiss and Roth, 1993). The second hypothesis,however, is not supported by the evidence. That hypothesis is that offeringcash for guns in a city will reduce the number of incidents in which gunsare used in crime in that city. Four evaluations reviewed in Figure 4 showno effects of gun buyback programs on guns. There are several reasons whybuyback programs may fail to reduce gun violence:

o they often attract guns from areas far from the program city

o they may attract guns that are kept locked up at home, rather thanbeing carried on the street

o potential gun offenders may use the cash from the buyback programto buy a new and potentially more lethal firearm; the buyback cash valuefor their old gun may exceed market value substantially.

The enormous expense of these programs is instructive. When St. Louisinvested $250,000 in gun buybacks in 1994, the same funds could have beenused to match 250 children with Big Brothers/Big Sisters. Those 250 childrenwould then have enjoyed about half the risk of becoming drug users, atleast for the first year (Tierney and Grossman with Resch, 1995). But theopportunity cost of the programs never entered into the debate.

The scientific rigor of the buyback evaluations is not great. They canbe summarized as providing moderate evidence of no effect. Theyfail to show effects on gun crimes relative to a comparison of trends inthe same types of crimes committed without guns. Given their high costand weak theoretical rationale, however, there seems little reason to investin further testing of the idea.

Table 3-4

Gun Buyback Evaluations

Source Scientific Rigor Program Content Program Effects Score Rosenfeld 1995 3 1991 Gun Buyback in No reduction in St. Louis of 7,500 homicides or gun guns assaults relative to same offenses, no guns Rosenfeld 1995 3 1994 Gun Buyback in No reduction in St. Louis of 1200 homicide or gun guns assaults relative to same offenses, no guns Callahan et al 1995 3 1992 Seattle Gun No effect on crime Buyback reports or medical records of gun injuries 

CONCLUSIONS

This chapter has shown that there is a substantial disconnection betweenwhat is known about community causes of serious violence and what thisnation is doing about those causes. The scientific evidence that communitiesmatter is strong. The evidence that serious crime is concentrated in avery small number of communities is even stronger. But the link betweenthose facts and the design of prevention programs is very thin indeed.Instead, a National Academy of Sciences report concludes there is evidencethat federal and local transportation and housing policies over the pasthalf-century have substantially contributed to the causation of seriouscrime, especially in the hypersegregated inner cities where over half ofall homicides occur.

Despite the past gap between causation and prevention, there are manyas-yet unevaluated new efforts on the horizon attempting to bridge thatgap. There is also promising evidence that some programs can be successfulwithout addressing the root causes diagnosis of causation. Thus the prospectsfor progress in community-based prevention may be stronger than the currentevaluation record suggests.

By the criteria used in this report, there are no community based programsof "proven effectiveness" by scientific standards to show withreasonable certainty that they "work" in certain kinds of settings.There are programs for which we can conclude the evidence shows with reasonablecertainty that they do not work, at least in the settings where they havebeen evaluated. But even these programs might be found effective if variedin significant ways and rigorously evaluated. Moreover, there is both empiricalevidence and theoretical reason to conclude that some programs are promisingenough to merit further replication and evaluation.

What's Promising

o Gang violence prevention focused on reducing gang cohesion,but not increasing it

o Volunteer mentoring of 10 to 14 year-olds by Big Brothers/BigSisters is promising for the reduction of substance abuse, but not delinquency

What's Doesn't Work

o Community mobilization against crime in high-crime inner-citypoverty areas

o Gun buyback programs operated without geographic limitationson gun sources

THE EFFECTIVENESS OF DOJ PROGRAMS

These findings offer some answers to the Congressional question aboutthe effectiveness of DOJ crime prevention programs. Perhaps most importantis the scientific support for the growing emphasis on comprehensive programsfor high crime communities found throughout the Office of Justice Programs(OJP). With the advent of the Enterprise Zone/Empowerment Communities (EZ/EC)initiative, the emphasis on comprehensive risk factor strategies is spreadingto the entire federal executive branch. The scientific evidence supportsthis approach, especially to the extent that it actually concentrates onthe specific neighborhoods in which serious crime is most heavily concentrated--notjust the cities in which those neighborhoods are located. Because thisreview finds no community-based programs of scientifically proven effectivenessto employ in those high-crime communities, however, there is a criticalneed for further research and development to help focus that funding moreeffectively. And because the statutory plan allows states to expend DOJfunds in communities with moderate to low rates of serious youth violenceand risk factors for crime and delinquency, the expenditure of the fundsis not yet optimal for discovering programs of proven effectiveness inthose areas.

Several DOJ funding programs provide support for community-based localprevention programs. The major funding areas are Byrne Grants, Weed andSeed, Local Law Enforcement Block Grants, and the Title V Delinquency PreventionGrants. Most important, however, may be the DOJ funding for rigorous programevaluations of community-based prevention.

Byrne Grants

The Byrne Formula Grant program (as distinct from discretionary grants--seeChapter One) awarded $1.8 billion through the states and territories from1989 through 1994 (Dunworth, et al, 1997: 5). Community crime prevention,property crime prevention, and public housing are three of the twenty-oneoriginal (now 26) "Purpose Areas" for the program. Grants fundedunder these purpose areas could generally fall in the institutional settingaddressed by this chapter. Together the three purpose areas received approximately$68 million, or less than four percent of the total funding. Drug treatmentis a fourth Purpose Area operating at the community level, receiving $107million in those years or 6 percent of total formula grants.

As noted in Chapter One, the broad diversity of programs funded andgeneral absence of scientifically rigorous impact evaluations makes itimpossible to assess the effectiveness of the Byrne funding stream as asingle policy. Even the specific Byrne Purpose Areas cover a broad rangeof local programs. The scientific evidence reviewed in this chapter, however,strongly supports the statutory language calling for "strategic plansto target resources on geographic and substantive areas of greatest need"(Dunworth, et al, 1997: 3). The key question raised by this chapter isthe best criteria for selecting the areas of greatest need. A related questionis the most appropriate definition of "area." Absent a clearfocus on the geographic areas with the most serious crime, community-basedprograms offer little scientific basis for claims of effectiveness at preventingsuch crime.

The evidence suggests that community-based Byrne grants may be mosteffective if concentrated on the small number of census tracts (often contiguous)where the majority of homicides in each state are clustered. The scientificevidence on the geographic distribution of homicides shows strong concentrationswithin high risk-factor census tracts. While a decade ago it would havebeen difficult for many states to analyze homicide data statewide by censustract, recent advances in microcomputers and computerized crime mappingmakes such analysis feasible. Not every high homicide area may be appropriatefor Byrne funding, given the difficulties of implementing community-basedprograms. But a statutory plan to focus a substantial percentage--perhapsfifty percent or more--of community-based Byrne Grant programming withinsuch communities could speed the process of discovering what works. Thiswould be especially likely if coupled with a national plan for testingcommunity-based strategies across large samples of communities (see below).

The issue of concentration helps to interpret the evidence on communitymobilization. That evidence shows that, by itself, mobilization isineffective against serious crime in low-income communities. But it isfar to early to close the door on mobilization as a possible necessarycondition for other strategies. Many questions remain about whether mobilizationcan enhance a wide range of other specific efforts to attack serious crime,such as helping police reduce illegal gun carrying, reducing the availabilityof drugs and alcohol, and divert youth from gangs. Those questions, again,can only be answered by large sample community level studies as recommendedbelow. In the absence of such programming for the sake of discovering whatworks, however, community mobilization funding would be of doubtful effectiveness.

Concentration of funds on high-crime communities would also make itpossible to evaluate programs like drug treatment in a community-basedway. Rather than examining the effects of drug treatment on individual-levelcrime rates, a community-level concentration of drug treatment could measurethe community crime prevention effects of substantial increases in localtreatment slots. The individual-level evidence we do have on drug treatment(see Kinlock, 1991), however, is certainly supportive of the effectivenessof Byrne funding spent on that Purpose Area.

Local Law Enforcement Block Grants

This formula grant program newly established in 1996 is also more focusedon high-crime communities than other federal funding of local crime prevention.Most of the $404 million in 1996 funds were allocated on the basis of eachlocal police agency's level of reported Part I violent crimes. The statutorydistribution plan clearly places greater resources in the cities with themost serious problems of violence and youth violence. It does not, however,require that the funding be concentrated within those cities in the areasof greatest risk.

Like the Byrne Program, Local Law Enforcement Block Grants (LLEBG) couldbe focused more precisely on census tracts with highest homicide rates.And like the Byrne grants, LLEBGs have awarded substantial support forcommunity mobilization. The 1996 amount was $33 million, about nine percentof program funding. The comments above about further funding of communitymobilization programs under Byrne apply to LLEBG as well; more investmentin discovering what works seems justified, while unevaluated funding islikely to be ineffective at either preventing crime or increasing scientificknowledge about prevention.

Weed and Seed

Since 1991, the Weed and Seed program (see Chapter One) has been themost theoretically appropriate federal funding program for dealing withconcentrated inner-city violence. Based upon the available DOJ publications,Weed and Seed funding offers the clearest focus on the census tracts withvery high homicide rates; the initial program area in Kansas City had arate of 180 per 100,000, or twenty times the national average. As the firstof many comprehensive inner-city programs developed in recent years byOJP, Weed and Seed also offers the best evidence on the challenges of implementingand evaluating comprehensive programs, especially those in which DOJ becomesthe lead agency in mobilizing resources from other federal departmentsat a micro-local level.

Weed and Seed's rationale for preventing serious crime is a high concentrationof resources addressing a high concentration of risk factors in a smallgeographic area. The basic structure of this approach apparently differsfrom the majority of DOJ funding, which by statute cannot be focused uponthe highest-crime communities. Given enough evaluation evidence for programsof proven effectiveness in such places, there could be a strong rationalefor channeling the majority of DOJ crime prevention funding in ways similarto Weed and Seed. The challenge for Weed and Seed is therefore not justto prevent crime in the target communities, but to do so in a way thatallows scientific evidence to accumulate about program effectiveness. Theinitial history of the program in that regard is instructive.

The initial Weed and Seed target area in Kansas City was accompaniedby an NIJ evaluation grant that was almost equal to the amount of the programfunding. That evaluation found a 49 percent reduction in gun crime anda statistically significant reduction in homicide associated with a singleelement of the program that fell outside the community-based institutionalsetting of this chapter (see Chapter Eight): directed police patrols atcomputer-located "hot spots" of gun crime (Moore, 1980). Thesepatrols produced a 65 percent increase in gun seizures not found in thecomparison area, where gun crime remained stable (Shaw, 1994; Sherman,Shaw and Rogan, 1995). The single element could be evaluated because noneof the other elements had been implemented at that time. Had there beenother elements implemented, it would have been scientifically impossibleto isolate the effects of this element. Fortuitously, the delay in theother program elements allowed the evaluation to discover an apparent effectwith important implications.

Subsequent Weed and Seed sites did not have such intensive evaluations.The 50-50 ratio of evaluation to program dollars was tipped overwhelminglyin favor of program dollars. In the five years since the subsequent sitefunding was awarded, no impact evaluation has been completed. A processevaluation published by NIJ (Roehl, et al, 1996) illuminated the complexityof the program, which has now attracted substantial state and private fundingin some sites. A second multi-site evaluation is now in progress, whichis slated to produce site-specific impact evaluations at a Scientific MethodsScore of either 2 or 3. The ability of that retrospective design to isolateprogram elements in relation to crime prevention will be difficult giventhe problem of multiple treatments (Cook and Campbell, 1979). Thus as theprogram currently stands, there is good scientific theory but no scientificdata to show the effectiveness of the program.

The most challenging theoretical element for any inner-city crime preventionprogram is raising the community rate of adult labor force participation(Wilson, 1996). Chapter Six discusses the evidence on that point in detail.Labor force programs have suffered from a lack of focus on the Weed andSeed strategy, scattering resources across individuals spread out overmany disparate communities. More recent private and public efforts to changecommunity labor markets, rather than personal labor skills, fit right intoWeed and Seed (see Bloom, 1996). They can easily become an integral partof its multi-risk factor reduction strategy, coupling high enforcementwith greater opportunity.

Comprehensive Communities Program

Similar in conception to Weed and Seed, the Comprehensive CommunitiesProgram (CCP) is an effort to integrate social programs and policing, publicand private organizations to control crime and improve the quality of life.The major difference is a lower funding level (see Chapter One) and a lessclear-cut focus on addressing the highest-crime, highest risk factor areas.CCP is more flexible about specific priorities set by city-wide leadershipfor specific programs and areas in which to operate them. The scientificevidence is thus less helpful in assessing such a program, given its greatervariability. An intensively measured level 2 process and impact evaluationis currently under way (Rocheleau, et al, 1996), but there is no well-controlledtest of its crime prevention effectiveness in progress. To the extent thatsome sites rely on gang programs that are of uncertain safety and effectiveness,as this chapter has shown, controlled tests of those specific program elementswould be a high priority.

Title V Community Prevention Grants Program

Since 1992, this program has assisted local juvenile justice agenciesto collaborate with other youth-serving agencies to develop an integratedsystem of services designed to prevent delinquency (see Chapter One). Amajor prevention component of this strategy is based on the CommunitiesThat Care model (CTC; Hawkins, Catalano, & associates, 1992). Consistentwith the scientific evidence of concentrated risk factors, but not withthe micro-local focus discussed in this chapter, the CTC model recommendsa flexible plan for reducing risk factors. The plan is for local jurisdictionsto identify risk factors known to be associated with delinquent behavior,to identify protective factors that buffer the effects of the identifiedrisk factors operating within the communities, and to target program interventionson those factors. Like Weed and Seed, this program has a firm foundationin indirect empirical evidence and theoretical support. What it lacks todate is scientifically rigorous crime prevention impact evaluations.

The Title V program is implemented in two phases. During phase one,the assessment and planning phase, communities (defined here as entirejurisdictions, not neighborhoods) interested in participating in the TitleV program must form a local prevention policy board and conduct an assessmentto identify and prioritize the risk factors operating in their community.On the basis of this assessment, the applicant community then must developa specific, comprehensive 3-year delinquency prevention plan. This planserves as the basis for the community's application to the state's juvenilejustice advisory group for Title V funding. Phase two of the process involvesthe implementation, monitoring, and evaluation of the programs and services.A 1996 survey administered by GAO showed that most of the 277 local projectssupported by this program appeared to be designed in accord with the CTCmodel.5 For example, 78% reported addressing multiple risk factorsin three or more substantive problem areas, and about 90% reported thatthey used two or more strategies identified in the CTC materials as "promising."Common prevention activities include parent training in effective techniquesof conflict resolution and after-school programs.

The CTC model recommends local monitoring of changes in risk and protectivefactors at the community (city or county) level, but that will yield limitedinsights on crime prevention effectiveness. A national evaluation of TitleV is being planned, but its scientific strength will be limited in theabsence of random assignment of funding, or at least of different preventionstrategies, to some communities and not others (Farrington, 1997). Thescientific possibilities for comparing two different approaches consistentlyapplied within two equivalent groups of communities, especially at theneighborhood level, would appear to be quite strong (Boruch, 1996). Butwhether it will happen depends in large part on the future of the issuesand recommendations presented in Chapter 10.

Based on our review of the evaluations of the programs in the OJJDP"menu" for Title V (Howell, 1995) in Chapters 2,3,4,7 and 8,we can make a limited assessment of the potential effectiveness of thiscrime prevention program. The framework provided for the Title V incentivegrants focuses local jurisdictions on selecting prevention strategies thathave some basis in research. It is possible, however, that the array of"promising" activities allowed under the model is too broad,encompassing some ineffective strategies along with the more effectiveones. The GAO report describes activities undertaken with Title V fundsin six jurisdictions. These descriptions are too general to support a judgementof the delinquency prevention potential of any particular activity, butthey seem to encompass a wide range of activities. Some of these, suchas social skills training (see Chapter Five) mentoring programs, appearpromising. Others, such as peer mediation and sports programs, do not.

Gang Prevention and Intervention

Funding for gang prevention and intervention programs is provided byBJA's Byrne formula grants, OJJDP, and potentially by Weed and Seed andLocal Law Enforcement Block Grants. There are currently no restrictionson the kinds of gang programs that are eligible for support. The scientificliterature suggests, but at a moderately low level of certainty, that theapproach taken to gangs is critically important. It is possible that DOJfunding is supporting programs that reduce gang cohesion, in which casethey are more likely to be effective. It is also possible that DOJ fundssupport programs that work with gangs in ways that may increase their cohesion,in which case they are less likely to be effective. Since the results ofthe available evidence cannot yet be generalized at a very high level ofcertainty, it is fairer to say that absent further evaluation evidence,the effects of DOJ-funded anti-gang programs are unknown.

JUMP:Juvenile Mentoring Program

This national discretionary program is a line-item Congressionally earmarkedappropriation for both schools and nonprofit organizations to establishmentoring programs for juveniles (See Chapter One). The school-based mentoringevidence discussed in Chapter 5 is less encouraging than the findings fromthe Big Brothers and Sisters experiment reviewed in this chapter, but theschool-based studies were also less rigorous. The $4 million annual appropriationsince 1994 was increased to $15 million in FY 1997. No impact evaluationsof JUMP have been completed, but one was solicited in 1996.6Based on the available scientific evidence, the drug abuse prevention effectivenessof the Big Brothers/Big Sisters model is promising, but the school basedmodel is of unknown effectiveness.

Based on the 1996 evaluation solicitation, it seems unlikely that theeffectiveness of JUMP will be measured scientifically in the near future.JUMP is yet another rapidly developing program that would benefit morefrom Congressional appropriations for evaluation than for expanded operations.The 1996 evaluation was budgeted at $150,000 per year to assess the effectivenessof a $4 million annual appropriation covering 41 separate grantees, orabout $3600 of evaluation funding per program grantee. While JUMP is idealfor the kind of level 5 evaluation conducted in the private sector usingrandomized controls (Tierney and Grossman, with Resch, 1995), the under-fundedDOJ evaluation clearly made controlled testing by independent evaluatorsimpossible. The design's reliance on program grantees for data collectioncompromises the independence and reliability of the data, and probablyprecludes such methods as obtaining police records on juvenile arrestsas an outcome measure. The Congress could correct these limitations byproviding twenty percent of program funds for a more limited number ofJUMP sites to be evaluated using the same design as the Tierney et al (1995)study.

STOP Formula Grants to Combat Violence Against Women

This program requires that states spend 25% of their funds to preventviolence against women on each of three priority areas (see Chapter One):law enforcement, prosecution, and victim services. None of these fall intocommunity-based crime prevention, but grants under the remaining 25 percentmay well do so. The purpose of the money is not just to combat domesticviolence (see Chapter Four), but also to prevent stranger violence againstwomen in the community. Hence community-based programs to reduce rape,stalking, purse-snatchings and carjackings would also be relevant here.The initial NIJ process evaluation of the program did not identify anycommunity-based programs (Burt, 1996), nor was our review able to identifyany impact evaluations of community prevention programs for stranger violenceagainst women.

IMPROVING EFFECTIVENESS THROUGH BETTER EVALUATIONS

Community-based programs are among the most difficult to evaluate. Theymay also be the most important. The "small science" approachto evaluations of community programs has prevented the discovery of programsof proven effectiveness in this vital institutional setting. The effectivenessof community prevention might be greatly increased by a substantial investmentin more controlled testing of program effects on serious crime. The Departmentof Labor has invested $15 million in a randomized test of a single jobtraining program. The prevention of serious crime in communities whereit is heavily concentrated should warrant at least that much.

A fast-track strategy for advancing knowledge about community crimeprevention is a multi-level randomized trial, with experiments imbeddedin experiments. Mentoring programs, for example, can be randomly assignedto half the communities. Then within communities, the program can be providedto half the applicants. Gang prevention strategies for reducing cohesioncan be randomly assigned to half of the communities, and then within halfof the communities receiving the program it can be randomly assigned tohalf of the gangs. If "communities" are defined at the levelof Census tract, there could be several hundred units of analysis availablefor this kind of multi-level research design.

The design could also embody elements that would always be deliveredto the entire community. Substantial increases in police patrol, for example,could greatly reduce the crime rate in the short run. That, in turn, couldassist efforts to attract new employers to the community, creating long-termemployment opportunities. That, in turn, could diversify the class andrace composition of the neighborhood, reducing hypersegregation on bothvariables as a risk factor. Drug prevention programs, recreation centers,school and family-based programs could be added as well. While many ofthese elements are already part of OJP funding plans, the method of testingthem in randomly assigned combinations is not.

A broader experiment in community-based mentoring could draw separatesamples from systematically different communities, chosen on theoreticalgrounds. A contemporaneous trial in two segregated inner-city communitiesof concentrated poverty, two predominantly white but high single-parentfamily suburban areas, and two racially and economically mixed areas wouldanswer a key question: is whether the effects of the mentoring programvary by community context. An added comparison of Hispanic and African-Americanpoverty areas would also illuminate the role of ethnicity, if any, in conditioningthe effects of community-based mentoring. Separate random assignment schedulesin each location would allow a strong test of interaction effects, ratherthan the multivariate correlational methods used in the Tierney and Grossmanwith Resch (1995) test.

The importance of testing mentoring in different communities is clear.Many prevention strategies evaluated in this report produce different effectsfor different kinds of people, and in different community contexts. TheCambridge-Somerville experiment is a caution that mentoring, like gangintervention, may well backfire. It would be a mistake of both scienceand policy to support community-based mentoring for all communities ona one-size-fits-all basis. While that may well be the ultimate result ofsuch a research program, the possibility of differential effects must becarefully examined.

Additional elements for a national experiment for dealing with highcrime communities are suggested in the following chapters. Regardless ofthe specific elements included, the scientific basis for such an experimentremains the same. While scientists clearly disagree over the best way tohandle the difficulties of community-level prevention (Bloom, 1996; Farrington,1997), there is substantial agreement that we are not learning enough aboutthe relative effectiveness of different strategies for community-basedcrime prevention.

NOTES

1Sampson and Lauritsen, 1993: 89.

2While community crime rates have clear correlations withrisk factors, there is still no scientifically conclusive evidence of causation,for reasons summarized in Sampson and Lauritsen (1993) at pp. 75-83. Thusthe term "cause" in this section is used flexibly to denote ahigh priority target for a public policy intervention, a risk factor whoseelimination might reduce crime.

3Whether this program is properly characterized as a mentoringprogram or something else is an issue debated within the University ofMaryland team, one that illustrates the difficulty of characterizing multi-dimensionalprograms on the basis of any one dimension.

4Control groups and randomized experiments are generallyfar more possible ethically than many public officials are willing to concede,giving the scarcity of resources. Waiting lists are an excellent opportunityfor controlled experiments. In this case, the control group males waitedno longer than the applicants on the waiting list.

5This section is based largely on a recent (August, 1996)G.A.O. report entitled “Status of Delinquency Prevention Program and Descriptionof Local Projects.”

6Office of Juvenile Justice and Delinquency Prevention, FY1996 Discretionary Competitive Program Announcements and Application Kit,p. 25.

REFERENCES

Anderson, Elijah

1978 A Place on the Corner. Chicago: University of Chicago Press.

1990 Streetwise: Race, Class and Change in an Urban Community. Chicago:University of Chicago Press.

Battin, S., K.G. Hill, J.D. Hawkins, R.F. Catalano, and R. Abbott

1996 Testing Gang Membership and Association with Antisocial Peers asIndependent Predictors of Anti-Social Behavior. Paper Presented to theAmerican Society of Criminology, Chicago, November.

Bernard, Thomas

1990 Angry Aggression Among the Truly Disadvantaged. Criminology 28:73-96.

Bibb, M.

1967 Gang-Related Services of Mobilization for Youth. In M. Klein andB.G. Myerhoff, Eds., Juvenile Gangs in Context. Englewood Cliffs, NJ: Prentice-Hall.

Bloom, Howard S.

1996 Building A Convincing Test of Public Housing Employment ProgramUsing Non- Experimental Methods: Planning for the Jobs-Plus Demonstration.Unpublished MS, The Wagner School, New York University.

Boruch, Robert and Ellen Foley

1996 Testing A Program Model in Many Sites: The Site and Other Entitiesas the Unit of Allocation and Analysis in Controlled Experiments. DraftPaper prepared for the US Department of Education.

Braithwaite, John

1989 Crime, Shame and Reintegration. Cambridge: Cambridge UniversityPress.

Brown, R.C., Jr. and D.W. Dodson

1959 The Effectiveness of a Boy's Club in Reducing Delinquency. Annalsof the American Academy of Political and Social Sciences 322: 47-52.

Bursik, Robert

1989 Political Decision-Making and Ecological Models of Delinquency:Conflict and Consensus. In S. Messner, M. Krohn, and A. Lisska, TheoreticalIntegration in the Study of Deviance and Crime. Albany: SUNY Press.

Bursik Robert and Harold G. Grasmick

1993 Neighborhoods and Crime. New York: Lexington

Burt, Martha

1996 The Violence Against Women Act of 1994: Evaluation of the STOPBlock Grants to Combat Violence Against Women. Washington, DC: Urban Institute.

Callahan, Charles, Frederick Rivara and Thomas Koepsell

1995 Money For Guns: Evaluation of the Seattle Gun Buy-Back Program.In Martha Plotkin, Ed., Under Fire: Gun Buy-backs, Exchanges, and AmnestyPrograms.

Carnegie Council on Adolescent Development

1995 Great Transitions. Washington, DC: Author.

Carney, Frank, Hans W. Mattick and John D. Callaway

1969 Action on the Streets. NY: Association Press.

Clark, Kenneth

1965 Dark Ghetto: Dilemmas of Social Power. NY: Harper and Row.

Cohen, Albert K.

1955 Delinquent Boys: The Culture of the Gang. Glencoe: Free Press.

Cohen, Lawrence and Marcus Felson

1979 Social Change and Crime Rate Trends: A Routine Activities Approach.American Sociological Review 44: 588-607.

Collins, James F.

1989 Alcohol and Interpersonal Violence: Less Than Meets the Eye. InN. Weiner and M. Wolfgang, Eds., Pathways to Criminal Violence. NewburyPark, CA: Sage.

Cook, Thomas D. and Donald T. Campbell

1979 Quasi-Experimentation: Design and Analysis Issues for Field Settings.Chicago: Rand-McNally.

Cook, Philip J.

1991 The Technology of Personal Violence. In Michael Tonry and NorvalMorris, Eds., Crime and Justice, Vol. 14. Chicago: University of ChicagoPress.

Cook, Philip J. and Mark H. Moore

1995 Gun Control. In James Q. Wilson and Joan Petersilia, Eds., Crime.San Francisco: ICS Press.

Curry, G. David and Irving A. Spergel

1988 Gang Homicide, Delinquency and Community. Criminology 26: 381-405.

1993 Gang Involvement and Delinquency Among Hispanic and African-AmericanAdolescent Males. Journal of Research in Crime and Delinquency. 29: 273-91.

Dicken, C., R. Bryson and N. Kass

1977 Companionship Therapy: A Replication in Experimental CommunityPsychology. Journal of Consulting and Clinical Psychology 45: 637-646.

Dunworth, Terence, Peter Haynes and Aaron J. Saiger

1997 National Assessment of the Byrne Formula Grant Program. ResearchIn Brief. Washington, DC: National Institute of Justice

Esbensen, Finn-Aage and David Huizinga

1993 Gangs Drugs and Delinquency in a Survey of Urban Youth. Criminology4: 565-589.

Farrington, David P.

1997 Evaluating a Community Crime Prevention Programme. Evaluation 3(forthcoming).

FBI (Federal Bureau of Investigation, U.S. Department of Justice)

1994 Crime in The United States, 1993. Washington, DC: USGPO.

Felson, Marcus

1994 Crime in Everyday Life. Thousand Oaks, CA: Pine Forge Press.

Fingerhut, Lois and J.C. Kleinman

1990 International and Interstate Comparisons of Homicides Among YoungMales. Journal of the American Medical Association 263: 3292-3295.

Fo, W.S.O, and C.R. O'Donnell

1974 The Buddy System: Relationship and Contingency Conditioning ina Community Intervention Program for Youth with NonProfessionals as BehaviorChange Agents. Journal of Consulting and Clinical Psychology 42: 163-169.

1975 The Buddy System: Effect of Community Intervention on DelinquentOffenses. Behavior Therapy 6: 522-524.

Fowler, Floyd J., Jr. and Thomas Mangione

1986 A Three-Pronged Effort to Reduce Crime and Fear of Crime: The HartfordExperiment. In Dennis Rosenbaum, editor, Community Crime Prevention: DoesIt Work? Beverly Hills, CA: Sage.

Gold, Martin and Hans W. Mattick

1974 Experiment in the Streets: The Chicago Youth Development Project.Ann Arbor, MI: Institute for Social Research.

Goldstein, Arnold P., B. Glick, & W. Carthan

1989 The Prosocial Gang: Implementing Aggression Replacement Training.Thousand Oaks, CA: Sage.

Goldstein, Paul

1989 Drugs and Violent Crime. In N. Weiner and M. Wolfgang, Eds., Pathwaysto Criminal Violence. Newbury Park, CA: Sage.

Goodman, G.

1972 Companionship Therapy: Studies in Structured Intimacy. San Francisco,CA: Jossey-Bass.

Green, B.C.

1980 An Evaluation of a Big Brothers' Program for Father-Absent Boys:An Eco- Behavioral Analysis. PhD Dissertation, New York University.

Hagedorn, John

1988 People and Folks: Gangs, Crime and the Underclass in a RustbeltCity. Chicago: Lakeview Press.

Hawkins, J. David, Richard F. Catalano and Associates

1992 Communities That Care: Action for Drug Abuse Prevention. San Francisco,CA: Jossey-Bass.

Hawkins, J. David, Michael W. Arthur, and Richard Catalano

1995 Preventing Substance Abuse. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Hill, K.G., J D. Hawkins, R. Catalano, R. Kosterman, R. Abbott and T.Edwards

1996 The Longitudinal Dynamics of Gang Membership and Problem Behavior:A Replication and Extension of the Denver and Rochester Gang Studies inSeattle. Paper Presented to the Anerican Society of Criminology, Chicago.

Hirsch, A.

1983 Making the Second Ghetto: Race and Housing in Chicago, 1940-1960.Chicago: University of Chicago Press.

Homel, Ross

1994 Personal Communication

Hope, Tim

1995 Community Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Howell, James C., Editor

1995 Guide for Implementing the Comprehensive Strategy for Serious,Violent, and Chronic Juvenile Offenders. Washington, DC: OJJDP.

Forthcoming Promising Programs for Youth Gang Violence Prevention andIntervention. (Draft) In Rolf Loeber and David P. Farrington, Eds., Reportof the Study Group on Serious, Violent, Chronic Juvenile Offenders. Washington,DC: OJJDP

Huff, Ronald C.

1989 Youth Gangs and Public Policy. Crime and Delinquency 35: 524-537.

Jackson, Pamela Irving

1991 Crime, Youth Gangs, and Urban Transition: The Social Dislocationof Postindustrial Economic Development. Justice Quarterly 8: 379-97.

Jones, M.B and D.R. Offord

1989 Reduction of Anti-Social Behavior in Poor Children by NonschoolSkill Development. Journal of Child Psychology and Psychiatry and AlliedDisciplines 30: 737-750.

Kanter, Rosabeth Moss

1977 Men and Women of the Corporation. New York: Basic Books.

Kelling, George L. and Catherine M. Coles

1996 Fixing Broken Windows: Restoring Order and Reducing Crime in OurCommunities. NY: Free Press.

Kennedy, David M., Anne M. Piehl, and Anthony A. Braga

1996 Youth Gun Violence in Boston: Gun Markets, Serious Youth Offenders,and a Use Reduction Strategy.

Kinlock, Timothy

1991 PhD dissertation, University of Maryland.

Klein, Malcolm

1968 From Association to Guilt: The Group Guidance Project in JuvenileGang Intervention. Los Angeles: University of Southern California.

1971 Street Gangs and Street Workers. Englewood Cliffs, NJ: Prentice-Hall.

1995 The American Street Gang: Its Nature, Prevalence and Control. NY:Oxford University Press.

Lavrakas, P.J. and S.F. Bennett

1989 A Process and Impact Evaluation of the 1983-1986 Neighborhood Anti-CrimeSelf- Help Program: Summary Report. Evanston, Ill: Northwestern University,Center for Urban Affairs and Policy Research.

Liebow, Elliott

1967 Tally's Corner. Boston: Little, Brown.

Lemann, Nicholas

1991 The Promised Land: The Great Black Migration and How It ChangedAmerica. New York: Alfred A. Knopf.

Lizotte, Alan, et al.

1996 Paper Presented to the American Society of Criminology, Chicago,November.

Logan, J. and H. Molotch

1987 Urban Fortunes: The Political Economy of Place. Berkeley, CA: Universityof California Press.

McCord, Joan

1978 A Thirty-Year Followup of Treatment Effects. The American Psychologist33: 284- 289.

1992 Understanding Motivations: Considering Altruism and Aggression.In J. McCord, Ed., Facts, Frameworks, Forecasts: Advances in CriminologicalTheory. New Brunswick, NJ: Transaction.

Massey, Douglas S. and Nancy A. Denton

1993 American Apartheid: Segregation and the Making of the Underclass.Cambridge, Mass: Harvard University Press.

Messner, Steven and K. Tardiff

1986 Economic Inequality and Levels of Homicide: An Analysis of UrbanNeighborhoods.

Criminology 24: 297-318.

Miczek, Klaus A., Joseph F. DeBold, Margaret Haney, Jennifer Tidey,Jeffrey Vivian, and Elise M. Weerts.

1993 Alcohol, Drugs of Abuse, Aggression, and Violence. In Reiss, AlbertJ., Jr. and Jeffrey Roth, Eds., Understanding and Preventing Violence.Vol. 3. Washington, DC: National Academy of Sciences.

Miller, Walter B.

1962 The Impact of a "Total-Community" Delinquency ControlProject. Social Problems 9: 168-191.

Moore, Mark H.

1980 The Police And Weapons Offenses. Annals of the American Academyof Political and Social Sciences 452: 22-32.

Moore, J. P.

1996 The 1995 Youth Gang Survey. Report to the OJJDP. Tallahasee: NationalYouth Gang Center.

Powers, E. and H. Witmer

1972 An Experiment in the Prevention of Delinquency: The Cambridge-SomervilleYouth Study. Montclair, NJ: Patterson Smith [Original publication 1951].

Quetelet, L.A.J.

1842 A Treatise on Man and the Development of His Faculties. Edinburgh:W. and R. Chambers.

Reiss, Albert J., Jr.

1986 Why Are Communities Important in Understanding Crime? In AlbertJ. Reiss, Jr. and Michael Tonry, Communities and Crime. Crime and Justice:A Review of Research. Vol. 8. Chicago: University of Chicago Press.

1988 Co-Offending and Criminal Careers. In Michael Tonry and NorvalMorris, Eds., Crime and Justice, Vol. 10. Chicago: University of ChicagoPress.

Reiss, Albert J., Jr. and Jeffrey Roth, Editors

1993 Understanding and Preventing Violence. Washington, DC: NationalAcademy of Sciences.

Rocheleau, Ann Marie, George L. Kelling, Wesley G. Skogan, Jeffrey R.Roth, and Dennis J. Rosenbaum

1996 Evaluation of the Bureau of Justice Assistance Comprehensive CommunitiesProgram: Interim Report. Cambridge, Mass.: BOTEC Analysis Corp.

Roehl, Janice A., Robert Huitt, Mary Ann Wycoff, Antony Pate, DonaldRebovich, and Ken Coyle

1996 National Process Evaluation of Operation Weed and Seed. Researchin Brief. Washington, DC: National Institute of Justice.

Roncek, Dennis and P. A. Maier

1991 Bars, Blocks and Crime Revisited: Linking the Theory of RoutineActivities to the Empiricism of Hot Spots. Criminology 29: 725-53.

Rosenfeld, Richard

1995 Gun Buybacks: Crime Control or Community Mobilization. In MarthaPlotkin, Ed., Under Fire: Gun Buy-backs, Exchanges, and Amnesty Programs.Washington, DC: Police Executive Research Forum.

Sampson, Robert

1986 Crime In Cities: The Effects of Formal and Informal Social Controls.In Albert J. Reiss, Jr. and Michael Tonry, Communities and Crime. Crimeand Justice: A Review of Research. Vol. 8. Chicago: University of ChicagoPress.

Sampson, Robert and Casey Groves

1989 Community Structure and Crime: Testing Social-Disorganization Theory.American Journal of Sociology 94: 774-802.

Sampson, Robert and Janet Lauritsen

1993 Violent Victimization and Offending: Individual, Situational andCommunity-Level Risk Factors. In Reiss, Albert J., Jr. and Jeffrey Roth,Eds., Understanding and Preventing Violence. Vol 3. Washington, DC: NationalAcademy of Sciences.

Schinke, Steven P., Orlandi, Mario A. and Kristin C. Cole

1992 Boys & Girls Clubs in Public Housing Developments: PreventionServices for Youth at Risk. Journal of Community Psychology OSAP SpecialIssue: 118-128.

Schuerman, L. and Solomon Kobrin

1986 Community Careers in Crime. In Albert J. Reiss, Jr. and MichaelTonry, Eds., Communities and Crime. Crime and Justice: A Review of Research.Vol. 8. Chicago: University of Chicago Press.

Shaw, James W.

1994 Community Police Against Crime: Violence and Guns. Ph.D. Dissertation,University of Maryland.

Shaw, Clifford R. and Henry McKay

1942 Juvenile Delinquency and Urban Areas. Chicago: University of ChicagoPress.

Sherman, Lawrence W.

1995 The Police. In James Q. Wilson and Joan Petersilia, Eds., Crime.San Francisco: ICS Press.

Sherman, Lawrence W., James W. Shaw and Dennis P. Rogan

1995 The Kansas City Gun Experiment. Research in Brief. Washington,DC: National Institute of Justice.

Skogan, Wesley

1986 Fear of Crime and Neighborhood Change. In Albert J. Reiss, Jr.and Michael Tonry, Eds., Communities and Crime. Crime and Justice: A Reviewof Research. Vol. 8. Chicago: University of Chicago Press.

1990 Disorder and Decline. NY: Free Press.

Spergel, Irving A.

1986 The Violent Gang in Chicago: A Local Community Approach. SocialService Review 60.

1995 The Youth Gang Problem: A Community Approach. NY: Oxford UniversityPress.

Spergel, Irving and S.F. Grossman

1995 Little Village Gang Violence Reduction Program. Paper Presentedat the National Evaluation Conference, National Institute of Justice, Washington,DC.

Taylor, Ralph, Stephen Gottfredson and Sidney Brower

1984 Block Crime and Fear: Defensible Space, Local Social Ties and TerritorialFunctioning. Journal of Research in Crime and Delinquency 21: 303-331.

Taylor, Ralph and Stephen Gottfredson

1986 Evironmental Design, Crime, and Prevention: An Examination of CommunityDynamics. In Albert J. Reiss, Jr. and Michael Tonry, Eds., Communitiesand Crime. Crime and Justice: A Review of Research. Vol. 8. Chicago: Universityof Chicago Press.

Thompson, D.W. and L.A. Jason

1988 Street Gangs and Preventive Interventions. Criminal Justice andBehavior 15: 323- 333.

Thornberry, Terence P.

Forthcoming Gangs and Serious, Chronic and Violent Offenders. Draft.In Rolf Loeber and David P. Farrington, Eds., Report of the Study Groupon Serious, Violent, Chronic Juvenile Offenders. Washington, DC: OJJDP

Thornberry, Terence P., Marvin D. Krohn, Alan J. Lizotte, and DeborahChard-Wierschem.

1993 The Role of Juvenile Gangs in Facilitating Delinquent Behavior.Journal of Research in Crime and Delinquency 30: 55-87.

Thrasher, Frederic

1936 The Boys' Club and Juvenile Delinquency. American Journal of Sociology41: 66-80.

Tierney, Joseph P. and Jean Baldwin Grossman, with Nancy L. Resch

1995 Making a Difference: An Impact Study of Big Brothers/Big Sisters.Philadelphia, PA: Public/Private Ventures.

Toby, Jackson

1957 Social Disorganization and Stake in Conformity: Complementary Factorsin the Predatory Behavior of Hoodlums. Journal of Criminal Law, Criminologyand Police Science 48: 12-17.

Torres, Dorothy

1981 Gang Violence Reduction Project: Fourth Evaluation Report. Sacramento,CA: California Youth Authority.

Tyler, Tom

1990 Why People Obey the Law. New Haven: Yale University Press.

Wilson, James Q. and Kelling

1982 Broken Windows: The Police and Neighborhood Safety. Atlantic Monthly249: 29-38.

Wilson, William Julius

1987 The Truly Disadvantaged. Chicago: University of Chicago Press.

1996 When Work Disappears: The World of the New Urban Poor.

Woodson, Robert L.

1981 A Summons To Life: Mediating Structures and the Prevention of YouthCrime.

Cambridge, Mass.: Ballinger.

Chapter Four

FAMILY-BASED CRIME PREVENTION

by Lawrence W. Sherman

Family risk factors have a major effect on crime. Family based crimeprevention can directly address those risk factors, with substantial success.The more risk factors they address, perhaps, the better. The earlier theystart in life, it seems, the better. Programs for infants and young childrenmay be most cost-effective in the long run, even if they are expensivein the short run. Combining home-visit parental support with preschooleducation reduces crime committed by children when they grow up. Rigorouslyevaluated pilot projects with tightly controlled prevention services areconsistently effective. Family problems later in life are more difficultto address, especially family violence by adults. But it is still possible.The potential of early, adolescent and adult family-based crime preventionis held back only by our failure to invest in more research and development.The need for testing programs that can work on a large scale is particularlygreat.

Most of these conclusions have been reached independently by diversescholars from diverse disciplines (Yoshikawa, 1994; Tremblay and Craig,1995; Hawkins, Arthur and Catalano, 1995; Crowell and Burgess, 1996; Kumpfer,Molgaard and Spoth, 1996; Wasserman and Miller, forthcoming). Given thenormal disagreements among social scientists, the level of consensus aboutthese conclusions is striking. But of all these conclusions, the need forfurther careful evaluations is the strongest point of agreement. Evaluatingthe varieties of possible transitions from a small pilot test of a programto a large-scale operation is a step that is frequently left out, as itwas in the case of Head Start (Lazar, 1992, and Zigler, 1992, both as citedin Yoshikawa, 1994). There is no government institution fully preparedto deliver family-based prevention of the kind found effective in the scientificliterature. Making the most out of what we know already will require evenmore knowledge about how to go from pilot tests to full operations.

Much more is known about making families better at child-raising thanabout preventing family violence. A recent review of the effectivenessof criminal sanctions in combatting domestic abuse concludes that the evidencein favor of these programs is either weak or absent (Fagan, 1995). Batterer'scounseling, mandatory arrest, special prosecution and victim advocacy programsall remain essentially unevaluated. While theoretical inferences supportsuch programs as battered women's shelters to reduce danger during thehigh-risk aftermath of an incident reported to police, there is no assurancethat any of these programs actually increase long-term victim safety. Courtorders of protection and other legal steps advised by victims' advocatesmay even increase risk of serious injury to victims. Mandatory arrest formisdemeanor spouse assault without prosecutorial action or court treatmenthas been found to be either ineffective or criminogenic in repeated controlledtrials, although it is effective in communities of strong social capital.

Perhaps least is known about the extent to which the same family-basedprograms can prevent both family violence and delinquent acts by childrenin the family. One home-visit program for infants, for example, reducedchild abuse, which is both a crime of domestic violence and a risk factorfor later delinquency of abused children. The potential for broadeningthe outcome measures and objectives of family-based crime prevention isimportant for public policy analysis. It has great potential, for example,in helping to design a program that might work on a much broader scalethan the pilot tests to date, most of which are limited to a few hundredparticipants or less. It is also one more good reason to invest more heavilyin research and development.

This chapter briefly reviews the variety of family-based crime preventionprograms. It then considers a few of the major research issues in evaluatingand designing family-based prevention. Five major areas of research arethen examined in detail, each in relation to an ecological context wherefamilies seek or receive help affecting crime and risk factors: homes,pre-schools and schools, clinics, courts, and other contexts. The chapterconcludes with a scientific summary of what works, what doesn't, and what'spromising, with assessments of what is known about the effectiveness offederally funded programs and suggestions for improving effectiveness throughbetter evaluations.

VARIETIES OF FAMILY-BASED CRIME PREVENTION

Family-based crime prevention is an unintended beneficiary of the vastresearch enterprise on human development. Much of what we know about itcomes from evaluations of programs established for other purposes. Manyof these human development programs are highly elaborated, each with itsown terminology, literature, and professional community. As programs intendedto improve parents' child-rearing skills, children's academic skills, orchildren's mental health, they have often resulted--almost coincidentally--inreduced crime. This fact underlines the importance of defining preventionnot as intention, but as result. It also shows how basic to human experiencethe factors affecting the risk of crime can be.

Several analyses of risk factors for both serious and general delinquencyconclude that family factors are important. While serious crime is geographicallyconcentrated in a small number of high crime communities, it is individuallyconcentrated in families with anti-social parents, rejecting parents, parentsin conflict, parents imposing inconsistent punishment, and parents whosupervise their children loosely (Tremblay and Craig, 1995: 158). Severalanalysts conclude that these risk factors are cumulative, and that themore of them a prevention program can address the better (Coie and Jacobs,1993; Yoshikawa, 1994; Tremblay and Craig, 1995; Wasserman and Miller,forthcoming). This hypothesis is consistent with much of the literature,and not falsified by any direct test. Perhaps the best way to explore itis to evaluate rigorously prevention programs addressing different numbersand combinations of risk factors.

Risk Levels and Prevention Strategy

The basic structure of family-based prevention programs depends uponstrategic choices with public safety, budgetary and political consequences.The basic choice is between universal and targeted programs(Institute of Medicine, 1994). Universal programs are offered to, or evenimposed upon, all families. In several European countries, for example,all families with newborn children are required to admit trained nursesto their homes to visit the baby. This program applies to everyone withoutregard to any risk factors. Targeted programs are of two kinds.One kind is "selective," in which families (or individuals)identified as being at high risk are offered or mandated to receive a serviceintended to prevent the onset of harm. The other kind of targeted programis called "indicated." In the case of crime and delinquency,indicated programs are offered to prevent recurrence of crime by childrenalready manifesting crime or crime risk factors. Because the term "targeted"in crime prevention policy is increasingly unacceptable to African-Americansas too resonant of racially discriminatory practices, this report willsubstitute the term "focused" to denote the same concept.

The choice between universal and focused programs is complex. Focusedprograms may make more efficient use of scarce resources, but universalprograms may attract greater resource levels per family. It may not benecessary to allocate resources equally to all families within a program.But it may well be necessary to have the program itself be universal inorder to make a very high cost investment politically palatable. The failureof Head Start to obtain full funding, for example, may be linked directlyto the fact that it is seen as a program for poor children, rather thanfor all children.

Families with high levels of crime risk factors may also be more likelyto accept universal programs than focused ones. This may be particularlyimportant for more intrusive interventions into family life, such as frequenthome visitation. Any possible stigma of such intrusion may be limited bythe universal character of the program. To the extent that risk factorsin some geographic areas are correlated with race, focused programs maybe even more problematic. But programs applying to all children and allfamilies avoid any implication of discrimination.

Even though this report generally concludes that crime prevention canbe most effective when scarce resources are focused on concentrations ofrisk factors, family-based crime prevention provides an important exception.What makes sense across cities and even schools may not work at the levelof family life. The state's relationship to the citizenry is most sensitivein the institutional setting of the family. Interpreting the policy implicationsof the scientific evidence reviewed in this chapter can be accomplishedmost usefully with the issue of universal versus focused programming inmind. The "elasticity" of demand for such programs may besuch that the more expensive they become through universal access, themore likely they are to be fully funded.

Figure 4-1

Family-Based Crime Prevention by Ecological Context

Ecological Context Program Prevention Delivery Agent (s) HOME Regular visits for Nurses, Universal or emotional, informational, Teachers, Selective instrumental and Para- educational support for professionals, Rarely parents of preschool (or Preschool indicated older) children Teachers Foster care outplacement Family Indicated for the prevention of services, physical, sexual abuse or Social worker neglect Family preservation of Private Indicated families at risk of Family, outplacement of child preservation teams Personal alarm for victims Police Indicated of serious domestic violence In-home proactive Police, Indicated counseling for domestic Social Workers violence PRESCHOOL Involvement of mothers in Preschool Universal or parent groups, job teachers selective training, parent training SCHOOL Parent training Psychologists. Indicated or Teachers Selective; some universal Simultaneous Parent and Psychologist, Indicated or Child Training Child Care selective Workers, Social Workers CLINICS Family Therapy Psychologists, Indicated, Psychiatrists, Selective Social Workers Medication--psychostimulants Psychiatrists, Indicated for treatment of Psychologists, hyperactivity and other Pediatricians childhood conduct disorders HOSPITALS Domestic Violence Nurses, Indicated Counseling Social Workers Low-Birthweight Baby Nurses, Indicated Mothers' Counseling & Social Workers Support COURTS Prosecution of Batterers Police, Indicated Prosecutors Warrants for Unarrested Police, Indicated Batterers Prosecutors Restraining Orders or Police, Indicated "Stay-Away" Orders of Prosecutors, Protection Judges, Victims' Advocates Hotline Notification of Probation, Indicated victim about Release of Victim Incarcerated Domestic Advocates Batterer BATTERED WOMEN'S SHELTERS Safe Refuge during Volunteers; Indicated high-risk 2-7 days staff aftermath of domestic assault; counseling; hotlines 

The Ecology of Family-Based Prevention

Despite the potentially greater appeal of universal programs, Figure1 reveals a striking fact: almost all family-based crime prevention iscurrently offered on a focused basis. Absent an indicated reason to intervenein family life, American government generally leaves families alone. Incontrast to many other western nations, the United States performs almostno universal monitoring of families in the home.1

This pattern creates a distinct ecology of prevention which treats familiesvery differently in different places (Stinchcombe, 1963). The state imposesrequirements on the disease-prevention vaccinations children must receivein hospitals and medical clinics, for example, but does not generally empowerpublic health agents to invade the home to deliver vaccinations. The authorityof the school teacher is great in a school building, but ambiguous whenthe teacher visits a private home by parental consent. The realm of thepossible in family-based crime prevention programs is defined largely bythe ecological context in which the programs might be delivered, and theauthority vested in the government to intervene in family life associatedwith each of those contexts.

These contexts, as presented in Figure 1, include schools, preschools,hospitals, clinics, courts and battered women's shelters, as well as thehome itself. All other contexts are in some sense merely windows on thehome, opportunities for dialogue between the state and the family thatcan shape the results of family life for public safety. Hospitals and schoolsare places where crimes in the home are often detected and reported topolice, who then have legal standing to investigate events in the home.They are also places where advice and instructions about reducing riskfactors can be given. Absent the indication of existing problems or highrisk, however, there are no universal crime prevention mechanisms comparableto medical vaccines.

This chapter is therefore a review of the effectiveness of programswithin one strategic realm of family-based crime prevention: focusedinterventions. This represents an existing choice not to develop universalprograms. It does not, of course, show whether focused programs are moreor less effective than universal programs might be. In order to answerthat question, it is necessary for a large-scale program of research anddevelopment to compare universal and targeted programs for their relativeeffectiveness. To the extent that universal programs might detect and preventmore problems than targeted programs, their value remains a major untestedhypothesis in family-based crime prevention.

EVALUATING FAMILY-BASED CRIME PREVENTION

Scientific evaluations of family-based crime prevention programs faceat least three distinctive problems, compared with other institutionalsettings. Perhaps foremost is the long time horizon often needed to measurethe effectiveness of prevention programs. Also important is the possiblevariation in effectiveness by intensity or accumulation of risk factors.There are also unique problems in measuring crimes committed by familymembers against one another, in relation to both privacy and safety forresearch subjects and accuracy of measurement.

Long Time Horizon

A basic premise of developmental crime prevention is that what happensduring infancy can affect the odds of crime two or three decades later.Giving this theory a fair test requires a very long time horizon. Sustainingthe test over the time required creates problems of cost, management, andinterpretation.

The problem of cost is not as great as it seems. Numerous birth cohortstudies of delinquency have been funded intermittently over decades, keepingtrack of where to find the research subjects for repeated interviews andofficial record checks (Farrington, Ohlin and Wilson, 1987). The currentOJP limitation of grant periods to two years poses more of a managementproblem than a cost problem, creating uncertainty about commitments toemploy key staff and other planning issues. Relaxing that limitation forfive- and ten-year projects would ease those difficulties, and help encouragemore tests of developmental crime prevention strategies. The major problemthis creates in interpreting available evidence is that there are so fewlong-term studies to examine.

The problem of management is perhaps more critical to interpretationof long-term findings. The two longest running tests of developmental crimeprevention are both reputed to be very well managed programs (Berrueta-Clementet al, 1985; Lally, et al, 1987). Critics have raised the problem of generalizingfrom the results of small, well-managed programs to large, bureaucraticallyadministered programs. The key question is how accurately we can predictthat a long-term program serving tens of thousands of families will havethe same effects as a short-term test program serving several hundred familiesfor 3 to five years. In order to answer that question, we require researchdesigns testing much larger scale programs over a longer period of time.That requires not only much greater cost, but a separate political processnecessary to sustain the resources for the time horizon required. For example,ten years worth of birth cohorts might be needed to see if the long-termeffects of a program operating during the enthusiasm (or confusion!) ofan initial launch were the same as a program that was three, five, eightor ten years old.

Finally, the issue of interpretation is compounded by the speed withwhich our society is changing. By the time the results are in from a two-decadeold test, the context of the program may have changed in important ways.Perhaps more qualified preschool teachers were available in the early 1960sthan today, for example. Or perhaps the concentration of poverty in innercities is so much worse in the 1990s than in the early 1960s (Wilson, 1996)that crime prevention benefits found in an earlier study would not standup to today's more intense risk factors. Early feedback from measures ofprotective factors (like school conduct assessment) and child abuse mighthelp solve this problem, providing both short- and long-term feedback.Conversely, short-term child abuse interventions such as Olds et al (1986)provide excellent opportunities for long-term followup of delinquency prevention,and even domestic violence prevention. Generating and funding such followupresearch should be a high priority for OJP. Similarly, short-term followupsof drug abuse prevention programs merit much longer term followups, tosee whether other factors cancel out early effects of interventions.

Cumulative Risk Factors and Contextual Data

This report's concern for the interdependency of crime prevention institutionsis not widely shared in crime prevention research. Many clinic-based studies,for example, do not report precise data on the neighborhoods from whichthe research subjects are drawn. It is one thing to say that the childrenare from families on welfare or have teenage mothers. It is another thingaltogether to report that 35% of the families in the sample reside in neighborhoodswith adult unemployment rates in excess of 70%, and with 60% of householdsin the census tract below the poverty line (see Chapter two). Very fewindividual-level experiments report community-level data in the degreeof specificity needed to begin to synthesize results and draw broader conclusionsabout program effectiveness.

Family-based prevention programs may work well in areas of high risk,but only up to a point. For example, clinic-based parent training for parentsof aggressive elementary school children may work in all neighborhoodsin Oregon, but not in many neighborhoods in Chicago. If there is a tippingpoint beyond which a parentally focused program may not work, it cannotbe identified from the literature without more precise measurement. Thereis also a problem of consistency of the treatment itself across citiesand treatment staff. That may interact, in turn, with the accumulationof risk factors. Some treatment staff or clinics may have greater capacityor experience to deal with concentrated risk factors than others.

Resolving the interaction of risk level with treatment effectivenessrequires systematic attention and costly cross-site scientific designs.Planned variations in staff capacity, neighborhood social factors and familyvariables must be structured into the research design. Controlled experimentationwith treatments across sites, as distinct from comparing naturally occurringvariation in local treatment capacity, is required to bring a scientificmethods score up to level 4 or 5. There is little precedent for this kindof research. But without it there will remain major limitations in generalizingfrom single-site experiments.

Measuring Crime in the Family

The issues of privacy and retaliation in measuring crimes within familiespose a great challenge for research. Continuing disagreements about theinterpretation of existing measures have afflicted even the strongest ofresearch designs (Fagan, 1996). The central problems are low completionrates of personal interviews with victims of family crimes who have beentreated, low or inconsistent reporting rates of subsequent crimes to police,and unwillingness to disclose crimes committed in the family during interviewsin the home while other family members are present (NCVS study).

In several sites of the NIJ spouse assault replication project (SARP),for example, there are different results found from victim interviews andofficial reports to police. While victim interview data showed that arrestedoffenders had committed fewer repeat offenses than offenders randomly assignedto a warning, the official data showed the opposite (Dunford, et al, 1990;Berk et al, 1992). In other cities the victim data showed no effect ofarrest while the official data showed some evidence of a backfiring effect(Hirschel, et al, 1990; Sherman, et al, 1991). But a major difference betweenthese data was the completion level: official data covered 100% of thesample while the victim interview rates were as low as 23%, and averaged41% in sites reporting a deterrent effect from victim interviews. Thusthe effects of arrest may have interacted with victim willingness to beinterviewed, biasing the sample towards victims who had enjoyed a protectiveeffect from arrest.

The measurement theory challenging official data on family violenceis that experimentally assigned criminal sanctions may encourage victimsto call police more readily, whereas experimentally assigned warnings maydiscourage victims from calling police. Thus the higher rates of reportedreoffending with the arrested subjects is arguably due to a measurementartifact. This theory does not explain why there are fewer repeat offensesreported about employed offenders randomly assigned to arrest comparedto those assigned to a warning, and why the measurement artifact wouldonly apply to unemployed offenders. A further theory could suggest thatpartners of employed males are less likely to call police than partnersor unemployed males after an arrest has been made for fear of the employedbatterer's losing his job. But none of these theories have been testeddirectly.

Possible solutions to these problems may lie in focusing scarce resourceson prevention and measurement of injuries treated in hospital emergencyrooms. Hospital cooperation with data collection on an anonymous basiscould then provide more reliable measures of domestic violence (Shermanand Strang, 1996), although even then questions will remain.

PREVENTION AT HOME

Perhaps the most promising results in all areas of crime preventionare found in the evaluations of home visitation programs. While these programsare often combined with other institutional elements, such as preschool,there is a large and almost uniformly positive body of findings on thispractice. Other prevention programs delivered in the home context, suchas personal alarms for domestic violence victims and family preservationservices, have been subject to far less research. These programs, however,generally operate on an indicated basis after crime problems have developedrather than on the selective basis of the home visitation programs. Combiningthese two findings may suggest even more reason for testing universal home-basedprevention programs, to see if possible benefits of child-centered programsmay be extended to family crimes involving adults.

Home Visitation Programs

Home visitation varies enormously in dosage levels, content, skill,and context. Yet there are common effects reported across all these variations.These common effects may be linked to a common core of treatment content,for which dosage levels may matter more than other dimensions. The commoncore of home visitation is a visitor who cares about child-raising sittingdown in a home with a parent and a child. Visitors can be nurses, socialworkers, preschool teachers, psychologists or paraprofessionals. They canprovide cognitive information, emotional support, or both. They can activelyteach parents, with hands on the children. Or they can passively watchand listen, merely giving parents a good listening to. They can be trainedin health (like nurses), human development (like psychologists and socialworkers), cognitive and social skills instruction (like preschool teachers)or some mixture of these subjects (like paraprofessionals). They can beexperienced or novice, enthusiastic or burned out, assertive or hesitant.But no matter who they are or what they do, they provide a bridge betweenthe parent, usually a mother, and the outside world.

Figure 2 summarizes the results of 18 different evaluations of programsthat included a home visitation component. The Figure and this discussiondraws primarily on the material in Yoshikawa's (1994) review, as well asTremblay and Craig's (1995) and the draft OJJDP review prepared by Wassermanand Miller (forthcoming). Based on the limited information provided inthe secondary reviews, the primary studies appear to merit level 4 to 5scientific methods scores by the standards of this report, although somemight drop to a 3 if they suffer large attrition problems. All of themshow positive effects of home visits on either some measure of crime bychildren when they enter adolescence (N = 2 experiments), child abuse duringor shortly after the period of home visits (N = 5 experiments), or riskfactors for delinquency (N = 10 experiments, 1 meta analysis). While themeta-analysis of Head Start evaluations (McKey, et al, 1985) shows thatthe measured effects wear off, that analysis includes the lowest dosageof home visits of any of the experiments: as few as two per year. In contrast,the substantial reductions in later delinquency in the two long-term followupstudies are associated with weekly home visits for periods up to five years.

Figure 4-2

Evaluations of Home Visitation Programs

(All studies ranked Level 4 or 5 on Scientific Methods Score)

(Secondary Review Sources: Yoshikawa, 1994 unless otherwise indicated;Tremblay & Craig, 1995; Wasserman & Miller, Forthcoming)

Primary Source Effects N of Visits, Visitors, Other Age of (Secondary Time Visited Service Child source if not Yoshikawa) EFFECTS ON CRIME 1)Berrueta-Clement Lower adult Weekly, 2-3 Teachers, High Pre- 3-5 yrs et al 1984 arrests by age years, 30 weeks risk school; High/Scope Perry 24 Exp= 7% per yr African-American Parent Preschool Control = 31% children & their Groups ( N = 121) (60 to 90 mothers visits) 2) Lally, et al Lower arrests by Weekly, 5 years Paraprofs, Low Pre-scho 0-5 1987 age 15 income, mostly ol; Syracuse Exp = 6% African-American pre-nata University Control = 22% children & their l Family (N = 119) (260 visits) mothers Development Research Program 3) Olds, et, al, Lower Child Bi-weekly over Nurses, first Doctor 0-2 1986, 1988 Abuse by age 2 122 weeks from born infants of Visits University of Exp = 19% late pregnancy high-risk low Rochester Control = 4% income white Prenatal/Early (N = 300) (up to 60 mothers Infancy Project visits) 4) Barth, Lower child Bi-weekly over Paraprof, Taught 0-6 mos Hacking & Ash abuse removals 26 weeks after children of Parent 1988 from home of birth mothers at risk Skills exps. for abusing them (N = 50) (12 visits) 5) Gray, et al, Fewer Injuries Weekly over an Nurses, children Doctor 0-2.5 1979 of Experimentals average of 130 and high risk visits yrs (N = 50) weeks mothers (130 visits) 6) Infant Health Less child abuse ? 3 years ? , High risk pre- 0-3 yrs Program & neglect children school (Tremblay & of experimentals Craig) (N = 985) 7) Larson 1980 Fewer Injuries 10 visits, most BA Psychologist, -- 0- 15 Montreal Home of Experimentals effect from 1 in infants of mos. Visitation Study (N = 95) pregnancy, 9 Canadian mothers (Wasserman & over 15 mos. in Montreal Miller) EFFECTS ON CRIME Effects N of Visits, Visitors, Other Age of RISK FACTORS  Time Visited Service Child 8) Seitz, et al Less anti-social Mean = 28 visits nurse, social Doctor 0-2.5 1982 behavior in over 2.5 years worker or Visits years Yale Child school at age 10 psychologist, Welfare Project by exp boys low ses (N = 30) first-borns and mothers 9) Johnson & Less anti-social 25 visits first Paraprofessional, Pre- 1-3 YRS Walker 1987 behavior in year of life for school Houston school at age 10 experimentals Low ses only and Parent-Child by exp children children of parent Development (N = 113) Mexican-American Center families classes 2d YR 10) Wasik et al Higher cognitive Biweekly from Teachers and see 0-5 mos 1990 scores up to 54 0-3; monthly 4-5 paraprofs, column Project Care mos. with Home months of age infants of low 2 visits + ses parents cognitive day care than with only home visits (N = 62) 11) Achenbach et Experimental 11 home visits, Reg. Nurse, Low None 0-3 mos al 1990 children had 0-3 mos Birth weight Vermont greater children Intervention cognitive skills Project by age 7 (N = 56) 12) McKey et al Head Start Varies, minimum Preschool Pre 3-4 yrs 1985 Meta-analysis 2 visits per teachers; school shows effects year children of wear off families in (N= 26 studies) poverty 13) Gutelius et Experimental Yr 1=18+ visits Nurses, first None pre-natal al 1977 children higher Yr 2=12+ visits children of to 3 on cognitive Yr 3= 8+ visits unmarried years scores to 3 yrs. mothers (N = 95) 14) Barrera et Experimental Weekly 0-4 mos Paraprof, None 0-1 Yr al 1986 mothers more Biweekly 5-9 Mothers of LBW responsive to mos infants age 1 LBW child Monthly 10-12 (N = 83) mos 15) Ross 1984 Mothers more Biweekly 0-3 Nurses, low ses None 0-1 Yr responsive, mos families with children better Monthly 4-12 LBW infants cognition age 1 mos (N = 80) 16) Jacobson & Exp. Mothers and Monthly in Paraprof, None Pre-natal Frye 1991 Infants more pregnancy firstborn to attached at age Weekly 0-2 mos children of 1 Yr 1 Monthly 3-12 low-ses mothers (N = 46) mos 17) Lieberman et Exp. children Weekly (52) Social Worker None 1-2 al 1991 less anxious at (MA, MSW), low years age 2 ses anxious and (N = 93) secure Hispanic children 18) Lyons-Ruth Exp. mothers and Weekly from Paraprof and MA None 9-18 mos et al 1990 infants more intake at 0-9 level; children attached at 18 mos up to of high risk mos completion at mothers (N = 76) 18 mos 

While the two long-term experiments both included preschool programs(also called "day care" in some studies), positive effects werefound in 11 of the experiments from home visitation without preschool.Some of the home visitations included doctor's office visits or some othercontexts for instruction and observation outside the home, but most didnot. None of the five experiments showing that home visitation reducedchild abuse included involvement in preschool.

The consistent finding of beneficial effects of home visits withoutpreschool is important for several reasons. One reason is theoretical:it shows that the visits are not simply a spurious correlate of the effectsof preschool programs on both the children and their mothers, who in somestudies are heavily involved in the preschool programs and who show beneficialeffects themselves in reduced welfare support and longer time between pregnancies.The fact that one trial (Wasik, et al, 1990) found stronger effects fromhome visits with cognitively oriented day care than from home visits tocomparison families (of which over half were in some other kind of daycare) does not contradict the independent effects of home visits. Yoshikawa(1994) and others have concluded that home visits are likely to be moreeffective in combination with early education, but the empirical evidencemay be still too preliminary to reach a conclusion either way.

Even if home visits were more effective in combination with other preventionefforts, the evidence of their independent effect has practical implications.The Hawaii state Healthy Start program, for example (U.S. Advisory Board,1995: 129), which reaches over half of all Hawaiian newborns, operateson a $7 million annual budget as a home visit program only. The evidencereviewed in Figure 2 suggests that the Hawaiian program is likely to beeffective at reducing child abuse, as would federal funding of home visitprograms nationally. Whether they would be effective at preventing delinquencyor serious crime in later life by the children visited cannot be determinedwithout longer-term studies. Child abuse and neglect is a risk factor fordelinquency, however, associated in one prospective study with a 50 percentincrease in prevalence and a 100% increase in frequency of adolescent arrests(Widom, 1989). Thus if the results of the home visitation experiments canbe generalized to other settings, they could clearly reduce a delinquencyrisk factor.

The effect sizes in these evaluations are particularly impressive. Bothof the long-term delinquency prevention effects are on the magnitude ofa relative reduction of three-quarters less prevalence of official criminalhistories. Similarly, the Rochester University study found a 79% relativereduction (4% compared to 19%) in child abuse. It is unlikely that an effectof this magnitude could be replicated nationally across all child abusecases because the same effect size is not observed in low-risk as in highrisk families. Such large effects are also unlikely to persist beyond thefirst two years of life. But applying the effect size to the estimated675,000 physical child abuse cases annually would reduce that number to142,000, or prevent 533,000 serious crimes (Reiss and Roth, 1993: 228).If the 1 million neglect cases are included as well, then an additional800,000 serious crimes might be prevented by home nurse visitation. Perhapsthe most immediate question in advancing the capacity to generalize fromcontrolled trials to national effects is the generalizability of the RochesterUniversity results from a rural white upstate New York sample. A long-termtrial of a similar approach among 1,100 African-American families in Memphis(National Research Council, 1993: 172) may soon report crucial resultson this point.

Foster Care and Family Preservation

Families in which child abuse is proven pose a major dilemma betweenfamily preservation and prevention of recidivism. The many documented deathsand injuries of children after prior reports of abuse underline the seriousnessof the dilemma. But the potential benefits of keeping thousands of familiestogether must be weighed against the cost. The current state of the evaluationscience of these two alternatives does not allow precise estimation ofthe costs and benefits. But a recent review of the evidence by a NationalAcademy of Sciences panel finds that the larger problems is not the choicebetween family preservation and foster care. The problem is that in somany cases neither course is taken.

The review found a national survey showing that more than one-thirdof confirmed cases of child maltreatment received no therapeutic or supportservices (McCurdy and Daro, 1993, as cited in National Research Council,1993: 268). This result occurs after 50 percent of the reported cases ofmaltreatment are found unsubstantiated, and the child protective servicesagency is required to decide whether children can remain safe in the homeduring treatment of the family. The officials making these decisions areoften understaffed, with poor training and high turnover. In 1991 in NewYork City, for example, 77 percent of the workers investigating child abusereports transferred to other agencies, resigned or were laid off (Dugger,1992, as cited in National Research Council, 1993: 268).

When children are placed in foster care due to abuse, it is not clearwhat their risks of further abuse become. Few studies of abuse rates ofthe estimated 200,000 children placed in foster care each year distinguishbetween abuse of the estimated 50% of children who were maltreated beforegoing into foster care and the other half who were not (Tatara, 1989, 1992,as cited in National Research Council, 1993: 271). Studies comparing ratesof abuse in foster care to other settings are methodologically weak. Onestudy almost two decades old did find that reported abuse by all fosterparents is lower than that by the general population, and much lower thanrates of re-abuse by abusive parents (Bolton, et al, 1981, as cited inNational Research Council, 1993: 230). But even if foster care createsa protective factor against further abuse, many cities report major shortagesin the availability of foster parents relative to the numbers of childrenjudged to need it (Kammerman and Kahn, 1989, as cited in National ResearchCouncil, 1993: 271).

When children are left in their family homes after documented maltreatment,they may or may not be at higher risk of further abuse and later delinquency.A review of four major federally funded studies of the effectiveness oftreatment across 3,253 families with abuse and neglect problems found thateven early and costly services are "not very successful" (Cohnand Daro, 1987, as cited in National Research Council, 1993: 255). Yetthe scientific literature in this area is characterized by many of thelimitations of general concern in this report (National Research Council,1993: 254):

    the research generally does not include controlled experiments, haslimited sample size, uses questionable measures to assess performance,and common assessment strategies have not been used across different interventions,making it difficult to know what works for whom.

The scientific methods used to evaluate family preservation programshave been stronger, but the results have been no more encouraging thanfor standard in-home treatment. Family preservation are often intense (20to 30 hours per week), brief (often 6 weeks) programs designed to preventfoster care placement through a variety of strategies. These include strengtheningfamily bonds, improving family skills, and providing stability in crisissituations. Rigorous experimental and quasi-experimental designs evaluatingthese programs show equivocal results, both on prevention of outplacementand longer-term outcome measures (National Research Council, 1993: 264-65).The studies have not yet disaggregated the problem by different kinds offamily problems, which could produce different results. The National ResearchCouncil Panel on Child Abuse and Neglect concluded that these programsare of unknown effectiveness. But the strategy remains popular becauseof its significant costs savings, an estimated $27,000 in tax dollars foreach outplacement prevented. No estimate of the risks of death and injuryassociated with that cost saving are available.

Domestic Violence Alarms and Visitation

Two home-based strategies for secondary prevention of domestic violencehave shown increasing use over the past decade. Personal radio alarms areindicated for extremely serious cases, while home visitation has been employedas a followup strategy after police response to a domestic disturbancecall.

The personal alarm is usually a small panic button worn as a necklace.Pressing the button directly activates a message at police headquartersto dispatch a police car on an urgent basis to the home of the wearer,who uses it to signal that a batterer is on the premises (Sherman, 1992:242; Farrell, 1995: 518-19). While the system is expensive to maintain,it can be allocated rationally based upon known risk factors. Police servingthe Liverpool, England area rotate the available alarms across the mostrecent and highest-risk victims of serious attacks, based on their findingthat repeat attacks were most likely to occur within thirty days afterthe last attack. This finding of highest risk of repeat victimization inthe first 24 hours and first 30 days after the last incident has been replicatedin a sample of 40,000 cases in an around Melbourne, Australia (Strang andSherman, 1996), and is an important basic research finding of indirectevidence in support of the use of personal alarms. Unfortunately, the manydocumented cases of domestic homicide of women who had been issued alarmsshows that the system is not foolproof. While it seems unlikely to increasethe risk of attack, there is no impact evaluation presently available toaddress the question of whether alarms are safe and effective.

The strategy of home visitation after a police contact for domesticviolence or disturbances also focuses on the high-risk time period in theimmediate aftermath of a police response to a domestic disturbance in thehome. The strategy has been evaluated in three tests using strong scientificmethods. An NIJ-funded Dade County (Florida) police experiment in the late1980s randomly assigned four responses to misdemeanor assault cases inwhich there was legally sufficient evidence to make an arrest: arrest,warning, arrest with followup visitation, and warning with followup visitation.The design was thus two separate controlled tests of followup visitationby police, one test following an arrest and one test following a warning(Pate, et al, 1991). The home visits consisted of a police detective reviewingthe family's history of domestic violence problems, their legal options,and social service agencies to which the detective could refer them forfurther assistance. The visit was a one-time treatment, with no attemptto provide a theoretically based psychological treatment. The very rigoroustest of the strategy found no effects of home visits on several diversemeasures of repeat domestic violence over a six-month followup period,including police offense reports, arrest reports, and victim interviews,analyzed by prevalence, frequency, and time to failure. The results werethe same for visits after an arrest and visits after a warning.

A second controlled experiment included both arrest cases (21%) andnon-arrest cases (79%) in the same sample randomly assigned to receivehome visitation (or not) by two person police-social worker teams (Davisand Taylor, forthcoming). The home visits were observed by researchersas lasting from ten to thirty minutes, depending on the victim's receptivenessand whether the batterer was present. The team tried to educate the victim,and the batterer if present, about the seriousness of domestic violenceand encourage the family to seek change through the courts or other services.Specific information was provided about how to go to court for restrainingorders, and to social services including battered women's shelters, substanceabuse treatment, relocation to another address, and home security. No differencein repeat violence between experimentals and controls were reported invictim interviews (response rate = 72%), but homes assigned to the experimentalgroup generated twice as many domestic calls to police. The authors interpretthis as evidence that visitation increases reporting but not violence;an alternative interpretation (untested in the analysis) is that visitationincreased repeat calls, with the homes with no victim interviews accountingfor a substantial portion of the total increase in the experimental group.

However the data are interpreted, there are now three strong tests ofthe police home visits strategy for preventing domestic violence. All threeof the tests falsify the hypothesis that this strategy is effective.

Figure 4-3

Effects on Domestic Violence of Proactive Home Visitation after ReactivePolice Contacts

Study Scientific Home Visitation Results Providers Methods Score Pate et al 1991 5 Police Detectives Visits after a warning had no effect on repeat violence (N= 447) over a 6 month followup period as reported by victim interviews or documented in official records Pate et al 1991 5 Police Detectives Visits after an arrest had no effect on repeat violence (N= 442) over a 6 month followup period as reported by victim interviews or documented in official records Davis and Taylor 5 Police-social Visits in domestic violence forthcoming worker teams public housing "hot spots" had no effect over a six (N= 436) month followup period on repeat violence reported by victims; calls to police about domestic incidents from experimental group almost twice as high as from control homes 

PREVENTION LINKS BETWEEN PARENTS AND PRESCHOOL OR SCHOOL

Outside the home, the preschool and the school provide major opportunitiesfor family-based crime prevention. Many of the prevention effects associatedwith early infancy home visits are impossible to separate from the simultaneousprovision of a strong linkage between parents and preschool. As childrenage, the school takes over more of the child's day (see Chapter Five),but many schools continue to seek parental involvement in reducing a child'sbehavioral risk factors for delinquency. Without duplicating the coverageof school-based prevention in the next chapter, this section explores theevidence on family-based prevention delivered through school settings.

Developmentally, the family-school linkage can begin as early as infantsare left in educationally enriched day care for even part of the day. Forchildren whose parent or parents are employed, the availability of suchcare can be a crucial factor allowing the parents to work. For childrenwho have at least one parent out of the labor force, the link to day careor preschool can be an important means of helping that parent find work.The daily structure of commuting to a child care center, and of spendingpart of each day or week there, can help establish patterns essential forparticipation in mainstream society. Effects of maternal participationin preschool in studies reviewed by Yoshikawa (1994) included increasedemployment, reduced welfare dependency, and increased time between givingbirth. To the extent that these effects were also linked to home visitation,however, the greatest certainty about generalizing from these results liesin framing them as a combined preschool-home visitation effect.

School setting programs for parent training and family-based preventionwith older children also combine several different treatments. The recentreview by Tremblay and Craig shows generally positive effects of theseprograms on delinquency or, more often, risk factors for delinquency withindicated or selective samples. Many of the evaluations suffer from smallsamples, short (or no) followup periods, and other methodological weaknesses.But the consistency of the results suggests that school-family outreachto train parents of problem children could be an effective means of preventingdelinquency in certain kinds of areas.

Children at Risk. Unfortunately, the results of the moderatelystrong evidence in Figure 4-4 were not confirmed by a very strong testof a very expensive program linking schools and families of very high-riskyouth to a wide range of services in very high risk neighborhoods. TheUrban Institute's four-year NIJ-funded evaluation of the Children at Riskprogram in Austin (TX),

Figure 4-4

Effects of Parent Training in School Settings

(Secondary Review Source: Tremblay and Craig, 1995; Scientific Methods Not Scored) 
Primary Source Type Sample Treatments Effects Tremblay et al 1994 Indicated 160 boys 2 Years of 6-Year Followup aged 7 parent showed lower years at training, self-reported outset social (ES= .25) and skills official (ES = training .07) delinquency, better school adjustment Hawkins et al 1992 Universal 1,659 boys 4 years of 5 month followup and girls training of showed lower aged 6 at parents, self-reported outset teachers, delinquency (ES= students .16), better parenting, attachment to family & school Pepler et al 1991 Indicated 40 boys & 12 weeks of 3 month followup girls aged parent and showed better 8 years student control over some training disruptive behaviors, not others Horn et al 1990 Indicated 42 boys & 12 weeks of 8 month followup girls aged parent showed better 7 to 11 training social control, years and child less self-control hyperactivity and therapy conduct problems Kolvin et al 1981 Selective 574 3 to 15 20 to 32 month Children months of followup showed age 7 years parent less anti-social counseling, behavior and group neurotic problems therapy 

Bridgeport (Conn), Memphis, Savannah and Seattle was a randomized trialwith 671 experimentals and controls, plus 203 youth in comparison neighborhoods(Harrell, 1996). Eligible subjects were referred to the program betweenages 11 and 13 while attending 6th or 7th grade at the middle school inthe study neighborhood in each city, where they were required to live.Referrals from school, police or courts were based on indicators of atleast three school risk factors (such as truancy), one family risk factor(such as parental violence), or one personal risk indicator (such as priorarrests or gang membership). Service protocols were locally determinedin each site, including some help from each of the following services:social work, family services, tutoring or educational services, recreationalafter-school and summer programs, mentoring, gifts and special events,community policing and juvenile courts. Half the sample was African-Americanand one third was Hispanic. Funding from private and DOJ sources for theprogram cost between $11 and 20 million.

The preliminary findings from the evaluation so far have shown thatthese intensive and expensive interventions combined had virtually no effect.The findings are based on self-reported behavior by the experimental andcontrol adolescents, with a 75% response rate after four years from theoriginal randomly assigned sample. No differences were detected in attritionpatterns by treatment group, which gives the analysis a scientific methodsscore = 5. The interviews show no difference within the high-risk areasbetween experimentals and controls on self-reported delinquency, drug usein the past month or entire lifetime, or sexual activity. A small differencein weapon carrying favored the treatment group. Further analyses stillto be reported include officially measured crime and delinquency from policeand court records, which will cover 100% of the experimental sample andnot just the survey respondents (Harrell, 1996). Thus the conclusions couldchange. Even with the best possible results from official data, however,further findings on the effectiveness of services costing about $35,000per child will be unable to provide clear evidence of effective crime prevention.

The CAR findings from self-reported delinquency do not provide muchguidance on how to prevent crime effectively in the places where preventionis needed the most. But the negative findings may not generalize to lower-riskfamilies, adolescents, schools or neighborhoods. Figure 4-3 suggests thatmulti-treatment school outreach to parents might be effective with othersamples. Similar results suggest the same about family therapy clinicsworking with families of children showing risk factors, either in the clinicalsetting or with the clinicians working with families in the home.

PREVENTION IN CLINICS

One key factor in the Children at Risk evaluation may have been thelow parental involvement with the adolescent (Harrell, personal communication,1996). Where parents can be successfully engaged in the question of howto raise their children more effectively, the results may be more encouraging.Figure 4-5 summarizes Tremblay and Craig's review of twelve evaluationsof family therapy. Only one of these has a delinquency measure, but thatone finds a prevention effect of moderate effect size. The other studies,while weaker, consistently report reductions in risk factors associatedwith family therapy by clinics.

Figure 4-5

Effects of Family Therapy Interventions By Clinical Staff

(Secondary Review Source: Tremblay and Craig, 1995; Scientific MethodsNot Scored)

Primary Source Type Sample Treatment Effects Therapy Delivered in Clinics Kazdin et al 1992 Indicated 97 boys & 6 to 8 1-Year Followup girls months of showed lower around 10 cognitive-be self-reported years old havioral delinquency (ES= at outset parent .25), anti-social training behavior and parental stress Dishion et al 1992 Selective 58 boys and 12 weeks of During treatment 61 girls parent child's aged 10 to training, Anti-social 14 at self-regulat conduct and outset ion parent's negative discipline declined; home conduct worse Yu et al 1986 Indicated 35 boys 20 weeks of During treatment aged 7 to parent and boys improved on 12 years student problem-solving, training in externalizing and problem-solv social competence ing Horn et al 1990 Indicated 42 boys & 12 weeks of 8 month followup girls aged parent showed better 7 to 11 training social control, years and child less self-control hyperactivity and therapy conduct problems Kolvin et al 1981 Selective 574 3 to 15 20 to 32 month Children months of followup showed age 7 years parent less anti-social counseling, behavior and group neurotic problems therapy Clinical Therapy Delivered at Home McNeil et al 1991 Indicated 30 children 14 weeks less aggression & X = 4.9 parent opposition by years old training children during treatment Packard et al 1983 Indicated 34 2 weeks of 11 week followup mother-child parent showed less pairs, training problem behavior child age X = 4.3 Shure & Spivak 1979 Indicated 10 boys, 10 3 months of Less impulsivity, girls age Social better X = 4.3 problem-solv problem-solving ing and during treatment parent training Webster-Stratton et Indicated 171 fathers 4 months of 3 year followup al 1988, 1990 & mothers parent showed better of children training parenting, less aged child X = 4.5 hyperactivity Strain et al 1982 Indicated 40 boys & 17 weeks 3 to 9 year girls aged child & followup showed 3 to 5 parent less oppositional training behavior and more compliance Dadds et al 1987 Indicated 24 families 6 weeks of 6 month followup with parent showed less children training, oppositional, aged problem-solv more compliance X = 4.2 ing behaviors by children Strayhorn and Selective 84 children 5 months 1 year followup Weidman 1991 aged X = parent shows better 3.7 years training parenting, less hyperactivity, no effect on hostility 

A recent analysis by Kumpfer (forthcoming) also shows beneficial effectsof parent training in "clinics" more broadly defined, includingrecreation rooms of public housing and other apartment complexes. Kumpfer'swork attends to the practical issues of incentives and transportation inobtaining high parental attendance rates at training sessions focused onprevention of substance abuse by both parents and children; when such issuesare properly addressed, she even finds high attendance rates in high-riskareas.

PREVENTION IN COURT

When prevention practices in all other settings fail, families oftenrely on the criminal justice system to stop the crime. This is especiallytrue for problems of family violence. Compared to what is known about humandevelopment and developmental crime prevention, however, the science ofdomestic violence has little knowledge to offer for effective policymaking.But the opportunities for advancing evaluations of legal efforts at violenceprevention are great, once the limitations of the current state of knowledgeare fully understood.

The basic science of domestic violence and the law offers several well-knownfacts (Crowell and Burgess, 1996): domestic violence is widespread andhighly under-reported to authorities. When police are called, they findno evidence of actual physical violence in over half of all "domestic"calls, and make no arrests in the majority of cases where such evidenceis available. The vast majority of arrests that are made are for misdemeanorassaults with limited evidence of injury, for which prosecutors drop chargesin the majority of the cases (Sherman, 1992). While the suspect is gonefrom the scene when police arrive in 40% of the cases in which police dohave sufficient evidence to arrest, few courts or police agencies botherto issue arrest warrants unless the victim requests one by making a burdensometrip to court. Rising arrest rates for simple assault in the early 1990shas placed even more workload pressure on courts and prosecutors, for whichthere is some evidence that the odds of prosecution per arrest will decline.Odds of conviction per arrest for misdemeanor domestic assault are as lowas 1 percent, with odds of incarceration per arrest as low as zero per400 cases (Sherman, 1992: 337).

The prevention program often recommended in response to these factsof under-enforcement of the law is full, or fuller, enforcement. The premiseof this policy is two-fold, both moral and empirical. The moral premiseis that full enforcement is the proper response to all crimes, from drugpossession to homicide, even though there is ample evidence that under-enforcementof the law by 50 percent or more cuts across both felonies and misdemeanorsof almost all kinds (Reiss, 1971; Black, 1980; Smith and Visher, 1981).From this perspective, the crime prevention effects of fuller enforcementare not dispositive.

Fuller enforcement is also claimed, however, to have preventive effects.The empirical premise is that increasing certainty and severity of punishmentwill create either general or specific deterrence of domestic violence."General" deterrence refers to prevention of crimes by peoplein the community generally regardless of whether they have been caughtand punished for a crime. "Specific" deterrence denotes the preventiveeffects of punishment on people who have been caught. Both hypotheses arewidely accepted as true by legislators, but hotly debated by evaluationscientists (Zimring and Hawkins, 1973; Blumstein, et al, 1978).

Rigorous scientific impact evaluation evidence is unavailable aboutmost of the criminal law strategies for preventing domestic violence (Crowelland Burgess, 1996; Fagan, 1996). Police have been the component of thelegal system most willing to engage in rigorous impact evaluations. Otheragencies of the criminal justice system have repeatedly refused to allowcareful testing of their effectiveness; prosecutors in Milwaukee and judgesin Minneapolis are just two examples over the past decade. As a result,a great deal is known about the effects of one police decision, while littleis known about most other criminal justice practices.

The National Institute of Justice has pioneered in supporting rigoroustests of domestic violence responses. This include the six offender-presentand one offender-absent experiment in arrests for misdemeanor domesticassault (Scientific Methods Score = 5), reviewed in Chapter Eight. Thesestudies find no consistent support for the specific deterrent hypothesis,in the general absence of any referrals, prosecutions or convictions afteran arrest; they do find arrest is effective for employed offenders (Sherman,1992) and absent offenders for whom police issue a warrant (Dunford, 1991).A frequent conclusion from these findings is that arrest must have followupactions in order to be effective. That hypothesis, however, remains untested.So does the general deterrence hypothesis that mandatory arrest in a citywill prevent domestic violence city-wide. The hypothesis that allowingvictims to decide whether or not an arrested batterer should be prosecutedwill prevent violence, however, has also been tested by an NIJ-funded controlledexperiment (Scientific Methods = 5). The Indianapolis Domestic ViolenceProsecution Experiment (Ford, 1993) randomly assigned cases in the prosecutor'soffice to a policy of either "no-drop" or victim decision. Thevictim decision policy produced a lower repeat violence rate, also falsifyingthe hypothesis that full enforcement offers greater prevention.

The hypothesis that mandatory referral of arrested batterers to counselingor therapy will help prevent repeat violence has also been tested withNIJ support, although with weaker scientific methods than the evaluationsdescribed above. This test provides moderately strong evidence of a negativeeffect. Harrell (1991) found in a matched comparison of arrested batterersreferred to court ordered treatment and those not referred to treatmentthat the treated group had higher repeat violence rates. Crowell and Burgess(1996: 122), however, cite several weaker studies that find the oppositeconclusion. The strongest design appears to be Goldkamp's (1996) evaluationof the Dade County Domestic Violence Court program combining substanceabuse treatment with domestic violence counseling, a randomized experimentnot yet reported with significance tests or other statistics (SMS = 3);preliminary results suggest a reduction in same-victim domestic violenceby offenders in the combined treatment, compared to offenders given onlyone or the other treatment approaches. The effects of court-ordered treatmentseem likely to vary widely by the specific approach to treatment, the skillsof the individual therapists, the background of the batterers, and otherfactors making it difficult to generalize from a few weak evaluation designs(Fagan and Browne, 1993).

Most domestic violence evaluations have been focused on noninjuriousviolence, and very little is known about the prediction or prevention ofserious injury. One of the major practices to be evaluated is the effectivenessof court orders of protection. According to an NIJ-funded study by theNational Center for State Courts (1996) in Wilmington, Denver and the Districtof Columbia, women who seek orders of protection suffer very high ratesof serious injury prior to obtaining the order. According to a matchedcontrol evaluation of women granted orders in Denver and Boulder, the one-yearrecidivism rates are lower against women who obtain the orders (Harrell,et al, 1993), thus supporting the full enforcement deterrence hypothesis.In the absence of any other reported impact evaluations of restrainingorders, this level three study makes the use of such orders at least "promising."

PREVENTION IN OTHER SETTINGS

The effects of practices in other settings on families and their crimerisks may be quite substantial. Churches, employers, landlords and neighborsmay all play roles that are not yet well understood. This section addressesonly a few of the other settings affecting families: battered women's shelters,hospitals, and gun shops.

Battered Women's Shelters. The number of battered women's sheltersin the US was recently estimated at 1,200 (Plichta, 1995, cited in Crowelland Burgess, 1996: 101). These shelters, and 600 other related programs,offer a wide array of services to families and women suffering intimateviolence. The core of a shelter's service, however, is providing a safehaven during the high risk period in the immediate aftermath of a domesticviolence incident (Farrell, 1995; Strang and Sherman, 1996). There is evidencethat current levels of this service are insufficient to meet the demand;an estimated 300 women and children per week were turned away from NewYork City shelters in March of 1995 due to lack of space (O'Sullivan etal, 1995, as cited in Crowell and Burgess, 1996: 102).

Whether shelters actually reduce violence against women is an importantquestion for evaluation. The logical basis for predicting that result isthe reduction of risk after the passage of time with the offender unableto gain access to the victim. Berk et al (1986), however, found quasiexperimental(Scientific Methods Score = 4) evidence that unless the shelter clientstook other steps to seek help beyond staying in the shelter, their ratesof repeat violence after leaving the shelter were actually higher thana similar group who had not gone to a shelter. Among women who did takeadditional steps, however, the shelter stay had a measured protection effectagainst repeat violence lasting about 6 weeks. The relatively small samplesize (N = 155) and Santa Barbara (CA) site for this analysis (N = 155)may limit the generalizability of the findings, but the results suggestthe clear need for impact evaluations of all crime prevention programs.

Hospitals. Little is known about the identification and reportingof family violence in hospitals. A recent NIJ grant to the Illinois CriminalJustice Information Authority will examine the possible data collectionopportunities in hospital emergency rooms, which could lead to operationalindicators as well as research findings. A clear interpretation of thenumber domestic violence cases reported to police is impossible as longas increased reports might reflect growing confidence in the police, ratherthan more violence in the home (Davis and Taylor, forthcoming; Shermanand Strang, 1996). Hospital measures over time may provide a communitywith its most reliable indicator of progress or decline in the effectivenessof its efforts to deal with the problem.

Gun Shops also play a crucial role in family violence, and mostof some 2,000 domestic homicides a year. The 1996 Lautenberg Act imposeda federal ban on gun ownership among persons convicted of domestic violencemisdemeanors. We may estimate the likely effect of implementing this lawby noting that an estimated 100,000 to 150,000 persons are convicted annuallyof domestic violence misdemeanors.2 Moreover, the risk of adomestic homicide is approximately eight times higher among people whohave had police encounters for misdemeanor offenses than among people whohave not in Milwaukee, and 18 times higher in Victoria (Melbourne), Australia(Strang and Sherman, 1996). While this risk is nonetheless a very low 1in 33,000 person-years, it still amounts to 5 murders per year among peoplenewly convicted of domestic violence. If the prior convictions were includedfor 20 years, that could amount to 100 murders per year committed by personspreviously convicted of domestic violence misdemeanors. How many of thosemurders would be prevented by the Lautenberg law is impossible to predict.But the indirect evidence on risk factors suggests that the law does addressa major risk factor for serious domestic injury and death.

CONCLUSIONS: WHAT WORKS, WHAT DOESN'T, WHAT'S PROMISING

This section discusses the following conclusions, and their researchand policy implications:

What Works

o Long-term frequent home visitation combined with preschool preventslater delinquency

o Infant weekly home visitation reduces child abuse and injuries

o Family Therapy by clinical staff for delinquent and pre-delinquentyouth

What Doesn't

o Home visits by police after domestic violence incidents fails to reducerepeat violence

What's Promising

o Battered women's shelters for women who take other steps to changetheir lives

o Orders of Protection for battered women

The Effectiveness of DOJ-Funded Local Prevention Programs

Over the last three decades, the Congress has left family-based crimeprevention largely in the hands of other federal agencies besides the Departmentof Justice (DOJ). This began to change with the rising concern over domesticviolence in the 1980s.3 The passage of the Violence AgainstWomen Act (VAWA) as Title IV of the 1994 Crime Act was a major increasein the role of DOJ in the family (although the VAWA also addresses crimescommitted by strangers). Most recently, the Office of Justice Programshas identified infant home visitation as an important strategy to includein comprehensive community prevention programs such as Weed and Seed andvarious OJJDP initiatives. The evidence suggests that DOJ's increasingresponsibility for national crime rates logically draws it to the majorrisk factors for crime, which must clearly include the family.

In what may be a period of transition towards more explicit focus onfamily-based prevention, Congress has created a number of funding programsthat offer opportunities to develop that role. These may be divided intodevelopmental and family violence prevention. The developmental programsare funded primarily by OJJDP and the Executive Office of Weed and Seedwith discretionary funds, while the family violence funding is concentratedin the Violence Against Women Grants Office.

Safe Kids, Safe Streets (OJJDP, with VAWGO and EOWS). This fundingprogram will provide about $1.4 million per year for five years to eachof six communities. Informed by much of the research reviewed in this chapter,the program is specifically aimed at prevention of child abuse and neglectand related risk factors for delinquency. The strategies supported by theprogram include family strengthening, mental health services and treatment.A national process evaluation is underway to determine exactly what strategieseach site selects, and a national impact evaluation is planned for futureyears.4 To the extent that the local grantees elect to employapproaches to family based prevention reviewed in this chapter, there isevidence that the funding can be effective in preventing crime. To theextent that the local grantees focus on the highest risk pre-adolescentsin the highest-risk neighborhoods, however, the preliminary results fromthe Children at Risk Program may indicate that the state of the preventionart is not yet up to such a sever challenge (Harrell, 1996).

Title V Incentive Grants for Local Delinquency Prevention (OJJDP).This program distributed $20 million in FY 1995 for local programs encouragedto adopt the Communities That Care model (Hawkins, et al, 1992). The programwas initially developed and field tested by OJJDP in the early 1980s, andhas established a substantial record of evaluation results. The CTC modelrecommends consideration of parent training as well as family therapy forhigh-risk adolescents and early childhood home-based and center-based strategies.This review finds all those approaches can be effective.5

Operation Weed and Seed (EOWS). This program is currently planningto conduct a field test of the Rochester University model of early infancyhome nurse visitation. The location of such a test within Weed and Seedneighborhoods would provide an excellent replication of the original Elmirastudy. Results from the Memphis replication currently underway could alsoinform the Weed and Seed approach to this model, which has such strongevidence of reducing child abuse among high-risk rural white families.

Congressional Action on Universal Home Visitation. The evidencereviewed in this chapter suggests that substantial crime prevention effectscould be obtained from greater federal investment in early infancy andpre-school home visitation. For reasons discussed in this chapter, a universalapproach to such a program is more likely to succeed than a selective approachbased upon risk factors. The latter approach is more cost-efficient butpotentially stigmatizing. While further research is needed to compare thecrime prevention benefits of early prevention to costly federal programssuch as prison construction, such research could inform the Congress ofwhere it can find the maximum crime prevention for each taxpayer dollar.While appropriations for Head Start have never been able to meet the demandfor the program, that may reflect its use on a selective basis. A universalhome visitation program that promises to reduce crime may be more feasiblethan fully funding Head Start. Controlled testing of visitation with andwithout Head Start, however, is required in order to determine whethervisitation alone can create lasting benefits without reinforcement forboth parent and child through the pre-school environment.

Universal home visitation for children may also have the benefit ofhelping to prevent or at least detect domestic violence. Visitation hasbeen found ineffective in the immediate aftermath of a police response.But it may well be effective at reducing unreported cases, especially infamilies where police are never called. While this would not be a centralgoal of universal early infancy visitation, it could be a side benefit.That hypothesis also provides a linkage between DOJ efforts to preventcrime developmentally and among members of the family.

STOP (Services, Training, Officers, Prosecutors) Formula Grants(VAWGO). By far the largest OJP expenditure on issues affectingcrime in families is the STOP Grant funding program, which distributed$23 million in fiscal 1995 and has been appropriated $145 million for FY1997.6 This money, which addresses all violence against womenand not just family violence, is appropriated on the basis of population.How the money is used is up to the states, within the broad initial guidelinesof 25% allocations to each of three areas (Burt, 1996: vi): law enforcement,prosecution, and victim services. Much of it appears to go for training,model policies, equipment, and other support materials.

To the extent that this funding can be effective in reducing familyviolence, it could be more so if the funds were allocated on the basisof some crime risk indicator. Possible criteria include the number of womenmurdered by men in each state, or total women murdered (which would haveless reporting bias than other crimes against women like rape). Like policepatrol funding (see Chapter Eight), the population based formula may putthe money equally in places that need it desperately and places that donot.

As the major source of federal funds that could be used to combat familyviolence, STOP might provide a vehicle for increasing prosecution and adjudicationof domestic violence arrests. The full enforcement hypothesis remains anunanswered question, even though there is clear evidence that it is notsupported with certain kinds of offenders. In order to test the effectsof higher levels of prosecution and sentencing, the funding required forthe extra courtroom work must be provided. A review of the FY 1995 grantawards made by the States, however, suggests that the funds are not beingused to support increased volume of court case processing--unlike the competitiveGrants to Encourage Arrest Policies. Most of the purposes are for supportservices such as training. The effects of training of police and prosecutorson crime prevention have not yet been evaluated. But the effects of increasedprosecution are also unknown. The general absence of scientific tests ofmost local practices in domestic violence prevention provides very littleguidance to Congress, DOJ and the states on how this funding can be spentmost effectively to prevent domestic violence.

Grants to Encourage Arrest Policies (VAWGO). A review of grantaward abstracts for the FY 1996 grants suggests that these grants are supportingdiverse local programs. The most direct operational activity is increasedcapacity for prosecution, with DOJ funds used to hire prosecutors and bringcharges in cases that would otherwise be dropped. Some jurisdictions evencommit to 100% prosecution. Thus the program may provide a realistic possibilityin many communities to link arrest to a high certainty of prosecution,a response that has never been evaluated but which could be very differentfrom arrest alone. Until evaluations of that kind are conducted, the effectivenessof increased prosecution a crime prevention practice will remain unknown.

These grants also support training, data bases and other approachesdesigned to increase arrests made by police officers. Here again, the currentstate of evaluation science has little guidance to offer one way or anotherabout any expenditures to encourage domestic violence arrests. The potentialvalue for impact studies across a range of options for such programs wouldbe to identify those which appear most cost-effective.

Rural Domestic Violence and Child Abuse Enforcement Assistance grants(VAWGO). The absence of scientific evidence on the effectiveness oflocal practices in rural areas, and with child abuse cases, also limitsthe assessment that can be made of this funding based on scientific evidence.The 1993 National Academy of Sciences Panel found the problem of childabuse to lack rudimentary science on many of these questions. To the extentthat the Olds et al (1986) experiment prevented rural child abuse successfully,the Congress may wish to open the scope of acceptable funding for thisprogram to include prevention as well as enforcement. Alternatively, theuse of nurses legally obligated to report abuse might qualify as childabuse enforcement. If local programs funded by DOJ use their money in thatfashion, it seems reasonably likely to be effective in rural white low-incomecommunities or families.

National Stalker and Domestic Violence Reduction. This$6 million three-year program establishes a data base as part of the NationalCrime Information Center that will cover various offenses and offendersin domestic and family violence and stalking. In addition to the data basefunding from the STOP Block grants, these funds will help create the capacityfor implementing the 1996 Lautenberg Act extending the Brady Bill to misdemeanordomestic violence. While the latter Act prohibits persons convicted ofsuch misdemeanors from owning a gun, there is currently no data base availablein most states to identify such persons. This gap results from the absenceof special statutes for "domestic" offenses, which are generallyprosecuted under generic laws against assault. Whether a misdemeanor assaultconviction reflects domestic violence is not a part of the court record,and can only be determined retrospectively by examining police records.The latter are often kept in paper files rather than computers, makingthe task very difficult in retrospect. But if new data bases can capturethe data prospectively, it may be possible to implement the law with thesefunds by the 21st Century. It seems unlikely to happen without these DOJfunds.

No empirical test of the effect of a handgun ban for domestic violencemisdemeanant has ever been conducted. Ongoing NIJ evaluations of the Bradybill may provide some idea. Other uses of the data bases created by VAWAfunding could have even greater preventive effect, such as public accessto a registry of convicted batterers. Such a registry could have a fargreater deterrent effect than arrest alone, and could also help warn potentialvictims to avoid relationships with previously convicted batterers. Whetherany of these hypothesized effects would occur, however, can only be determinedby a program of rigorous research and development.

Office of Victims of Crime. This office, funded by fines collectedby federal courts, provides grants in support of some of the local practicesreviewed in this chapter. Support for battered women's shelters is a notableexample. The potential value of these programs in preventing crime suggeststhat this Office might be included in the overall scope of DOJ crime preventionactivity.

Improving Funding Effectiveness Through Better Evaluations

As the Congress recognized in its passage of the Violence Against WomenAct of 1994, the research agenda for family-based crime prevention is substantial.A great many key questions about local practices remain unanswered, whiletens of millions of cases are processed annually. This final section considersthree high priority areas: home visitation, police arrest policies, andorders of protection.

Early Infancy Home Visitation. This Chapter's primary recommendationis the same as the 1993 Report of the National Research Council (1993)on Child Abuse and Neglect:

    "Research on home visiting programs focused on the prenatal, postnatal,and toddler periods has great potential for enhancing family functioningand parental skills and reducing the prevalence of child maltreatment."(National Research Council, 1993: 191-92). "The panel recommends thatevaluations of home visiting programs include descriptions of what goeson in visits..and direct observations of home visitors in action."(NRC 1993: 193).

The theoretically powerful early infancy visitation model raises a hostof unanswered questions about its effectiveness. Before formulating orproposing a national policy, DOJ needs to procure randomized experimentstesting the basic model under different conditions: high and low crimeneighborhoods, different training for the visitors, different frequencyand length of visitation, and different combinations of other interventionssuch as preschool with parental involvement. The funding of visitationprograms as part of existing DOJ programs creates an opportunity to implementthis proposal. The absence of a randomized controlled trial, however, wouldgravely limit what can be learned from an impact evaluation. The feasibilityof a rigorous experiment has been demonstrated in Elmira and Memphis, andDOJ can build upon that precedent.

Police Arrest Policies. Given the growing use of arrest for domesticviolence and the continuing debate over the interpretation of the previousNIJ experiments, it would be very helpful to continue the program of researchthat produced them. Collaborative experiments with prosecutors and courtswould seem to be the highest priority, to test the hypothesis that fullenforcement by the criminal justice system is an effective prevention approach.Alternative sanctions, such as reintegrative shaming conferences (Braithwaiteand Daly, 1993), also need to be tested against more customary measureslike probation and fines. Even stigmatic shaming such as court-ordereddisplay of bumper stickers or t-shirts proclaiming the offender to be abatterer (Kahan, 1997) could be tested against its theoretical competitionin reintegrative shaming (Braithwaite, 1989). More sophisticated researchdesigns can also now be employed to control for contextual effects of neighborhoodlabor force participation rates, rather than the less policy-relevant individualemployment status.

Orders of Protection. Given the high risks of serious injurysuffered by many domestic violence victims who receive orders of protection,the need for further research is great. The most theoretically promisingstrategy for further testing would be a randomized trial of the personalpanic alarm in a big city jurisdiction. A large city would minimize theethical problems with the creation of a control group, since there wouldbe far too many victims for most jurisdictions to give them all a panicalarm. Randomized tests of women who volunteer for an evaluation of a randomizedtrial based upon informed consent may also lead to a strong test of ordersof protection without any additional tools, which is by far the most commoncondition under which they are issued.

NOTES

1This discussion is limited to government, rather than abroader range of institutions, by the content of the available research.All of the available program evaluations examine the effects of governmentprograms, broadly defined to include schools and publicly supported healthcare. Other institutions, such as churches and charities, no doubt providecrime prevention services (also broadly defined) to families. But in thecurrent social structure of American life, it seems unrealistic to expectprivate resources to fund the level of intervention that research suggestsis needed to appreciably reduce serious crime rates. While churches andother private groups may be ideal for administering such efforts undergovernment contracts, the level of resources associated with the evaluatedprograms far exceeds those likely to be raised from solely non-public sources.

2This calculation employs the FBI count of 1.86 million arrestsfor all assaults in 1995, less 75% for non-domestic assaults, adjustedby the arrest probability of 22% for domestic and 13% for non-domesticassaults observed in the Indiana University police observation study (Oppenlander,1982), and multiplied by a conviction probability estimate of 20% givena domestic arrest (Sherman, 1992).

3Attorney General's Task Force on Family Violence, Report.1984.

4OJJDP Fact Sheet #38, June 1996.

5OJJDP, 1995 Report to the Congress: Title V Incentive Grantsfor Local Delinquency Prevention Programs.

6This program is not just focused on families, but aims toprevent all forms of violence against women including stranger violence.Since most violence against women is caused by relatives and intimates,however, much of these funds are appropriately focused on family violence.

REFERENCES

Achenbach, T.M, V. Phares, C. Howell, V.A. Raugh, and B. Nurcombe

1990 Seven Year Outcome of the Vermont Intervention Program for Low-BirthweightInfants. Child Development 61: 1672-1681.

Barth, R.P., S. Hacking and J.R. Ash

1988 Preventing Child Abuse: An Experimental Evaluation of the Child-ParentEnrichment Project. Journal of Primary Prevention 8: 201-217.

Berk, Richard, Phyllis J. Newton, and Sarah F. Berk

1986 What A Difference A Day Makes: An Empirical Study of the Impactof Shelters For Battered Women. Journal of Marriage and the Family 48:431-490.

Berk, Richard A., Alec Campbell, Ruth Klap, and Bruce Western

1992 The Deterrent Effect of Arrest in Incidents of Domestic Violence:A Bayesian Analysis of Four Field Experiments. American Sociological Review57: 698-708.

Barrera, M.E., P.L. Rosenbaum, and C.E. Cunningham

1986 Early Home Intervention With Low Birth-Weight Infants and TheirParents. Child Development 57: 20-33.

Berreuta-Clement, J.R., L.J. Schweinhart, W.S. Barnett, A.S. Epstein,and D.P. Weikart

1985 Changed Lives: The Effects of the Perry Preschool Program on YouthsThrough Age 19. Ypsilanti, MI: High Scope Press.

Black, Donald

1980 The Manners and Customs of the Police. NY: Academic.

Blumstein, Alfred, Cohen, Jacqueline, and Daniel Nagin (eds).

1978 Deterrence and Incapacitation: Estimating The Effects of CriminalSanctions on Crime Rates. Washington, DC: National Academy of Sciences.

Bolton, F.G., R.H. Lanier, and D.S. Gia

1981 For Better or Worse? Foster Parents and Foster Children in an OfficiallyReported Child Maltreatment Population. Children and Youth Services Review3: 37-53.

Braithwaite, John

1989 Crime, Shame and Reintegration. Cambridge: Cambridge UniversityPress.

Braithwaite, John and Kathleen Daly

1993 Masculinities, Violence and Communitarian Control. Paper Presentedto the National Conference on Violence. Canberra: Australian Instituteof Criminology.

Burt, Martha

1996 The Violence Against Women Act of 1994: Evaluation of the STOPBlock Grants to Combat Violence Against Women. Washington, DC: Urban Institute.

Coie, J.D. and M.R. Jacobs

1993 The Role of Social Context in the prevention of Conduct Disorder.Development and Psychopathology 5: 263-275.

Crowell, Nancy A. and Ann W. Burgess

1996 Understanding Violence Against Women. Washington, DC: NationalAcademy of Sciences.

Dadds, M.R., S. Schwartz, and M. Sanders

1987 Marital Discord and Treatment Outcome in Behavioral Treatment ofChild Conduct Disorders. Journal of Consulting and Clinical Psychology55: 396-403.

Davis, Robert C. and Bruce G. Taylor

Forthcoming A Proactive Response to Family Violence: The Results ofa Randomized Experiment. Criminology 1997.

Dishion, T.J., G.R. Patterson, and K.A. Kavanagh

1992 An Experimental Test of the Coercion Model: Linking Theory, Measurementand Intervention. In J. McCord and R.E. Tremble, Eds., Preventing Anti-SocialBehavior: Interventions from Birth Through Adolescence. NY: Guilford.

Dugger, C.W.

1992 Shortage of Trained Caseworkers Imperils Young Victims of Abuse.New York Times. December 28.

Dunford, Franklyn W.

1990 System-Initiated Warrants for Suspects of Misdemeanor DomesticAssault: A Pilot Study. Justice Quarterly 7: 631-653.

1992 The Measurement of Recidivism in Cases of Spouse Assault. Journalof Criminal Law and Criminology 83: 120-136.

Dunford, Franklyn W., David Huizinga, and Delbert S. Elliott

1990 The Role of Arrest in Domestic Assault: The Omaha Police Experiment.Criminology 28: 183-206.

Fagan, Jeffrey

1996 The Criminalization of Domestic Violence: Promises and Limits.Research Report. Washington, DC: National Institute of Justice.

Fagan, Jeffrey and Angela Browne

1993 Violence Between Spouses and Intimates: Physical Aggression BetweenWomen and Men in Intimate Relationships.

Farrell, Graham

1995 Preventing Repeat Victimization. In Michael Tonry and David P.Farrington, Eds., Building A Safer Society. Crime and Justice, Vol. 19.Chicago: University of Chicago Press.

Farrington, David P., Lloyd Ohlin and James Q. Wilson

1986 Understanding and Controlling Crime: Toward a New Research Strategy.New York: Springer-Verlag.

Ford, David

1993 The Indianapolis Domestic Violence Prosecution Experiment. Departmentof Sociology: Indiana University, Indianapolis.

Goldkamp, John S.

1996 The Role of Drug and Alcohol Abuse in Domestic Violence and ItsTreatment: Dade County's Domestic Violence Court Experiment. ExecutiveSummary. Philadelphia, PA: Crime and Justice Research Institute.

Gray, J.D., C.A. Cutler, J.G. Dean, and C.H. Kempe

1979 Prediction and Prevention of Child Abuse and Neglect. Journal ofSocial Issues 35: 127-139.

Gutelius, M.F., A.D. Kirsch, S. MacDonald, M.R. Brooks, and T. McErlean

1977 Controlled Study of Child Health Supervision: Behavioral Results.Pediatrics 60: 294-305.

Harrell, Adele

1991 Evaluation of Court-Ordered Treatment of For Domestic ViolenceOffenders. Washington, DC: Urban Institute.

1996 Findings of the Evaluation of the Children at Risk Program. PaperPresented at the University of Maryland, College Park.

Harrell, Adele, B. Smith and L. Newmark

1993 Court Processing and the Effects of Restraining Orders for DomesticViolence Victims. Washington, DC: The Urban Institute.

Hawkins, J. David, Michael W. Arthur and Richard F. Catalano

1995 Preventing Substance Abuse. In Michael Tonry and David P. Farrington,Eds., Building A Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Hirschel, David, Ira W. Hutchison III, Charles W. Dean, Joseph J. Kelley,and Carolyn E. Pesackis

1990 Charlotte Spouse Assault Replication Project: Final Report. Charlotte,NC: University of North Carolina at Charlotte.

Horn, W.F., N. Ialongo, G. Greenberg, T. Packard, and C. Smith-Winberry

1990 Additive Effects of Behavioral Parent Training and Self-ControlTherapy With Attention Deficit Hyperactivity Disordered Children. Journalof Clinical Child Psychology 19: 98-110.

Institute of Medicine

1994 Reducing Risks of Mental Disorders: Frontiers for Preventive InterventionResearch. Washington, DC: National Academy of Sciences.

Jacobson, S.W. and K.F. Frye

1991 Effect of Maternal Social Support on Attachment: Experimental Evidence.Child Development 62: 572-582.

Johnson, D.L., and T. Walker

1987 Primary Prevention of Behavioral Problems in Mexican-American Children.American Journal of Community Psychology 15: 375-385.

Kahan, Donald

1997 Its A Shame There is None. Wall Street Journal, January.

Cohn, A.H. and D. Daro

1987 Is Treatment Too Late: What Ten Years of Evaluative Research TellsUs. Child Abuse and Neglect 11: 433-442.

Kammerman, S.B., and A.J. Kahn

1989 Social Services for Children, Youth and Families in the U.S. NY:The Annie E. Casey Foundation.

Kazdin, A.E., T.C. Siegel, and D. Bass

1992 Cognitive Problem-Solving Skills Training and Parent ManagementTraining in the Treatment of Anti-Social Behavior in Children. Journalof Consulting and Clinical Psychology 60: 733-747.

Kolvin, I., R.F. Garside, A.R. Nicole, A. MacMillen, F. Wolstenhome,and I.M. Leitch

1981 Help Starts Here. NY: Tavistock.

Kumpfer, Karol

Forthcoming Report of Family-Based PEPS, Center for Substance AbusePrevention

Kumpfer, Karol, Virginia Molgaard and Richard Spoth

1996 The Strengthening Families Program for the Prevention of Delinquencyand Drug Use. In Ray DeV Peters and Robert J. McMahon, Eds., PreventingChildhood Disorders, Substance Abuse and Delinquency. Thousand Oaks: Sage.

Lally, J.R., P.L. Mangione and A.S. Honig

1988 The Syracuse University Family Development Research Project: Long-RangeImpact of an Early Intervention with Low-Income Children and Their Families.In D.R. Powell, Ed., Annual Advances in Applied Developmental Psychology,Vol. 3: Parent Education as Early Childhood Intervention: Emerging Directionsin Theory, Research, and Practice. Norwood, NJ: Ablex.

Lazar, I.

1992 Discussion. In F.L. Parker, R. Robinson, S. Sambrano, C. Piotrkowski,J. Hagen, S. Randolph, and A. Baker, Eds., Conference Proceedings: NewDirections in Child and Family Research: Shaping Head Start in the 90s.NY: National Council of Jewish Women Center for the Child.

Lieberman, A.F., D.R. Weston, J.H. Pawl

1991 Preventive Intervention and Outcome With Anxiously Attached Dyads.Child Development 62: 199-209.

Lyons-Ruth, K., D.B. Connell, H.U. Grunebaum and S. Botein

1990 Infants at Social Risk: Maternal Depression and Family SupportServices as Mediators of Infant Development and Security of Attachment.Child Development 61: 85-98.

McCurdy, K. and D. and Daro

1993 Current Trends in Child Abuse Reporting and Fatalities: The Resultsof the 1992 Annual Fifty State Survey. Working Paper 808. April. Chicago,IL: National Committee for Prevention of Child Abuse.

McKey, R.H., L. Condelli, H. Ganson, B.J. Barrett, C. McConkey, andM.C. Plantz.

1985 The Impact of Head Start on Children, Families and Communities:Final Report of the Head Start Evaluation, Synthesis and Utilization Project[DHHS Publication No. 85- 31193]. Washington, DC: USGPO.

McNeil, C.B., S. Eyberg, T.H. Eisenstadt, K. Newcomb, and B. Funderburk

1991 Parent-Child Interaction Therapy With Behavioral Problem Children:Generalization of Treatment Effects to the Social Setting. Journal of ClinicalChild Psychology 20: 140-51.

National Center for State Courts

1996 The Effectiveness of Civil Protection Orders: Executive Summary.Williamsburg, VA: National Center for State Courts.

National Research Council

1993 Understanding Child Abuse and Neglect. Washington, DC: NationalAcademy of Sciences.

O'Sullivan, C., J. Wise, and V. Douglass

1995 Domestic Violence Shelter Residents in New York City: Profile,Needs and Alternatives to Shelter. NY: Victim Services Agency.

Olds, D.L., C.R. Henderson, R. Chamberlin, and R. Tatelbaum

1986 Preventing Child Abuse and Neglect: A Randomized Trial of NurseHome Visitation. Pediatrics 78: 65-78.

1988 Improving the Life-Course Development of Socially DisadvantagedMothers: A Randomized Trial of Nurse Home Visitation. American Journalof Public Health 78: 1436-1445.

Packard, T., E.A. Robinson, and D. Grove

1983 The Effect of Training Procedures on the Maintenance of ParentalRelationship Building Skills. Journal of Clinical Child Psychology 12:181-186.

Pate, Antony, Edwin E. Hamilton and Sampson Annan

1991 Metro-Dade Spouse Abuse Replication Project: Draft Final Report.Washington, D.C.: Police Foundation.

Pepler, D.J., G. King and W. Byrd

1991 A Social Cognitively-Based Social Skills Training Program for AggressiveChildren. In D.J. Pepler and K.H. Rubin, Eds., The Development and Treatmentof Childhood Aggression. Hillsdale, N.J.: Erlbaum.

Plichta, S.B.

1995 Domestic Violence: Building Paths for Women to Travel to Freedomand Safety. College of Health Sciences, Old Dominion University.

Reiss, Albert J. Jr.

1971 The Police and the Public. New Haven: Yale University Press.

Reiss, Albert J. Jr., and Jeffrey A. Roth, eds.

1993 Understanding and Preventing Violence. Washington, DC: NationalAcademy of Sciences.

Ross, G.H.

1984 Home Intervention for Premature Infants of Low-Income Families.American Journal of Orthopsychiatry 54: 263-70.

Schure, M.B. and G. Spivak

1979 Interpersonal problem-Solving, Thinking and Adjustment in the Mother-ChildDyad. In M.W. Kent and J.E. Rolf, Eds., Primary Prevention of Psychopathology.Hanover, NH: University Press of New England.

Seitz, V., L.K. Rosenbaum, and N.H. Apfel

1985 Effects of Family Support Intervention: A Ten-Year Follow-Up. ChildDevelopment 56: 376-391.

Sherman, Lawrence W.

1992 Policing Domestic Violence: Experiments and Dilemmas. NY: FreePress.

Sherman, Lawrence W., Janell D. Schmidt, Dennis P. Rogan, Patrick R.Gartin, Ellen G. Cohn, Dean J. Collins and Anthony R. Bacich

1991 From Initial Deterrence to Long-Term Escalation: Short-CustodyArrest for Poverty Ghetto Domestic Violence. Criminology 29: 821-50.

Sherman, Lawrence W. and Heather Strang

1996 Policing Domestic Violence: The Problem-Solving Paradigm. PaperPresented to the Conference on Problem-Oriented Policing as Crime Prevention,Stockholm, Swedish National Police College.

Smith, Douglas and Visher, Christy

1981 Street-Level Justice: Situational Determinants of Police ArrestDecisions. Social Problems 29: 167-78.

Stinchcombe, Arthur

1963 Institutions of Privacy in the Determination of Police AdministrativePractice. The American Journal of Sociology 69: 150-160.

Strain, P.S., P. Steele, T. Ellis, and M.A. Timm

1982 Long-Term Effects of Oppositional Child Treatment With Mothersas Therapists and Therapist-Trainers. Journal of Applied Behavior Analysis15: 163-169.

Strang, Heather and Lawrence W. Sherman

1996 Paper Presented to the American Association for the Advancementof Science, Baltimore, MD, February.

Strayhorn, J.M. and C. Weidman

1991 Follow-up One Year After Parent-Child Interaction Training: Effectson Behavior of Preschool Children. Journal of the American Academy of Childand Adolescent Psychiatry 30: 138-143.

Tatara, T.

1989 Characteristics of Children in Foster Care. Division of Child,Youth and Family Services Newsletter (American Psychological Association)12: 16-17.

1992 Characteristics of Children in Substitute and Adoptive Care: AStatistical Summary of the VCIS National Child Welfare Data Base. Basedon FY 82 through FY 8 Data. March.

Tremble, R.E., L. Kurtz, L.C. Masse, F. Vitaro, and R.O. Pihl

1994 A Bimodal Preventive Intervention for Disruptive Kindergarten Boys:Its Impact Through Mid-Adolescence. Unpublished MS. Montreal: Universityof Montreal Research Unit on Children's Psycho-Social Maladjustment.

Tremble, Richard and Wendy Craig

1995 Developmental Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building A Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

US Advisory Board on Child Abuse and Neglect

1995 Annual report. Washington, DC: DHHS.

Wasik, B.H., C.T. Ramey, D.M. Bryant and J.J. Sparling

1990 A Longitudinal Study of Two Early Intervention Strategies: ProjectCARE. Child Development 61: 1682-1696.

Wasserman, Gail. A. and Laurie S. Miller

forthcoming The Prevention of Antisocial Behavior. Draft. In Rolf Loeberand David P. Farrington, Eds., Report of the OJJDP Study Group on Serious,Chronic, and Violent Juvenile Offenders. Washington, DC: OJJDP.

Webster-Stratton, C.

1990 Long-Term Follow-Up of Families With Young Conduct Problem Children:From Preschool to Grade School. Journal of Clinical Child Psychology 19:144-49.

Webster-Stratton, C., M. Kolpacoff, and T. Hollinsworth

1988 Self-Administered Videotape Therapy for Families With Conduct ProblemChildren: Comparison With Two Cost-Effective Treatments and a Control Group.Journal of Consulting and Clinical Psychology 56: 558-66.

Widom, Cathy Spatz

1989 Child Abuse, Neglect, and Violent Criminal Behavior. Criminology27: 251-271.

Wilson, William Julius

1996 When Work Disappears: The World of the New Urban Poor. NY: AlfredA. Knopf.

Yoshikawa, Hirokazu

1994 Prevention as Cumulative Protection: Effects of Early Family Supportand Education on Chronic Delinquency and Its Risks. Psychological Bulletin115: 28-54.

Yu, P., G. E. Harris, B.L. Solovitz, and J.L. Franklin

1986 A Social Problem-Solving Intervention for Children at High Riskfor Later Psychopathology. Journal of Clinical Child Psychology 15: 30-40.

Zigler, E.

1992 Head Start Falls Behind. The New York Times, June 27, p. 23.

Zimring, Franklin E. and Gordon Hawkins

1973 Deterrence. Chicago: University of Chicago Press.

Chapter 5

SCHOOL-BASED CRIME PREVENTION

by Denise C. Gottfredson1

Schools have great potential as a locus for crime prevention. They provideregular access to students throughout the developmental years, and perhapsthe only consistent access to large numbers of the most crime-prone youngchildren in the early school years; they are staffed with individuals paidto help youth develop as healthy, happy, productive citizens; and the communityusually supports schools' efforts to socialize youth. Many of the precursorsof delinquent behavior are school-related and therefore likely to be amenableto change through school-based intervention.

Figure 5-1 shows several school-related precursors to delinquency identifiedby research. These factors include characteristics of school and classroomenvironments as well as individual-level school-related experiences andattitudes, peer group experiences, and personal values, attitudes, andbeliefs. School environment factors related to delinquency include availabilityof drugs, alcohol, and other criminogenic commodities such as weapons;characteristics of the classroom and school social organization such asstrong academic mission and administrative leadership; and a climate ofemotional support. School-related experiences and attitudes which oftenprecede delinquency include poor school performance and attendance, lowattachment to school, and low commitment to schooling. Peer-related experiences,many of which are school-centered, include rejection by peers and associationwith delinquent peers. And individual factors include early problem behavior,impulsiveness or low levels of self-control, rebellious attitudes, beliefsfavoring law violation, and low levels of social competency skills suchas identifying likely consequences of actions and alternative solutionsto problems, taking the perspective of others, and correctly interpretingsocial cues. Several recent reviews summarize the research literature linkingthese factors with crime (Gottfredson, Sealock, & Koper, 1996; Hawkins,Catalano, & Miller, 1992; Howell, Krisberg, Wilson & Hawkins, 1995).

What Works, What Doesn't, What's Promising (4)

Figure 5-1 also draws attention to fact that schools operate in largercontexts which influence their functioning as well as their outcomes. Byfar the strongest correlates of school disorder are characteristics ofthe population and community contexts in which schools are located. Schoolsin urban, poor, disorganized communities experience more disorder thanother schools (Gottfredson & Gottfredson, 1985). Research has alsodemonstrated that the human resources needed to implement and sustain schoolimprovement efforts -- leadership, teacher morale, teacher mastery, schoolclimate, and resources -- are found less often in urban than in Figure1: Christina's Figure 5-1 other schools (Gottfredson, Fink, Skroban, andGottfredson, in press). It is precisely those schools whose populationsare most in need of prevention and intervention services that are leastable to provide those services. Although schools can not be expected toreverse their communities' problems, they can influence their own ratesof disorder. Controlling on relevant characteristics of the larger community,characteristics of schools and the way they are run explain significantamounts of variation in school rates of disorderly behavior (Gottfredson& Gottfredson, 1985).

National priorities for children focus on schools as a locus for theprevention of diverse social problems including crime. The Department ofHealth and Human Services' Healthy People 2000 goals include increasinghigh school graduation rates and reducing physical fighting, weapon-carrying,substance use, and pregnancy among adolescents. National Education Goal6 states that every school will be free of drugs, violence, and the unauthorizedpresence of firearms and alcohol, and will offer a disciplined environmentconducive to learning by the year 2000. The 1986 Drug-Free Schools andCommunities Act legislation provided substantial funds to states to developand operate school-based drug prevention programs. In 1994 this legislationwas modified to authorize expenditures on school-based violence preventionprograms as well.

This substantial national interest in schools as a prevention tool isnot matched by federal expenditures in this area. Table 5-1 shows thatfederal expenditures on school-based substance abuse and crime preventionefforts are modest,2particularly when compared with federal expenditures on control strategiessuch as policing and prison construction.3Perhaps more troubling, the meager federal expenditures on school-basedprevention are not well spent. The single largest federal expenditure onschool-based prevention (Safe and Drug-Free Schools and Communities moniesadministered by the U.S. Department of Education) funds a relatively narrowrange of intervention strategies, many of which have been shown eithernot to work

Table 5-1. Partial List of Federal Expenditures on School-based Prevention

Federal Program Agency Funding Strategies level Safe and Drug-Free DOE FY95: State and local education agency programs: instruction, Schools & 466.98M student assistance programs, teachers and staff training, Communities curriculum development and acquisition; red-ribbon week; Program before-after-school programs and community service. Note: Prior to Governor's state and local programs: Instruction 1994, this program (D.A.R.E.), replication of other drug education programs, funded drug high-risk youth programs programs in schools. The 1994 legislation authorized expenditures on violence prevention programs and curricula as well. High-Risk Youth DHHS FY95: 65.2M Various. In-school and after-school programs; violence Demonstration [CSAP] and drug prevention. Program Youth Violence DHHS FY95: 10.7M Various. Projects include instruction (violence Prevention Program [CDC] prevention, self-control, social competency; cognitive behavioral methods, tutoring, mentoring, recreation, campaigns to change norms, peer mediation and conflict resolution, changes in school management processes, parent training) Community Schools DHHS FY95: 10M Various. Prevention and academic achievement enhancement Youth Services and [Admin-ist during the non-school hours. Supervision ration Program for Children, Youth, & Families] Learn & Serve Corporatio FY95: 32M Community service tied to the school curriculum. Attempt America n for FY96: 32M to engage youths in school to prevent dropout. Character Program National education. Service D.A.R.E. (Drug DOJ/DOI FY95: 1.75M Instruction (core program and booster lessons); Abuse Resistance [BJA] FY96: 1.75M A recent extention of the program (D.A.R.E. + PLUS; Play Education) (To and Learn under Supervision) also includes and DOE D.A.R.E. after-school program. America) Plus annual funds from Byrne Block Grant Plus approx. 10M annually through Safe and Drug Free Schools program G.R.E.A.T. (Gang DOJ/ FY95: 16.2M Instruction Resistance TREAS Plus 265K Education and [ATF/NIJ] (eval) Training) C.I.S. (Cities in DOJ FY95: 592K School-based supportive services for at-risk students and Schools) [OJJDP] FY96: 340K their families JUMP (Juvenile DOJ FY96: 15M Mentoring Mentoring Program) [OJJDP] L.R.E. DOJ FY95: 2.7M Instruction, character education (Law-related [OJJDP] FY96: 1.2M education) Note: M=million; K=thousand 

(e.g., counseling) or to have only small effects (e.g., drug instruction).School-based prevention monies administered by OJP also fail to capitalizeon the full range of empirically-tested, effective strategies.

This chapter is intended to provide information for use in setting federalresearch agendas and guiding funding decisions about what works, what doesnot work, what is promising, and how delinquency prevention efforts canbe strengthened. It begins by clarifying the outcomes sought in school-basedprevention programs. It then classifies school-based prevention activitieswithin two broad approaches -- environmental and individual-focused --into more specific program types. Next it reviews research related to eachtype of activity, comments on the quality of the available informationabout the efficacy of each type of activity, and summarizes knowledge aboutwhat works, what does not work, and what is promising. It ends with a summaryof findings and recommendations for OJP funding of school-based preventioninterventions and further research.

The Nature of School-Based Prevention

Measures of effectiveness. School-based prevention programs includeinterventions to prevent a variety of forms of "problem behavior,"including theft, violence, illegal acts of aggression, alcohol or otherdrug use; rebellious behavior, anti-social behavior, aggressive behavior,defiance of authority, and disrespect for others. These different formsof delinquent behavior are highly correlated and share common causes. Manyof the programs considered in this chapter were not specifically designedto prevent the problem behaviors, but instead to affect presumed causalfactors such as school drop-out, truancy, or other correlates which areexpected to increase protection against or decrease risk towards engagingin problem behaviors at some later date. This focus on non-crime programoutcomes is entirely appropriate given the young ages of many of the targetedstudents. Different outcomes have different saliencies for different agegroups. Positive program effects on reading skills for six-year-olds maybe as important in terms of later crime prevented as reducing marijuanause for sixteen-year-olds. Many prevention researchers and practitionersalso assume a link between less serious problem behaviors and later moreserious crime. They are satisfied when their interventions demonstrateeffects on the early forms of problem behavior. This developmental perspectiveunderlies many school-based prevention efforts today and may explain thewide variety of outcome measures used to assess the effectiveness of theseprograms, some of which are summarized in Figure 5-2.

Studies of the effects of school-based prevention on serious violentcrime are rare. Of the 149 studies examined for this review, only 9 measuredprogram outcomes on murder, rape, robbery or aggravated assault. Only 15measured outcomes on serious property crimes such as burglary, larceny-theft,and motor vehicle theft. More (25) measured less serious or unspecifiedcriminal behavior. Far more common are studies assessing program effectson alcohol, tobacco, or other drug use (77 studies) and other less seriousforms of rebellious, anti-social, aggressive, or defiant behaviors (79studies). Most studies measure the risk or protective factors directlytargeted by the program (e.g., academic achievement, social competencyskills).

Figure 5-2: Common Outcome Measures for School-based Prevention ProgramsAlcoholand other drug use: Ingestion of alcoholic beverages and ingestionof any illicit drug are considered substance abuse. Dimensions of use thatare often measured distinctly in evaluations of prevention programs includeage of first use (age at onset); status as having used alcohol or anotherdrugs at least once; and current use, including frequency of use and amounttypically used. Substance use is most often measured using youth self-reportsin evaluations of school-based prevention programs.

Delinquent and criminal behavior: Delinquent or criminal behavioris any behavior which is against the law. Delinquency is criminal behaviorcommitted by a young person. Laws, and therefore the precise definitionof behaviors in violation of the law, vary slightly from state to state.Crime and delinquency includes the full range of acts for which individualscould be arrested. It includes crimes against persons ranging in seriousnessfrom murder to robbery to minor assault. It includes an array of crimesagainst property ranging from arson to felony theft to joyriding. Crimeand delinquency also includes possession, use, and selling of drugs. Forjuveniles, it includes status offenses such as running away. Dimensionsof crime that are often measured distinctly in evaluations include ageof first involvement, status as a delinquent ever in one's life, currentcriminal activity, and frequency of delinquent involvement. Delinquencyis more often measured using youth self-reports than official records ofarrest or conviction in evaluations of school-based prevention programs.

Withdrawal from school: Leaving school prior to graduation fromthe 12th grade and truancy are often used as measures of success in preventionprograms. The precise definition of truancy differs according to location.For practical purposes it is often measured as the number of days absentfrom school.

Conduct problems, low self-control, aggression: These characteristicsare so highly related to delinquent behavior that they may be consideredproxies for it. Studies of school-based prevention often measure thesecharacteristics in addition to or in lieu of actual delinquent behaviorbecause (1) the subjects are too young to have initiated delinquent behavior,(2) the questions are less controversial because they are not self-incriminating,or (3) teachers and parents are more able to rate youth on these characteristicsthan on actual delinquent behavior, which is often covert. Conduct problembehavior subsumes a variety of behaviors: defiance, disrespect, rebelliousness,hitting, stealing, lying, fighting, talking back to persons in authority,etc. Low self control is a disposition to behave impulsively, and aggressioninvolves committing acts of hostility and violating the rights of others.

Risk and protective factors: As noted in the text, the effectivenessof prevention programs is often assessed by examining program effects ofa variety of factors which are known to elevate or reduce risk for delinquentinvolvement at a later date. These factors are discussed above and shownin Figure 1.

Because Congress has asked for a review of scientific literature oncrime prevention, studies including evaluations on crime, delinquency,alcohol or other drug use, or other forms of antisocial behavior are highlighted.Studies with demonstrated effects on risk and protective factors relatedto delinquency are also mentioned. Many substance abuse prevention programsare summarized in the chapter because substance use is one aspect of theadolescent problem behavior syndrome, is itself a form of criminal behaviorfor adolescents, and is highly correlated with more serious forms of criminalbehavior. A distinction between substance use (including alcohol, marijuana,and harder drug use) and all other forms of delinquency is maintained throughoutthe report. Programs are considered to influence substance use or delinquentbehavior if their evaluations demonstrate effects on any measure of eachoutcome, regardless of its type or seriousness level.

Categories of school-based prevention. Programs included in thischapter are located primarily in school buildings (even if outside of schoolhours) or are implemented by school staff or under school or school systemauspices. Programs targeting all grade levels -- kindergarten, elementary,and secondary -- are included. Excluded from this chapter are school-basedprograms intended to alter family conditions or practices (these are coveredin the family chapter), and school-based attempts to secure the schoolboundaries from intruders, weapons, and drugs. These are considered inthe chapter on place-based strategies.

Figure 5-3 describes four categories of school-based prevention focusingon altering school or classroom environments and Figure 5-4 describes fivecategories of school-based prevention focusing on changing the behaviors,knowledge, skills, attitudes, or beliefs of individual students. Classifyingany particular school-based prevention activity is a difficult task becausemost school-based prevention programs contain a mix of different typesof activities. In the 149 studies examined for this review, most (94%)contained multiple components (i.e., components falling into more thanone of the major categories of program activity shown in the figures).About 40% of the studies contained components in four or more differentcategories. Table 5-2 shows the major types of activities and the percentageof studies whose evaluated programs contained each type of activity. Itshows that the school-based programs described in most studies includean instructional component and a component intended to alter classroommanagement strategies. These common strategies are often combined withattempts to teach students new ways of thinking and dealing with potentialsocial problems. Other fairly common approaches in these studies are behaviormodification and attempts to change the normative climate of the school.

The multi-component strategy found in most studies of school-based preventionis perfectly reasonable given the nested nature of the schooling experienceand the multiple routes to problem behavior. Student behavior is most directlyinfluenced by the attitudes, beliefs, and characteristics of the studentand his or her peers. Individually-targeted interventions such as instructionalor behavior modification techniques that teach students new ways of thinkingand acting may be effective in changing these individual factors. But severalof these individual factors (e.g., low self-control, academic failure experiences,and attitudes favorable to drug use) are likely causes of problem behaviorand are best targeted through a set of inter-related program componentsrather than through a single intervention. Moreover, students interactin the context of classrooms, each of which has its own normative climateencouraging or discouraging certain behaviors. And classrooms exist inschool environments which establish larger contexts for all activitiesin the school. An instructional program teaching students to resolve conflictsnon-violently is not likely to be as effective for reducing violence ina school or classroom setting in which fights are regularly ignored asin one which immediately responds to such incidents. The interconnectionsamong different prevention components and the interdependence of differentcontexts should be considered in the design of prevention programs (Elias,Weissberg, et al., 1994).

Most recent reviews of school-based prevention are organized by developmentallevel (e.g., elementary, junior high, senior high) rather than by programtype. Despite the difficulties inherent in classifying prevention activities,it is nevertheless a useful activity because only by decomposing differentsets of activities into their major parts can we (a) describe the activities;(b) describe how the mix of activities varies across location (e.g., urban,suburban, rural) and developmental level; and (c) design evaluations ofspecific constellations of components. Also, several evaluations of relativelynarrow programs are available and can provide information about the potentialof each activity as a piece of a larger, more potent, prevention strategy.Ongoing research jointly sponsored by the Bureau of Justice Assistanceand National Institute of Justice will cross-classify program types bydevelopmental level and school location to provide a more comprehensivepicture of which school-based prevention activities are used in which locationsfor which grade levels.

Figure 5-3: Environmental Change Strategies for School-Based Prevention
Environmental Change Strategies

Building School Capacity

: Interventions to change the decision-makingprocesses or authority structures to enhance the general capacity of theschool. These interventions often involve teams of staff and (sometimes)parents, students, and community members engaged in planning and carryingout activities to improve the school. They often diagnose school problems,formulate school goals and objectives, design potential solutions, monitorprogress, and evaluate the efforts. Activities aimed at enhancing the administrativecapability of the school by increasing communication and cooperation amongmembers of the school community are also included.

Setting Norms for Behavior, Rule-Setting

: School-wide efforts to redefinenorms for behavior and signal appropriate behavior through the use of rules.It includes activities such as newsletters, posters, ceremonies duringwhich students declare their intention to remain drug-free, and displayingsymbols of appropriate behavior. Some well-known interventions in thiscategory are "red ribbon week" sponsored through the Departmentof Education's Safe and Drug-Free Schools and Communities program and school-widecampaigns against bullying. The category also includes efforts to establishor clarify school rules or discipline codes and mechanisms for the enforcementof school rules.

Managing Classes: Using instructional methods designed to increasestudent engagement in the learning process and hence increase their academicperformance and bonding to the school (e.g., cooperative learning techniquesand "experiential learning" strategies); and classroom organizationand management strategies. The latter include activities to establish andenforce classroom rules, uses of rewards and punishments, management oftime to reduce "down-time," strategies for grouping studentswithin the class, and use of external resources such as parent volunteers,police officers, or professional consultants as instructors or aides.

Regrouping Students: Reorganizing classes or grades to create smallerunits, continuing interaction, or different mixes of students, or to providegreater flexibility in instruction. It includes changes to school schedule(e.g., block scheduling, scheduling more periods in the day, changes inthe lengths of instructional periods); adoption of schools-within-schoolsor similar arrangements; tracking into classes by ability, achievement,effort, or conduct; formation of grade level "houses" or "teams;"and decreasing class size. Alternative schools for disruptive youths arealso included in this category.

Figure 5-4: Individual-Change Strategies for School-Based Prevention
Individual-Change Strategies:
Strategies to Change Student Knowledge, Skills, Attitudes, Beliefs, orBehaviors

Instructing Students

: The most common strategy used in schools. Theseinterventions provide instruction to students to teach them factual information,increase their awareness of social influences to engage in misbehavior,expand their repertoires for recognizing and appropriately responding torisky or potentially harmful situation, increase their appreciation fordiversity in society, improve their moral character, etc. Well-known examplesinclude Drug Abuse Resistance Education (D.A.R.E.), Law-related Education(L.R.E.), and Gang Resistance Education and Training (G.R.E.A.T.).

Behavior Modification and Teaching Thinking Strategies: Behaviormodification strategies focus directly on changing behaviors and involvetimely tracking of specific behaviors over time, behavioral goals, anduses feedback or positive or negative reinforcement to change behavior.These strategies rely on reinforcers external to the student to shape studentbehavior. Larger or more robust effects on behavior might be obtained byteaching students to modify their own behavior using a range of cognitivestrategies research has found lacking in delinquent youth. Efforts to teachstudents "thinking strategies" (known in the scientific literatureas cognitive-behavioral strategies) involve modeling or demonstrating behaviorsand providing rehearsal and coaching in the display of new skills. Studentsare taught, for example, to recognize the physiological cues experiencedin risky situations. They rehearse this skill and practice stopping ratherthan acting impulsively in such situations. Students are taught and rehearsedin such skills as suggesting alternative activities when friends proposeengaging in a risky activity. And they are taught to use prompts or cuesto remember to engage in behavior.

Peer Programs: Peer counseling, peer mediation, and programs involvingpeer leaders.

Other Counseling and Mentoring: Individual counseling and case managementand similar group-based interventions, excluding peer counseling. Counselingis distinguished from mentoring, which is generally provided by a lay personrather than a trained counselor is not necessarily guided by a structuredapproach.

Providing Recreational, Enrichment, and Leisure Activities: Activitiesintended to provide constructive and fun alternatives to delinquent behavior.Drop-in recreation centers, after-school and week-end programs, dances,community service activities, and other events are offered in these programsas alternatives to the more dangerous activities. The popular "MidnightBasketball" is included here.

Table 5-2. Percentage Studies Including Each Intervention Strategy

 Program Strategy Percentage Studies Including Instructing Students 78 Managing Classrooms 66 Teaching Thinking Strategies 49 Setting Norms for Behavior, Rule-Setting 33 Behavioral Modification 27 Peer Counseling, mediation, and leaders 16 Counseling 14 Providing Recreational, Enrichment, and 10 Leisure Activities 10 Building School Capacity 5 Regrouping Students 3 Mentoring 

Methods

Search and summary methods used in this chapter are described in moredetail in the methods appendix. Briefly, a library search was conductedto locate all published studies of school-based prevention programs. Thislist was augmented with additional studies cited in recent reviews of preventionprograms. In all, 149 studies were located and classified into the programcategories described above. Studies of multi-component programs were assignedto the category which best described the program. For categories containinga manageable number of studies, all studies were coded for methodologicalrigor and effect sizes were computed4(when possible) for measures of delinquency and substance use. For categoriescontaining more studies than could be coded in the short time availableto produce this report, recent high-quality secondary reviews were summarizedand two or three of the most rigorous studies were coded using the sameprocedures as for the smaller categories.

The following paragraphs discusses in more detail three issues specificto this chapter.

Effect sizes. Program effects are expressed whenever possiblein this chapter as "effect sizes" (ES), a measure of change dueto the treatment as a proportion of the standard deviation for each measureemployed. ESs usually range from -1 (indicating that the treatment groupperformed one standard deviation lower than the comparison group) to +1(indicating that the treatment group performed one standard deviation higherthan the comparison group). Rosenthal and Rubin (1982) show that ESs canbe translated for ease of interpretation into the equivalent of percentagedifferences by simply dividing the ES by 2 and multiplying by 100. Theresulting figure represents the relative percentage difference in success(or failure) rates between the experimental and control groups. For example,an ES of .5 might indicate that the success rate for the treatment groupis 25 percentage points above that of the comparison group. Lipsey &Wilson (1993), summarizing effect sizes from 156 reviews of 9,400 interventionsin the social and behavioral sciences and education, reported an averageeffect size of .47 (SD=.28) for many different types of programs and manydifferent outcomes. By comparison, Lipsey (1992) showed the average effectsize in 397 studies of delinquency treatment and prevention was .17 (SD=.44).Delinquent behavior appears more difficult to change than more conventionalbehaviors. The practical significance of an effect size depends largelyon the seriousness of the outcome for the population. Lipsey argues thateven small ESs (e.g., .10) for serious crime have practical significance.

Level of analysis. Most studies of school-based prevention sharea methodological shortcoming: Data that should be analyzed at the classroomor school level are instead analyzed at the individual level. School-basedprevention programs are usually administered to intact classrooms or schoolsand these larger units are usually assigned to treatment and control conditions.But most studies, conducted with limited funding, involve relatively smallnumbers of classes or schools. The largest study reviewed in this chapterinvolved only 56 schools, and most involve fewer than 10. Investigatorsusually analyze their data as though individuals were assigned to treatmentand comparison conditions. Resulting estimates of the effects of school-basedprevention practices are imprecise. Corrections are seldom or never madefor the correlated error terms that result when observations are clusteredin larger units. Effect sizes are usually underestimated because they usethe larger individual-level standard deviation estimates rather than thesmaller standard error estimates for classrooms or schools. This shortcomingcan be corrected in future studies only with increased funding for studiesto allow for larger numbers of schools and classrooms.

Scientific vs. programmatic rigor. The scientific rigor of studiessummarized in this chapter was classified using the coding scheme describedin the methods appendix. The programmatic rigor of prevention programsis not as easily quantified because the same level of consensus does notexist about the elements of programmatic rigor. We can be reasonably certain,however, that longer-term, multi-component strategies located in naturalschool settings, using staff readily available to the schools, employingmethods that are acceptable to regular school staff are most likely toproduce the strongest and most durable effects. A conundrum for school-basedprevention research is that such rigorous programs are the most difficultto study using rigorous methods. Long-term interventions are more likelyto suffer from attrition problems. In natural setting it is not alwayspossible to randomly assign subjects to treatment and control conditions,thus lowering confidence in the interpretation of any differences observedas due to the effects of the intervention. The most rigorous programs,therefore, are usually not studied with the highest level of scientificrigor.

Studies of School-based Prevention

The Office of Juvenile Justice and Delinquency Prevention (OJJDP) launcheda large-scale school-based demonstration project in the early 1980s, fundingeighteen different school-based delinquency prevention models in fifteencities. Program models ran the gamut from alternative schools employingbehavior modification for high-risk youths, to counseling classes, to enhancingmanagement processes in schools. Seventeen of the projects were includedin the national evaluation of the initiative, also funded by OJJDP. Gottfredson(1987), summarizing the evaluation, concluded that the initiative was successfulin demonstrating that some school-based preventive interventions reducedelinquency. Schools in the initiative became significantly safer and lessdisruptive over the course of the initiative. The initiative as a wholedemonstrated that school-based prevention can work, but evaluations ofspecific program models showed great variability in their effectiveness.Reports on many of the specific program models included in the initiativehave made their way into the scientific research literature and will besummarized at appropriate points later in this chapter.

Changing School and Classroom Environments

Correlational evidence suggests that the way schools are run predictsthe level of disorder they experience. Schools in which the administrationand faculty communicate and work together to plan for change and solveproblems have higher teacher morale and less disorder. These schools canpresumably absorb change. Schools in which students notice clear schoolrules and reward structures and unambiguous sanctions also experience lessdisorder. These schools are likely to signal appropriate behavior for students(Corcoran, 1985; Gottfredson, 1987; Gottfredson & Gottfredson, 1985;Gottfredson, Gottfredson, & Hybl, 1993). Schools in which studentsfeel as though they belong and that people in the school care about themalso experience less disorder (Duke, 1989). These schools are probablybetter at controlling behavior informally. Intervention studies have testedfor a causal association between each of these factors and delinquencyor substance use among students. Four major strategies for changing schooland classroom environments are summarized below: (1) building school capacityto manage itself; (2) setting norms or expectations for behavior and establishingand enforcing school rules, policies, or regulations; (3) changing classroominstructional and management practices to enhance classroom climate orimprove educational processes; and (4) grouping students in different waysto achieve smaller, less alienating, or otherwise more suitable micro-climateswithin the school.

Building School Capacity

Program Development Evaluation (PDE; G. Gottfredson, 1984a; Gottfredson,Rickert, Gottfredson, and Advani, 1984) is a structured organizationaldevelopment method developed to help organizations plan, initiate, andsustain needed changes. Researchers and practitioners collaborate, usingspecific steps spelled out in the program materials, to develop and implementprograms. A spiral of improvement is created as researchers continuouslyprovide data feedback during the implementation phase to the practitionersand work with them to identify and overcome obstacles to strong programimplementation. The method -- first developed for use with schools participatingin the OJJDP alternative education initiative -- was intended to solvethe problem that evaluations up until that time had found few efficaciousdelinquency prevention models. The developer assumed that the poor showingwas due to weak evaluations, failure to inform program design with researchknowledge and social science theory, and weak program implementation.

PDE was used in a comprehensive school improvement intervention -- projectPATHE -- that altered the organization and management structures in sevensecondary schools between 1981 and 1983 as part of OJJDP's alternativeeducation initiative (D. Gottfredson, 1986; scientific methods score=4).District-level administrators used PDE to develop a general plan for allseven schools, and then used PDE to structure specific school-level planninginterventions. These efforts increased staff and student participationin planning for and implementing school improvement efforts. Changes resultingfrom the planning activity included efforts to increase clarity of rulesand consistency of rule enforcement and activities to increase students'success experiences and feelings of belonging. These activities targetedthe entire population in each school.

The evaluation of the project compared change on an array of measuresfrom the year prior to the treatment to one year (for four high schools)5and two years (for five middle schools) into the intervention. One schoolat each level was a comparison school selected from among the non-participatingschools to match the treatment schools as closely as possible. The studentsin the participating high schools reported significantly less delinquentbehavior6 (ES=-.16)and drug use (ES=-.19), had fewer suspensions (ES=-.27), and fewer schoolpunishments (ES=-.18) after the first year of the program. Students inthe comparison high school did not change significantly on these outcomes.A similar pattern was observed for the middle schools after two years.As serious delinquency increased significantly in the comparison school,it decreased (nonsignificantly) in the program middle schools (ES=-.27).Changes in drug use (ES=-.13) and school punishments (ES=-.15) also favoredthe program schools. Suspensions also declined significantly in the programmiddle schools, but a similar decline was observed in the comparison school.Several indicators of the school climate directly targeted by the program(e.g., safety, staff morale, clarity of school rules, and effectivenessof the school administration) also increased in the program schools, witheffect sizes ranging form .16 to .63.

D. Gottfredson (1987; scientific methods score=4) reported the resultsof a similar effort -- The Effective Schools Project -- in a difficultBaltimore City junior high school. PDE was used with a team of school anddistrict-level educators to plan and implement changes to instructionaland discipline practices. School-wide and classroom-level changes weremade to the disciplinary procedures to increase the clarity and consistencyof rule enforcement, and to substitute positive reinforcement strategiesfor strategies that relied solely on punishment. Instructional innovationsincluding cooperative learning and frequent monitoring of class work andhomework were put in place, an expanded extracurricular activities programwas added, and a career exploration program which exposed youth to positiverole models in the community, took them on career-related field trips,and provided instruction on career-related topics was undertaken.

The evaluation of the project involved a comparison of pre-treatmentmeasures to post-treatment measures taken two years later for the one treatmentschool and a second school which was intended to receive the program butinstead chose to develop a school improvement plan with minimal assistancefrom the researchers (and without using the PDE method). Indicators oforganizational health (e.g., staff morale, cooperation and collaborationbetween faculty and administration, and staff involvement in planning andaction for school improvement) improved dramatically in the treatment school.Only the Planning & Action scale improved in the comparison school.Significant reductions from pre- to post-treatment on delinquency (seefootnote 3, ES=-33) and increases in classroom orderliness (ES=.57) wereobserved for the treatment school. A reduction in student reports of rebelliousbehavior in the treatment school was observed (not significant) while asignificant increase was observed in the comparison school (ES=-.22).

Kenney & Watson (1996; scientific methods score = 3) report on anintervention to empower students to improve safety in schools. This study,funded by NIJ in 1993, involved 11th grade students (N's range from 372to 451) in the application of a problem-solving technique to reduce problemsof crime, disorder, and fear on the school campus. As part of their governmentand history class, students implemented a four-step problem-solving methodcommonly used in problem-oriented policing interventions to identify problems,analyze possible solutions, formulate and implement a strategy, and evaluatethe outcomes of the intervention. The investigators anticipated that empoweringstudents to serve as change agents in the school would produce safer schools.Among the problems selected by the students to work on were streamlininglunch-room procedures and monitoring the restrooms. These place-orientedstrategies are discussed in Eck's chapter in this volume.

Baseline surveys used by the planning groups to identify school problemswere used also as baseline measures for the evaluation of the project.Change over a two-year period was examined for the treatment and one comparisonschool. The study found that students in the treatment school reportedsignificantly less fighting and less teacher victimization and were lessfearful about being in certain places in the school at the end of the two-yearperiod compared with their baseline. Students in the comparison schooldid not change on these outcomes. A few of the items measuring teacherfear and victimization experiences were significantly lower at the endof the program, but positive effects were more evident in student thanon teacher reports. The positive findings for this program on measuresof fighting, fear, and victimization experiences are consistent with theGottfredson et. al. research showing that building school capacity forinitiating and sustaining change reduce delinquency and drug use. All threestudies were of acceptable methodological rigor, with scientific methodsscores of 3 or 4. The size of the effects on delinquency and substanceuse ranged from small (-.13) to moderate (-.33), with larger effects (upto .57) observed for less serious forms of misbehavior.

Norms for Behavior and Rule-Setting

Research on the correlates of school disorder summarized earlier inthis chapter suggests that a constellation of discipline management-relatedvariables -- clarity about behavioral norms, predictability, consistencyand fairness in applying consequences for behaviors -- are inversely relatedto rates of teacher and student victimization in schools. Several studieshave attempted to intervene in schools to increase the clarity and consistencyof rule enforcement. Others have deliberately involved students in thedevelopment and enforcement of the rules in an attempt to increase theperceived validity and fairness of the rules. Still others have attemptedto establish or change school norms using campaigns, ceremonies, or similartechniques.

Gottfredson, Gottfredson, & Hybl (1993; scientific methods score=4)tested a discipline management intervention in six urban middle schools.This program (BASIS) included the following components:

Increasing clarity of school rules and consistency of rule enforcementthrough revisions to the school rules and a computerized behavior trackingsystem;

Improving classroom organization and management through teacher training;

Increasing the frequency of communication with the home regarding studentbehavior through systems to identify good student behavior and a computerizedsystem to generate letters to the home regarding both positive and negativebehavior; and

Replacing punitive disciplinary strategies with positive reinforcementof appropriate behavior through a variety of school- and classroom-levelpositive reinforcement strategies.

School teams of administrators, teachers, and other school personnelwere responsible for implementing the program. When all six participatingschools were compared with the two non-randomly selected comparison schools,significant changes in the expected direction were observed from the beginningto the end of the program on the measures most directly targeted: classroomorderliness, classroom organization, classroom rule clarity, and fairnessof school rules. Student reports of rebellious behavior, a scale measuringminor delinquent acts, increased significantly over the three yeartime frame for students in both treatment and comparison schools, and slightlymore so in treatment schools (ES=.27) than in the comparison schools (ES=.19).This increase was probably due to the county-wide aging of the middle schoolstudent population which resulted when the implementation of higher grade-to-gradepromotion standards resulted in a huge increase in grade retentions. Implementationdata showed that the components of the program were implemented with highfidelity to the original design in only three of the six program schools.In these three schools, teachers reports of student attention to academicwork increased significantly (ES=.09) and their ratings of student classroomdisruption decreased significantly (ES=-.12). The increase in rebelliousbehavior was smallest (ES=.11) in the these schools, although the differencebetween these "high implementation" treatment schools and thecontrol schools was small (difference in ES=.08).

In another three-year discipline management study implemented in nineschools, Mayer, Butterworth, Nafpaktitus, & Sulzer-Azaroff (1983; scientificmethods score=5) demonstrated positive effects for a program that trainedteams of school personnel to use behavioral strategies for reducing studentvandalism and disruption. Each team also met regularly to plan and implementprograms on a school-wide basis that would teach students alternative behaviorto vandalism and disruption. These included lunch-room and playground managementprograms and classroom management programs that stressed the use of specificpositive reinforcement. Graduate student consultants worked with each teacherabout twice per week and conducted about two team meetings per month duringthe school year. The study showed that rates of student off-task behaviordecreased significantly and vandalism costs plummeted in the project schools.These results replicated results from an earlier pilot study (Mayer &Butterworth, 1978; scientific methods score=4). Note that the school teamapproach used in this study resembles that used in the PDE method describedabove.

An impressive program of research on an intervention designed to limitconflict in schools undertaken in Norway (Olweus, 1991, 1992; Olweus &Alsaker, 1991; scientific methods score=3) suggests that school-wide effortsto redefine norms for behavior reduce delinquency. Olweus noted that certainadolescents -- "bullies" -- repeatedly victimized other adolescents.This harassment was usually ignored by adults who failed to actively interveneand thus provided tacit acceptance of the bullying. A program was devisedto alter environmental norms regarding bullying. A campaign directed communicationto redefining the behavior as wrong. A booklet was directed to school personnel,defining the problem and spelling out ways to counteract it. Parents weresent a booklet of advice. A video illustrating the problem was made available.Surveys to collect information and register the level of the problem werefielded. Information was fed back to personnel in 42 schools in Bergen,Norway. Among the recommended strategies to reduce bullying were: establishingclear class rules against bullying; contingent responses (praise and sanctions);regular class meetings to clarify norms against bullying; improved supervisionof the playground; and teacher involvement in the development of a positiveschool climate.

The program was evaluated using data from approximately 2,500 students(aged 11 to 14) belonging to 112 classes in 42 primary and secondary schoolsin Bergen. The results indicated that bullying decreased by 50 percent(exact ESs can not be computed from the information provided in the publishedreports, but they appear to range from approximately -.10 to -.50 for differentgrade levels, genders, and measures of bullying). Program effects werealso observed on self-reports of delinquent behavior -- including truancy,vandalism, theft. These effects on delinquency were smaller in magnitude(ESs below -.2 except for one of the 10 comparisons whose ES was approximately-.42).

Encouragement to adopt norms against drug use during adolescence hasalso been identified as an essential element of drug abuse prevention (Instituteof Medicine [IOM], 1994). Curricula that promote norms against drug useoften include portrayals of drug use as socially unacceptable, identificationof short-term negative consequences of drug use, provision of evidencethat drug use is less prevalent among peers than children may think, encouragementfor children to make public commitments to remain drug-free, and the useof peer leaders to teach the curriculum (IOM, 1994, page 264). These activitiesare present in 29% of drug prevention curricula (Hansen, 1992), but alwaysin conjunction with other components such as conveying information aboutrisks related to drug use and resistance skills training. Norm-settingand public pledges to remain drug-free are usually elements of the mosteffective drug education curricula, but meta-analyses have not been ableto disentangle the effects of the various components. In a study designedto do just that, Hansen & Graham (1991; scientific methods score=4)found that positive effects on marijuana use and alcohol use were attributablemore to a normative education than to a resistance skills training component.

In summary, programs aimed at setting norms or expectations for behavior,either by establishing and enforcing rules or by communicating and reinforcingnorms in other ways (e.g., campaigns), have been demonstrated in severalstudies of reasonable methodological rigor to reduce alcohol and marijuanause and to reduce delinquency. Note, however, that studies in which schoolrules were manipulated also used school teams to plan and implement theprograms, so it is not possible to separate the specific effects of theschool rule and discipline strategies from the more general effects ofencouraging teams of school personnel to solve their schools' problems.

Managing Classes

Effective Instructional Practices Summarized in Brewer et al. (1995)Smaller kindergarten and first grade classrooms
Within-class and between-grade ability grouping in elementary grades
Nongraded elementary schools
Behavioral techniques for classroom management
Continuous progress instruction (e.g., instruction in which students advancethrough a defined hierarchy of skills after being tested for mastery ateach level usually with teachers providing instruction to groups of studentsat the same instructional level)
Computer-assisted instruction
Tutoring
Cooperative learning

Most of students' time in school is spent in classrooms. How these micro-environmentsare organized and managed may influence not only the amount of disorderlybehavior that occurs in the class but also important precursors of delinquencyand drug use, including academic performance, attachment and commitmentto school, and association with delinquent peers.

Classroom organization and management strategies are found in most school-basedprevention studies. They are usually incorporated into both the school-wideinterventions summarized above and (less often) into the instructionalinterventions described later. For example, cooperative learning strategieswere used in Project PATHE (Gottfredson, 1986), the Effective Schools Project(Gottfredson, 1987), and Project STATUS (Gottfredson, 1990), all of whichdemonstrated reductions in delinquent behavior. Classroom management techniqueswere used in Project BASIS (Gottfredson, Gottfredson, & Hybl, 1993).In all of these projects, the classroom instruction and management strategieswere elements of broader, school-wide organization development or disciplinemanagement projects (or in the case of STATUS, a law-related educationcurricular intervention), thus making it impossible to isolate the effectsof the classroom strategies. Classroom management innovations constitutethe major intervention in the studies summarized in this section.

The literature on effective instructional processes is vast. Most ofthis literature assesses effectiveness on academic outcomes rather thanon behavioral outcomes. Brewer et al. (1995) summarize existing meta-analysesof instructional strategies and conclude that the strategies shown in thebox on the preceeding page increase academic performance, which is relatedto delinquency and drug use. These instructional strategies should be consideredpromising elements of prevention efforts at the classroom level, althoughtheir effects on delinquency and substance use have not been demonstrated.

Table 5-3 summarizes evidence from two long-term interventions intendedto test the efficacy of upgrading classroom instructional and managementmethods on subsequent substance use and delinquent behavior. The SeattleSocial Development Project (Hawkins et al., 1988; 1991; 1992; O'Donnellet al, 1995) used cooperative learning strategies, proactive classroommanagement, and interactive teaching. Proactive classroom management consistedof establishing expectations for classroom behavior, using methods of maintainingclassroom order that minimize interruptions to instruction, and givingfrequent specific contingent praise and encouragement for student progressand effort. Interactive teaching involved several instructional practicesgenerally accepted as effective (e.g., frequent assessment, clear objectives,checking for understanding, and remediation). Cooperative learning usedsmall heterogeneous learning groups to reinforce and practice what theteacher taught. Recognition and team rewards were provided to the teams,contingent on demonstrated improvement. Parent training in family managementpractices was also provided. This program was implemented with supportfrom OJJDP continually from first through sixth grades in several elementaryschools beginning in 1981. In addition, the classroom management strategieswere implemented without the parent training in a one-year study of seventhgraders (Hawkins, Doueck, & Lishner, 1988). Several of the projectreports are summarized in Table 5-3. The evaluations demonstrated consistentsignificant positive effects on attachment and commitment to school, andthe absence of such effects on belief in moral order and attitudes aboutsubstance use. For the long-term project including parent training, measuresof alcohol and marijuana use generally favored the treatment students,but were marginally significant and sometimes significant only for girls.Measures of aggressive behavior favored the treatment group in second grade,but only for males. By fifth grade, measures of school misbehavior andminor delinquency initiation showed no significant effects for the fullsample. By sixth grade, a lower delinquency initiation was observed forthe treatment group, but only for low income males participating in theprogram. For low-achieving seventh graders who received the classroom portionof the program with no parent training, no significant effects were observedon measures of delinquency and drug use, although the treatment group hadsignificantly fewer suspensions from school. Table 5-3. Studies of ClassroomManagement

Author Scientific Effect size for measure Effects on risk and protective factors (year) methods of problem behavior score/ Number of cases Hawkins, Von Cleve, & 3 Aggressive behavior Internalizing problem behaviors, anxiety, Catalano (1991) (teacher reports) social withdrawal [NS] N=458 boys & [favors treatment, [results for second girls significant for males graders after two years only, ES=-.34 for of program] males] Externalizing problem behavior (teacher reports) [favors treatment, significant for males only, ES=-.29 for males Hawkins, Catalano, 2 Alcohol use Attachment to school, Commitment to school, Morrison, O'Donnell, [favors treatment, Attachment to family, Family management Abbott, & Day (1992) N= 853 boys almost significant ( [significantly favors treatment] & girls p<.1), ES=-.12] [results for fifth Achievement test scores [significantly favors graders after four Minor delinquency control] years of program] initiation [NS; ES=-.11] Belief in moral order, Attitudes favoring substance use [NS] School misbehavior [NS] O'Donnell, Hawkins, 2 Alcohol use Attachment to school [significantly favors Catalano, Abbott, & Day [favors treatment for treatment. p<.05 for girls, p<.10 for boys] (1995) N= 49 boys girls only; almost and 57 girls significant (p<.1) for Commitment to school [significantly favors [results for sixth (analyzed girls only, ES=-.40 for treatment for boys and girls] graders after six years separately girls] of program] by gender) Grades [favors treatment, significant for Marijuana use boys only] [favors treatment for girls only; almost Achievement test scores [favors treatment, significant (p<.1) for significant for boys only] girls only, ES=-.34 for girls] Belief in moral order [NS] Minor delinquency Attitudes favoring substance use [NS] initiation [favors treatment, almost significant (p<.1) for boys only, ES= -.54 for boys] Hawkins, Doueck, & 3 Self-reported Achievement test scores [NS] Lishner (1988) delinquency [NS; ES's N=160 range from .04 to .14 School attachment [of 6 items, 2 [results for seventh low-achieving favoring control] significantly favor treatment group] graders after one year boys and of program] girls Drug use [NS; ES=-.11 Commitment to school [significantly favors favoring treatment] treatment ] Times suspended [significantly favors treatment, ES=-.37] Battistich, Schaps, 3 Alcohol use NA Watson, & Solomon [significantly favors (1996) N=1479 - treatment, ES=-.12] 1745, [fifth and sixth depending on Marijuana use [NS] graders assessed after the year each year of a two-year Delinquency [10 items, program] NS] Solomon, Watson, 3 Negative behaviors Supportive and friendly behaviors Delucchi, Shaps, & observed in classrooms [significantly favors treatment] Battistich (1988) N=67 [NS] class-rooms Spontaneous pro-social behavior [Kindergarten through [significantly favors treatment] fourth grade classrooms assessed after each year of a five-year program] 

A second major classroom intervention (CDP, the Child Development Project)was conducted with several cohorts of elementary school students in 12elementary schools for 2 consecutive years beginning in 1992 (Battistichet al, 1996). It included the following components:

"Cooperative learning" activities intended to encourage studentdiscussion, comparison of ideas, and mutual challenging of ideas on academicand social topics;

A "values-rich" literature-based reading and language artsprogram intended to foster understanding of diversity;

"Developmental discipline," a positive approach to classroommanagement that stresses teaching appropriate behavior rather than punishment,involving students in classroom management, and helping them to learn behaviormanagement and conflict resolution skills;

"Community-building" activities aimed at increasing appreciationfor diversity or students' sense of communal involvement and responsibility;and

"Home-school" activities to foster parent involvement in theirchildren's education.

A similar program was conducted in three elementary schools for 5 consecutiveyears beginning in 1982 (Solomon et al, 1988). The evidence from evaluationsof these efforts is also summarized in Table 5-3. The program increasedpro-social behaviors but did not decrease negative behavior among studentsin grades K though 4. It had no effect on delinquency or marijuana use,but alcohol use among the treatment youths in grades 5 and 6 was significantlylower than among the control students (Battistich et al, 1996; ES=-.12).In this study, supplementary analyses which take into account varying levelsof implementation across schools showed that marijuana use and two of theten delinquency items were significantly lower among treatment youths inthe schools with the highest level of implementation, but these resultsare ambiguous because the high implementation schools also have strikinglyhigher levels of marijuana use and delinquency at all time-points. Regressionto the mean is not ruled out as an alternative explanation for the observedpattern of results.

In all but one study, classroom management strategies were combinedwith family-based strategies, making it impossible to determine the uniqueeffects of the classroom intervention. Program effects were not as positivein the one study that used only the classroom strategies. Both the CDPand Seattle projects found evidence of positive effects on substance useinitiation, but the effects were sometimes only marginally significantand were not as consistent across different substances and gender groupsas would be expected. Also, although these strategies appear effectivefor increasing positive behaviors and a number of protective factors, littlepromise for reducing delinquency is demonstrated. Classroom organizationand management strategies should be combined with other more potent componentsand tested more rigorously.

Regrouping Students

Four studies have examined interventions which group students to createmore supportive or challenging environments for high-risk youths. Felner,Ginter & Primavera (1982) and Felner & Adan (1988) studied theSchool Transitional Environment Project (STEP), a one-year program forstudents making the transition to high school. Incoming students were assignedto small "schools within the school" consisting of 65 to 100students. Students remained in intact small groups for their home roomperiod and their academic subjects, and these classrooms were physicallyclose together. The role of the home room teacher was redefined so as toinclude more responsibility for meeting the administrative, counseling,and guidance needs of the students. Reyes & Jason (1991) implementeda similar program which also contained an attendance monitoring component.D. Gottfredson (1990) studied another school-within-a-school intervention-- Student Training Through Urban Strategies (STATUS), one of the programsin OJJDP's alternative education initiative. This program grouped high-riskyouths to receive an integrated social studies and English program whichinvolved a law-related education curriculum and used instructional methodsemphasizing active student participation. Students stayed together fortwo hours each day. These studies are summarized in Table 5-4.

STEP increased protective factors (school attendance, persistence, andachievement) in the Felner studies, but its replication in Reyes &Jason was largely a failure. STATUS reduced delinquency and drug use (ESsrange from -.07 to -.42) and changed in the desired direction several riskand protective factors related to delinquency. STATUS involved innovativeteaching methods (many of which are reviewed in the classroom managementsection above), a law-related education curriculum, and the innovativeschool-within-a-school scheduling. It is not possible to disentangle theeffects of these components. However, the major intermediate outcome throughwhich the law-related education curriculum was expected to reduce delinquency-- belief in the validity of laws -- was the only outcome that did notfavor the treatment group. We have seen above that classroom managementstrategies alone or in combination with family interventions do not reducedelinquency. It is unlikely, therefore, that the positive effects foundin the STATUS program were due solely to the instructional and classroommanagement methods or to the law-related education curriculum. The studysuggests that the combination of innovative grouping and scheduling withthe other two components is promising.

Table 5-4. Summary of Studies using Reorganization of Grades or Classes

 Author Scientific Effect size for Effects on risk and protective factors (year) methods measure of problem score/ behavior Number of cases Felner, Ginter & 4 NA School dropout during three years following Primavera (1982); the program -- 43% of controls vs. 21% of Felner & Adan (1988) N=172 treatment dropped out (significant students difference) [results for ninth graders directly Absenteeism and grade-point-average -- following a one-year significantly favors treatment at end of one program, with follow-up year of treatment and at end of year one and three years following treatment following program] Gottfredson (1990) 4 Delinquency -- favors Negative peer influence, grades, and treatment group in both attachment to school -- significantly favor [results at end of N=123 junior schools, significant treatment group in both schools one-year program in one high and 124 for high school only junior and one senior senior high [ES's -.33 and -.42] School attendance -- favors treatment high school] students students in both schools, NS Drug Involvement -- significantly favors Number of months enrolled in school -- treatment group in both significantly favors treatment students, high schools [ES's -.42 and school only -.35] Belief in rules -- favors control students, Court contacts -- both schools, NS favors treatment group in both schools, NS Educational expectations -- favors treatment [ES's -.07 and -.18] students in high school, control students in junior high, NS. Reyes & Jason (1991) 4 NA Achievement test scores -- one of three tests significantly favors treatment [results for ninth N= 154 graders at end of Grade point average, absences, and dropout -- one-year program] At the end of one year of treatment, NS 

In summary, programs which group high-risk students to create smaller,more tightly-knit units for instruction show promise for reducing delinquency,drug use and drop-out. These programs are risky in light of other researchthat shows negative effects of grouping high-risk youths for peer counselingor other therapeutic services (to be reviewed shortly), but the studiessummarized in this section suggest that it may be beneficial to group high-riskfor instruction in the context of "schools-within-schools" whichoffer a strong academic program, use effective instruction and classroommanagement strategies, and supportive staff.

A note on alternative schools. Alternative schools for disruptiveyouths are often proposed as a solution to the problem of disorder in schools.OJJDP's alternative education initiative sponsored five such schools, allsmall schools for students who had not flourished in the regular schoolsetting. After reviewing the content of these programs, G. Gottfredson(1987) concluded that they are far too variable in nature, student composition,structure, and purpose to warrant any blanket statement about their effectiveness.He reviews two of the five models -- one based on a theory that intensepersonal involvement of the educators with the youth would reduce delinquencythrough increased bonding, and the other based on the theory that rigorousdiscipline and behavior modification techniques would result in decreaseddelinquency. The evaluation of the first program found remarkable improvementsin several risk factors for delinquency, including commitment to school,attachment to school, and belief in rules. It also found significantlyless self-reported drug use (but not self-reported delinquency or arrestrecords) among alternative school students than among controls. The evaluationof the second alternative school implied that the program was effectivefor increasing several measures of academic persistence, but that studentsliked school less and reported significantly more delinquent behavior thanthe comparison students. The varied models employed in alternative schoolssuggest that the question, "are alternative schools effective?"is too simplistic. The components of the interventions involved in alternativeschools must be disentangled in future evaluations.

Individual-Change Strategies

Strategies that aim to alter students' delinquent behavior or theirknowledge, skills, beliefs, behaviors or attitudes directly related todelinquent behavior are summarized below. These strategies include instructionwith specific content related to delinquency or drug use; methods aimedat changing thinking strategies (cognitive or cognitive-behavioral training);behavior modification; peer counseling, mediation, and leaders; other counseling;mentoring; and "alternatives" programs which provide opportunitiesfor recreation, enrichment or leisure.

Instructing Students

The most common school-based prevention strategy is instruction. Mostschools provide instruction aimed at reducing drug use or delinquency,often in the form of the programs like Drug Abuse Resistance Education(D.A.R.E.), Law-related Education (L.R.E.), and Gang Resistance Educationand Training (G.R.E.A.T.), which enjoy substantial federal subsidy. Thecontent of interventions that provide instruction to students is varied.The box at the right shows some of the topics covered in instructionalprograms.

Topics Covered in Instructional Programs General health or safety;
Alcohol, tobacco, and other drugs: information about and consequences ofuse;
Violence prevention;
Character/moral development;
Law;
Recognizing and resisting social influences to engage in misbehavior andrisky situations, being assertive;
Identifying problem situations, generating alternative solutions, evaluatingconsequences;
Setting personal goals, self-monitoring, self-reinforcement, self-punishment;
Attributing the cause of events or circumstances to ones own behavior;
Interpreting and processing social cues, understanding non-verbal communication,negotiating, managing anger, controlling stress, anticipating the perspectivesor reactions of others.

The following pages summarize what is known about the effectivenessof drug education, broader social competency development curricula, violenceprevention curricula, and law-related education. The first two of thesecurriculum types have been studied extensively and several excellent secondaryreviews are available. These secondary reviews will be summarized and onlythe most rigorous studies will be singled out for discussion. Instructionalprograms funded by OJP (D.A.R.E. and Law-related education) and a gangprevention program recently evaluated with N.I.J. funding (G.R.E.A.T.)will also be summarized here.

Alcohol and other drug education. Several meta-analyses and reviewsof the effectiveness of school-based drug prevention instruction have beenconducted (Botvin, 1990; Botvin et al, 1995; Dryfoos, 1990; Durlak, 1995;Hansen, 1992; Hawkins, Arthur, & Catalano, 1995; Institute of Medicine,1994; Tobler, 1986, 1992). Botvin (1990) traces the historical developmentof these programs. He shows that "information dissemination"approaches which teach primarily about drugs and their effects, "feararousal" approaches that emphasize the risks associated with tobacco,alcohol, or drug use, "moral appeal" approaches which teach studentsabout the evils of use, and "affective education" programs whichfocus on building self-esteem, responsible decision-making, and interpersonalgrowth are largely ineffective for reducing substance use. On thecontrary, approaches which include resistance-skills training to teachstudents about social influences to engage in substance use and specificskills for effectively resisting these pressures alone or in combinationwith broader-based life-skills training do reduce substance use. The boxto the right shows the typical content of these instructional programs.Curricula which focus on general life-skills are typically longer thanthose which focus only on social resistance skills.

Typical Content of Social Influence and Life-Skills InstructionComponentsof Social Resistance Skills Instruction:
Increasing student awareness of the social influences promoting substanceuse
Teaching skills for resisting social influences from peers and the media
Correcting normative expectations concerning the use of substances

Additional Skills Targeted in Life-Skills Instruction:
Problem-solving and decision-making
Self-control or self-esteem
Adaptive coping strategies for relieving stress or anxiety
Interpersonal skills
Assertiveness

This section summarizes substance abuse curricula having an emphasison social competency skill development. Two such school-based instructionalprevention programs which have been scrutinized using rigorous methodsare ALERT (Ellickson & Bell, 1990, Ellickson, Bell, & McGuigan,1993) and Life Skills Training (L.S.T., Botvin & Eng, 1982; Botvin,Baker, Botvin et al, 1984; Botvin, Baker, Renick et al, 1984; Botvin, Batsonet al, 1989). ALERT is essentially a social resistance-skill curriculumconsisting of eight lessons taught a week apart in the seventh grade, followedby three eighth grade "booster" lessons. L.S.T. is a more comprehensiveprogram focusing on resistance skills training as well as the general lifeskills mentioned above. This program consists of 16-sessions deliveredto seventh grade students followed by eight session "boosters"in grades eight and nine. This section ends with a discussion of D.A.R.E.,an OJP-funded substance abuse prevention program whose content is not asfocused on social competency development as the other programs summarized.

The ALERT study (scientific methods score=5) was a multi-site experimentinvolving the entire seventh grade cohort of 30 junior high schools drawnfrom eight urban, suburban, and rural communities in California and Oregon.These 30 schools were randomly assigned to treatment and control conditions.Results are reported using individuals as the unit of analysis, althoughthe investigators reported that results from school-level analyses supportedthe same conclusions with more positive results. Program effects were assesseddirectly after the seventh grade programs as well as before and directlyafter the eighth grade booster. Students were followed up again when theywere in 9th, 10th, and 12th grades. The program had positive effects forboth low- and high-risk students and was equally effective in schools withhigh and low minority enrollment. The program's most consistent effectswere found for marijuana use. It reduced the use of marijuana among studentsat each risk level, with the strongest effects for the lowest risk group:those students who had not initiated either cigarette or marijuana useat the time of the baseline measurement. In this group, 8.3% of the ALERTstudents compared with 12.1% of the control students (ES=-.08) had initiatedmarijuana use by the end of the eighth grade booster. Small but statisticallysignificant positive effects on the amount of marijuana used were observedfor the other risk groups directly after the seventh grade sessions, butthese effects were no longer statistically significant (and were not practicallymeaningful) by the end of the booster session. For all groups, small positiveprogram effects were initially observed for alcohol use, but they too erodedby grade 8. The follow-up studies showed that once the lessons stop, sodid the program's effects on drug use. Although some effects on cognitiverisk factors persisted through grade 10, they were not sufficient to producereductions in drug or alcohol use.

L.S.T. has also undergone rigorous test in an ongoing series of studiesfirst published in 1980, conducted by Botvin and his colleagues. The morerecent studies examined the effect of the program on alcohol and marijuanause (in addition to cigarette use) and tracked long-term program effects.Botvin, Baker, Renick, Filazzola & Botvin (1984; scientific methodsscore=3) examined the effectiveness of a 20-session course delivered to7th graders from 10 suburban New York junior high schools. The subjectswere primarily white, from middle-class families. Schools were randomlyassigned to receive the program as implemented by older students, by regularclassroom teachers, or to serve as controls. All analyses were reportedusing individuals as the unit of analysis. Results measured immediatelyafter the program showed that program students compared with control studentswere significantly less likely to report using marijuana (ES=-.10) andengage in excessive drinking, but these positive effects were found onlyfor the peer-led condition. Botvin, Baker, Filazzola & Botvin (1990;scientific methods score=4) reported on the one-year follow-up of thisstudy. This study contrasts not only the teacher- and peer-led conditions,but also the presence or absence of a 10-session booster course deliveredduring eighth grade. As with the ALERT study, the results showed that theeffects of the program diminished without the booster. In the peer-ledcondition with the booster session, significant effects were maintainedat the end of the eighth grade on the amount of alcohol used and marijuanause (ESs ranged from .04 for used in last day to .16 for used in last month).Again, positive effects were found only for the peer-led condition.

In a larger study involving 56 public schools, the same 20-session 7thgrade program, 10-session booster session in eighth grade, and an additional5-session booster in the ninth grade was studied for long term effectson substance use at 12th grade (Botvin, Baker, Dusenbury, Botvin, &Diaz, 1995; scientific methods score=5). In this study, the 56 schools(serving mainly white, middle-class populations) were stratified accordingto baseline levels of cigarette smoking and geographic location and randomlyassigned to experimental conditions. All results were reported using individualstudents as the level of analysis. This study involved only teacher-ledclassrooms. The 12th grade results for the full sample of 3,597 subjectsrevealed significant positive effects on the prevalence of drunkenness(ESs range from -.08 to -.10), but not for other measures of alcohol use.Significant effects were not reported for marijuana use, although the effectsize for the prevalence of weekly marijuana use is as large (-.09) as theeffects sizes for the significant effects on excessive drinking. The lowerbase rate for marijuana use reduces the likelihood of finding statisticallysignificant results for this outcome. When only subjects who received areasonably complete version of the program were examined, the results weremore positive. Additional research (Botvin, Batson, Witts-Vitale, Bess,Baker, & Dusenbury, 1989; Botvin, Dusenbury, James-Ortiz, & Kerner,1989) showed that the positive effects generalize to African American andHispanic American populations.

D.A.R.E., developed in 1983 by the Los Angeles Police Department andthe Los Angeles Unified School District, is the most frequently used substanceabuse education curriculum in the United States. According to D.A.R.E.America (Law Enforcement News, 1996), the program is now used by seventypercent of the nation's school districts and will reach 25 million studentsin 1996. About 25,000 police officers are trained to teach D.A.R.E. Itis also popular in other countries, forty-four of which have D.A.R.E. programs.The complete array of D.A.R.E. activities currently on the market includes"visitation" lessons in which police officers visit studentsin kindergarten through fourth grade for brief lessons on topics such asobeying laws, personal safety, and the helpful and harmful uses of medicinesand drugs; a 17-week core curriculum for fifth or sixth graders (to bedescribed shortly); a 10-week junior high school program focusing on resistingpeer pressure, making choices, managing feelings of anger and aggression,and resolving conflicts; and a 10-week senior high program (co-taught withthe teacher) on making choices and managing anger. In addition, D.A.R.E.offers an after-school program for middle-school-aged students, calledD.A.R.E. + PLUS (Play and Learn Under Supervision). This provides a varietyof fun activities for students during the after-school hours. Programsfor parents and special education populations are also available.

The core 17-lesson curriculum delivered to students in grades 5 or 6has always been the most frequently used form of the program. The greatmajority (81%) of school districts with D.A.R.E. implement the core curriculum,while 33% use the visitations, 22% the junior high, 6% the senior high,and 5% the parent curriculum (Ringwalt et al, 1994). The core curriculumis the only part of the program that had undergone rigorous outcome evaluation.

The core D.A.R.E. program is taught by a uniformed law enforcement officer.The original 17-lesson core curriculum focuses on teaching pupils the skillsneeded to recognize and resist social pressures to use drugs. It also containslessons about drugs and their consequences, decision-making skills, self-esteem,and alternatives to drugs. Teaching techniques include lectures, groupdiscussions, question and answer sessions, audiovisual materials, workbookexercises, and role-playing. The curriculum was revised in 1993 to substitutea lesson on conflict resolution and anger management skills for one onbuilding support systems.

Several evaluations of the original 17-lesson core have been conducted.7Many of these are summarized in a meta-analysis of D.A.R.E.'s short-termeffects (Ringwalt et al, 1994), sponsored by NIJ. This study located 18evaluations of D.A.R.E.'s core curriculum, of which 8 met the methodologicalcriterion standards for inclusion in the study. The study found:

  1. Short-term effects on drug use are, except for tobacco use, nonsignificant.
  2. The sizes of the effects on drug use are slight. Effect sizes average.06 for drug use and never exceed .11 in any study. The effects on knownrisk factors for substance use targeted by the program are also small:.11 for attitudes about drug use and .19 for social skills.
  3. Certain other programs targeting the same age group as D.A.R.E. --upper elementary pupils -- are more effective than D.A.R.E. "Interactive"programs which emphasize social skill development and social competenciesand use interactive teaching strategies have effect sizes for increasingsocial skills, reducing attitudes favorable to use, and reducing drug useat least three times as large as D.A.R.E. Other programs which emphasizeknowledge about drugs and affective outcomes (such as self-esteem) andare primarily delivered by an expert are no more effective than D.A.R.E.Note, however, that even the more effective programs show only small effectsizes (ES=.18) for reducing drug use.

Four more recent reports, three of them longitudinal, have also failedto find positive effects for D.A.R.E. Lindstrom (1996), in a reasonablyrigorous study (scientific methods score= 3) of approximately 1,800 studentsin Sweden, found no significant differences on measures of delinquency,substance use, or attitudes favoring substance use between students whodid and did not receive the D.A.R.E. program. Sigler & Talley (1995)(scientific methods score= 2) found no difference in the substance useof seventh grade students in Los Alamos, New Mexico who had and had notreceived the D.A.R.E. program 11 months before. Rosenbaum, Flewelling,Bailey, Ringwalt, & Wilkinson (1994; scientific methods score= 4) reporton a study in which twelve pairs of schools (involving nearly 1,600 students)were randomly assigned to receive or not receive D.A.R.E. Although somepositive effects of the program were observed immediately following theprogram, by the next school year no statistically significant differencesbetween the D.A.R.E. and non-D.A.R.E. students were evident on measuresof the use of cigarettes or alcohol. Also, only one of thirteen interveningvariables targeted by the program showed a positive effect. Clayton, Cattarello,and Johnstone (1996; scientific methods score= 4) reported on long-termeffects for D.A.R.E. Thirty-one schools were randomly assigned to receiveor not receive D.A.R.E. All students in the sixth grades in these schoolswere pre-tested prior to the program, post-tested shortly after the program,and resurveyed each subsequent year through the 10th grade. Although positiveeffects were observed during the seventh grade on some risk factors forsubstance use, no significant differences were observed between the D.A.R.E.and control schools on measures of cigarette, alcohol, or marijuana useeither during seventh grade or at any later point. These studies and recentmedia reports have criticized D.A.R.E. for (a) focusing too little on socialcompetency skill development and too much on affective outcomes and drugknowledge; (b) relying on lecture and discussion format rather than moreinteractive teaching methods; and (c) using uniformed police officers whoare relatively inexperienced teachers and may have less rapport with thestudents.

To the untrained eye, the content and methods used in D.A.R.E. are notstrikingly different from those used in the more effective programs suchas Life Skills Training (summarized above) and Social Problem Solving (summarizedbelow). But more subtle differences exist: L.S.T. and S.P.S. provide broaderand deeper coverage of and more practice for students in the developmentof social competency skills. For example, while all three programs containlessons on identifying social influences to use drugs and problem-solving,the non-D.A.R.E. programs provide more lessons on these topics and alsoinclude lessons on communication skills or emotional perspective taking.Weissberg's S.P.S. program is able to address self-control skills in greaterdepth because it completely omits lessons on self-esteem and factual informationabout drugs. The instructional methods are also different: L.S.T. and S.P.S.were carefully designed to make use of cognitive-behavioral methods includingfrequent role-playing, rehearsal of skills, and behavioral modeling.These methods are main features of the programs. D.A.R.E., even with theaddition of more "interactive" techniques, lacks a major emphasison the use of these carefully developed, research-based teaching techniques.

Although the content and method differences described above probablyaccount for some of the discrepancy between the effects found for the differenttypes of instructional programs, the largest difference among the programsis D.A.R.E.'s use of uniformed officers to deliver the program, a featurethat remains in the revised D.A.R.E. and whose effects on the efficacyof the program are unknown.

D.A.R.E. proponents challenge the results of the scientific D.A.R.E.evaluations. Officials of D.A.R.E. America are often quoted as saying thatthe ample public support for the program is a better indicator of its utilitythan scientific studies. They criticize D.A.R.E. studies for (a) lookingonly at the original D.A.R.E. model; (b) focusing on the absence of effectson alcohol and drug use among fifth and sixth graders when the base ratesare so low that effects would naturally be difficult to detect; and (c)failing to study the longer term effects of D.A.R.E. which are expectedto be more substantial. Each of these points is addressed below.

In 1993, D.A.R.E. added more coverage of social competency skills andmore interactive teaching techniques to its core curriculum (Ringwalt etal, 1994). These changes were expected to bring the program more in linewith the competition. No outcome evaluation of this revised curriculumhas been reported, but it appears unlikely that the revision will changethe results much because the largest difference between the earlier andrevised program is the substitution of a single lesson on reducing violencefor one on building support systems. Ringwalt et al. (1994) show that evenin the revised core curriculum for D.A.R.E., only 9 of the 17 lessons coversocial skill development.

D.A.R.E. is indeed atypical in its focus on elementary school-aged youths.As Hansen (1992) demonstrated, the percentage of fifth graders estimatedto have used tobacco, alcohol or marijuana in the past month ranges betweenabout 1 and 8 percent nationally. While lifetime use estimates (the outcomemeasure often used in D.A.R.E. evaluations) are certainly higher, the relativelylow prevalence rates mean that larger samples may be required in studiesof D.A.R.E. than in studies of programs targeting slightly older students.But D.A.R.E. evaluations can not be summarily dismissed on the basis ofthese criticisms because some have involved samples whose base rates forsubstance use are much higher than the national average and others haveinvolved samples with sufficient power to detect meaningful differenceseven in low-base-rate populations. For example, the Rosenbaum et al. (1994)study involved nearly 1,600 students in a sample whose base rate for lifetimealcohol use was 55%. Half of the studies summarized in the Ringwalt etal. (1994) study had sample sizes larger than 1,000, and none could bedescribed as small-sample research. Also, the Ringwalt et al. (1994) meta-analysisrelied not only on statistical significance tests, which are misleadingwhen the number of cases is not sufficiently large to detect the expectedeffect, but also on effect sizes to assess the magnitude of the effectsregardless of statistical significance. Inferences based on effect sizesare not as prone to misinterpretation as those based on significance levels.

D.A.R.E. proponents also argue that D.A.R.E.'s effects are delayed --i.e., that effects appear when students reach higher grades. The threerecent longer-term evaluations of D.A.R.E. (Clayton, Cattarello, &Johnstone, 1996; Sigler & Talley, 1995; Rosenbaum, Flewelling, Bailey,Ringwalt, & Wilkinson, 1994; summarized above) do not support thiscontention. The absence of long-term effects is not surprising given themore general finding that effects for instructional substance use preventionprograms decay rather than increase over time in the absence of continuedinstruction.

In summary, using the criteria adopted for this report, D.A.R.E. doesnot work to reduce substance use. The programs's content, teaching methods,and use of uniformed police officers rather than teachers might each explainits weak evaluations. No scientific evidence suggests that the D.A.R.E.core curriculum, as originally designed or revised in 1993, will reducesubstance use in the absence of continued instruction more focused on socialcompetency development. Any consideration of the D.A.R.E.'s potential asa drug prevention strategy should place D.A.R.E. in the context of instructionalstrategies in general. No instructional program is likely to have a dramaticeffect on substance use. Estimates of the effect sizes of even the strongestof these programs are typically in the mid- to high-teens. D.A.R.E.'s meagereffects place it at the bottom of the distribution of effect sizes, butnone of the effects are large enough to justify their use as the centerpieceof a drug prevention strategy. Rather, such programs should be embeddedwithin more comprehensive programs using the additional strategies identifiedelsewhere in this chapter.

Broader social competency development curricula. Other curriculafocus specifically on social competency development, without an emphasison substance abuse prevention per se. Weissberg's social competence promotionprogram, for example, covers the entire array of social competency skillswithout tying them directly to any specific problem behavior. Problem-specificmodules aimed at preventing anti-social and aggressive behavior, substanceuse, and high-risk sexual behavior are available. The program ranges inlength from 16- to 29-sessions, depending on the version.

Caplan, Weissberg, Grober, Sivo, Grady, & Jacoby (1992; scientificmethods score=4) studied the effect of a 20-session version of Weissberg'ssocial competence promotion program aimed at stress management, self-esteem,problem-solving, substances and health information, assertiveness and socialnetworks on 282 sixth and seventh graders in an inner-city and a suburbanmiddle school in Connecticut. Classrooms were randomly assigned to receivethe program or not. Results were reported using individuals as the unitof analysis. Students in program classes improved relative to studentsin the control classrooms on measures of problem-solving ability and stressmanagement. Teacher ratings of the participating students improved relativeto the controls on measures of conflict resolution with peers and impulsecontrol, both important protective factors for later delinquency, and popularity.Students' self-reports of their behavioral conduct were not affected bythe program, and effects on self-reports of intentions to drink alcoholand use drugs were mixed. No significant difference was found for a self-reportmeasure of frequency of cigarette, alcohol, and marijuana use, but programstudents reported significantly less excessive drinking than controls (ESsrange from .26 to .32). The program was as effective for students in theinner-city and the suburban schools. The sample size in this study waslikely too small to detect as statistically significant any small differencesbetween the treatment and comparison students.

In another study involving 447 students from 20 classes in four urban,multi-ethnic schools, Weissberg & Caplan (1994; scientific methodsscore=4) evaluated a similar 16-session social competence promotion programfor students in grades five through eight. This version of the programdid not include lessons on substance use. It focused on teaching students:

impulse-control and stress-management skills,

thinking skills for identifying problem situations and associated feelings,

establishing positive pro-social goals,

generating alternative solutions to social problems, anticipating thelikely consequences of different actions, choosing the best course of action,and successfully enacting the solution.

Random assignment to treatment and control conditions was not accomplishedin this study. Program students improved more than controls on problem-solvingabilities and pro-social attitudes towards conflict resolution. Teacherratings indicated that the training improved impulse control, problem-solving,and academic motivation and decreased teasing of peers, important riskand protective factors for later delinquency. Self-reported delinquencyof a relatively minor form (stealing, starting fights, vandalism, skippingschool, etc.) also increased less for the program participants (2.8% increase)than for comparison students (36.8% increase) between the beginning andthe end of the program. No significant effects were observed for self-reportsof substance abuse in this study. Weissberg & Greenberg (in press)summarize another study which shows that the positive effects of the programare maintained in the year after the program only when the training iscontinued into the second year

.

Greenberg, Kusche, Cook & Quamma (1995; scientific methods score= 4) report on the PATHS (Promoting Alternative Thinking Strategies) curriculumon emotional competence for elementary school-aged children. This projectused a 60-lesson version of the curriculum composed of units on self-control,emotions, and problem-solving. Lessons were sequenced according to increasingdevelopmental difficulty and included didactic instruction, role-playing,class discussion, modeling by teachers and peers, social and self-reinforcement,and worksheets. Extensive generalization techniques were included to assistteachers in applying skills to other aspects of the school day. Specifically,the curriculum included:

A Feelings and Relationships Unit -- 35 lessons on emotional and interpersonalunderstanding. The lessons cover approximately 35 different affective statesand were taught in a developmental hierarchy beginning with basic emotions(e.g., happy, sad, angry) and proceeding to more complex emotional states(e.g., jealous, guilty, proud).

Self-control and initial problem-solving -- The development of self-control,affective awareness and communication, and beginning problem-solving skillswere integrated during the Feelings Unit with the introduction of the ControlSignals Poster (CSP), which had a red light to signal "Stop - CalmDown," a yellow light for "Go Slow - Think," a green lightto signal "Go - Try My Plan," and at the bottom, the words "Evaluate- How Did My Plan Work?" In a series of lessons, the children weretaught skills to use with the different signals of the poster. For purposesof generalization, a copy of the CSP was placed in the classroom and teacherswere coached on how to use this model for active problem-solving duringthe classroom day.

Interpersonal Cognitive Problem-Solving -- 20 to 30 lessons sequentiallycovering eleven problem-solving steps, similar to those discussed aboveas part of Weissberg's program above.

Generalization procedures -- A variety of generalization techniqueswere included throughout the curriculum to foster transfer of the skillsand ideas taught.

The intervention teachers attended a 3-day training workshop and receivedweekly consultation and observation from project staff. The PATHS lessonswere taught approximately

3 times per week, with each lesson lasting 20-30 minutes. The weeklyconsultations were intended to enhance the quality of implementation throughmodeling, coaching, and providing ongoing feedback regarding program delivery.

The social competency promotion intervention was field-tested in Washingtonstate using random assignment of schools serving "regular education"students to treatment and control conditions as well as random assignmentof classrooms of "special needs" children (in different schoolthan the regular education students) to treatment and control conditions.In all, 286 students participated in the study. Students were in the firstand second grades at the time of the pre-test, and in the 2nd and 3rd gradesat the time of the first post-test, which occurred approximately one monthafter the end of the intervention. Two additional follow-up assessmentswere conducted to examine maintenance of effects one and two years afterthe intervention.

Immediate positive effects of the program were observed for both regularand special education students on measures of the specific social competencyskills targeted. Greenberg (1996) reports on the longer-term effects ofthe program. At the final follow-up, significant differences favoring theregular education treatment students emerged on teacher ratings of externalizingbehaviors, a measure of serious conduct problems highly related to laterdelinquent behavior. Intervention students in both groups also self-reportedsignificantly lower rates of conduct problems at the later follow-up points.

Violence-prevention instruction. Brewer, Hawkins, Catalano, &Neckerman (1995) provide a comprehensive summary of conflict resolutionand violence prevention curricula. These instructional programs are designedto improve students' social, problem-solving, and anger management skills,promote beliefs favorable to nonviolence, and increase knowledge aboutconflict and violence. Brewer et al. (1995) summarize evaluations of eightviolence prevention curricula. Target populations for these programs rangefrom pre-K through grade 10. The quality of the evaluations of these programsis uniformly poor. No study used random assignment of subjects to treatmentand comparison conditions. Only four of the studies assessed program effectson aggressive or violent behavior, and two of these studies suffered fromserious methodological flaws. The other two studies reported positive resultson measures of aggressive behavior, but no corresponding positive changeson attitudes towards violence.

Perhaps the most rigorous evaluation is for the Washington (DC) CommunityViolence Prevention Program (Gainer, Webster, & Champion, 1993; scientificmethods score=3), a 15-session curriculum focusing on social informationprocessing deficits and belief systems associated with aggressive behavior,modeled after the Viewpoints program that had received positiveevaluations in a correctional institutional setting (Guerra & Slaby,1990). The program was evaluated with 5th and 7th graders in three inner-cityschools. Students receiving the course were compared with students fromthe same schools and grade levels during the following year. Program effectson violent behavior were not assessed, and effects on social problem solvingskills and attitudes about violence were mixed. Some measures showed significantlypositive effects, some significantly negative effects, and some no difference.

Gang Resistance Education And Training (G.R.E.A.T.) was developed in1991 by the Phoenix Police Department to reduce adolescent involvementin criminal behavior and gangs. Although not specifically designed as aviolence prevention program, its emphasis on gang membership, a major correlateof violent crime, justifies its inclusion here. The Bureau of Alcohol,Tobacco, and Forearms has funded officer training for this program, andas of July, 1996, more than 2,000 officers from 47 states and the Districtof Columbia had completed training. In 1994, NIJ began funding an evaluationof G.R.E.A.T.. It currently supports a three-year study to assess the short-and long-term effects of the program on students in six sites. A less rigorouspreliminary assessment of effects one-year following the program in 11cities was also recently completed with NIJ funding. Results from thispreliminary study are summarized below.

G.R.E.A.T. is a brief (9-week) instructional program taught to middleschool students by trained, uniformed law enforcement officers. The programteaches students about the impact of crime on its victims and the community;cultural differences; conflict resolution skills; how to meet basic needswithout joining a gang; and responsibility to the school and neighborhood.The program ends with a lesson in which students are taught the importanceof goal-setting. The G.R.E.A.T. program differs from instructional programsknown to be effective for reducing drug use or delinquency by being (a)less intensive; (b) almost entirely devoid of content and methods focusingon teaching students social competency skills; and (c) lacking follow-upsessions. It is taught by uniformed law enforcement officers -- a featurewhose costs and benefits as a crime prevention strategy are unknown.

The preliminary evaluation of the program (Esbensen & Osgood, 1996)compared the survey responses of approximately 2,600 eighth grade studentswho said they had completed G.R.E.A.T. with those of approximately 3,200eighth students who said they had not. The investigators attempted to shoreup the weak evaluation design (post-test only for non-equivalent treatmentand comparison groups) by statistically controlling for differences betweenschools and demographic characteristics of participants and non-participants,but the scientific methods score of the study remains only a 2 on our 5-point-scale.The study found several statistically reliable differences favoring theG.R.E.A.T. participants, including less delinquency (ES=-.07) and druguse (ES=-.04). Nineteen of the thirty-one outcomes examined significantlyfavored the G.R.E.A.T. participants, and none significantly favored thenon-participants. The investigators cautioned that the magnitudes of theeffects were very small and the design of this preliminary study is tooweak to warrant confident conclusions about the effects of the program.The effect sizes for the significant delinquency and drug use outcomesare all less than .10 (e.g., the difference between the participants andnon-participants on outcome measures is less than one-tenth of one standarddeviation), suggesting that even if the effects could be safely attributedto the program they are small. Such small differences between groups areoften detected as statistically significant in large studies. For thisreason, the effect size is a more meaningful indicator of program effects.

Law-related education (L.R.E.). Schools have implemented law-relatededucation curricula for nearly three decades. These curricula are designedto familiarize youths with the country's laws, develop appreciation ofthe legal process, encourage responsible political participation, developmoral and ethical values, and develop analytical skills. Lack of knowledgeabout the law, citizenship skills, and positive attitudes about the lawand the role of the government are cited in L.R.E. materials as causesof juvenile crime.

In 1979, the justice department's National Institute for Juvenile Justiceand Delinquency Prevention (NIJJDP, OJJDP's research arm) funded five organizationsto develop and demonstrate L.R.E. methods. An evaluation of these efforts,also funded by NIJJDP, examined the effects of the program on delinquencyand factors related to delinquency. Most of the results of this evaluationare summarized in Johnson & Hunter (1985). The evaluation included61 L.R.E. elementary, junior, and senior high classrooms and 44 comparisonclassrooms in 32 schools in 6 states. The results for 1981, the first yearof the evaluation, were regarded as formative. It showed that L.R.E. didnot always produce positive effects, and that the quality of implementationwas correlated with the amount of positive change from pre- to post-teston many measures. Results for the second year of the evaluation (1982)were more positive, but the effects were, according to the authors, "severelydiminished" except in one site in Colorado site in which generallypositive outcomes were observed. The strongest program implementation occurredin 1983. Johnson & Hunter (1985) summarize the results comparing outcomes,separately by teacher, for students in 21 L.R.E. classes and 14 comparisonclasses (most of which were non-randomly assigned). Out of 132 effectsreported for the 11 delinquency items, 15 showed a significant effect (13would have been expected by chance using the one-tailed test of significancereported). Nine of these differences favored the L.R.E. students, and sixfavored the comparison students. Significant program effects on attitudestowards deviance and violence favored the comparison students. Many positiveeffects were found for outcomes measuring knowledge about the law and legalpractices and other outcomes that might be expected from improved classroommanagement techniques (such as reduced "clock watching").

Johnson (1984) focused on the nine L.R.E. classes in the site for whichrandomization to treatment and control conditions was obtained. He showedthat the nine L.R.E. classes fared significantly better than the two controlclasses on more than half of the forty-one possible measures. Three ofthe eleven items measuring delinquency were reported as significantly favoringthe L.R.E. group. The effect sizes for all eleven items ranged from 0 (forviolence against other students) to .66 (for school rule infractions suchas cheating on tests and skipping school). The average effects size forthe eleven delinquency items was .22.

In summary, these evaluation activities from the early 1980s showedclear program effects on law-related factual knowledge. Effects on otheroutcomes were minimal. In one particularly strong site, consistent positiveeffects were observed on certain risk factors for delinquency (e.g. attachmentto school and attitudes towards violence and deviance), but not others(e.g., association with delinquent peers) and small positive effects werefound on certain measures of delinquency but not others.

This extensive national evaluation produced no bottom line. The partof the evaluation focusing on the entire national sample was the weakestmethodologically (scientific methods score= 3) and showed no reason foroptimism about L.R.E.'s effect on delinquency. The "sub-study"of Colorado sites was stronger methodologically, and more positive outcomeswere observed. What is not clear, however, is the extent to which resultsfor these "well-implemented" schools can be generalized to otherschools implementing L.R.E. programs. Because the L.R.E. intervention atthis site included a large dose of general instructional and classroommanagement training for teachers in addition to law-related activitiesit is not possible to rule out the possibility that any positive effectsof the program are due to these general techniques rather than to the law-relatedcontent of the curriculum. Because L.R.E. programs are not necessarilyaugmented with these additional strategies, it is not clear that the positiveevaluations are relevant to understanding the effects of typical L.R.E.programs.8

Law-related education curricula, like other forms of instruction, willprobably not reduce delinquency significantly when used in isolation. TheL.R.E. program evaluators found that when the program is embedded in amore comprehensive program of improved classroom organization and managementprocesses, the outcomes are better. Gottfredson (1990) also found thatwhen an L.R.E. curriculum was enriched with state-of-the-art classroominstructional and organization methods and implemented in the contextof a school-within-a-school model, it reduced delinquency. More work isnow required to isolate the working parts of these multi-component programsinvolving L.R.E..

Statements found in materials published by the organizations that continueto develop and disseminate L.R.E. using OJJDP funding -- "Researchindicates that properly implemented law-related education changes attitudesand reduces crime" (National Institute for Citizenship Education inthe Law, 1988) -- are at best misleading because they ignore the resultsobtained for most of the sites in the national study. More rigorous evaluationis needed.

Summary. Certain instructional programs to reduce drug use haveproduced consistent evidence of positive effects on substance use in rigorousstudies, and others have consistently shown no effects. "Informationdissemination" instructional programs which teach primarily aboutdrugs and their effects, "fear arousal" approaches that emphasizethe risks associated with tobacco, alcohol, or drug use, "moral appeal"approaches which teach students about the evils of use, and "affectiveeducation" programs which focus on building self-esteem, responsibledecision-making, and interpersonal growth are largely ineffectivefor reducing substance use. D.A.R.E. as it is most commonly implementedis largely ineffective for reducing substance use. Approaches which includeresistance-skills training to teach students about social influences toengage in substance use and specific skills for effectively resisting thesepressures alone or in combination with broader-based life-skills trainingdo reduce substance use. But the effects of even these programs are smalland short-lived in the absence of continued instruction. Hansen and O'Malley(1996) report average effect sizes for social influence training programssuch as ALERT ranging from .14 to .27 (on alcohol, marijuana, and cigaretteuse), but Gorman (1995) shows these programs have little or no effect ondrinking behavior. More comprehensive programs such as L.S.T. and Weissberg'sSocial-Problem Solving have effect sizes ranging from .08 to .37.

More comprehensive social competency promotion programs work betterthan programs which do not focus on social competencies and those thatfocus more narrowly on resistance skill training. Also, the more extensivethe reliance on cognitive-behavioral training methods such as feedback,reinforcement, and behavioral rehearsal (as in the Greenberg and Weissbergprograms) rather than traditional lecture and discussion, the more effectivethe program. The Weissberg and Greenberg works are also important becausethey demonstrate that social competency promotion programs works for reducingdelinquency or early conduct disorder leading to delinquency as well asdrug use.

Some violence prevention programs teach interpersonal skills and behaviorssuch as communicating, making eye-contact, cooperating, and sharing. Othersuse the same cognitive-behavioral strategies used in the most effectivesocial competency promotion programs summarized above. These programs seemplausible, but until they are rigorously evaluated they should be usedwith caution. Just as the first-generation substance abuse prevention programswere found to increase rather than decrease drug use (Botvin, 1990), somight these early violence prevention efforts increase violence. Althoughdescribed by some as "promising," the G.R.E.A.T. program doesnot meet the criteria necessary to earn this descriptor in our review.Until the outcome of the more rigorous evaluation now underway is complete,the effects of the program remain unknown.

The effects of law-related education curricula as typically implementedalso remain unknown. Evaluations have supported their effectiveness whenimplemented as part of a more comprehensive program, but it is not clearto what extent the law-related curriculum contributes to the effectiveness,if at all. Rigorous research is needed.

Modifying Behavior and Teaching Thinking Skills

Behavior modification interventions focus directly on changing behaviorsby rewarding desired behavior and punishing undesired behavior. Severalwell-known programs for delinquent youths (e.g., Achievement Place) relyon these methods, as do many educational programs -- especially those servingspecial education populations. Many programs for delinquent and "at-risk"populations also attempt to alter thinking skills. These "cognitive-behavioraltraining" interventions are based on a substantial body of researchindicating that delinquents are deficient in a number of thinking skillsnecessary for social adaptation. Delinquents often do not think beforethey act, believe that what happens to them is due to fate or chance ratherthan to their own actions, misinterpret social cues, fail to consider alternativesolutions to problems, and lack interpersonal skills necessary for effectivecommunication. Programs often combine behavioral and cognitive methodsin an attempt to alter immediate behavior and promote the generalizationof behavior change to other settings.

As indicated above, instructional programs that teach social competencyskills and rely on cognitive-behavioral methods such as feedback, reinforcement,and behavioral rehearsal are the most effective for reducing substanceuse in general populations. Meta-analyses (Garrett, 1985; Izzo & Ross,1990; Lipsey, 1992) have also concluded that the most effective delinquencyprevention and treatment programs incorporate strategies aimed at developingsocial skills and using cognitive-behavioral strategies. Forman (1980;scientific methods score=4) showed that both cognitive training and behavioralinterventions decrease aggressive behavior in elementary school children,although the behavioral intervention decreased disruptive behavior to asomewhat greater extent.

The programs reviewed below incorporate many of the same principlesfound in the more effective instructional programs. These programs differin that they are often targeted at students identified as at especiallyhigh-risk for engaging in delinquent activities, are delivered in smallgroups or individually, and provide more intensive intervention than ispossible with classroom-based instructional programs. Only three of themany high-quality studies of interventions using behavioral and cognitive-behavioralmethods are reviewed here.

Elements of Lochman's Anger-Coping Intervention Establishing group rulesand contingent reinforcements;
Using self-statements to inhibit impulsive behavior;
Identifying problems and social perspective-taking;
Generating alternative solutions and considering the consequences to socialproblems;
Modeling videotapes of children becoming aware of physiological arousalwhen angry, using self-statements, and using a set of problem-solving skillsto solve social problems;
Having the boys plan and make their own videotape of inhibitory self-statementsand social problems solving;
Dialoging, discussion, and role-playing to implement social problem solvingskills with children's current anger arousal problems.

Lochman's work with highly

aggressive boys is reported in a series of research articles beginningin the mid-80's. Lochman's anger-coping intervention is based on researchthat shows that aggressive children tend to attribute hostility to otherpeople's intentions and to mis-perceive their own aggressiveness and responsibilityfor conflict. In addition to targeting specific cognitive skills (shownin the box), the intervention uses behavioral techniques (operant conditioning)to reward compliance with group rules. The program is targeted at boysin grades four through six who are identified as aggressive and disruptiveby their teachers. A school counselor and a mental health professionalfrom a Community Guidance Clinic co-lead groups of aggressive boys for12 - 18 group sessions, each 45 minutes to an hour. Importantly, this cognitivetraining is augmented with teacher consultation in which the mental healthprofessional running the children's group assists the childrens' regularteachers in classroom management in general and in helping the targetedyouths generalize new skills to the regular classroom.

The effectiveness of this "anger coping" intervention wasinvestigated in a series of studies which systematically varied featuresof the program to learn more about its essential elements. In one study(Lochman, Burch, Curry, & Lampron, 1984; scientific methods score=4),76 boys from eight elementary schools ranging in age from 9 to 12 werestudied. They were not randomly assigned to experimental conditions, butpre-treatment measures showed the groups to be equivalent on the outcomesmeasures of interest. In comparison to aggressive boys receiving no treatmentor minimal treatment, aggressive treatment group boys reduced their disruptive-aggressiveoff-task behavior in school (ES=-.55) and their aggressive behavior asrated by their parents (ES=-.61) directly after the intervention. A three-yearfollow-up study was conducted when these and some boys from other earlierstudies were 15-years old (Lochman, 1992; scientific methods score=4).The study found that the intervention had a significant effect on self-reportedalcohol and substance abuse (ES=-.38) but no significant effect on self-reportedcriminal behavior (ES=-.11). It can be argued that a reduction in delinquencyof this magnitude (approximately equivalent to a 5 percentage point differencein crime rate between the treatment and control group) in a highly delinquentpopulation is practically meaningful even if it is not statistically significant.Also, the treatment group in this follow-up study was significantly youngerthan the comparison group, which worked against finding program effectsas younger age was associated with higher rates of delinquency.

Rotheram also demonstrated the efficacy of cognitive behavioral trainingin a primary prevention program for upper elementary school youths. Inone study (Rotheram, 1982; scientific methods score=4) eight 4th through6th grade classes were randomly assigned to participate in a social skillstraining intervention or to serve as control classes. Students in eachclass were randomly assigned to small training groups led by graduate andundergraduate students. A drama situation game was conducted in each groupfor a one-hour session twice a week for twelve weeks. Each "game"involved teaching a specific assertiveness concept to help children think,act, or feel assertive; presentation of specific problem situations; groupproblem solving in which the students generated alternative solutions tothe problem and evaluated the solutions; and behavioral rehearsal and feedback.Although all students in the treatment classes were included in the intervention,only the 101 subjects identified (prior to the intervention) as being disruptive,under-achieving or exceptionally high in terms of comportment and achievementwere included in the evaluation. Students in the social skills trainingcondition generated significantly more assertive and significantly fewerpassive and aggressive problem-solving responses than did the control groupdirectly after treatment, and had larger increases in their grade-point-averagesover pre-treatment one year after the treatment. Teacher ratings of comportmentalso improved significantly more from pre-treatment to immediately followingthe treatment (ES=.42) as well as one year after the treatment (ES=.40).

Interventions relying solely on behavior modification strategies havealso been successful. Brewer et al. (1995) summarize two highly effectiveprograms that monitored school attendance and provided contingent rewardsfor good attendance. Both studies used rigorous evaluation methods andproduced positive outcomes on attendance. These results are important becausetruancy is an important risk factor for delinquency.

Bry's work also used behavioral monitoring and reinforcement with highrisk youths. Students were randomly assigned to the treatment and controlconditions in this study. Students' tardiness, class preparedness, classperformance, classroom behavior, school attendance, and disciplinary referralswere monitored weekly for two years. Students met with program staff weeklyand earned points contingent on their behavior which could be used fora class trip of the students' choosing. Frequent parent notification wasused. Experimental students had significantly better grades and attendanceat the end of the program than did controls, but the positive effects didnot appear until the students had been in the program for two years (Bry& George, 1979; scientific methods score=5; Bry & George, 1980;scientific methods score=4). Bry (1982; scientific methods score=4) reportsthat in the year after the intervention ended, experimental students displayedsignificantly fewer problem behaviors at school than did controls and inthe 18 months following the intervention, experimental students reportedsignificantly less substance abuse (ES=-.44) and criminal behavior (ES=-.30).Five years after the program ended, experimental youth were 66% less likelyto have a juvenile record than were controls (ES=-.50)

These rigorous studies of targeted behavior modification and cognitiveskill-training demonstrate clear positive effects on drug use and aggressive,anti-social behavior. Effect sizes are among the highest observed for anyschool-based strategy. Only Bry's work demonstrated a statistically significantreduction in actual criminal behavior (other than drug use), but the directionand size of the effect in the Lochman work provide additional support fora positive effect on criminal activity.

Peer Counseling, Peer Mediation, and Peer Leaders

Peer group counseling is popular in schools and is often used in preventionprograms for at-risk youths and adjudicated delinquents. This type of counselingusually involves an adult leader guiding group discussions in which participantsare encouraged to recognize problems with their own behavior, attitudes,and values. Peer pressure to adopt pro-social attitudes is expected tooccur. G. Gottfredson (1987) reviewed these approaches to delinquency preventionand evaluated a large-scale school-based program which was one of severalprograms included in OJJDP's alternative education initiative in the 1980s.This study (scientific methods score=3, involving random assignment ofsubjects to experimental conditions) "lends no support to any claimof benefit of treatment, with the possible exception that the treatmentmay enhance internal control for elementary school students. For the highschool students, the effects appear preponderantly harmful." (G. Gottfredson,1987, p 708). Specifically, high school treatment youths reported significantlymore delinquent behavior, more tardiness to school, less attachmentto their parents, and more "waywardness," a scale measuring aconstellation of anti-social attitudes, beliefs, and behaviors includingrebelliousness, lack of attachment to school, low beliefs in rules, delinquency,and association with delinquent peers. The effect sizes for these differenceswere small (less than .05). Presumably, these interventions backfire whenstudents are brought into closer association with negative peers duringthe peer counseling sessions. Gottfredson also notes that frequent discussionsof parent/home issues in the groups may have led to a weakening of parentalbonding and a subsequent increase in delinquency.

Peer mediation programs rose in popularity in the 1980s. These programsuse students to assist in dispute resolution when conflicts arise amongstudents. Trained peer mediators assist in developing alternative solutionsto fighting and provide an alternative to traditional interventions bya school administrator (e.g., warnings, suspensions, or demerits). Lam(1989, cited in Brewer et al, 1995) reviewed 14 evaluations of peer mediationprograms. The methodological rigor of all but three of the programs wastoo weak to justify any conclusions about the effect of the programs. Accordingto Brewer, none of the three studies in the Lam review employing quasi-experimentaldesigns showed significant effects on observable student behavior (e.g.,fighting, disciplinary referrals). One additional study of peer mediationpublished after Lam's review (Tolson, McDonald, and Moriarty, 1992; scientificmethods score=3) suggested that students assigned to receive peer mediationhave fewer interpersonal conflicts in the 2.5 months following the program,but the study was small and the outcome measure (referrals to the officefor interpersonal conflict) was weak.

Students have also been used as peer leaders in substance use preventionprograms. The rationale for this approach is that anti-drug messages willbe more credible when delivered by a peer than an adult. Although somestudies (e.g., Botvin, Baker, Renick, Fillazzola & Botvin, 1984; Perry,Grant, et al, 1989) have found that substance abuse prevention programsfocusing on skill development are more effective when led by peers thanby teachers, other studies (e.g., Ellickson & Bell, 1990) find no suchadvantage for peer-led programs. Tobler's (1992) meta-analysis also foundno evidence that programs with peer leaders produce better outcomes thanprograms of similar content led by adults.

The overall patterns of results for programs involving peers in thedelivery of services is not promising. Peer mediation programs are notpromising, although they have not been sufficiently evaluated. These programsare likely to be ineffective interventions when implemented as stand-aloneprograms rather than as part of broader attempts to improve disciplinarypractices. Peer counseling interventions for high-risk youths are contraindicated,and studies using peer leaders to lead substance abuse prevention programshave produced mixed results.

Counseling and Mentoring

Many studies have examined the effect of counseling interventions ondelinquency. Lipsey's (1992) meta-analysis of juvenile delinquency treatmenteffects shows that, for juvenile justice and non-juvenile justice interventionsalike, counseling interventions are among the least effective for reducingdelinquency. Twenty-four studies of individual counseling in non-juvenilejustice settings yielded an effect size of -.01 on measures of recidivism.

A popular form of school-based counseling is the Student AssistanceProgram (SAP). These programs are among the most common programs foundin schools, accounting for approximately half of the expenditures of Drug-freeSchools and Communities funds (Hansen & O'Malley, 1996, citing GAO,1993) administered through the U.S. Department of Education. These programsinvolve group counseling for students with alcoholic parents, counselingfor students who are using drugs or alcohol or whose poor academic performanceplace them at risk for substance abuse, and work with parent and communitygroups to develop ways of dealing with substance abuse problems. Oftenthe peers of student clients are involved as crisis managers, group facilitators,and referral agents. SAP counselor's are school-based but employed by mentalhealth departments or other outside agencies. After surveying the scantliterature on the effectiveness of SAP programs, Hansen & O'Malley(1996) concluded that evaluations are "universally absent." Theseprograms must be evaluated if federal funding for them is to be continued.

Gottfredson (1986; scientific methods score=5), in a study sponsoredas part of OJJDP's alternative education initiative, examined effects ondelinquent behavior of a program of services provided to high risk secondaryschool students. Students' behavioral and academic problems were diagnosed,and individual plans were developed by school specialists (either teachersor counselors assigned to work individually with the high risk studentsfor this project). Counseling and tutoring services were provided consistentwith the individual plans, and the specialists also acted as advocatesfor the students, worked with the students' parents, and tried to involvethe students in extracurricular activities to increase bonding to the school.On average, school specialists met twice per month directly with the targetstudents and the students also participated in peer counseling and "rap"sessions with other students. Random assignment of 869 eligible high-riskyouths to treatment and control conditions yielded equivalent groups. Aftertwo year of treatment, the targeted youths were significantly better offthan the control students on several measures of academic achievement andeducational persistence. Students were promoted to the next grade at ahigher rate after the first year in the program (ES=.15), drop-out rateswere significantly lower for students in some of the schools (ES=.09 overall),graduation rates were higher (ES=.68), and the percentage of students scoringin the bottom quartile of a standardized achievement tests scores was lower(ES=-.19). However, the services did not result in a reduction in delinquency.Gottfredson (1986) examined six indicators of delinquent behavior, includingself-reports, school records, and police records. For only one of the measureswere significant differences observed. Treatment students reported significantlymore drug use (ES=.23). In all, two measures showed no difference,two favored the treatment group (ES's=-.08 and -.14) and two favored thecontrol students (ES's=.02 and .23). The study suggests that even relativelysmall doses of tutoring lead to improvements in academic outcomes. It isprobable that the poor showing on the delinquency measures was due to thecounseling intervention which brought high-risk youths together to discuss(and therefore make more salient to others) their poor behavior.

Mentoring -- one-on-one interaction with an older, more experiencedperson to provide advice or assistance -- is an increasingly popular delinquencyprevention strategy. OJJDP has invested $19 million in juvenile mentoringprograms, as mandated by Congress. Our review uncovered four studies ofschool-based mentoring (See Table 5-5). Chapter 2 reviews additional studiesof community-based mentoring. The results of the studies can be summarizedas follows: (1) The methodological rigor of the studies is generally poor.Only one study received a scientific methods score of three or more, andthis study did not assess the programs' effect on crime outcomes. (2) School-basedmentoring programs appear promising for increasing school attendance. (3)The effectiveness of school-based mentoring for reducing delinquency anddrug use is not known. See Chapter 2 for a summary of one rigorous studyof a particularly well-implemented community-based mentoring program whichfound positive effects on substance use, bearing in mind that the resultsfrom that study may not generalize to mentoring programs run in or by schools.

In summary, counseling interventions for high-risk youths are contraindicated,and school-based mentoring programs appear promising for reducing nonattendancebut have not been studied with sufficient rigor to justify confident conclusionsabout its effectiveness for reducing delinquency or substance use.Table5-5. Summary of Mentoring Studies

 Author Scientific Effect size for Effects on risk and protective factors (year) methods measure of problem score/ behavior Number of cases Higgins (1978) 2 Offenses, weighted by School persistence -- NS severity -- not [results for high N=106 significantly different School performance -- significantly better school students [ES=.25 males; .23 for mentored, males only) returning from females, favoring correctional mentored group] institution after approximately one year of program] McPartland & Nettles 3.5 NA Absences -- significantly fewer absences (1991) [ES=-.18] N=334 [results for middle (approx.) English grades -- significantly better school students [ES=.14] directly after two years of program] GPA and grade promotion -- NS Slicker & Palmer (1993) 2 NA Drop-out and GPA-- NS [results for 10th grade N=64 "at-risk" students directly after six months of program] LoSciuto, Rajala, 2 Frequency of substance Days absent -- significantly fewer for Townsend, & Taylor use in past 2 months -- mentoring group (1996) N=562 almost significantly lower (p=.056) among [results for sixth mentored students graders directly after [ES=-.22] one school year of program] 

Recreational, Enrichment, and Leisure Activities

Some programs offer recreational, enrichment or leisure activities asa delinquency prevention strategy. These programs historically have beenbased on one of the following assumptions: (1) "idle hands are thedevil's workshop;" (2) children -- especially those who do not fitthe academic mold -- will suffer from low self-esteem if they are not ableto display their other competencies; or (3) students need to vent theirenergy. With the rise in violent crime, the typical rationale for alternativeactivities programs is that occupying youth's time will keep them out ofharm's way -- the "safe haven" theory. Drop-in recreation centers,after-school and week-end programs, dances, community service activities,and other events are offered as alternatives to the more dangerous activities.After-school programs have enjoyed a recent boost in popularity in lightof evidence that 22% of violent juvenile crime occurs between 2 p.m. and6 p.m. on school days (Snyder, Sickmund, & Poe-Yamagata, 1996). Thisis more than would be expected if juvenile crime were uniformly distributedacross the waking hours.

Relevant research on alternative activities is found both in basic researchon the causes and correlates of delinquency and in evaluations of preventionprograms involving these activities. Basic research has examined the plausibilityof the "idle hands is the devil's workshop" rationale for explainingdelinquency and found it lacking. Several studies have found that timespent in leisure activities is unrelated to the commission of delinquentacts (Gottfredson, 1984b; Hirschi, 1969). Time spent on activities whichreflect an underlying commitment to conventional pursuits (e.g., hoursspent on homework) is related to the commission of fewer delinquent acts,while time spent on activities which reflect a (premature) orientationto adult activities (e.g., time spent riding around in cars) is relatedto the commission of more delinquent acts. But the myriad activities ofadolescents that have no apparent connection to these poles (e.g., clubs,volunteer and service activities, youth organizations, sports, hobbies,television, etc.) are unrelated to the commission of delinquent acts. Simplyspending time in a these activities is unlikely to reduce delinquency unlessthey provide direct supervision when it would otherwise be lacking.

Alternative activities programs have been found to not preventor reduce alcohol, tobacco, and other drug use in several reviews of theeffectiveness of drug prevention (Botvin, 1990; Hansen, 1992; Schaps, Bartolo,Moskowitz, Palley, and Churgin, 1981; Schinke, Botvin, and Orliandi, 1991).More recent evidence of the impotence of alternative activities programscomes from the National Structured Evaluation (NSE; Stoil, Hill, and Brounstein,1994), a major study of the effectiveness of prevention activities initiatedin 1991 by the Center for Substance Abuse Prevention (CSAP), which examinedhundreds of different program models in operation during or after 1986.The NSE found that alternative activities alone do not reduce alcoholand other drug use, alcohol and other drug -related knowledge and attitudes,or other risk and protective factors related to alcohol and other druguse. However, when these drug-free activities appeared as secondary componentsin programs primarily aimed at psycho-social skill development, they wereeffective for reducing alcohol and other drug use and related risk andprotective factors. Note that the reviews and the NSE summarize evidencerelated to broadly-defined alternative activities programs operating inboth school and community contexts. They do not tell us whether the null

Table 5-6. Summary of Recreation, Enrichment, and Leisure ActivitiesStudies

 Author Scientific Effect size for Effects on risk and protective factors (year) methods measure of problem score/ behavior Number of cases Thompson & Jason (1988) 2 NA Gang membership -- favors experimental, p=.06; [ES=-.16] [results for eighth N=117 grade students at-risk for gang membership directly after one school year of program] Ross, Saavedra, Shur, 4 NA Achievement test scores -- no significant Winters, & Felner difference overall (1992) N=667 Risk-taking -- significantly favors control [results for low-income group elementary school children directly after Impulsiveness -- significantly favors control 144 days of program] group Cronin (1996) 4 Rebellious Behavior -- Grade-point average, Attachment to school, NS Commitment to school, Belief, Attitudes [results for at-risk N=508 favoring drug use, Attendance -- NS sixth grade students Drug Use in Last Year directly after one -- significantly favors school year of program] control group (ES=.47) Drug Use in Last Month -- NS 

findings apply equally to programs in these different settings. Fewevaluations of the effect of these recreation, leisure, and enrichmentactivities on delinquency other than substance use are available. Theyare summarized in Table 5-6. These studies all combine an emphasis on alternativeactivities with other components such as instruction in skills relatedto the alternative activity. One program (Ross et al, 1992) involved instructionand supervised homework and self-esteem building exercises in a school-basedafter-school program. The study did not assess program effects on actualdelinquent behavior due to the young age of the children, but it did measurelow self-control, a potent risk factor for later delinquency. The Thompsonand Jason (1988) study reported on a gang prevention program involvinginstruction plus an after school program involving a sports clinic, socialand recreational activities, job-skills and educational assistance. Cronin(1996) reported on a community service program which also involved reflection/discussionsessions for "processing" the service experience. As Table 5-6shows, the results are unfavorable to alternative programs, except forone study which shows a marginally significant (p=.06; ES=-.16) positiveeffect on a risk factor for delinquent behavior, gang membership. The otherstudies suggest that these alternative activities programs may actuallyincrease the risk for delinquent behavior.

These studies of alternative activities do not specifically addressthe crime prevention potential of recreational strategies such as "midnightbasketball" which are designed to keep the most crime-prone segmentof the population off the streets during peak crime hours (i.e., to providea "playground for ... idle hands") and to enhance positive youthdevelopment through mandatory attendance at workshops covering topics suchas job development, drug and alcohol use, safe sex, GED preparation andcollege preparation, and conflict resolution. These programs have receivedmedia attention and public support in recent years. Midnight Basketballwas praised in 1991 by President George Bush as one of his "thousandpoints of light." The "Crime Bill" signed into law by PresidentClinton in 1994 featured alternative activities prominently among its variouscrime prevention strategies. Early versions of the bill included a lineitem for Midnight Basketball, and although the line item was eventuallyeliminated when it became the symbol of pork-barrel spending among conservativesin and out of Congress , alternative activities strategies still figureprominently among its prevention strategies. Midnight Basketball is mentionedexplicitly as one of the preferred Local Crime Prevention Block Grant Programstrategies, along with other supervised sports and recreation programs;non-school recreation strategies are included in the Ounce of PreventionGrant Program; supervised sports and extracurricular programs includingarts and crafts and dancing during non-school hours are included in theCommunity Schools Youth Services and Supervision Grant Program; and parkand recreation programs in high risk areas are called for in the UrbanRecreation and At-Risk Youth Grants to local governments (Youth Today,Nov/Dec, 1994).

Midnight basketball programs are not likely to reduce crime. The evidencefrom meta-analyses of drug prevention programs suggests no behavioral effectof such programs, and the few studies that have examined effects on delinquencyor anti-social behavior suggest no effect. The only compelling argumentfor continuing to consider this approach is that they may be able to provideadult supervision when it would otherwise be lacking. But research (Rosset al, 1992, summarized in Table 5-6) indicates that programs intendingto provide such supervision for unsupervised youth in the after-schoolhours may actually increase risk for delinquency. These investigators foundthat (1) the students most in need of after-school supervision chose notto participate in the program, (2) the program increased risk-takingand impulsiveness, and (3) the program worked no better for latch-key childrenthan for children who had access to other supervision during the afterschool hours. These unfortunate outcomes make sense in light of other evidence(e.g., G. Gottfredson, 1987) demonstrating that interventions that grouphigh-risk youths with lower-risk youths in the absence of a strong interventionto establish pro-social group norms often backfire.

In summary, research clearly supports the crime-prevention potentialof providing direct adult supervision of high-risk juveniles when theywould otherwise be unsupervised, but designing such interventions so thatthey will reach the intended population and counteract potential negativeeffects of grouping high-risk youths remains a challenge. The chapter oncommunity programs finds reason for guarded optimism about the crime preventionpotential of after-school recreation programs operating in high-crime areasby community-based organizations such as Boys and Girls Clubs. It is possiblethat such programs are more effective than the more broadly defined alternativeactivities programs summarized here. It is also possible that featuresof the implementing organization and the community context within whichthe programs operate moderate the programs' effectiveness. Better researchis clearly needed to isolate these characteristics of programs and contexts.At this point in time, expectations for these programs far exceed theirempirical record. Because some studies have found backfire effects, itis particularly important to proceed with due caution.

A Comprehensive OJP-Funded Program: Cities in Schools (C.I.S.)

C.I.S. is a comprehensive dropout prevention program which combinesseveral individual-level prevention strategies within a broader effortto alter the school environmental to facilitate the delivery of servicesto high-risk youths. Its breadth defies the program categorization adoptedfor this report. C.I.S. operates in 665 sites in 197 communities nationwide(OJJDP, 1995). It is operated by Cities in Schools, Inc., a nonprofit organizationheadquartered in Alexandria, VA. Regional and state-level offices bridgethe gap between the national office and local programs. Regional staffare the primary providers of technical assistance and training to new andexisting programs. State office functions parallel those of the regionaloffices.

The C.I.S. model utilizes the school as a site for service coordinationand integration. It is more a strategy for service delivery than a program.It is based on the belief that the "existing human services deliverysystem is fragmented, categorical and uncoordinated, and that the clientsof the system have multiple problems that extend beyond the relativelynarrow agendas of particular agencies (Rossman and Morley, 1995)."Several different strategies are used to address the problems of youthat risk for drop out. The central feature of C.I.S. is the assignment ofcaseworkers to groups of problem students at inner city schools. Commonstrategies include: 1) case management (often focusing on obtaining neededservices such as health and dental screening, bus tickets, clothing, etc.),2) individual or group counseling, 3) assistance with academic subjects,4) attendance monitoring, and 5) activities to promote self-esteem andteam building. A "C.I.S. class", although not required, is recommendedby the national organization. No standard curriculum exists for the C.I.S.classes, but many focus on life-skills education and contain an emphasison building students' self esteem and encouraging prosocial attitudes andbehaviors. The activities are loosely structured. Tutoring and mentoringare among the most commonly provided services, but individual sites areencouraged to develop special services and arrangements according to theirlocal needs, resources, and constraints.

Two evaluations of the C.I.S. program have been conducted. The first(Murray, Bourque & Mileff, 1981) reviews program outcomes from 1978-1980,the second (Rossman and Morley, 1995) outcomes from 1989-1991. The methodologicalrigor of both studies (2 and 1, respectively) falls below the cut-pointestablished in this report for scientific credibility. Conclusions regardingprogram efficacy cannot be drawn based on either evaluation.

Murray et al. (1981) showed that the services delivered were not asstrong as anticipated by the C.I.S. model. Rossman and Morley (1995) wereunable to quantify the level of program implementation because the systematicrecords were not kept by the program. Analysis of drop out and absencesincluded in the first evaluation suggested that C.I.S. did not have thedesired effect on students. Analysis of absences included in the secondevaluation generally showed that C.I.S. students with the most severe problemsdemonstrated improvement over time. Whether this is attributable to theprogram or to regression to the mean is not known. Analysis of drop outin the second evaluation suggested that the dropout rate for C.I.S. studentscompared favorably to other at-risk populations in the nation but offeredno evidence about the comparability of these other populations to the C.I.S.population on other variables that would place students at risk for droppingout. An examination of the effect of the C.I.S. program on a variety ofproblem behaviors was included in the second evaluation. C.I.S. studentsare asked to report how big of a problem a behavior used to be and whetheror not this has changed. Results indicated that students were more likelyto experience improvement or no change as opposed to getting worse. Thedesign (lack of comparison group, retrospective self-report) tells us nothingabout the effects of C.I.S. on these behavioral outcomes.

In summary, although several aspects of the C.I.S. strategy resemblecomponents shown in other work to have promise for reducing delinquencyand substance use, the effects of C.I.S. on these behaviors is unknownbecause its evaluations have lacked the rigor necessary to justify anyconclusions about its effectiveness. Mentoring and the "school-within-a-school"structure used in some of the C.I.S. sites are promising for reducing delinquencyor substance use. On the other hand, counseling, unstructured life skillsclasses, and community service activities have been shown to be ineffectivefor reducing these problem behaviors, and grouping high-risk students togetherin the absence of a structured program appears to increase delinquency.C.I.S. has been successful in accessing a large number of at risk students,establishing a service delivery mechanism for them, and generating funds(both federal and other) to initiate and sustain interventions. The programsneeds to be rigorously evaluated.

Scientific Conclusions

What Works? Strategies for which at least two different studieshave found positive effects on measures of problem behavior and for whichthe preponderance of evidence is positive are:

Crime and delinquency:

(1) Programs aimed at building school capacity to initiate and sustaininnovation.

(2) Programs aimed at clarifying and communicating norms about behaviors-- by establishing school rules, improving the consistency of their enforcement(particularly when they emphasize positive reinforcement of appropriatebehavior), or communicating norms through school-wide campaigns (e.g.,anti-bullying campaigns) or ceremonies; and

(3) Comprehensive instructional programs that focus on a range of socialcompetency skills (e.g, developing self-control, stress-management, responsibledecision-making, social problem-solving, and communication skills) andthat are delivered over a long period of time to continually reinforceskills.

Substance use:

(1) Programs aimed at clarifying and communicating norms about behaviors;

(2) Comprehensive instructional programs that focus on a range of socialcompetency skills (e.g, developing self-control, stress-management, responsibledecision-making, social problem-solving, and communication skills) andthat are delivered over a long period of time to continually reinforceskills; and

(3) Behavior modification programs and programs that teach "thinkingskills" to high-risk youths.

What does not work? Strategies for which at least two differentstudies have found no positive effects on measures of problem behaviorand for which the preponderance of evidence is not positive are:

(1) Counseling students, particularly in a peer-group context, doesnot reduce delinquency or substance use.

(2) Offering youths alternative activities such as recreation and communityservice activities in the absence of more potent prevention programmingdoes not reduce substance use. This conclusion is based on reviews of broadly-definedalternative activities in school- and community settings. Effects of theseprograms on other forms of delinquency are not known.

(3) Instructional programs focusing on information dissemination, feararousal, moral appeal, and affective education are ineffective for reducingsubstance use.

What is promising? Several strategies have been shown in onlyone rigorous study to reduce delinquency or substance use. If the preponderanceof evidence for these strategies is positive, they are regarded as "promising"until replication confirms the effect. These strategies are:

Crime and delinquency:

(1) Programs that group youths into smaller "schools-within-schools"to create smaller units, more supportive interactions, or greater flexibilityin instruction; and

(2) Behavior modification programs and programs that teach "thinkingskills" to high-risk youths.

Substance use:

(1) Programs aimed at building school capacity to initiate and sustaininnovation; (2) Programs that group youths into smaller "schools-within-schools"to create smaller units, more supportive interactions, or greater flexibilityin instruction; and

(3) Programs that improve classroom management and that use effectiveinstructional techniques.

Effectiveness of DOJ Programs

With the notable exception of D.A.R.E. evaluations, the evaluationsof school-based prevention programs funded by OJP are generally too weakto justify conclusions about the effectiveness of the programs.

D.A.R.E. Evaluations show that as it is most commonly implemented,D.A.R.E. does not reduce substance use appreciably. But the revised D.A.R.E.curriculum with its follow-up sessions in later grades has not been evaluated.Given the more general finding that instructional drug prevention programsare most effective when delivered over extended periods of time, a reasonablecourse of action would be to conduct a rigorous study to compare the revisedD.A.R.E. program including its follow-up sessions with other plausible,long-term drug prevention curricula containing more social competency content.This study should randomly assign fifth or sixth grade classrooms to receiveeither D.A.R.E. with its booster sessions or a non-D.A.R.E. program ofequal length and intensity and its booster sessions. Long-term effectsshould be assessed in a longitudinal study and care should be taken toensure sufficient statistical power to detect small differences in effectiveness.

L.R.E. The national evaluation of L.R.E. was inconclusive. Asdetailed above, L.R.E. has theoretical promise only when the law-relatedcurriculum is embedded in a more comprehensive program of improved classroomorganization and management. A stand-alone law-related education curriculumis no more likely to reduce delinquency than a stand-alone drug educationprogram is to reduce substance use. More rigorous evaluation is neededto evaluate L.R.E. as it is typically implemented, and to isolate the effectiveingredients in the multi-component L.R.E. interventions that have resultedin positive evaluations.

C.I.S. Evaluations of C.I.S. have not been of sufficient methodologicalrigor to justify conclusions about its crime prevention potential. C.I.S.represents a vehicle through which a variety of prevention services couldbe effectively delivered. But as currently implemented, the mix of servicesprovided is as likely to contain ineffective as effective ones. If Congressis to continue to mandate these programs, rigorous tests of evaluationsshould now be conducted.

Several additional categories of school-based programs are supportedfrom time-to-time by OJP. These include "Midnight basketball"and other recreational activities intended to reduce crime, peer mediationprograms; and violence prevention curricula. A variety of after-schoolprogram models are also being developed. None of these program types havebeen studied with sufficient rigor to justify conclusions about their effectiveness,but some evaluations have produced disappointing results. Rigorous evaluationof the OJP-funded programs is required.

Byrne Funds. Little is known about the specific school-basedprograms supported by Byrne Block Grant Funding. One of the purpose areasfor this funding is education, however, and $74.7 million was spent between1989 and 1994 for these education programs. Some of this funding is knownto support local D.A.R.E. programs, known to be ineffective as most commonlyimplemented. The block grant program as it is currently organized mightbe strengthened through federal efforts to disseminate information to stateand local agencies about what school-based strategies work to reduce delinquency.

Improving Effectiveness Through Evaluation and Research

The studies reviewed in this chapter have demonstrated that school-basedprevention can work. With few exceptions, the different categories of preventionactivities have been shown to reduce delinquency or substance abuse inat least one rigorous study. The magnitude of the effects of these strategiesranges from small (e.g., for instructional drug prevention programs andclassroom management interventions) to moderate (e.g., for a behavior modificationintervention and some of the more comprehensive programs such as STATUS,that combined a school-within-a-school structure with an innovative curriculumand effective instructional methods). Yet the magnitude and durabilityof effects of school-based prevention efforts, although at least comparableto those of delinquency prevention and treatment efforts in other settings,are low relative to the theoretical promise and anticipated potential ofthese programs. More important than the question of which individual strategies"work" is the question of how the promising strategies can bestrengthened to improve their yield. These efforts should focus on twobroad areas: Specifying theories underlying school-based prevention andimproving the level of implementation of prevention programs.

Specifying theories of school-based prevention. Much school-basedprevention is guided by the following general notions about the natureand causes of problem behaviors: (1) Different problem behaviors are highlyrelated; (2) different problem behaviors share common antecedents; (3)the common antecedents are the risk and protective factors identified inresearch as correlates of problem behavior (e.g., as summarized in reviewssuch as Hawkins, Catalano, & Miller, 1992 and Loeber & Dishion,1983); and (4) prevention efforts aimed directly at these risk and protectivefactors will reduce problem behavior. Specific school-based delinquencyprevention practices are often justified on the basis of demonstrated effecton one or more known risk or protective factors for delinquency.

The prevention focus on risk and protective factors is enormously popularamong practitioners and has succeeded in pushing practice away from strategieswith no basis in research and towards strategies with plausibility. Atthe same time, accumulated evidence has raised questions about the relativepotency of different risk and protective factors and their possible differentialeffects on various problem behaviors. Some risk-based strategies show promisefor reducing substance use but not other forms of delinquency (e.g, mentoringprograms and the classroom organization and management strategies summarizedearlier). Other programs have clear effects on aggressive behavior andschool conduct problems, but the evidence for an effect on measures ofcriminal activity is less convincing (e.g., cognitive training strategiesand social competency instruction). Many programs have large effects onacademic achievement, commitment to school, or attachment to school, butno effect (Hawkins, Catalano, et al, 1992; Hawkins, Doueck & Lishner,1988) or even negative effects (D. Gottfredson, 1986) on delinquency andsubstance use. Clearly, enhancing protective factors or reducing risk factorsdoes not ensure a large reduction in delinquency. The focus on risk andprotective factors has been and no doubt will continue to be a valuablecontribution to the prevention field. But more productive theory-buildingand testing is now required to make significant progress. School-basedprevention efforts would benefit from the development and testing of multi-leveltheories that specify how environmental features of schools interact withindividual-level processes generating delinquent behavior. Efforts to clarifythe causal processes linking school characteristics and schooling experiencesto delinquency can be expected to lead to refined program designs whichtarget the most potent theoretical variables.

Improving implementation of school-based prevention programs.Researchers have recently turned their attention to better understandingthe conditions which may impede the implementation of prevention programsand therefore limit their effectiveness. Elias, Weissberg, et al. (1994)recommend comprehensive, multi-year, multi-component approaches over moretraditional single-intervention ones. This idea is also supported by meta-analysisresults showing that programs using multiple interventions work betterthan those using a single intervention strategy (Tobler, 1986) and by resultssummarized above. Some of the more comprehensive programs reviewed above(e.g., Olweus' bullying intervention in Norway schools; Gottfredson's school-capacitybuilding interventions) are among the more potent programs for reducingdelinquency. Given that the single largest federal expenditure on school-basedprevention (Safe and Drug-Free Schools and Communities monies administeredby the U.S. Department of Education) funds primarily narrower strategiessuch as student assistance programs (a form of counseling) and drug instruction,this recommendation alone, if heeded, can be expected to boost the effectivenessof school-based prevention activities.

Elias, Weissberg et al. (1994) also advocate strategies to strengthenthe "host environment." These strategies include working withstaffs in the schools to ensure goal consensus and fluency in the theoryunderlying the prevention approach, and using an action research modelto clarify expectations, monitor progress, and identify and resolve problemswhich present obstacles to implementation as they arise. Support for thesecapacity-building strategies is summarized above.

Gottfredson, Fink, Skroban & Gottfredson (1996) summarize literatureon factors related to successful educational reform in general. The capacityof schools to initiate and sustain reform, and consequently the strengthand fidelity of those reforms, varies considerably across geographic areas,with schools in urban areas most likely to lack the infrastructure necessaryto support change. Many features of school organizations shown to be relatedto successful reform -- quality leadership, teacher morale, teacher mastery,school climate, and resources -- are lower on average in urban than inother schools. The literature on school reform suggests that the strengthand durability of school-based prevention programs can be increased byembedding specific program components within a broader capacity-buildingeffort that attends to these larger organizational issues.

The recommended direction for school-based prevention -- towards multi-faceted,longer-term, and broader-reaching programs embedded in school capacity-buildingactivities-- presents a challenge to researchers and policy-makers alikebecause the "user-friendliness" of programs is related to thefidelity of their implementation. More complex programs are more likelyto be watered down or "reinvented" by school staff. Indeed, experienceworking with a troubled urban middle school to implement a multi-componentprevention program over a four-year period (Gottfredson, Gottfredson, &Skroban, 1996) illustrated the challenge. The program included severalcomponents aimed at increasing social competency skills as well as componentsaimed at increasing social bonding and school success. Most pieces hadbeen demonstrated in prior "single intervention" research toreduce problem behavior or factors leading to it, and are included amongthe program strategies that "work," summarized above. The five-yearstudy tested the transportability of these intervention strategies intoa more comprehensive program that could be implemented in a natural schoolsetting as part of a multi-year school-based prevention demonstration.The evaluation of the five-year effort showed that the program never reachedits expected level of implementation and no reliable effects on youth behaviorsor attitudes were observed. The organization proved incapable of absorbingthis ambitious program.

The question of what it will take to initiate and sustain meaningfulchange in schools is the highest priority question for researchers andpolicy makers at this time. We know from research summarized in this chapterthat a variety of strategies can reduce delinquency or substanceuse. But the conditions under which much of the research -- particularlythe research on individually-focused interventions -- was conducted donot resemble real-world conditions in schools where programs are most needed.Tobler (1992) shows, for example, that among the top ten most effectivedrug prevention programs identified in the literature, only onewas implemented by classroom teachers, and even that intervention was unusualbecause extraordinary amounts of training and consultation was providedfor the teachers. When school-based programs are implemented under lessthan ideal conditions results have not been as positive. In a study ofHispanic students in eight urban schools in the New York area, Botvin,Dusenbury, Baker, James-Ortiz, & Kerner (1989), reported that the amountof the L.S.T. program material covered by teachers ranged from 44% to 83%.When the experimental sample was divided into high implementation (witha mean completion rate of 78%) and low implementation (mean of 56%), positiveeffects of the program were found only in high implementation group. Thisaccords with more general findings from Lipsey's (1992) extensive meta-analysisof prevention and treatment programs which found that programs deliveredby researchers were more effective than those delivered by the typicalpractitioner, presumably because researchers attended more to issues ofstrength and integrity of program implementation.

These facts must be understood if we are to strengthen prevention programming.Several of the studies summarized above (e.g., Botvin et al, 1995; Gottfredson,Gottfredson, & Hybl, 1993; Johnson, 1984) reported effects separatelyfor groups of schools or teachers that differed on the strength and fidelityof program implementation. The evidence always suggests that more delinquencyis prevented when strategies are implemented with greater fidelity overprolonged periods and that these conditions are met more easily in someschools than in others. Additional research is now needed to increase ourunderstanding of how the potential of strategies we already know aboutcan be realized in real-world settings.

An example of a comprehensive, theory-based, well-implemented school-basedintervention. A recent example of a school-based intervention to reduceconduct disorder that addresses the shortcomings of prevention programmingsummarized earlier is the FAST Track (Families and Schools Together; ConductProblems Prevention Research Group, 1992) Program, currently being testedin four cities with support from the National Institute of Mental Health.The program was developed by a consortium of social scientists on the basisof developmental theory about the causes of conduct disorder in childrenand previous evaluations of specific, theory-based program components.It integrates five intervention components designed to promote competencein the family, child, and school and thus prevent conduct problems, poorsocial relations, and school failure -- all precursors of subsequent criminalbehavior -- during the elementary school years. The program involves trainingfor parents in family management practices; frequent home visits by programstaff to reinforce skills learned in the training, promote parental feelingsof efficacy, and enhance family organization; social skills coaching forchildren delivered by program staff and based on effective models describedearlier; academic tutoring for children three times per week; and a classroominstructional program focusing on social competency skills coupled withclassroom management strategies for the teacher. The program thereforeincludes several of the most effective school-based strategies summarizedearlier as well as the most effective strategies from the family domain.

The participating schools and families work closely with the researchteam to implement the program in a strong fashion and support its evaluation.Only preliminary data are available from the rigorous evaluation of thisongoing project. Dodge (1993) reported that after one year of this intensiveprogram, clear positive effects were evident on several of the intermediatebehaviors targeted by the program (e.g., parent involvement in the child'seducation and child social-cognitive skills) and significantly less problembehavior (ES=-.25) was recorded by trained observers for the treatmentthan for the comparison children. These positive results for such a difficultpopulation are encouraging and attest to the need for more comprehensive,theory-based, preventive interventions implemented with careful attentionto strength and fidelity. The cost of such high-quality program developmentis high compared with typical expenditures on program development and evaluationfor OJP programs: FAST Track's budget exceeds $1 million per year for eachof the four program sites.

These comments are intended to stimulate thinking about what Congressand OJP can do to contribute to the development of stronger school-baseddelinquency prevention efforts. Specific recommendations for strengtheningprograms are:

1. Increase Congressional appropriations for school-based preventionactivities. OJP funding for school-based crime prevention is meager comparedwith its expenditures in other domains within OJP and compared with expendituresby other agencies on school-based prevention. Total expenditures on school-basedprevention (partially summarized in Table 5-1) are less than $25 millionper year,9 comparedwith $1.4 billion for the extra police programs and $617 million for prisonconstruction. This limited investment in school-based crime prevention,in light of its promise demonstrated in this chapter, represents a lostopportunity for preventing crime.

2. Support multi-year prevention efforts (e.g., programs that span theelementary school years, the middles school years, and the high schoolyears rather than single-year programs);

3. Support multi-component prevention efforts that include the environmental-changeand individual strategies that have been shown to work in some settingsunder some conditions and whose positive results have been replicated:

(a) Programs aimed at building school capacity to initiate and sustaininnovation;

(b) Programs aimed at clarifying and communicating norms about behaviors;

(c) Comprehensive instructional programs that focus on a range of socialcompetency skills (e.g, developing self-control, stress-management, responsibledecision-making, social problem-solving, and communication skills) andthat are delivered over a long period of time to continually reinforceskills; and

(d) Behavior modification programs and programs that teach "thinkingskills" to high-risk youths.

4. Reduce funding for program categories (counseling students for delinquencyprevention, alternative activities such as recreation and community serviceactivities in the absence of more potent prevention programming for drugprevention, and instructional drug prevention programs focusing on informationdissemination, fear arousal, moral appeal, and affective education) knownto be ineffective.

5. Support activity to disseminate information about effective and ineffectiveschool-based strategies to practitioners and to local- and state-levelprogram managers and policy-makers.

Additional recommendations for evaluation and research needed to improvethe effectiveness of school-based prevention include:

1. Require (and provide the substantial financial investment to enable)rigorous evaluation of the long-term multi-component models recommendedabove, insisting that studies of the effectiveness of strategies aimedat altering school and classroom environments be conducted using schoolsor classrooms as the unit of analysis, and testing the generalizeabilityof effects across different types of communities.

2. Support replication studies of the promising strategies identifiedin the summary section above;

3. Support theory-building and testing efforts which seek to clarifythe causal models relating school experiences and delinquency;

4. Support research to investigate school conditions conducive to high-qualityimplementation of prevention programs; and

5. Support the development and rigorous testing, especially in urbanareas, of strategies designed specifically to improve the level of implementationof prevention programs.

NOTES

1The editorial assistanceof Roger Weissberg and the research assistance of Todd Armstrong, VeronicaPuryear, John Ridgely, Stacy Skroban, and Shannon Womer are gratefullyacknowledged.

2Of course, moremoney is spent on maintaining basic educational services. The largest proportionof spending for children and youth in all states is tied to schools (Holmes,Gottfredson, & Miller, 1992) -- mostly to maintain basic educationprocesses. An argument can be made for counting these large basic educationexpenditures as prevention expenditures because they are directed at improvingthe social capital of the citizenry (e.g., education and proper conduct)which protects youths from later involvement in a variety of problem behaviors.Because the evidence for a connection between basic education programsand practices and crime is largely indirect, such basic education functionswill be given short shrift in this chapter. Researchers and policy-makersshould devote more attention, however, to understanding the crime preventionpotential of large federal entitlement programs such as Chapter I of TitleI of the Elementary and Secondary Education Act, which distributes approximately$6.7 billion in federal funds to local school districts to enhance basiceducational processes.

3OJP spends approximately$1.4 billion on extra policing programs and $617 million on prison constructionprojects per year.

4Code sheets usedto code methodological rigor and gather information for the computationof effect sizes are show in the methods appendix. All coding was done bytwo trained graduate students. All discrepancies were discussed and resolved.Seven aspects of the methods used in each study were rated to arrive atan overall rating of methodological rigor ranging from “1" (for studieshaving no controls for plausible alternative explanations for observedeffects, insufficient power to detect program effects, or inadequate measurementof key outcome variables) to “5" (for studies employing random assignmentto treatment and control conditions, sufficient power, and reliable andvalid measurement).

5A district consolidationof high schools prevented continued evaluation at the high school level.

6Effect sizes reportedhere are the effect sizes for treatment school change from pre-interventionto post-intervention reported in the original report minus the same effectsizes reported for the comparison schools.

7Evaluations ofD.A.R.E. are too numerous for detailed summary of each. The Bureau of JusticeAssistance has identified 23 D.A.R.E. evaluations conducted between 1991and 1996, several of which are included in the summary below. Others arenot included because they are primarily descriptive evaluations of state-levelefforts which have not appeared in the scientific literature. An assessmentof this fugitive literature seems unneccesary given the consistency offindings in the published literature. At any rate, such an effort is beyondthe scope of this review.

8The researcherswho conducted the national evaluation for OJJDP have continued to developand write about the program. Later reports contain the same ambiguity asthe earlier study of the Colorado sites.

9This figure doesnot include Byrne Block Grant monies, some of which fund local D.A.R.E.programs. But even with the Byrne funds, expenditures on school-based preventionare meager.

References

Battistich, V., Schaps, E., Watson, M., and Solomon, D.

1996 Prevention effects of the child development project: Early findingsfrom an ongoing multi-site demonstration trial. Journal of Adolescent Research11:12-35.

Botvin, G. J.

1990 Substance abuse prevention: Theory, practice, and effectiveness.In M. Tonry and J. Q. Wilson (eds.), Drugs and Crime. Chicago: The Universityof Chicago Press.

Botvin, G. J., Baker, E., Botvin, E. M., Filazzola, A. D., and Millman,R. B.

1984 Prevention of alcohol misuse through the development of personaland social competence: A pilot study. Journal of Studies on Alcohol 550-552.

Botvin, G. J., Baker, E., Dusenbury, L., Botvin, E. M., and Diaz, T.

1995 Long-term follow-up results of a randomized drug abuse preventiontrial in a white middle-class population. Journal of the American MedicalAssociation 273:1106-1112.

Botvin, G. J., Baker, E., Filazzola, A. D., and Botvin, E. M.

1990 A cognitive-behavioral approach to substance abuse prevention:One-year follow-up. Addictive Behaviors 15:47-63.

Botvin, G. J., Baker, E., Renick, N. L., Filazzola, A. D., and Botvin,E. M.

1984 A cognitive behavioral approach to substance abuse prevention.Addictive Behaviors 9:137-147.

Botvin, G. J., Batson, H. W., Witss-Vitale, S., Bess, V., Baker, E.,and Dunesbury, L.

1989 A psychosocial approach to smoking prevention for urban black youth.Public Health Reports 12:279-296.

Botvin, G. J., Dusenbury, L., James-Ortiz, S., and Kerner, J.

1989 A skills training approach to smoking prevention among Hispanicyouth. Journal of Behavioral Medicine 12:279-296.

Botvin, G. J., and Eng, A.

1982 The efficacy of a multicomponent approach to the prevention ofcigarette smoking. Preventive Medicine 11:199-211.

Brewer, D. D., Hawkins, J. D., Catalano, R. F., and Neckerman, H. J.

1995 Preventing serious, violent, and chronic juvenile offending: Areview of evaluations of selected strategies in childhood, adolescence,and the community. In J. C. Howell, B. Krisberg, J. J. Wilson, and J. D.Hawkins (eds.), A Sourcebook on Serious, Violent, and Chronic JuvenileOffenders. Newbury Park, CA: Sage Publications.

Bry, B. H.

1982 Reducing the incidence of adolescent problems through preventiveintervention: One- and five-year follow-up. American Journal of CommunityPsychology 10:265-276.

Bry, B. H., and George, F. E.

1979 Evaluating and improving prevention programs: A strategy from drugabuse. Evaluation and Program Planning 2:127-136.

Bry, B. H., and George, F. E.

1980 The preventative effects of early intervention on the attendanceand grades of urban adolescents. Professional Psychology 11:252-260.

Caplan, M., Weissberg, R. P., Grober, J. S., Sivo, P. J., Grady, K.,and Jacoby, C.

1992 Social competence promotion with inner-city and suburban youngadolescents: Effects on social adjustment and alcohol use. Journal of Consultingand Clinical Psychology 60:56-63.

Clayton, R. R., Cattarello, A. M., and Johnstone, B. M.

1996 The effectiveness of Drug Abuse Resistance Education (Project DARE):Five-year follow-up results. Preventive Medicine 25:307-318.

Conduct Problems Prevention Research Group

1992 A developmental and clinical model for the prevention of conductdisorder: The FAST Track Program. Deveopment and Psychopathology 4:509-527.

Corcoran, T. B.

1985 Effective secondary schools. In R. M. J. Kyle (ed.), Reaching forExcellence: An Effective Schools Sourcebook. Washington, D.C.: U.S. GovernmentPrinting Office.

Cronin, J.

1996 An Evaluation of a School-based Community Service Program: TheEffects of Magic Me. Technical report available from Gottfredson Associates,Inc., Ellicott City, MD.

Dicken, C., Bryson, R., and Kass, N.

1977 Companionship therapy: A replication in experimental communitypsychology. Journal of Consulting and Clinical Psychology 45:637-646.

Dodge, K. A.

1993 Effects of intervention on children at high risk for conduct problems.Paper presented at the biennial meeting of the Society for Research inChild Development, New Orleans, March.

Dryfoos, J. G.

1990 Adolescents at Risk: Prevalence and Prevention. New York: OxfordUniversity Press.

Duke, D. L.

1989 School organization, leadership, and student behavior. In O. C.Moles (ed.), Strategies to Reduce Student Misbehavior. Washington, D.C.:U.S. Department of Education.

Durlak, J. A.

1995 School-based Prevention Programs for Children and Adolescents.Thousand Oaks, CA: Sage.

Elias, M. J., Weissberg, R. P., Hawkins, J. D., Perry, C. A., Zins,J. E., Dodge, K. C., Kendall, P. C., Gottfredson, D. C., Rotheram-Borus,M., Jason, L. A., and Wilson-Brewer, R.

1994 The school-based promotion of social competence: Theory, practice,and policy. In R. J. Haggerty, N. Garmezy, M. Rutter, and L. Sherrod (eds.),Risk and Resilience in Children: Developmental Approaches. Cambridge: Universityof Cambridge Press.

Ellickson, P. L., and Bell, R. M.

1990 Drug prevention in junior high: A multi-site longitudinal test.Science 247:1299-1305.

Ellickson, P. L., Bell, R. M., and McGuigan, K.

1993 Preventing adolescent drug use: Long-term results of a junior highprogram. American Journal of Public Health 83:856-861.

Esbensen, F. A. & Osgood, D. W.

1996 GREAT Program Effectiveness: Results from the 1995 Cross-SectionalSurvey of Eighth-Grade Students. Unpublished technical reprint availablefrom author.

Felner, R. D., and Adan, A. M.

1988 The school transitional environment project: An ecological interventionand evaluation. In R. H. Price, E. L. Cowen, R. P. Lorion, and J. Ramos-McKay(eds.), 14 Ounces of Prevention: A Casebook for Practitioners. Washington,D.C.: American Psychological Association.

Felner, R. D., Ginter, M., and Primavera, J.

1982 Primary prevention during school transitions: Social support andenvironmental structure. American Journal of Community Psychology 10:277-290.

Forman, S. G.

1980 A comparison of cognitive training and response cost proceduresin modifying aggressive behavior of elementary school children. BehaviorTherapy 11:594-600.

Gainer, P. S., Webster, D. W., and Champion, H. R. .

1993 A youth violence prevention program: Description and preliminaryevaluation. Archives of Surgery 128:303-308.

Garrett, C. J.

1985 Effects of residential treatment on adjudicated delinquents: Ameta-analysis. The Journal of Research in Crime and Delinquency 22:287-308.

General Accounting Office

1993 Drug Education: Limited Progress in Program Evaluation. Statementof Eleanor Chemlinshy before the Subcommittee on Select Education and CivilRights, Committee on Education and Labor, House of Representatives, GAO/T-PEMD-93-2.

Gorman, D. M.

1995 Are school-based resistance skills training programs effectivein preventing alcohol misuse? Journal of Alcohol and Drug Education 41(1):74-98.

Gottfredson, D. C.

1986 An empirical test of school-based environmental and individualinterventions to reduce the risk of delinquent behavior. Criminology 24:705-731.

1987 An evaluation of an organization development approach to reducingschool disorder. Evaluation Review 11:739-763.

1990 Changing school structures to benefit high-risk youths. In P. E.Leone (ed.), Understanding Troubled and Troubling Youth. Newbury Park,CA: Sage Publications.

Gottfredson, D. C., Fink, C. M., Skroban, S., and Gottfredson, G. D.

In press Making prevention work. In R. P. Weissberg (ed.), Issues inChildren's and Families' Lives (Volume 4): Healthy Children 2010: School-and Community-based Strategies to Enhance Social, Emotional, and PhysicalWellness.

Gottfredson, D. C., Gottfredson, G. D., and Hybl, L. G.

1993 Managing adolescent behavior: A multiyear, multischool study. AmericanEducational Research Journal 30:179-215.

Gottfredson, D. C., Gottfredson, G. D., and Skroban, S.

1996 A multimodal school-based prevention demonstration. Journal ofAdolescent Research 11:97-115.

Gottfredson, D. C., Sealock, M. D., and Koper, C. S.

1996 Delinquency. In R. DiClemente, W. Hansen, and L. Ponton (eds.),Handbook of Adolescent Health Risk Behavior. New York: Plenum PublishingCorp.

Gottfredson, G. D.

1984a A theory-ridden approach to program evaluation: A method for stimulatingresearcher-implementer collaboration. American Psychologist 39:1101-1112.

1984b The Effective School Battery: User's Manual. Odessa, FL: PsychologicalAssessment Resources, Inc.

1987 Peer group interventions to reduce the risk of delinquent behavior:A selective review and a new evaluation. Criminology 25:671-714.

Gottfredson, G. D., and Gottfredson, D. C.

1985 Victimization in Schools. New York: Plenum.

Gottfredson, G. D., Rickert, D. E., Advani, N., and Gottfredson, D.C.

1985 Standards for program development evaluation. Psychological Documents14:31 (ms. No. 2668).

Greenberg, M. T.

1996 The PATHS Project: Preventive Intervention for Children. FinalReport to NIMH (Grant Number R01MH42131).

Greenberg, M. T., Kusche, C. A., Cook, E. T., and Quamma, J. P.

1995 Promoting emotional competence in school-aged children: The effectsof the PATHS curriculum. Development and Psychopathology 7:117-136.

Guerra, N. G., and Slaby, R. G.

1990 Cognitive mediators of violence in adolescent offenders. DevelopmentalPsychology 26:269-270.

Hansen, W. B.

1992 School-based substance abuse prevention: A review of the stateof the art of curriculum: 1980-1990. Health Education Research 7:403-430.

Hansen, W. B., and Graham, J. W.

1991 Preventing alcohol, marijuana, and cigarette use among adolescents:Peer pressure resistance training versus establishing conservative norms.Preventive Medicine 20:414-430.

Hansen, W. B., and O'Malley, P. M.

1996 Drug use. In R. J. DiClemente, W. B. Hansen, and E. L. Ponton (eds.),Handbook of Adolescent Health Behavior. New York: Plenum Press.

Hawkins, J. D., Arthur, M. W., and Catalano, R. F.

1995 Preventing substance abuse. In M. Tonry and D. Farrington (eds.),Building a Safer Society: Strategic Approaches to Crime Prevention . Chicago:Univ. of Chicago Press.

Hawkins, J. D., Catalano, R. F., and Associates.

1992 Communities That Care: Action for Drug Abuse Prevention. San Francisco:Jossey-Bass.

Hawkins, J. D., Catalano, R. F., and Miller, J. L.

1992 Risk and protective factors for alcohol and other drug problemsin early adulthood: Implications for substance abuse prevention. PsychologicalBulletin 112:64-105.

Hawkins, J. D., Catalano, R. F., Morrison, D. M., O'Donnell, J., Abbott,R. D., and Day, L. E.

1992 The Seattle Social Developmental Project: Effects of the firstfour years on protective factors and problem behaviors. In J. McCord andR. E. Tremblay (eds.), Preventing Antisocial Behavior: Interventions fromBirth Through Adolescence. NewYork: The Guilford Press.

Hawkins, J. D., Doueck, H. J., and Lishner, D. M.

1988 Changing teaching practices in mainstream classrooms to improvebonding and behavior of low achievers. American Educational Research Journal25:31-50.

Hawkins, J. D., Von Cleve, E., and Catalano, R. F.

1991 Reducing early childhood aggression: Results of a primary preventionprogram. Journal of the American Academy of Child and Adolescent Psychiatry30:208-217.

Higgins, P. S.

1978 Evaluation and case study of a school-based delinquency preventionprogram: The Minnesota Youth Advocate Program. Evaluation Quarterly 2:215-234.

Hirschi, T.

1969 Causes of Delinquency. Berkeley, CA: University of California Press.

Holmes, A. B., Gottfredson, G. D., and Miller, J.

1992 Resources and strategies for findings. In G. D. Hawkins and R.F. Catalano (eds.), Communities That Care. San Francisco, CA: Jossey Bass.

Howell, J. C., Krisberg, B., Wilson, J. J., and Hawkins, J. D.

1995 A Sourcebook on Serious, Violent, and Chronic Juvenile Offenders.Newbury Park, CA: Sage Publications.

Institute of Medicine

1994 Reducing Risks for Mental Disorders: Frontiers for Preventive InterventionResearch. Washington, D.C.: National Academy Press.

Izzo, R. L., and Ross, R. R.

1990 Meta-analysis of rehabilitation programs for juvenile delinquents.Criminal Justice and Behavior 17:134-142.

Johnson, G.

1984 When law-related education is a deterrent to delinquency: Evaluationmethods and findings. Paper presented at Rocky Mountain Regional Conferenceof the National Council for the Social Studies.

Johnson, G., and Hunter, R.

1985 Law-related education as a delinquency prevention strategy: A three-yearevaluation of the impact of LRE on students. Paper adapted from a reportsubmitted to the National Institute for Juvenile Justice and DelinquencyPrevention on June 6, 1984.

Kenney, D. J., and Watson, T. S.

1996 Reducing fear in the schools: Managing conflict through studentproblem solving. Education and Urban Society 28:436-455.

Lam, J. A.

1989 The Impact of Conflict Resolution Programs on Schools: A Reviewand Synthesis of the Evidence. Amherst, MA: National Association for Mediationin Education.

Law Enforcement News

1996 Truth and DARE: Washington cities shelve anti-drug curriculum.NY: John Jay College of Criminal Justice.

Lindstrom, P.

1996 Partnership in crime prevention: Police-school cooperation. Paperpresented at the annual meeting of the American Society of Criminology,Chicago.

Lipsey, M. W.

1992 Juvenile delinquency treatment: A meta-analytic inquiry into thevariability of effects. In T. D. Cook, H. Cooper, D. S. Cordray, H. Hartman,L. V. Hedges, R. V. Light, T. A. Louis, and F. Mosteller (eds.), Meta-Analysisfor Explanation. Beverly Hills, CA: Sage.

Lipsey, M. W., and Wilson, D. B.

1993 The efficacy of psychological, educational, and behavioral treatment:Confirmation from meta-analysis. American Psychologist 48:1181-1209.

Lochman, J. E.

1992 Cognitive-behavioral intervention with aggressive boys: Three-yearfollow-up and preventive effects. Journal of Consulting and Clinical Psychology60:426-432.

Lochman, J. E., Burch, P. R., Curry, J. F., and Lampron, L. B.

1984 Treatment and generalization effects of cognitive-behavioral andgoal-setting interventions with aggressive boys. Journal of Consultingand Clinical Psychology 52:915-916.

Loeber, R. T., and Dishion, T.

1983 Early predictors of male delinquency: A review. Psychological Bulletin93:68-99.

LoSciuto, L., Rajala, A. K., Townsend, T. N., and Taylor, A. S.

1996 An outcome evaluation of Across Ages: An intergenerational mentoringapproach to drug prevention. Journal of Adolescent Research 11:116-129.

Malvin, J. H., Moskowitz, J. L., Schaps, E., and Schaeffer, G. A.

1982 Evaluation of Two School-based Alternative Programs. The Napa Project,Napa, CA: Pacific Institute for Research and Evaluation.

Malvin, J., Moskowitz, J., Schaps, E., and Schaeffer, G.

1985 Evaluation of two school-based alternatives programs. Journal ofAlcohol and Drug Education 30:98-108.

Mayer, G. R., and Butterworth T. W.

1979 A preventive approach to school violence and vandalism: An experimentalstudy. Personnel and Guidance Journal 57:436-441.

Mayer, G. R., Butterworth, T. W., Nafpaktitis, M., and Sulzer-Azaroff,B.

1983 Preventing school vandalism and improving discipline: A three-yearstudy. Journal of Applied Behavior Analysis 16:355-369.

McPartland, J. M., and Nettles, S. M.

1991 Using community adults as advocates or mentors for at-risk middleschool students: A two-year evaluation of Project RAISE. American Journalof Education 568-586.

Murray, C. A., Bourque, B. B., and Mileff, S. J.

1981 The National Evaluation of the Cities in Schools Program. ReportNo. 4: Final Report (Contract No. 400-77-0107). Washington, D.C.: NationalInstitute of Education.

National Institute for Citizen Education in the Law

1988 Annual Report 1987-1988: Expanding Programs to Reduce Crime andPromote Citizenship. Washington, D.C.: NICEL.

O'Donnell, J., Hawkins, J. D., Catalano, R. F., Abbott, R. D., and Day,L. E.

1995 Preventing school failure, drug use, and delinquency among low-incomechildren: Long-term intervention in elementary schools. American Journalof Orthopsychiatry 65:87-100.

Office of Juvenile Justice and Delinquency Prevention

1995 Delinquency Prevention Works: Program Summary. Washington, D.C.:OJJDP.

Olweus, D.

1991 Bully/victim problems among schoolchildren: Basic facts and effectsof a school based intervention program. In D. J. Pepler and K. H. Rubin(eds.), The Development and Treatment of Childhood Aggression. Hillsdale,NJ: Lawrence Erlbaum.

1992 Bullying among schoolchildren: Intervention and prevention. InR. D. Peters, R. J. McMahon, and V. L. Quinsey (eds.), Aggression and ViolenceThroughout the Life Span. Newbury Park, CA: Sage Publications.

Olweus, D., and Alsaker, F. D.

1991 Assessing change in a cohort-longitudinal study with hierarchicaldata. In D. Magnusson, L. R. Bergman, G. Rudinger, and B. Torestad (eds.),Problems and Methods in Longitudinal Research: Stability and Change. Cambridge:Cambridge University Press.

Perry, C. D., Grant, M., Ernberg, G., Florenzano, R. U., Langdon, M.C., Myeni, A. D., Waahlberg, R., Berg, S., Andersson, K., Fisher, J., Blaze-Temple,D., Cross, D., Saunders, B., Jacobs, D. R., and Schmid, T.

1989 WHO collaborative study on alcohol education and young people:Outcomes of a four-country pilot study. The International Journal of theAddictions 24:1145-1171.

Reyes, O., and Jason, L. A.

1991 An evaluation of a high school dropout prevention program. Journalof Community Psychology 19:221-230.

Ringwalt, C., Greene, J., Ennett, S., Iachan, R., Clayton, R. R., andLeukefeld, C. G.

1994 Past and Future Directions of the DARE Program: An Evaluation Review:Draft Final Report (Award # 91-DD-CX-K053). Washington, D.C.: NationalInstitute of Justice.

Rosenbaum, D. P., Flewelling, R. L., Bailey, S. L., Ringwalt, C. L.,and Wilkinson, D. L.

1994 Cops in the classroom: A longitudinal evaluation of drug abuseresistance education (DARE). Journal of Research in Crime and Delinquency31:3-31.

Rosenthal, R., and Rubin, D. B.

1982 A simple, general purpose display of magnitude of experimentaleffect. Journal of Educational Psychology 74:166-169.

Ross, J. G., Saavedra, P. J., Shur, G. H., Winters, F., and Felner,R. D.

1992 The effectiveness of an after-school program for primary gradelatchkey students on precursors of substance abuse. Journal of CommunityPsychology, OSAP Special Issue 22-38.

Rossman, S. B., and Morley, E.

1995 The National Evaluation of Cities in Schools: Executive Summary(Cooperative Agreement #91-JN-CX-K001). Office of Justice and DelinquencyPrevention. Washington, D.C.: U.S. Department of Justice.

Rotheram, M. J.

1982 Social skills training with underachievers, disruptive, and exceptionalchildren. Psychology in the Schools 19:532-539.

Schaps, E., Bartolo, R. D., Moskowitz, J., Palley, C. S., and Churgin,S.

1981 A review of 127 drug abuse prevention evaluations. Journal of DrugIssues 11(1):17-43.

Schinke, S. P., Botvin, G. J., and Orliandi, M. A.

1991 Substance Abuse in Children and Adolescents: Evaluation and Interventions.Newbury Park, CA: Sage Publications.

Slicker, E. K., and Palmer, D. J.

1993 Mentoring at-risk high school students: Evaluation of a school-basedprogram. The School Counselor 40:327-334.

Snyder, H. N., Sickmund, M., and Poe-Yamagata, E.

1996 Juvenile Offenders and Victims: 1996 Update on Violence. Washington,D.C.: Office of Juvenile Justice and Delinquency Prevention.

Solomon, D., Watson, M. S., Delucchi, K. L., Schaps, E., and Battistich,V.

1988 Enhancing children's prosocial behavior in the classroom. AmericanEducational Research Journal 25:527-554.

Stoil, M., Hill, G., and Brounstein, P. J.

1994 The seven core strategies for ATOD prevention: Findings of theNational Structured Evaluation of What is Working Well Where. Paper presentedat the 12th annual meeting of the American Public Health Association, Washington,D.C.

Thompson, D. W., and Jason, L. A.

1988 Street gangs and preventive interventions. Criminal Justice andBehavior 15:323-333.

Tierney, J. P., Grossman, J. B., and Resch, N. L.

1995 Making a Difference: An Impact Study of Big Brothers/Big Sisters.Philadelphia, PA: Public/Private Ventures.

Tobler, N. A.

1992 Drug prevention programs can work: Research findings. Journal ofAddictive Diseases 11(3):1-28.

Tobler, N. S.

1986 Meta-analysis of 143 adolescent drug prevention programs: Quantitativeoutcome results of program participants compared to a control or comparisongroup. The Journal of Drug Issues 16:537-367.

Tolson, E. R., McDonald, S., and Moriarty, A. R.

1992 Peer mediation among high school students: A test of effectiveness.Social Work in Education 14:86-93.

Weissberg, R. P., and Caplan, M.

1994 Promoting social competence and preventing antisocial behaviorin young urban adolescents. Manuscript submitted for publication.

Weissburg, R. P., Caplan, M. Z., and Sivo, P. J.

1989 A new conceptual framework for establishing school-based socialcompetence promotion programs. In L. A. Bond and B. E. Compas (eds.), PrimaryPrevention and Promotion in the Schools. New York: Sage.

Weissberg, R. P., and Greenberg, M. T.

In press School and community competence-enhancement and preventionprograms. In W. Damon (Series ed.) and I. E. Sigel and K. A. Renniger (Vol.eds.), Handbook of Child Psychology: Vol. 5. Child Psychology in Practice(5th ed.). New York: John Wiley and Sons.

Youth Today

1994 Youth crime prevention.Youth Today 3(6):6-7.

Chapter Six

LABOR MARKETS AND CRIME RISK FACTORS

by Shawn Bushway and Peter Reuter

I. INTRODUCTION

Employment and crime have a complex relationship. For an individual,they can be substitutes or complements. For example, some people choosecrime rather than legitimate work because of an expectation that they canmake more money from crime and/or because they find it more rewarding inother ways (Katz, 1989; Bourgois, 1995). On the other hand, the workplacecan offer opportunities for certain kinds of crimes that are more difficultto commit elsewhere, such as theft of inventory or selling of gamblingservices.

The relationship between employment and crime at the community levelthe relationship is equally ambiguous. Crime in a community is the outcomeof the intersection between the propensity to commit a crime and the opportunityto commit a crime. For example, in a given community over time, high employmentmay be associated with reduced presence of residents and greater wealth,thus increasing criminal opportunities. On the other hand, low employmentalso provides better legitimate work opportunities for potential offenders,thus reducing their propensity to commit crime. Looking across communities,one can see the same potentially countervailing influences; poor communitiesoffer weak job prospects but also (except for drug markets) financiallyunrewarding criminal opportunities. At this level, crime rates may dependnot on the level of employment but on a much more fundamental set of socialand individual characteristics.

Pure theory is not likely then to provide guidance about the strengthor direction of the relationship between employment and crime. Howeverit is at least plausible that a strong negative relationship exists. Atthe descriptive level, those who commit crimes tend to be out of the laborforce or unemployed. The communities in which crime, particularly violentcrime, is so heavily concentrated show persistently high jobless rates.Increasing employment and the potential for employment for individualsand communities currently at high risk of persistent joblessness may havea substantial preventive effect on crime. Thus a comprehensive assessmentof crime prevention programs should include those aimed at increasing employment.

Our review includes any program which aims to increase the employmentof individuals or populations at risk of serious criminal involvement.We exclude general economic stimulus policies, (e.g., looser monetary policyaimed at lowering interest rates) though these may in theory reduce crime;such policies are driven by other factors and in any case the evidenceon the aggregate relationship between employment and crime is very ambiguous.We include, however, a range of community and individual programs whichdo not specifically target crime, as indicated by the frequent omissionof crime, or even risk-factors for crime, as an outcome measure. Thus muchof this review assesses just how well such job-training and creation programs,distinct from those aimed at criminal-justice-involved offenders, actuallydo at increasing employment for the targeted community or individual. Thecrime consequences are inferred from our review of the relationship betweenemployment and crime at various levels.

For policy purposes the reciprocal relationship of crime and employmentpresents a major challenge. Areas of high crime are unattractive for investment.Both property and personnel are at risk; goods are stolen, premises damaged,employees assaulted and customers intimidated. Attracting capital requiresa reduction in crime so as to allay the legitimate concerns of investors/employers.On the other hand, crime reduction on a large scale may require the creationof employment opportunities for the large numbers of young adults thatare the source of so much of the crime in the area. At the same time, manyoffenders lack the skills needed to obtain and retain attractive jobs,that is positions that pay enough to avoid poverty (well above the minimumwage for a two-parent, two-child household with only one wage earner) andwhich offer potential progress and a sense of accomplishment. Thus improvingtheir work force skills may be essential even when capital can be attractedinto the community.

Existing programs aimed at reducing crime through employment and/orincreasing employment in high crime areas fall into the following two maincategories:

  • Supply-side programs aim to improve the attractiveness of individualsto employers. Mostly these programs increase the potential productivityof the worker through education or job training. However the category includesprograms that take account of the fact that many high risk individualsare handicapped by their location. These programs move people to jobs,either by transportation subsidies or by actually providing access to housingin lower crime communities nearer areas of high employment potential. Thelatter also may have crime prevention effects by allowing children to growup in communities with more employed adult role models.
  • Demand-side programs aim to reduce the costs of employment borne bythe employer. One way to do this is through wage supplements or subsidizedbonds (insuring the employer against theft by the employee) for ex-offenders.Another alternative is community development programs which lower costsfor businesses locating in particularly needy communities. The influx ofcapital into communities characterized by low employment and high crimeshould generate jobs and thus, by a variety of mechanisms, reduce crimein the community.

Section II briefly surveys the theoretical and empirical literatureon the relationship between crime and employment at various levels. SectionsIII and IV survey supply and demand side programs. Each examines the evaluationevidence on program outcomes: For only a very few evaluations do we haveexplicit findings on the crime consequences of the intervention; the restproviding only employment measures. Section V then offers an integrationof all these findings and Section VI offers conclusions and recommendations.

II. THE RELATIONSHIP BETWEEN EMPLOYMENT AND CRIME

The relationship between crime and employment has been a long standingissue in research, involving a range of paradigms.1Fagan (1995) and Freeman (1995) provide recent reviews, particularly focusedon understanding how the returns from crime and legitimate work jointlyaffect the decision to engage in crime. We propose here to give more attentionto the multiplicity of relationships between the criminal participationand work opportunities that operate at different levels (individual andcommunity) and at different points in an individual's life-span (school,young adult, adult). Our goal is not to make theoretical contributionsbut to give a better grounding to an analysis of programmatic and policyoptions.

Theoretical Perspective

Fagan (1995) and Uggen (1994) have identified four major theoreticalexplanations for the link between employment and crime: economic choice,social control, strain, and labeling theory.

Economic choice theory (Ehrlich, 1973) posits that an individualmakes choices between legal and illegal work based partly on the relativeattractiveness of the two options. Moral values still influence actionsbut are assumed not to change with economic factors. It is, like economictheory generally, about response to changes or differences. If legal workbecomes less rewarding or if illegal work becomes more rewarding, individualsmay shift to crime and away from legal work. Education plays a role inframing choices; low educational attainment, which now puts young malesat risk of frequent periods of unemployment and of achieving only low payingand unsatisfactory jobs, will be associated with high crime participation.This is exactly what Freeman claims happened in the late 1980's: "Giventhe well-documented growth of [legitimate] earnings inequality and fallin the job opportunities for less-skilled young men in this period, andthe increased criminal opportunities due to the growth of demand for drugs,the economist finds appealing the notion that the increased propensityfor crime is a rational response to increased job market incentives tocommit crime." (Freeman, 1995:177-178.)

Notice that within this theory, the crimes in question are income-generatingcrimes which are used to replace income gained from legitimate means. Thetheory offers no account of non-income generating crime. Much violent crimeis expressive (e.g., an enactment of drunken anger) rather than instrumental(e.g., aimed at ensuring success of a robbery). However economic theoryis not entirely silent on violent crime. Employment should raise the opportunitycost of incarceration (i.e. what the individual loses with his freedom),both through loss of earnings and the loss of work experience; this mightdeter acts that endanger the individual's freedom.

The economic choice framework allows individuals to engage in both legitimatework and crime simultaneously; this is appropriate as most offenders alsomaintain some relationship to the workplace over their criminal careers(Reuter et al., 1990). What may be affected by changes in the relativeattractiveness of crime and legitimate work is the allocation of time betweenthe two types of income generating activities; better employment opportunitiesreduce the fraction of time going to crime. Importantly, this theory hasfurther implications beyond a simple contemporaneous choice of legal versusillegal work. The individual, particularly in adolescent years, also hasto decide how much to invest in human capital (education and other workforcerelevant skills). If the legal labor market opportunities appear weak,a youth is less likely to make adequate investment in acquiring the humancapital necessary for success in the legal labor market. As a result, thistheory can explain both participation in income-generating crime and under-investmentin human capital which reduces legitimate income later.

Control theory claims that employment exerts social control overan individual (Gottfredson and Hirschi, 1990). The absence of employmentfor an individual leads to a breakdown of positive social bonds for thatindividual. That in turn is hypothesized to induce the individual to increasehis criminal activity, both violent and income related. This theory, expandednaturally to cover not just individuals but areas, is a key part of WilliamJulius Wilson's analysis of inner city problems. Using a series of carefullyconstructed studies of poverty areas in Chicago, he claims "many oftoday's problems in the inner-city ghetto neighborhoods - crime, familydissolution, welfare, low levels of social organization and so on are fundamentallya consequence of a disappearance of work" (Wilson, 1996: xiii). Employmentis seen as the main builder of pro-social bonds and institutions in a communityand its absence results in large scale disorder.

Anomie is another more aggregate level theory (see Uggen [1994]for a concise summary targeted to this issue). This theory suggests thatfrustration caused by income inequality and other aggregate level problemswill cause individuals to resort to crime out of frustration.

One small area of theory that explicitly includes the idea that crimeitself could be criminogenic is labeling theory (Lemert, 1951).Individuals who participate in crime acquire stigmatic labels (both toothers and to themselves) and are then denied opportunities because ofthese labels. What is intriguing about this theory is that it suggeststhe very real possibility of feedback between employment and crime. Thisfeedback suggests that cessation from crime will be difficult once criminalactivity has been initiated, particularly if the offender acquires an officialrecord (see Schwartz and Skolnick [1964], Nagin and Waldfogel [1994, 1995],Bushway [1996]).

Labeling theory points also to a community level connection betweencrime and employment- joblessness in an area may be caused by past criminalactivity of the residents, as well as the converse. In a sense the communityor area is "labeled", which makes it difficult for the communityto attract investment. This is a point first made forcefully by formerNIJ Director James K. Stewart (1986).

These theories, potentially complementary, point to important potentialfeedback between crime and unemployment. Programs aimed solely at improvingan individual's employability (motivated by economic choice) or solelyat increasing the number of jobs in an area (motivated by all four theories)are vulnerable, the first to the failure of program graduates to find jobsand the second simply to the difficulty of achieving the goal of providingjobs. In the extreme case, a community including many individuals withlow human capital, limited ties to positive social structures and institutions,and negative labels is likely to be characterized by both high crime andlow employment, with complex interaction between the two problems. Theorysuggests that areas characterized by both high crime and low employmentrequire attention to all three factors: weak social institutions, low humancapital and negative labels.

Research on Crime and Employment

We now review empirical research aimed at assessing the relationshipbetween crime and employment,2a necessary bridge between the theories and the program evaluations. Thisresearch has been conducted at many different levels of aggregation, includingnational time-series data, state and local cross-sectional data and individual-leveldata.

National level. A review by Chiricos (1986) finds that most nationallevel analyses have yielded weak results on the crime-employment relationship.Freeman (1994) claims that this is primarily because of the weakness inthe time-series statistical model with national data. One exception isa paper by Cook and Zarkin (1985). They report mixed results from an analysisof business cycles from 1933 to 1982. In general, crime has increased overthe last 50 years. However, homicide rates did not vary systematicallywith the business cycle while the rate of increase in burglary and robberyhas been higher during the economic downturns than during the upturns.This is consistent with the idea that low employment leads to an increasedpropensity to commit property crime while violent crime is driven by otherfactors. At the same time, they found that auto-theft was actually pro-cyclical---auto-theft increased faster when the economy improved and more slowly whenthe economy declined. This is consistent with the idea that the opportunityfor auto-theft increases when employment (and hence disposable income)increases. We shall present no other findings at this level of aggregationbecause it seems to provide least insight into those policy issues withwhich we are particularly concerned.

Community Level Chiricos does find, however, that at lower levelsof aggregation (states, counties and cities) roughly half of all reportedstudies show a positive and statistically significant relationship betweenemployment and crime, using post-1970 data.3The fraction of positive results increases to almost 75 percent of allstudies when property crimes are analyzed separately from violent crimes.

Individual level Analyses of individual level data have attractedmore attention as these data have become available. Studies of the 1945Philadelphia birth cohort have shown that unemployment is associated withcrime (e.g., Wolfgang, Figlio, Sellin, 1972), a finding that is reportedin numerous other studies. However the causality is uncertain. Sampsonand Laub (1993) argue that employment per se or by itself does not reducecrime or increase social control; it is only stability, commitment andresponsibility that may be associated with getting a job that has crimereducing consequences. Gottfredson and Hirschi (1990) argue that the relationshipis essentially spurious, reflection of a common third factor which theycall the level of individual social control.

Economic choice theory is further supported by evidence showing thathuman capital influences earnings, and earnings influence recidivism byex-offenders (Needels, 1996). Social control theory seems to have relevance,too, within the context of economic choice. Farrington et al. (1986) tiecrime more directly to employment by examining the timing of crime andemployment over almost 3 years for a sample of teenage males in England.They show that property crimes are committed more frequently during periodsof joblessness. However, this relationship held only for those who werepredisposed to crime (as reflected by self-reports on earlier criminalactivity and moral values); otherwise spells of joblessness did not inducemore criminal offending.

This brief review establishes that researchers have measured a relationshipbetween crime and employment, and that a number of mechanisms, operatingboth at the individual and community level, may explain the relationship.The key remaining question is whether or not programs aimed at increasingemployment for at-risk populations can attain that goal and reduce crime.

III. SUPPLY SIDE PROGRAMS

Job Training and Education

The earliest labor market-oriented crime prevention programs followedjust this logic -- providing legitimate employment or employment skillsto at-risk individuals would reduce their criminal participation. Numerousprograms were developed to provide basic education, vocational trainingand work experience for youth in high crime and high unemployment communities.The federal government spends large sums ($2.5 billion in 19944)on skills-developing programs aimed at increasing the employment prospectsof individuals who are at high risk of being persistently unemployed. Mostof these interventions target youth, particularly adolescents, on the reasonableassumption that early interventions have higher pay-off if successful.The other large set of interventions targets those already involved withthe criminal justice system, since they are also known to have low humancapital.

We will consider these two groups of interventions separately, sincethe division corresponds to differences in institutions and outcome measures.The programs for youth generally are provided by social service agencieswhile those for offenders frequently occur in correctional settings. Moreovercriminal justice program evaluations almost always include recidivism asan outcome measure, and sometimes do not include employment, while thegeneral population programs always include employment as an outcome measurebut rarely crime. Programs are further divided into those aimed at youth,broadly defined, and those aimed at adults; these have different theoreticaljustifications and programmatic content.

Job Training Programs Connected to the Criminal Justice System

Introduction

Targeting human capital development programs at offenders while in,or just leaving, the criminal justice system has the merit of focusingresources on the highest risk group. It is a human services equivalentof Willie Sutton's famous line about the banks; in this case, we are goingwhere the crime is. Like Sutton's strategy, it also has an obvious weakness;just as banks are well guarded, so offenders in the criminal justice systemhave already developed behavior patterns that are difficult to reversewith educational programs.

We divide programs by age of the target population: juvenile and adult.That reflects the fact that juveniles seem most suitable for programs thatfocus on the development of human capital, as is true of education generally;adult programs give more emphasis on reintegration into the workforce.We will also distinguish programs by whether they are in prison or post-release.

Juvenile Offenders5

Young offenders are confined in institutions which generally give moreemphasis to rehabilitation than do adult correctional facilities. Educationand training programs frequently fit into a broad array of habilitationand rehabilitation services generally. Indeed, it is difficult to identifythe main effects of these programs alone, precisely because they are imbeddedinto a bigger set (e.g., cognitive therapy, substance abuse treatment)which may interact with education and training.

Generally the findings are of negligible or modest effects; see Table6-1. All evaluations point to a problem getting participants to completethe program once started; high drop out rates indicate either that theprogram was poorly implemented or it was unattractive to many of the participants.Some of the programs also involved a very low level of services for theclients; even if they were well done it would seem implausible that theycould have large behavioral consequences.

TABLE 6-1

Criminal Justice System Programs

Studies  Scientific Method Description of Intervention and Findings   Score   (Number of cases Treatment/control) YOUTH Greenwood & 5 PCYC offers a comprehensive array of intervention services and activities Turner 1993, including counseling, peer support and skills training. One year follow-up Paint Creek (73/75) data showed no significant differences in arrests or self-reported Youth Center delinquency between experimental and control groups. Lattimore, et 5 VDS involved the use of vocational skills training, job readiness, and al. 1990, employment skills training. 36% of the experimental group, compared to 46% Vocational (154/130) of the control group, were re-arrested following release (statistically Delivery significant p<.10). System Leiber & 3 Rehabilitative strategy that uses social skills training, pre-employment Mawhorr 1995, training, and job placement opportunities (4 months). Youths who received Second Chance (57/56) the treatment intervention are as likely to be involved in official program offending as are the equivalent matched comparison (37% compared to 29%). Piliavin & 5 Low-skilled and low-wage rate jobs provided for participants for no longer Masters 1981, than 12-18 months. Found little effect on delinquents' post-program Supported Work (2200 ex-offenders1400 employment or on their criminal activity after program participation; for ex-addicts 1200 youth) adult offenders and drug addicts, especially those over 35, increased employment and reduced crime effects were found. ADULTS Adams, et al. 3 Participation in academic and vocational programs bore no relation to 1994, PERP reincarceration; the % of inmates who were returned to prison did not vary (5608/ 8001) significantly across groups of program and non-program inmates. Berk, et al. 5 Intervention included the eligibility for unemployment benefits at several 1980, TARP levels of the alternative of job counseling. Membership in any of the (775/200) three experimental groups eligible for payments of job counseling had no statistically significant impact on either property or non-property arrests. 1976, Baltimore 5 Treatment groups received either income maintenance(3 months), job LIFE placement or both. Financial aid treatment groups were re-arrested for (216/ 216) property crimes 8.3% less (statistically significant) than control and job assistance groups; they were re-arrested 7% less for other crimes (not statis. sign.) Finn & 3 Findings suggest that ex-offender status had no effect on employment at Willoughby 1996, termination or follow-up; only the barrier of being long-term unemployed JTPA (521/734) negatively influenced prospect of employment. Hartmann, et al. 3 Treatment included employment skills classes, job club peer support, life 1994, KPEP skills and GED training. Offenders who successfully completed the program (156) were significantly less likely to recidivate than those who did not (felony arrest p<.004; any arrest p<.005). Henry 1988, CADD 3 Provided inmates with job training and skills along with substance abuse counseling. No difference found between industry working inmates and (34/56) non-industry inmates with regard to the proportion of disciplinary reports per month in prison. Home Builders 1 Involves an 8 week pre-apprenticeship carpentry training program for 1996, TRADE incarcerated adult offenders. Well over half of program graduates were (219) placed in related jobs in 4 out of the 5 sites; 3 month. recidivism results (7.3%) are consistent or better than those of other vocational programs. Maguire, et al. 3 Intervention involved participation in prison industry for at least 6 1988, PIRP continuous months. After controlling for differences between the two (399/497) groups, the recidivism rates for industry and non-industry participants were virtually identical. Menon, et al. 3 RIO provides services such as educational and vocational training 1992, Project pre-release and job search and placement assistance post -release. It RIO (Evaluation not also uses vouchers from the Targeted Jobs Tax Credit program and federal clear) bonding as special incentives for prospective employers. Positive and significant impact on employment and negative and significant impact on recidivism, particularly for the high risk offenders. Piliavin & 5 Strong positive effect of program participation on ex-offender employment Gartner 1981, declines over time until the experimental-control differential in hrs. Supported Work (1117/1194) worked per month has disappeared (1 yr); no recidivism impact. Saylor & Gaes 3 Treatment group had either worked in prison industry, or had received 1996, PREP in-prison vocational instruction. Long-term findings (8 yrs.) show that (over 7000) male prison industry subgroup had 20% longer survival times (time before committing new offense) than comparison group; training program subgroup had 28% longer survival times: both results are statistically significant. Spencer 1980, 4 Treatment involved career counseling, job placement, and special Ex-Offender counseling services. Ex-offenders enrolled in the Clearinghouse program Clearinghouse (478/478) were significantly more likely to obtain employment and/or constructive activity than those not enrolled. Van Stelle 1995, 4 Provides in-prison training, as well as post-release transition services STEP such as job placement assistance. There were no significant differences (89/42) between graduates and controls with regard to arrest after release. Vera Institute 4 Offers counseling and vocational opportunities such as job training or 1972, academic placement for a period of 90 days in lieu of tradition court Manhattan Court disposition. During the initial 23 months of operation, the re-arrest Employment rate for the successfully dismissed group was about 50% less than that of Project (214/91) the terminated or control groups (p<.01). No reported results for entire treatment group vs. control. Baker and Sadd 5 Offers counseling and vocational opportunities such as job training or 1981, academic placement for a period of 120 days in lieu of traditional court Court Employment disposition. There was no difference in recidivism between the treatment Project and control groups initially, after 12 months or after 23 months. (410/256) 

For example, Lattimore, Witte and Baker (1990) report a randomized controltrial, one of the few in the literature, for 18-22 year old offenders intwo North Carolina prisons. We classify this as a juvenile population becausethe subjects are indeed early in their post-school careers but note thatthey have been serving time in adult correctional facilities. 295 inmateswere enrolled in a Vocational Delivery System (VDS) aimed at identifyingvocational interests and aptitudes, providing appropriate training forthe individual and then helping with post-release employment. Subjectswere picked from all inmates in the two institutions who were aged 18-22,committed for property offenses, had IQ no less than 70, were in good healthand within 8 to 36 months of an in-state release. Data were available for154 of the experimental and 130 of the controls at approximately the two-yearmark.6

No employment results were reported; thus the impact of the programon workplace performance must be inferred from the impact on crime. But" (t)hose participating in the program were more likely than controlgroup members to complete vocational training and other programs. . . .VDS participants were less likely to be arrested following release fromprison." (p.117) At 24 months the control group showed a 50 percentrecidivism rate (based on arrest records) compared to 40 percent for theexperimental group. The difference was only weakly significant (10 percentlevel) and barely that for tests on other outcome measures. This relativelylarge effect exists even though only 18% of the people assigned to theVDS program actually completed the program. The completers (i.e. thosewho received all the services included in VDS) were substantially lesslikely to be arrested. This combination of high dropout and excellent resultsfor completers is typical of other programs that strive to challenge enrollees.The problem is that researchers do not know whether or not the programcompleters are the same people who would have succeeded in the absenceof the program - therefore looking only at the program graduates leadsto selection bias. On the positive side, this study provides evidence thatvocational programs aimed at young property offenders could have positiveoutcomes if implementation and participation problems could be resolved.

Piliavin and Masters (1981) report similar weak findings for youth enrolledin the "Supported Work" program, using a randomized assignmentof 861 youth (average age 18) in five sites. The program lasted 12-18 monthsand provided work experience along with a stipend in a sheltered work environment.Although this program was not officially run through the criminal justicesystem, we include it in this section because two thirds of the youth hadan arrest before entry into the program and 28 percent had been incarcerated,for an average of 20 weeks; they were predominantly Black (78%) and Hispanic(16%).

Both employment and official criminal justice outcomes were reported.The labor market outcome differences were non-significant and small; e.g.,at 36 months the experimental group worked 83.3 hours per month, comparedto 75.8 for the control group. The crime differences were weakly significant(10 percent level). At 27 months, 30 percent of the experimental grouphad been arrested, compared to 39 percent of the control group; the differencewas larger and had greater statistical significance for those without priorarrest. Although this effect size is relatively large (a 30% differencebetween controls and experimentals more than 2 years after the programended), the evaluators concluded that there was no evidence of an effectfor youth. As in the VDS case, the evaluators point to failure of mostparticipants to complete the program as one of the sources of error inthe study.

Other programs tended to have fewer resources, and the evaluations haveweaker designs. Leiber and Mawhorr (1995) used a variety of matched controlgroups to assess the impact of the Second Chance program on youth who werein court but not yet sentenced to an institution. Second Chance involves16 weekly group meetings aimed at developing certain social skills, alongwith a pre-employment training program (including how to conduct an independentjob search, interview for a job and demonstrate good work habits). With85 program entrants (only 57 of whom completed it), the test does not havemuch statistical power. The findings were of no significant differencesin official arrests; the control group actually showed lower recidivismthan the experimental group (completers or drop-outs). Again the evaluationpointed to the lack of treatment integrity. Note that this program alsoinvolved more than just training and education.

A recent OJJDP review of correctional educational programs noted thelack of rigorous evaluation of juvenile vocational education programs withinthe criminal justice system (OJJDP, 1994).7The one "rigorous" evaluation cited by OJJDP is the New Prideprogram in Denver. New Pride is a community-based program that providesa year of intensive non-residential treatment and training, including participationin an on-site business run by the program. The evaluation consisted oftracking the success of the program participants without any comparisongroup. This is a poor evaluation design that does not meet minimal standards(less than a "1" on our scale). Widespread replication of thisprogram, while encouraged by its evaluators (James and Granville [1984]),does not appear to be justified by the quality of the evaluation.

Adult Offenders

Though both theory and political rhetoric emphasize juveniles as themost suitable targets for training and education, a large fraction of adultoffenders in the criminal justice system have poor educational and jobmarket records. That fact was the original source of interest in the early1960's in assessing whether recidivism might be reduced by providing theseadults with additional educational and job skills. Moreover the life coursemodel of crime suggests that many offenders may be more receptive to workthan adolescents.

Secondary reviews from the early 1970's, after these programs had beenaround for roughly 10 years, were uniformly negative. The Department ofLabor's Manpower Administration sponsored research on these programs, andprovided a comprehensive review of the research in 1973 (Rovner-Pieczenik,1973). Despite strong commitment and great enthusiasm by program operators,the study reluctantly reports that very few programs led to a substantialdecline in recidivism. By way of explanation, the report highlights problemsin persuading correctional institutions to focus on education and post-releaseobjectives. The report also highlights the great educational deficits ofthe offenders, who are generally high school dropouts reading several yearsbelow grade level with no discernible job skills. The author concluded"that we entertain no fantasies about the degree of change which manpowerprojects for the offender can help to bring about. Some offenders willremain unemployed and unemployable no matter what programs are available."(Rovner-Pieczenik, 1973:77)

These disappointing conclusions were communicated to a much broaderaudience with Martinson's (1974) widely read review of 231 rehabilitative(including employment-based) programs. Martinson concluded that "withfew and isolated exceptions the rehabilitative efforts that have been reportedso far have had no appreciable effect on recidivism" (p.25). Thisreport has often been held responsible for the decline of the rehabilitativemodel in corrections (see the chapter on Corrections in this volume) andhas limited the research done on these programs.

The sheer numbers of offenders, however, have led correctional officialsto continue their efforts to curtail recidivism by reintegrating ex-offendersinto the workforce. We found only one recent secondary review of adultoffender prison educational programs (Gerber and Fritsch 1993), which (unsurprisingly)is explicitly oriented at rebutting the pessimism of Martinson. Most ofthe evaluations reviewed compare program enrollees with a matched comparisongroup of people who did not enroll in the program. These evaluations aresubject to selection bias, since people who enroll in the programs arelikely to more motivated than those who do not enroll. These motivatedpeople might be expected to do better even without training programs.

Given this caveat, the review found that three out of six studies ofpre-college education programs consisting of classroom education had anegative and significant (but small) impact on post-release recidivism.Three out of four programs showed a statistically significant increasein post-release employment. Four out of six college education programs(again primarily classroom education) showed a statistically significantdecline in post-release recidivism, although the effect was small, andthere is no evidence that college education leads to increased employmentoutcomes. Finally, four out of six studies found that vocational educationconsisting of participation in training and prison industry programs leadsto a decline in post-release recidivism. Only two out of four studies showedthat vocational education actually led to improvement in post-release employmentoutcomes. In fact, the one study that had some random assignment (Markley1983) showed that vocational education had no effect on post-release recidivismor employment. The report concludes that while the evidence is mixed (andtherefore encouraging), better evaluations which control for selectionbias are needed.

Our own review of these programs found that it is difficult often totell exactly what is involved in a program. For example, one of the betterstudies of prison industry programs was done by the U.S. Bureau of Prisons(Saylor and Gaes, 1996), following 7,000 individuals. Inmates were consideredto have participated in the program if they had participated in industrialwork within the prison, or had received in-prison vocational training orapprenticeship training. This program participation is so broad that itis hard to determine which program or program element led to the observed35% decline in re-arrests (by federal authorities only) for program participantsrelative to the control group. One the plus side, this seems to be clearevidence that vocational education in federal prisons helps to reduce crime.This is an important positive contribution. However, the lack of precisionmakes replication in other prison systems difficult. What program worked,and did it work by increasing employment?8

This study also highlights the problem of selection bias. When programparticipation is open to everyone with no restrictions, it becomes difficultto claim that the non-participants are identical to the participants evenif regressions are used to control for observed differences between thegroups. Unobserved differences in motivation could account for much ofthe resulting change in behavior, otherwise attributed to the training/vocationalprogram.

We found one prison-based program which attempted to perform a truerandomized experiment to control for selection bias. This program -- SpecializedTraining and Employment Project or STEP -- was run by the Wisconsin Departmentof Corrections and evaluated by the University of Wisconsin Medical School(Van Stelle, 1995). This program randomly assigned a well-defined groupof offenders to a six-month program prior to release which included participationincentives, classroom and job training in the institution, and post-releaseemployment assistance. This project showed no decline in recidivism afterthe first year of the program, but the process evaluation stressed theextraordinary difficulty in implementing a program of this intensity withinthe prison system.

Another approach which avoided prisons entirely was the pre-trial intervention,a major movement during the 1970's. The concept of pre-trial diversionswas attached to the labor market in the Court Employment Project. Thiswas evaluated twice by the Vera Institute, first in the late 1960's (VeraInstitue 1970) , and then again during 1977-1979 (Baker and Sadd, 1981).In the first less rigorous study, non-serious offenders were offered theopportunity to participate in a 90-day job training and placement program.Successful completion of the program resulted in the dismissal of all charges.Less than half of the participants successfully competed the program. 12months after the completion of the program, only 15.8% of the successfulcompleters had recidivated, compared to 31% of the non-completers and thecontrol group. Again, the problem of selection bias eliminates the abilityto say for sure that the program worked -- the difference between all theprogram participants (23.6% recidivism rate) and the control group wasnot statistically significant. Low dosage, problems with implementationand data collection are again cited as part of the reason for the weakresults.

By the time the more rigorous study was undertaken almost 8 years later,the program had been assumed into the New York City government and hadgrown significantly. 410 arrestees were assigned to the program, while256 controls went through the normal court process. The evaluators foundno statistically significant difference between recidivism for the twogroups, during the diversion period, twelve months after the diversionor 23 months after the diversion. Partial explanations for the failureof the program include the large disturbance in the program immediatelybefore the evaluation due to New York City's budget crisis. However, theevaluators concluded that there were systematic problems with the structureof the pre-trial diversions. For example, counselors did not believe thatit was realistic to change the attitude of offenders towards work in 4months, especially since participants typically lived in criminogenic enviromentsremoved from the world of work. Therefore, the training program was notseen as a route to real employment (and hence non-recidivism) but ratheras a route away from jail time. In addition, the evaluators felt that theprosecutors had started using the program to control offenders who wouldotherwise have their cases dismissed, instead of diverting cases whichwould not be dismissed away from the courts (Hillsman, 1982).

Another approach concentrates on transitional assistance after an individualleaves prison. Job Training Partnership Act (JTPA) programs have attemptedto help ex-prisoners by giving them a) job search assistance, b) remedialeducation, c) occupational skills, d) work experience, e) on-the-job training,or f) customized training for a particular employer. The one evaluationof these programs (Finn and Willoughby, 1996) looked at all 521 ex-prisonerswho enrolled in JTPA training programs in the state of Georgia for oneyear starting in July 1989. These enrollees were compared to 734 non-offenderJTPA participants. The researchers found no sign of any difference in employmentoutcomes either at program termination or 14 weeks after termination betweenthe ex-prisoners and the non-offenders. This result is hard to interpret.Other studies have shown a consistent difference between ex-offenders andother workers. Perhaps the finding of no difference indicates that JTPAprograms have helped eliminate some of the stigma of offending. However,since JTPA programs are generally regarded as only minimally effectiveat improving employment outcomes, that conclusion is hypothetical at best.

Another large federally-funded program tried in the late 1970's involvedthe use of income supplements during post-release in order to lessen theneed to commit crime for money at a time when it may be particularly difficultto find a job. These randomized experiments known collectively as the TransitionalAid Research Project [TARP] (Berk, Lenihan and Rossi, 1980) showed thatno combination of job training and transitional income support could reducearrest rates. TARP built on a smaller Baltimore LIFE (Living Insurancefor Ex-Offenders) experiment, carefully designed and evaluated (Mallarand Thornton, 1978; Rossi, Berk and Lenihan, 1980; Myers, 1982): The LIFEevaluations found that even combinations of job assistance and counselingfor one year had no impact on recidivism but that the transitional paymentsdid make a statistically significant difference. Perhaps TARP could notmaintain the program integrity of LIFE once the program was expanded.

Despite the failure of TARP, long term follow-up of the Georgia TARPsubjects by Needels (1996) demonstrated that the intuition of these programsis still valid - Needels found that the ex-offenders with jobs commit fewercrimes than the ex-offenders without jobs, and those with higher earningscommit fewer crimes than those with lower earnings. Even after 30 yearsof trying, however, no program -- in-prison training, transitional assistance(both in kind and monetary assistance) or pre-trial diversion -- has consistentlyshown itself capable (through a rigorous random assignment evaluation)of decreasing recidivism through labor-market orientated programs, insideor outside of prison. These results might exist because offenders are eithertoo deeply entrenched in crime or the criminal justice system is not aneffective delivery system for these types of programs.

Offender-based programs come late in criminal careers, simply becauseincarceration or even conviction tends to come late. There are strong argumentsfor intervening early. The next subsection reviews programs that are aimedat high-risk youth before they become involved with the criminal justicesystem.

Job Training and Education Programs for At-Risk Youth

A large number of relatively well-funded governmental programs havetried to boost the labor market performance of at-risk youths (high schooldrop-outs, kids from poor households or poor communities). Although wedo not have total expenditures for all such job training programs, thelargest single program, Job Corps, enrolled 60,000 youth at a total costof $970 million in 1993, while Title II-C of the JTPA (Job Training andPartnership Act)9 enrolled360,000 youth at a total cost of $650 million. These programs have undoubtedlyattracted the largest amount of government spending of any single labormarket category in this review. Encouragingly, there are also many rigorousevaluations, with most studies using some form of randomized experiment(method score 4 or higher); see Table 6-2. In reviewing the findings ofthese evaluations, we rely primarily on three reviews of the literature:Donohue and Siegelman (1996), Heckman, (1994) and U.S. Department of Labor(1995).

Programs aimed at youth tend to take three forms, arrayed below in orderof increasing expense and program intensity.

1) The provision of summer work or other forms of subsidized employmentin either public or private sector organizations.10These programs typically cost about $1,000 (in terms of 1995 dollars) perparticipant and lasted about three months. The Summer Youth Employmentand Training Program [SYETP] is the Department of Labor's current summerjobs program, providing minimum wage summer jobs and some education tohundreds of thousands of disadvantaged youth, aged 14-21. Less typicalis the more intense Supported Work program from the late 1970s, which providedabout one year of full-time public sector employment to minority high schooldrop-outs aged 17-20, with job search assistance at the end of the workperiod.

TABLE 6-2

Non-Criminal Justice System: At-Risk Youth

 Studies  Scientific Method Description of Intervention and Findings   Score (Number of    cases  Treatment/control) Summer Jobs/ Subsidized Work Ahlstrom & 3 Combines work experience program with a modified academic program. Havighurst There appeared to a negative effect on arrest, as the experimental 1982, Kansas (~100/~100) group was more likely to be arrested by the age of 16 than was the City Work/Study comparison group (51% versus 36%). Cave & Quint 5 Services of Career Beginnings include summer jobs, workshops and 1990,Career classes, counseling and the use of mentors lasting from junior year of Beginnings (621/612) high school through graduation. Experimentals were 9.7% more likely to attend college than controls (stat. signif);they therefore worked less and earned less. Farkas, et al. 4 Guaranteed full-time summer jobs and part-time school year jobs to 1982, YIEPP disadvantaged youth who stayed in school. School year employment (2778/ 1255) doubled from 20% to 40%, while summer employment increased from about 35% to 45%; however, YIEPP was unable to attain its goals of increased school enrollment and success despite the school enrollment requirement. Grossman & Sipe 5 Program lasting 15 months which involves remediation, life skills, 1992, STEP summer jobs over two years and school-year support. STEP had little (1613/1613) or no impact on youth's educational experience and had not altered employment patterns for either in-school or out-of-school youth. Maynard 1980, 5 Structured transitional employment program which offers limited term Supported Work employment at relatively low wage rates for up to 12 or 18 months., (570/682) combined with peer group support and close supervision. Up to 18 months post-program, there was a significantly larger % of treatment group youth employed; there was no significant impact on arrest rate of youths. Summer Youth N/A Provides summer jobs for youth. Program appears to greatly increase Employment and summer employment rates among disadvantaged youth in sites where jobs Training are provided; have not investigated whether SYETP creates positive Program (SYETP) long-term impacts on employment after participants leave their summer jobs. Short-Term Training Programs Bloom 1994, 5 Federal government's major training program for disadvantaged youth JTPA which provides on average of 5 months of services including on-the-job (total of 4777) training, classroom training, and job search assistance (an average of 420 hrs of service). After 30 months no increase in earnings was found, and there was no decrease in crime rates. Cave, et al. 5 Provides instruction in basic academic skills, occupational skills 1993, JOBSTART training, training related support services and job placement (988/953) assistance. JOBSTART led to a significant increase in the rate of GED attainment, or completion of high school. In the final two years of the follow-up, experimentals' earnings appeared to overtake those of controls, but the magnitude of this impact was not significant. Wolf, et al. 4 Provides job search assistance, educational services and job 1982, 70001 preparation classes to high school dropouts (an average of 80-90 hrs. Ltd. (535/ 440) Of services are given). On long-term follow-up (24-40 months.), there were no significant earnings impact reported; however, significant positive impact on GED attainment. Intensive Residential Programs Mallar, et al. 4 Residential program that provides intensive skills training, basic 1982, Job Corps education, support services and job placement for one year. Average (4334/ 1457) over first 4 years after program exit of 15% earnings increase and a reduction in serious (felony) crime (both significant). Also, a large and significant increase in GED attainment and college enrollment. Wolf, et al. 3 Combines work sponsored by various public resource agencies with youth 1987, development activities for up to one year. CCC is not an effective California (943/1083) way of raising the earnings of all participants when they first enter Conservation the labor market; however, it did improve earnings of disadvantaged Corps. residential corps members and significantly increased their hours worked, post-program. 

2) Short-term training with job placement for out-of-school youth.These programs typically last about six months and cost $2,500 to $5,000per participant. For example, the federal government's principal programfor disadvantaged youth, JTPA, enrolled 125,000 out-of-school youth aged16 to 21 for five months, during which they received on-the-job training,classroom training and job search assistance. JOBSTART was a large scaledemonstration program, designed as a more intensive version of JTPA, lastingseven months and including more classroom training, at a cost of $5,000per participant.

3) Long-term, intensive residential programs providing vocationaland life skills training, general education and job placement after graduation.The most prominent of these programs is Job Corps, a residential programaimed at extremely disadvantaged populations. In 1993 Job Corps enrolled62,000 new youth in tailored one-year programs that included classroomtraining in basic education, vocational skills, and a wide range of supportiveservices (including health care), at a cost of roughly $15,000 per student.

Very few evaluations of these programs measure change in criminal behavior,simply because crime prevention is not generally a primary objective andrequires substantial and complex additional data collection.11Crime control is a secondary effect which may happen as the consequenceof increased employment, the primary objective. The remainder of this sectionwill briefly review the principal evaluations of these programs, startingwith the subsidized work programs.

Subsidized work programs are the cheapest and least intensive of anyof the training programs aimed at at-risk youth. Although all subsidizedwork programs show a decided increase in employment for the targeted populationover the time period of the subsidy, no evaluation has shown any long termeffect on employment. Not surprisingly, the one evaluation that lookedat crime (Supported Work) showed no sustained decrease in crime rates (Piliavinand Masters, 1981). Perhaps more damning, the crime rate of participantsin the Supported Work program did not decline while they were working inthe subsidized jobs. The conclusions seem robust -- subsidized work doesnot increase productivity in any appreciable way and these types of jobsdo not appear to have the necessary characteristics to be supportive ofnon-criminal behavior.12

The picture is only slightly less gloomy for short-term skill trainingprograms. None of the rigorous evaluations in this category have shownany lasting impact on employment outcomes, although some of the programsshow a short term gain in earnings. It is again not surprising then thatthe one evaluation that looked at crime shows no lasting impact (JOBSTART).A slightly more detailed look at the data show that while there are noemployment gains, there are some educational gains from these programs.JOBSTART and other programs effectively doubled the fraction of GED recipients.Although GED completion is in fact correlated with higher earnings, itapparently serves as a credentialling device rather than a training device;i.e. the fact of earning a GED indicates an ability to sustain consistenteffort but working toward the diploma does not actually develop skills.This helps explains why the earnings gains showed in these programs arenot long lasting. Eventually, those without GED's are also able to acquiresimilar jobs; it just takes them longer without the GED credentials. Theseprograms are generally unable to increase productivity in any meaningfulway within the constraints of a short-term non-intensive program.

The one positive result in this literature is from the long-term residentialtraining program, Job Corps. Job Corps is by far the most intensive andexpensive non-military training sponsored by the federal government. Thehigh cost is a consequence of the residential element of the program andits severely disadvantaged population (over 80% are high school dropouts).The most recent Job Corps evaluation in 1982 was not as rigorous as mostof the other evaluations in this literature because it was not randomizedexperiment. It had to use a comparison group drawn from people eligiblebut not likely to participate in Job Corps because of geographic location.Despite these limitations, the study was carefully done and generally regardedas credible, although Donohue and Siegelman (1996) raise serious questionsabout the magnitude of the decline in the homicide rate for enrollees.13

The evaluation found that four years after graduating, enrollees earnedon average $1,300 more per year than the control group, a difference of15%. Not surprisingly, these achievements corresponded with real increasesin educational achievement. Enrollees were 5 times as likely to get a GEDor finish high school, and twice as likely to go to college. Also, therewas a significant decline in arrests for serious crimes, especially theft.However, there was also an unexplained increase in minor arrests, especiallytraffic incidents.

The failure of all but the most intensive job training programs forat-risk youth to have an effect on either employment or crime raises someserious questions about this particular approach. There are several possibleexplanations:

1) The first, and simplest explanation is simply that low dosage programsover a six month period (or less) do not have enough statistical powerto make a measurable impact.

2) More substantively, these lower dosage programs might not simplybe enough to counterbalance a failed academic career that often finds 15and 16 year olds reading at the 5th grade level. A large amount of trainingmust be exerted in order to raise reading levels four, five and six gradelevels.

3) Perhaps Wiliams et al. (1996) have a point in that employment byitself is not enough to stop crime. In fact, employment for youth mightbe criminogenic -- the low paying, low skill jobs normally taken by youthdo not add significantly to human capital, but they do take time away fromschool activities which could increase human capital.

Points two and three, taken together, suggest that the real, long termanswer to this problem for the vast majority of at-risk youth lies notwith after-the-fact job training but rather with an effort that makes schoolingmore meaningful to students before they drop out of school. The Departmentof Labor, in following this logic, suggests that the answer lies in connectingtraining to real jobs to a school environment through the recently enactedSchool to Work Opportunities Act. The emphasis on the school-to-work transitionis supposed to make students and schools more motivated to learn, and decreasedropouts (Rosenbaum [1996]). This belief is based in part on the successof Job Corps in connecting education to success in the labor market. Thisphilosophy of using the school-to-work transition as the instrument forimproving the utility of regular schooling is untested. Given the administration'scommitment to the idea, we expect evaluations will be completed withinthe next two years.

It is valuable to note that there have been evaluations of school-basedanti-dropout programs that are not based on the school-to-work model. Evaluationsof these programs are neither as numerous nor as rigorous as those forjob training programs. The evidence also suggests that anti-dropout programs,because they involve working within the complex environments of schools(see the School Chapter in this volume) are extremely difficult to implement.However, two random assignment evaluations have shown that intensive anti-dropoutprograms have had substantial success in reducing drop-out rates and showinggains in human capital acquisition.

The strongest positive evaluation is for the Quantum Opportunities Program(QUOP), a demonstration program offering extensive academic assistance,adult mentoring, career and college guidance, a small stipend and moneyset aside for a college fund. Services totaling 1286 hours over four years(equivalent to about 6 hours per week) were provided to children from AFDCfamilies throughout high school, at a total cost per participant of $10,600.The rigorous evaluation of 100 students in four sites (random assignment,scientific method score = 5) found that 42% of the QUOP students were inpost-secondary education versus only 16% of the controls; a total of 63%of the QUOP students graduated from high schools, versus only 42% of thecontrol group (DOL, 1995). This evaluation has no long-term follow-up ofemployment outcomes. However, the increase in enrollment in college islikely to be a good predictor of improved labor market performance.

In this evaluation, adult mentors were assessed to be the most importantelement. Apparently the mentors provide the necessary focus and motivationfor students to change their behavior and perform better in school. Yetnotice that in QUOP, the key elements of the school-to-work philosophy-- direct connections to the labor market, and contextual learning -- werenot employed. As in Job Corps, QUOP students were in routine contact withadults who projected a positive attitude about meaningful employment.

It is impossible within the context of the current literature to determineif mentoring or a school-to-work program (or some combination) is betterable to change the motivation of the at-risk youth. However, it is clearthat individuals need to become focused on obtaining meaningful and productiveemployment before they will/can take advantage of job training or schooling.We will discuss other ways to change the orientation of youth later inthis section.

Job Training for Adults in the General Population

A narrow focus on job training for at-risk youth is perhaps justifiedwithin the context of a crime prevention program. Adults who have not offendedby age 25 are at low risk of offending. If they have offended by age 25,chances are they will be already be involved with the criminal justicesystem. However, some people will be out of the criminal justice system,yet still need training in order to find meaningful employment. These olderadults may have a reduced propensity to commit crime due to maturation.As a result, the number of crimes prevented by such a training programmight be lessened, but at the same time, these individuals may be finallyready to take advantage of training programs that are offered. In reviewingthe extensive literature on job training for the general population, Heckmanconcludes the following:

    Employment and training programs increase the earnings of female AFDCrecipients. Earnings gains are (a) modest , (b) persistent over severalyears, (c) arise from several different treatments, (d) are sometimes quitecost-effective. . . . For adult males the evidence is consistent with thatfor adult women. (Heckman, 1994: 112).

Consistent with these findings, older ex-offenders in the SupportedWork program appear more responsive to the program than do younger ex-offenders.In addition, older subjects in the Baltimore Life experiment also recidivatedless often relative to their controls than did younger subjects. The authorsof the Supported Work program conclude "the evidence in this experimentand elsewhere suggests older disadvantaged workers, including those whoare known offenders, may be much more responsive (than younger workers)to the opportunity to participate in employment programs (Piliavan andMasters, 1981:45)."

Housing Dispersal and Mobility Programs

Much of the above discussion has been focused on at-risk individuals,rather than places. But depressed urban areas deserve special attentionin this chapter given the simultaneous existence of high crime and lowemployment in these areas. A decade ago, William Julius Wilson (1987) identifiedthe movement of jobs from the inner city to the suburbs as the key factorin the growing concentration of African-American poverty and the socialproblems related to that hyper-segregation. More recently he has arguedthat only an employment oriented policy can reduce the social problemsof these communities (Wilson, 1996). Yet, as we will see in the followingsection, stimulating true economic development in the inner city throughtax incentives or direct capital subsidies has proven very difficult. Substantialeconomic forces14have led to the movement of businesses to the suburbs, and these forcesare extremely difficult to counteract (Hughes, 1993).

As a result, policy makers have recently begun to develop ways to changethe supply of labor by bringing the people in the inner city to the jobsin the suburbs, instead of bringing jobs to the people in the inner city.One way to do this is to physically relocate inner city residents to thesuburbs (housing dispersal programs).

The only published outcome evaluation of the housing dispersal conceptis based on what is known as the Gautreaux housing mobility program inChicago. Starting in 1979, the Gautreaux program has given 6,000 innercity families (primarily single mothers) vouchers that allow them to relocateto low poverty neighborhoods throughout a six county area in and aroundChicago. The program, started as the result of a federal court ruling ina housing discrimination case, also allowed families to move within thecity of Chicago. Families were assigned to the suburbs or the city basedon where there were apartment openings when they became eligible for theprogram. Because the waiting list was long, and because families were placedat the back of the list when they rejected an opening, very few familiesrejected an apartment when it was offered, regardless of the location.

Rosenbaum (1992) took advantage of this natural experiment to comparethe employment and educational outcomes of the city movers with the suburbanmovers (scientific method score 4). He found that women who moved to thesuburbs were 28% more likely to be employed than the women who moved insidethe city, on average 5.5 years after moving. This was true even thoughthe wage gains attributed to the move were the same for all women who worked,regardless of their location. In addition, he found that 9 years (on average)after the move, the children of the suburban movers were doing significantlybetter than the children of the city movers (scientific method score 315).Although criminal activity was not measured, the children of the suburbanmovers dropped out of high school only 25% as often as the city movers,were in college track courses 1.6 times as often as the city movers, were2.5 times as likely to attend college, were more than 4 times as likelyto earn $6.50 an hour if working, and only 38% as likely to be unemployed.These results suggest that for children in these environments, relocationcan be an effective tool to change their focus towards positive outcomeslike meaningful employment.

These large positive results led to significant optimism on the partof policy makers about the benefits associated with simply relocating poorfamilies to non-poverty areas. Several programs modeled on the Gautreauxprograms were spawned and now operate in Cincinnati, Memphis, Dallas, Milwaukeeand Hartford. In 1992, HUD provided $168 million to fund Moving to Opportunityas a demonstration program for the housing mobility concept. Moving toOpportunity has 5 sites in large cities -- Baltimore, Boston, Chicago,New York, and Los Angeles -- and is funded for at least 10 years. The projecthas been set up with a rigorous evaluation component (scientific methodscore = 4) -- households are randomly assigned to either placement in asuburban location with less than 10% poverty, placement in the centralcity, or no treatment. About 1,300 families will be given vouchers whichallow them to relocate in low poverty suburbs, along with extensive counselingabout relocation and assistance in finding a new apartment. Initial evaluationsshould be available by mid-year 1997.

Despite Gautreaux's apparent success, and the development of programslike Moving to Opportunity, housing dispersal programs have met significantopposition from suburban residents afraid of the impact of poor minorityfamilies on their communities. For example, the expansion of Moving toOpportunity to include more than 1,300 families was defeated after it becamea political issue in the 1994 election. The Mount Laurel decision in NewJersey, a two-decade-old, court-enforced dispersal strategy, is now beingundermined by legislators. In addition, minorities sometimes voice a concernthat the dispersal of minorities to the suburb will weaken minority politicalpower (Hughes, 1993). According to Kale Williams, former director of theGautreaux program in Chicago, part of the success of Gautreaux was because"it hasn't been large enough to threaten anyone and hasn't been concentratedenough to arouse apprehension." Given these problems, it seems politicallyunlikely that housing mobility programs will ever expand to any significantsize or at least cannot politically afford to move large numbers of poorpeople in specific non-poor neighborhoods.16

This reality, however frustrating, suggest that perhaps a strategy aimedat integrating workplaces instead of neighborhoods might be easier to implement.This argument suggests that the best approach to the problem of inner citypoverty are mobility programs which provide transportation for inner cityresidents to the suburbs (Hughes, 1993). Such a program recognizes (andtakes advantage of) the power of the suburban labor markets to increaseresidents' incomes while avoiding the political problems associated withhousing dispersal. This idea is relatively new, and as a result only asmall number of programs are in operation in the United States.17

However, HUD has funded an $18 million dollar demonstration programin five sites starting in 1996 and running for four years. The strategyhas three main components: a metropolitan-wide job placement service toconnect inner city residents with suburban jobs, a targeted commute mechanismto provide transportation to the jobs, and a support services mechanismwhich will try to ameliorate some of the problems that may arise due toa long-distance commute into a primarily white suburban location. Rigorousevaluation with random assignment will be undertaken by Public/PrivateVentures. If successful, this program will form a key component of thewelfare reform strategy.

The mobility programs are rooted in theoretically very different approachesto reducing central-city crime. Housing dispersal programs attempt to breakup the poverty community. Reverse commuting preserves the community butat a cost: the long commuting causes reduced guardianship and parentingthat have potentially negative effects in their home communities. Childrenalso do not benefit in the same way because they continue to live in thesame depressed environments. These reverse commuting programs might serveto increase employment and decrease the criminal activity of a particularperson, but the programs will probably not have the indirect anti-criminogeniceffects of housing dispersal programs.

IV. DEMAND-SIDE PROGRAMS

Bonding and Wage Supplements

All the programs described in the previous section focused on changingindividual behavior. Yet perhaps employers feel that certain individuals,particularly ex-offenders, represent a potential risk. A criminal historyrecord appears to be a predictor of low job attachment (in part becauseof the risk of future arrest and incarceration), poor performance, theftand malingering. To overcome these barriers, a number of demand-side programsoffer to compensate employers for incurring the risk of hiring workerswith a criminal record.

One class of program directly lowers the employer's wage payments, eitherwith a subsidy or through a targeted job tax credit (i.e. the employerof a particular class of worker is able to deduct the payments or someportion of them, from his taxable income. This reduces the amount thatan employer has to pay the worker, the difference being picked up by thegovernment. The programs are transitional and are intended to last justlong enough to allow the offender to acquire a work history that of itselfwill increase future prospects. The second class of program is more indirectand takes the form of subsidized bonding of offenders, thus reducing thecost for the employer of insuring himself against specific crimes, suchas inventory theft; such bonding is normally provided by private corporations.

The federal government , however, has offered a very low level of fundingfor these programs. The Department of Labor discontinued in 1995 the TargetedJobs Tax Credit, for which the annual budget never exceeded $10 million,with most of that targeted to other disadvantaged groups. Some state Departmentsof Corrections (e.g., Texas) do offer wage subsidies. However no evaluationidentifies the impact of these on either employment or crime. In addition,some researchers (DOL, 1995) feel that these programs actually hurt ex-offendersby clearly identifying their ex-offender status. While it might be worthwhileto fund an evaluation of the very small Federal Bonding program ($240,000total in 1996), the one independent review of the Targeted Jobs Tax Creditwas not optimistic that these programs improved employment among ex-prisoners(Jacobs, 1984).

Enterprise Zones

Community development programs use demand-side policies to help particularareas. Although these programs are focused on depressed areas, like housingdispersal programs, community development programs can be used in a widerarray of settings. These programs are of particular interest for crimeprevention because they propose to help both individuals and neighborhoods.New jobs present more opportunities for legitimate work to compete withillegitimate opportunities often present in these communities. Jobs visiblyavailable in an area would also provide motivation for education and skillstraining for young people. The economic activity that new or expanded businessesrepresent can also lead to increased social interactions among residentsand strengthen social institutions (churches, business organizations, schools)which can exert a positive influence on individuals who might otherwiserevert to crime.

Enterprise zones are one relatively new policy tool focusing tax incentivesat generally small, economically depressed geographic areas (Papke, 1993,Erickson and Friedman, 1991). According to Erickson and Friedman (1991)these programs typically employ three different types of program incentivesto encourage job development: investment incentives, labor incentives andfinancial incentives. The investment incentives include credits for propertytaxes, franchise taxes , sales taxes, investment taxes and other possiblystate-idiosyncratic employer taxes (e.g., inventory tax credits). The laborincentives include a tax credit for job creation, for hiring a zone residentor some other disadvantaged person, and for training expenditures . Finally,the finance incentives sometimes include an investment fund associatedwith the program and preferential treatment for federal bond programs.These programs are based on the assumption that employers are sensitiveto state and local tax incentives in their location decisions. The academicliterature shows mixed results about the validity of this claim, althoughrecent evidence suggests that investment is more responsive to state andlocal taxes than previously thought (Bartik, 1991).

As of 1995, 34 states had a total of 3,091 active enterprise zone programs(median = 16) and the Federal Empowerment Zone and Enterprise CommunityProgram has introduced 106 more zones (Wilder and Rubin, 1996). The statezones are limited in the value of the incentives they can offer, preciselybecause federal taxes (e.g., corporate profits tax) are so large and cannotbe waived by the state. According to Erickson and Friedman (1991), themedian zone population for the state programs is about 4,500 persons andthe median zone size is about 1.8 square miles. Zone designation is usuallybased on unemployment rates, population decline, poverty rates, medianincomes, the number of welfare recipients or the amount of property abandonment.The federal program amounted to $640 million in total tax credits in FY1995.

Since these programs are relatively new, (the median state began itsprogram in 1984) there are few outcome evaluations, most of which are reviewedin Wilder and Rubin (1996); no evaluations of the federal program haveyet been conducted. All evaluations consider only the immediate economicoutcomes of these programs, and do not examine the larger social implications(such as crime reductions) of the programs; see Table 6-3. Only Bartikand Bingham show an awareness that this is a shortcoming of these evaluations.The evaluations also do not attempt to determine the impacts of individualincentives. Although ideally researchers could identify the most effectivetax break, the incentives are typically used in concert, so that the economicgrowth in any given zone cannot be attributed to any one incentive; noris it possible to separate out component effects using econometric techniques.

The main theoretical concern about enterprise zones is that they willsimply relocate existing jobs rather than create new jobs. In fact, Britain,which pioneered these zones, abandoned its enterprise zone program afterresearchers found that nearly all jobs in enterprise zones (86%) were dueto relocation from neighboring communities. The US experience is somewhatmore optimistic - the literature seems to agree that, of all the new jobsfound in enterprise zones, roughly 25% are due to relocation, 25% are dueto new business and 50% are due to expansion of existing businesses (Wilderand Rubin, 1996). Of course, not all the jobs that appear in the enterprisezone should be attributed directly to the zone incentives. However, theprimary modes of evaluation in this field, correlation and before-and-afterwithout comparison group (scientific method score 1 and 2), do not allowresearchers to isolate the contribution of the zone incentives.

In addition, most of these studies use data from surveys of zone firmsor zone managers; these lack credibility as measures since both groupshave an incentive to place a positive bias on the outcomes.18These studies generally conclude that the zones increase jobs and investment,although results vary by zone.19

TABLE 6-3

Enterprise Zones

 Studies  Scientific Method Description of Intervention and Findings   Score (Number of cases Treatment/control) Boarnet & Bogart 3 New Jersey EZs have no impact on employment and business 1995 growth. (7/21) Papke 1992 3 Indiana EZs decrease zone unemployment by 19%. (15/24) Bostik 1996 3 California EZs in small cities increase business construction. (5/27) CA State Auditor 1 Survey of firms indicate small net increase in economic 1988 activity with wide variability across zones. (13) Dowall, et al. 2 Although employment growth and increased business activity 1994 increased in all CA zones, researchers concluded that zone (13) incentives could not be linked to growth. Erickson & 1 EZs in 17 states appear to create jobs in areas with Friedman 1991 development potential. EZs are ineffective in highly (35) distressed areas. GAO 1988 2 3 rural Maryland EZs zones showed significant increases in employment and investment after zone designation. (3) HUD 1986 1 Interviews with zone managers in 10 zones in 9 states responsible show zones lead to significant new investment (10) and job growth. Jones 1985 2 Connecticut EZ has no impact on building activity (1/1) Jones 1987 2 Illinois EZ has an impact on building activity (1/1) Wilder & Rubin 1 Firm-level survey data show increase in jobs due to Indiana 1989 EZ in Evanston. (1) 

Only three studies (Papke, 1994; Boarnet and Bogart, 1996; and Bostic,1996) attain a level 3 scientific method score; they are before-and-afterstudies of a particular state's enterprise zone program (Indiana, New Jerseyand California respectively) with comparison groups from other eligibleareas in the state. Each study also uses data collected by independentagencies, so the data is unlikely to be biased by EZ participants. Thefirst two studies used econometric methods to control for selection bias,the latter did not.

The results of the first two studies contrast strongly -- the New Jerseystudy found that the zones had no impact on total employment or propertyvalues in municipalities with zones,20while the Indiana study found that the zones led to a long term 19% declinein unemployment rates in municipalities with enterprise zones. The Indianaresearcher was somewhat surprised by the magnitude of this effect, giventhat the employment incentives were limited in the Indiana zones. But thestudy also found that firms responded to reductions in inventory taxesby increasing inventory by 8% and reducing capital machinery by 13%. Thesechanges in inventory and machinery may represent the conversion of firmsfrom manufacturing to more emphasis on distribution, generating a positiveimpact on employment.

Bostic's study used investment growth rather than employment as theprincipal outcome measure, He found that the EZs had a significant butsmall impact on commercial construction permits and an insignificant impacton the number of businesses in an area.

Clearly, additional research is needed to verify the positive impactof enterprise zones on employment and investment. Abt Associates have beencommissioned to do an evaluation of the Federal Empowerment Zones and Engberget al. at Carnegie Mellon are undertaking a comprehensive evaluation ofstate enterprise zones with controls for selection bias.

Community Development Block Grants

The 1974 Community Development Block Grant (CDBG) Program representsthe other major federally funded program aimed directly at revitalizingdistressed neighborhoods. Instead of relying on tax credits as incentives,this program provides direct funding to local governments. In 1992, CDBGsprovided local jurisdictions with $3.4 billion to be spent on activitiesthat support any one of three objectives: benefiting low-a and moderate-incomepersons, preventing or eliminating slums or blight, or addressing otherurgent community needs. The program funding breaks down broadly into 5main areas: housing (38%), public facilities (22%), economic development(12%), public services (9%) and acquisition and clearance (6%). Althoughthere are no outcome evaluations of this program,21the sheer size of the economic development component of this program ($251million in 1992) demands inclusion in this section.

Most of what follows is based on a 1995 funding process evaluation sponsoredby the Department of Housing and Urban Development (Urban Institute, 1995).The evaluation, like those for Enterprise Zones, considers only economicoutcomes. A full 78% of the $251 million economic development grant moneywas spent on loans and grants to private businesses. Most of the recipientbusinesses were small, and 37% of these businesses were minority owned.These loans seemed to perform better than the non-geographically targetedSmall Business Administration loans. According to the HUD report, theseloans were more important to the business activities of the recipientsthan the EZ tax incentives,22but neighborhood residents held a comparable number of the newly createdjobs under both programs (approximately 30%).

An effort was made to provide a before-and-after study of 250 censustracts in the CDBG program (scientific method score = 2), using a surveyon all CDBG funding and census data from 1980 and 1990. This study founda clear relationship between the level of funding and tract income: tractsthat saw an increase in income received $1,247 per capita, tracts thatwere stable between the two time periods received $844 per capita and tractsthat declined received $737 per capita. Improvement in low-income tractsusually only occurred through gentrification or out-migration of low incomepeople, but in several instances the arrival of major industrial facilitiesresulted in an increase in income for the tract residents.23

In more general terms, the researchers concluded that the existenceof an income-mix among neighborhood residents and a healthy commercialdistrict appeared to help development. Within the context of this review,these factors could signal the existence of a certain level of social controlwhich would allow community programs to be effective. Neighborhoods withoutthese factors may not have enough social capital to take advantage of anycommunity-based program.

V. A PROPOSED INTEGRATION

Existing evaluations of interventions aimed at increasing employmentfor high risk populations provide little positive guidance as to the appropriatedirection of labor market policies for more effective crime prevention.The demand-side programs (Community Development Block Grants and EnterpriseZones) have not been subject to rigorous evaluation. Moreover, they involvesuch a broad array of incentives and funds that it will be hard to determinewhat might explain any positive findings and thus what is worth replicating.Evaluations of training and education programs, aimed at labor supply,have shown little positive consistency; there are merely hints as to whatconstitutes a successful program and there are very few findings specificallyon crime reductions. Programs aimed at ex-offenders do a little betterbut again there is a lack of consistent positive findings that allow oneto say that specific interventions work for large segments of the eligiblepopulation.

The negative findings concerning job training and employment for highrisk groups may primarily be explained by the extremely limited natureof most of the interventions that have been tried. Heckman (1994) makesthis point nicely with respect to training programs generally. He suggestsa mind experiment. Assign a generous annual real rate of return of 10 percentto social programs, higher than is usually observed for investments ineducation. Interventions that cost $2,500 per client and are aimed exclusivelyat raising the earnings of participants would then be expected to produceannual earnings increases of $250. This is far too small an amount to liftsomeone out of poverty (being only about 3% of the required income foran individual) or to improve life prospects enough for most participantsthat they might make large changes in their decisions about schooling andwork.

Not only are small quantities of services provided but the range ofrisk factors targeted is very narrow. Providing jobs for adults will onlyweakly compensate for failure to invest in human capital when young. Sometheoretically promising interventions have not been tried but merit explorationas a major tool for reducing criminality among young adults. These interventionshave to address both the individual dynamics and social ecology of thecrime/work choice. The remainder of this section considers the specialchallenges associated with high crime/low employment areas.

The central argument is simple enough and not original to us. Crimeand unemployment are most strongly linked at the community level, at leastin urban areas. Persistently very high unemployment rates will generatehigh crime; that high crime will drive out capital and make jobs increasinglyremote. That produces a downward spiral to a low employment/high crimeequilibrium which is very stable and highly resistant to small increasesin employment or reductions in crime. Interventions have to explicitlydeal with crime and employment simultaneously. No evaluated interventionhas done so.24

The most important mediating factor in this story may be the motivationsof community residents. For example, the isolation of high poverty neighborhoodsfrom the legitimate job market may be critical in accounting for the lackof motivation among youth in these neighborhoods. Rosenbaum (1996), amongothers, makes this point by observing that youth have difficulty findingemployment when they live in impoverished neighborhoods without well-developedjob connections. That is exacerbated by geographic isolation from jobsand the possibility of racial discrimination. The perceived returns tocontinuing in school or in acquiring human capital in other ways is low.This leads to low high school graduation rates and high attrition in trainingprograms, maintaining the under investment in human capital of the previousgeneration in high poverty neighborhoods.

The claim that improving perceived legitimate employment opportunitieswill increase school attachment is still not well tested. The availableevidence does suggest, however, that school achievement is affected bythe achievement of others in the same community. For example, Case andKatz (1991) and Mayor (1991) found that youth are more likely to stay inschool or work if a large proportion of their peers do.25The Department of Labor (1995) review of evaluations used this findingas the basis for a claim that poor neighborhoods should be saturated witha range of interventions intended to alleviate poverty, so that "theemployment outcomes of some persons within a community can lead to `spillovereffects' as other people in the neighborhood are influenced by the positiveactions of their peers." (p.63).

There is some disagreement on the issue of the importance of neighborhoodeffects. The evaluation of the JTPA program claims that the "externalenvironmental factors - unemployment rates, population density,Â….had weakeffects, if any, on (individual JTPA program) success." (PPV, 1994:5). However, the same report claims that business involvement with trainingprograms is crucial because "it provides a built-in incentive forparticipants to feel that their participation is worthwhile." (PPV,1994, p.14). The tie to business is itself possibly a proxy for communityattachment to the labor market.

We believe the community level problem is compounded in two ways bydrugs. A large fraction of adult criminal offenders are substance abusers;their involvement with expensive illicit drugs, such as cocaine and heroin,is distinctive. This represents a major employment handicap. Thus workforceoriented interventions will have to deal with the substance abuse problemof potential workers if they are to increase employment and thus reducecrime. Employment is itself possibly a protective factor for substanceabuse, increasing the probability of desistance. High risk youth now showmore moderate rates of abuse of expensive illicit substances (as reflectedin the data from the National Institute of Justice's Drug Use Forecastingprogram for juveniles) but such risk still needs to be addressed in thecontext of labor market programs in high risk communities.

Illicit drugs are also a major problem because recent research showsthat drug markets in impoverished neighborhoods provide substantial alternativeemployment to legal markets. For example, Reuter, MacCoun and Murphy (1990)reported that drug selling earned its participants about $30 per hour inWashington in 1988. Saner, MacCoun and Reuter (1995) found that approximately30 percent of the 1967 cohort of black males resident in the District ofColumbia had been charged with drug selling between the ages of 18 and24. The existence of attractive alternatives outside the legitimate labormarket will complicate any program aimed at attracting individuals to legitimatework opportunity. Also, in contrast to the pre-cocaine-epidemic period,drug selling may now precede drug use; those who sell as juveniles becomeconsumers of their own drugs, making it still more difficult to maintainlegitimate employment as adults.

High unemployment neighborhoods generally show high levels of drug selling;this further weakens the ability of individually focused programs to increaseemployment prospects for men who continually have opportunities to earnsubstantial amounts in drug selling. In addition, neighborhoods where manymales support themselves through some drug selling will not have many ofthe social institutions that support legitimate work. This will make itmore difficult for individuals in these neighborhoods to make the transitionto legitimate work.

We have made this long digression about drugs as a reminder again howimportant it is to take the community as the central focus for programmaticintervention. Individually oriented programs cannot ameliorate many ofthe fundamental problems faced by program participants. Similarly, programslike reverse commuting, though they may bring important benefits for individuals,will generate few benefits for the most adversely affected communities.Indeed, as already mentioned, the long commutes involved in such programsreduce still further the extent of adult supervision of children that issuch an important component of effective community. Programs like Gautreauxwhich take households out of the community also paradoxically may worsenthe situation of those who remain, since the movers are likely to be amongthe more forward looking adults in these fragile inner city communities.This of course suggests the attractions of the converse, bringing somemiddle-class households back into the neighborhoods that are so devastated.But crime is as much an obstacle to that as it is to encouraging employersto relocate in the same communities.

VI. SCIENTIFIC CONCLUSION

We have reviewed here a large variety of programs that might reducecrime by increasing employment and labor market outcomes for high riskpopulations. Our assessments (using the criteria articulated in Chapter2) of which program types work, which programs do not work, which are promisingand for which we can venture no opinion are contained in Table 6-4.

Table 6-4

Program Conclusions

What works? 1) Short-term vocational training programs for older male ex-offenders no longer involved in the criminal justice system. What does not work? 1) Summer job or subsidized work programs for at-risk youth. 2) Short-term, non-residential training programs for at-risk youth. 3) Pre-trial diversions for adult offenders which make employment training a condition of case dismissal. What is promising? 1) Intensive, residential training programs for at-risk youth (Job Corps). 2) Prison-based vocational education programs for adults 3) Housing dispersion programs 4) Enterprise Zones What do not we not know enough about? 1) CJS-based programs for juvenile offenders 2) Post-release transitional assistance for offenders. 3) Reverse commuting 4) Wage subsidies 5) Bonding programs 6) Community development as done through the Community Development Block Grant Program. 7) School-to-Work programs funded by the School-to-Work Opportunities Act. 

Programs that Work

The one program type for which the evidence of effectiveness is fairlystrong is vocational programs aimed at older males ex-offenders who areno longer in the criminal justice system. We gave little attention to thisin the body of the chapter because these programs, though useful, comelate in criminal careers. Reducing crime by 35 year olds who have previouslybeen criminally active, will have a modest effect on serious violent crime,which is predominantly committed by younger males. The explanation forthe success of these programs may be found in theories of life course events.Age generally increases the desire for stability, lowers the desire forrisk and raises concern about the future. JTPA job training, (even forperiods as short as 90 days) may have an impact because many of those whoenroll are now motivated to seek employment.

What Does Not Work

In contrast, there have been numerous, well-conducted evaluations ofwell-executed, short-term (many of them summer only) programs aimed atat-risk youth, typically 15-21 year-olds. These have repeatedly found noeffect on earnings and crime rates. The high attrition rates we take tobe symptomatic of the lack of motivation for many participants, reflectingtheir perception of weak employment opportunities in early adult life.Without any evidence of impact on other risk or protective factors, webelieve that these programs cannot be justified on crime prevention grounds.The two serious evaluations of pre-trial diversion programs suggest thatpre-trial diversion programs do not work, at least in part because of theprograms tend to get co-opted by the prosecutors for purposes other thanthe intended purpose of rehabilitating offenders.

Promising Programs

The one class of program aimed at high risk youth for which positiveresults have been shown is Job Corps, which is both residential and, interms of expenditures, very intensive ($15,000 per youth). However thereis only one rigorous evaluation (moreover one that has some methodologicalweaknesses), although another major Job Corps evaluation is in process.Thus we can only classify this type of program as promising. The reasonsfor its possible success are multiple: it re-socializes the youth by breakingcommunity ties and presenting pro-social role models; its residential requirementreduces the intensity of contacts with anti-social groups and illegal earningsopportunities; its vocational focus and attachment to the labor marketprovides academic training in a supportive environment.

Prison-based vocational education programs aimed at adults, who constitutethe vast majority of the correctional population, are also promising. Thisagain may be explained by the life course model. The current evidence suggeststhat something works, but no random control trial has found an impact,and few studies have been able to pinpoint exactly what works. Programimplementation within a correctional facility and inmate motivation remainmajor problems. We suggest that the problem of motivation be dealt withby randomly selecting individuals for participation from within a pre-selectedpool of motivated individuals.

The last promising programs on the supply side of the labor market areprograms that provide dispersed housing for poverty-level households. TheGautreaux program has been found to have had a positive impact on bothmothers and children. The program has been operating for many years andfairly large numbers of households (over 6,000 families) have made useof it. Its apparent success in terms of improving educational and employmentoutcomes for both mothers and children is sufficiently strong that crimereductions can reasonably be inferred. We classify it only as promisingbecause, apart from our formal requirement that there be more than onerigorous evaluation before it be classified as working, the one existingevaluation has many weaknesses. Results from the multi-site, federally-fundedMoving to Opportunity program are scheduled to be available in 1997; thisshould shed more light on the efficacy of this type of program.

On the demand side, enterprise zones have a mixed record according tothe few moderately rigorous evaluations that have been conducted. We classifythe program as promising on the grounds that if designed with crime preventionobjectives in mind (i.e., as part of a more comprehensive effort) it mayinclude a critical ingredient of what is necessary for making high riskcommunities safer. The federal government is currently funding a largeenterprise zone program - it will be interesting to see if the programevaluators consider both the problems caused by crime in these areas, orlook at the crime reduction effects of some of these programs. Currentevidence suggests that areas with high crime are less likely to succeed.

Programs About Which Too Little is Known

Table 6-4 contains a list of seven program types about which too littleis known for any judgment beyond the broadest sense of theoretical plausibility.We include some programs that have frequently been evaluated and have producedmixed results.

The programs that have not been subject to any rigorous evaluation are:bonding programs, Community Development Block Grants, reverse commuting,school-to-work programs and targeted wage subsidies. Though theoreticalarguments can be made for each of them, those arguments seem strongestfor CDBGs (see Section V), and school-to-work. School-to-work's focus onavoiding dropout and developing human capital over the long run is an interestingresponse to the motivation and dosage problems we identified in SectionIII.

For the remaining three types of programs, a number of conflicting resultshave been produced. Criminal justice-based programs for improving the employmentprospects of juveniles seems to have shaky theoretical premises. Employmentconcerns are not strong for those under 17 and there is a small but growingliterature suggesting that early work gives youth too much autonomy attoo early an age by lessening their dependence on family. In addition,time spent working is time spent away from conventional schooling whichmight lead to more meaningful employment (see Cullen, 1996).

The mixed results from evaluations of transitional assistance for inmatesleaving the criminal justice system is harder to explain. Internal doubtsabout their ability to succeed in conventional society and the externalforces that limit them initially to fairly poor jobs combine to createa very difficult transition, especially for offenders without family orfriend networks. Transitional aid then seems particularly appropriate.Yet the results from TARP and Baltimore LIFE seem to suggest that motivation(or focus), rather than money is the important issue. Some programs likeProject Rio in Texas appear to have success reintegrating offenders usingcaseworkers, but the program has not been rigorously evaluated.

Bonding and wage subsidies are intended to help with the transitionfrom the criminal justice system to legitimate employment. These componentshave not been heavily evaluated. There is some concern that wage subsidiesand bonding tend to work against offenders by clearly identifying theirstatus to employers.

Program Recommendations

Our program recommendations for OJP are modest, principally becauseCongress has not directed the department to become involved in fundinglabor-market-oriented programs outside of the criminal justice system andbecause such programs clearly have much broader objectives than simplycrime prevention. Given the evidence summarized above, we believe thatCongress should encourage OJP to continue modest funding of programs thataim at improving the employment prospects of older ex-offenders. The programsdo not need to be intensive to be effective and these programs are generallyworking on the back end of the criminal career. The concept behind OperationWeed and Seed also has some merit -- we defer to Chapter 3 for recommendationson this program. The negative findings concerning the effects of short-termsubsidized work and non-residential training programs speak not to OJPprograms but rather to efforts funded principally by the Department ofLabor.

VII. IMPROVING EFFECTIVENESS THROUGH EVALUATION

We have more recommendations concerning research and evaluation. Congressshould encourage OJP to take advantage of on-going offender transitionprograms in state and federal systems to implement rigorous program evaluationsof some of the more promising programs. We also recommend that researchefforts be focused on the problem of implementing vocational programs withinthe prison system. Since it appears from some level 3 studies that vocationalprograms do have some effect, randomized experiments which control forselection bias while also isolating the effective characteristic of theprogram are clearly appropriate.

The remaining recommendations focus on opportunities for Congress toencgourage OJP collaboration with other federal agencies in examining whetherinterventions with other social aims also have crime prevention impacts.Other agencies will benefit in two general ways. First, crime reductionshould be an important element of cost/benefit estimates. Second, crimeprevention considerations may aid program design. We illustrate these byconsidering three important program classes: welfare reform, school-to-worktransitions, and enterprise zones.

Welfare Reform

Perhaps no policy innovation in recent times has attracted such intenseanalytic interest as the effort to fundamentally alter the long-standingbasic federal welfare program, Aid to Families with Dependent Children(AFDC), now converted to Transitional Aid for Needy Families (TANF) andmade principally a state responsibility. The center piece of welfare reformis an effort to move women at risk of becoming long-term welfare recipientsinto employment. This has potentially enormous consequences for their children.If it is successful, a large number of young males will grow up in householdsthat have regular contact with the workplace rather than with welfare checks.On the other hand, if welfare reform fails and large numbers of singlemothers become even poorer and more reliant on illegal earnings, this maywell have criminogenic effects on their children.

On a number of theoretical grounds, this may have an important impacton youth attitudes toward work and hence the prospect of becoming seriousviolent offenders. Successful welfare reform could turn out to be the mostimportant social program for crime prevention in recent decades, thoughthe effects will surely not show up for quite some time.

To our knowledge, little attention is being given to the crime preventionconsequences of this change. It is important the Congress direct OJP totake advantage of the many large scale research and evaluation effortsthat are now being put in place, both by the federal government and bymajor foundations (e.g., the Annie E. Casey Foundation, which is fundinga 5 year $30 million program at the Urban Institute) to ensure that theyassess the extent to which, and the mechanisms by which, welfare reformaffects the criminal behavior of the children of women at risk of long-termwelfare dependence. Though, as always, crime measures increase the complexityof the measurement task, it is likely that crime will play an importantrole in the final evaluation of the reform's success.

School-to-Work Transitions

There is growing interest in improving the flow of non-college boundhigh school graduates into the work force. This has been an area of programinnovation again in which Congress has not encouraged OJP to play a significantrole. Ensuring that evaluations include crime measures will increase thecomprehensiveness of the evaluations and may provide OJP with a major programopportunity.

Enterprise Zones and Community Development Block Grants

A major attribute, arguably the principal attribute, of inner-city communitiesthat makes them unattractive to employers is the high crime rates. Already,OJP is part of a working group cooperating with the consortium of federalagencies that operate the federal EZ and CDBG programs. We believe thiscooperation is important to the success of any such comprehensive programs,in part because reductions in crime may be able to explain variations inemployment outcomes. Understanding the role that crime control plays inattracting investment is another crucial, but understudied part of communitydevelopment programs.

One new multi-agency initiative might be an effort to assess whethera large scale job creation program, backed by other crime prevention measures,can make a substantial and lasting difference in high-crime communities.William Julius Wilson concluded that the lack of jobs was the principalsource of the decline of the neighborhoods that now account for such alarge share of American crime and outlined an ambitious program of interventionsto respond to this problem. As even he admits, it is unclear that sucha program can be implemented but there is certainly good theoretical argumentfor trying.

However, this proposition does force us to confront a central paradoxof prevention evaluation. Learning occurs through examination of variationsin one or a few components but successful interventions aimed at improvinglabor market outcomes for high risk individuals and communities are likelyto involve the simultaneous implementation of a large number of programs.But in a situation where individual interventions seem to have limitedpromise, testing whether a generous cocktail of programs can succeed maybe an important first step.

This insight lays the groundwork for what we believe must be the theoreticalbedrock of any successful program aimed at increasing labor market participationin order to decrease crime: A program must connect a community or individualsto the world of legitimate work so that residents will have the properincentives to acquire the necessary human capital needed for success inthat world. Without that connection, any work program is unlikely tosucceed in a substantial way.

NOTES

1Employment, likecrime, has many dimensions. Jobs vary in wage rates, work satisfaction,and duration. Measured correlation between employment rates and crime maybe confounded by failure to measure variation in job quality adequately.

2We focus here onemployment measures rather than unemployment because in many areas theproblem is less formally defined unemployment than low labor force participationrate. In the face of persistent unemployment, discouragement may lead todrop out even from job search.

3This may reflectthe higher quality of post 1970 data, itself a consequence of the activitiesof the Law Enforcement Assistance Administration and the investment incomputers, among other factors.

4It is difficultto classify all job training programs in terms of our programmatic interest.For example, Job Training and Partnership Act (JTPA)-Title IIA ($1 billionin FY 1993) is aimed at economically disadvantaged adults; some of thoseadults may be involved with the criminal justice system and others maystill be young enough to be reasonably classified as “youth” but many maybe at slight risk of serious criminal involvement. Given the large numberof other JTPA Titles that were more directly targeted at disadvantagedyouth, we did not include any of Title-IIA.

5We identify thetargets as offenders rather than ex-offenders because in fact what is knownis that they have committed a crime. The ex-offender status is a goal ratherthan a description.

6Differences inrelease date meant that a uniform follow-up period would have excludedsignificant periods of post-release exposure for some participants.

7Note once againthat the VDS and Supported Work programs cited above are not technicallypart of the juvenile CJS.

8These questionsare particularly hard to examine because the report provides no data onemployment outcomes.

9JTPA is the mainfederal funding source for job training programs in the U.S.. JTPA fundsa number of discrete program types including a) job search assistance,b) remedial education, c) occupational training d) work experience e) on-the-jobtraining or f) customized training for a particular employer.

10Strictly speakingthe provision of a job is not a job training or education program. Howevermany employment skills are learned on the job; employment increases futureemployability.

11Self-reportfrom program participants about crime involves inquiring about sensitivebehaviors. Official record checks of criminal histories requires obtainingPrivacy Act protected information from a different set of agencies thanthose providing the other outcome data.

12This resultis supportive of Sampson and Laub (1990) who claim that its not the jobbut the social bonds of the workplace, bonds that probably are absent ina short-term subsidized work environment.

13The reportedreduction in homicide rates suggests that the control group had extraordinarilyhigh homicide rates compared to their peers, thus making suspect the claimedreduction in homicides for the experimentals. Homicide reductions accountedfor a large share of the dollar benefits estimated in the evaluation. Onthe other hand the figure used for estimating the value of a life for homicideswas much lower than reported elsewhere in the literature; it is possiblethat the errors roughly cancel out.

14Massey and Denton(1993) argue that the strong desire for racial segregation has also beenan impetus for the exit of jobs.

15The sample isdifferent for the children and the mothers. The children come from a sampleoriginally composed in 1982. They were reinterviewed in 1989. Only 59%of the original sample could be relocated, and most of those relocatedhad not moved from the original location. The potential for bias existsbecause the harder to locate families might vary by suburban or urban location.

16Of course, manyof the same objectives met by housing dispersal programs could be met byencouraging gentrification of older depressed neighborhoods.

17Within thisarea, we noted the absence of any discussion of the role of crime in drivingbusiness to the suburbs, or the potential crime prevention effects of newjob connections in the suburbs.

18In an attemptto determine what would have happened if the zones had not existed, thesesurveys ask zone firms and zone managers how many of the jobs were duedirectly to the incentives. Obviously, it is in the self-interest of bothsets of agents to provide positive answers.

19The surveysdid provide useful insight into the elements of programs which seemed towork best. Bostic (1996) concludes that the incentives provide only marginalincentive for firms to locate in zoning areas. Program success in Californiadepends on supplementing the tax incentives with an active local governmentor community effort, mainly with marketing. Wilder and Rubin (1996) concludethat places with severe economic blight need additional assistance beyondenterprise zones, and autonomous management of the zone is effective. Finally,Erickson and Friedman (1991) conclude that the most successful state programsrestrict the number of zones, use a competitive award process (which pullstogether local resources), and provide significant incentives to theselimited, targeted areas.

20This resultis especially interesting given that a before and after study by Rubin(1990) found substantial effects in New Jersey.

21The lack ofoutcome evaluations is attributed to the flexibility of the programs, thelack of credible evidence about what would have occurred in the absenceof the program and the inability to conceptualize and measure clear outcomesat a neighborhood level.

22A full 80% ofrecipients said that the loan was crucial to their activity, while EZ incentivesare typically important for 30 to 40% of all EZ businesses (Wilder andRubin 1996).

23Although thesenumbers appear to suggest that higher CDGB funding generates improvements,this conclusion is not possible without some other comparison. For example,there may be selection bias, as the result of better organized communities,which are more likely to be improving economically anyway, may do betterin the grant application process.

24Operation Weedand Seed, a major OJP program described in Chapter 3, did make an effortto do so. As Chapter 3 discusses, the evaluation was not an integral partof the project, and no results are as of yet available. Difficulty obtainingbaseline data after program initiation has made the evaluation particularlydifficult.

25Analytically,the problem is to disentangle true peer impacts from the tendency of peoplewith similar unobservable characteristics to live near each other.

REFERENCES

Adams, Kenneth. 1994. "A Large Scale Multidimensional Test of theEffect of Prison Education Programs on Offenders' Behavior." ThePrison Journal. 74: 433-449.

Ahlstrom, Winton and Robert J. Havighurst. 1982. "The Kansas CityWork/Study Experiment." In Daniel J. Safer, ed., School Programsfor Disruptive Adolescents. Baltimore, MD: University Park Press.

Anderson, D.B. 1981. "The Relationship Between Correctional Educationand Parole Success." Journal of Offender Rehabilitation. 5:25.

Anderson, Stephen V. 1995. Evaluation of the Impact of Participationin Ohio Penal Industries on Recidivism. Ohio Department of Rehabilitationand Correction, Bureau of Planning and Evaluation.

Baker, Sally and S. Sadd. 1979. The Court Employment Project Evaluation,Final Report. Vera Institute of Justice.

Baker, Sally and S. Sadd. 1981. Diversion of Felony Arrests: An Experimentin Pretrial Intervention: An Evaluation of the Court Employment Project,Summary Report. U.S. Department of Justice, National Institute of Justice.

Bartik, Timothy J. 1991. Who Benefits from State and Local EconomicDevelopment Policies?. Kalamazoo, MI: W.E. Upjohn Institute for EmploymentResearch.

Bartik, Timothy J. and Richard D. Bingham. "Can Economic DevelopmentPrograms Be Evaluated?" In Richard Bingham and Robert Mier, eds.,Significant Issues in Urban Economic Development. Newbury Park,CA: Sage.

Becker, Gary. 1968. "The Economics of Crime and Punishment."Journal of Political Economy .76

Berk, R.A., K.J. Lenihan, and P.H. Rossi. 1980. "Crime and Poverty:Some Experimental Evidence from Ex-Offenders." American SociologicalReview. 45: 766-786.

Bloom, Howard. 1993. "National JTPA Study: Origins, Objectivesand Issues of Interpretation." Evaluation Forum. Summer: issue9.

Bloom, Howard, et al. 1994. The National JTPA Study: Overview ofImpacts, Benefits, and Costs of Title IIA. Bethesda, MD: Abt Associates.

Boarnet, Marlon G. and William T. Bogart. 1996. "Enterprise Zonesand Employment: Evidence from New Jersey." forthcoming in the Journalof Urban Economics.

Bostic, Raphael W. 1996. Enterprise Zones and the Attraction of Businessesand Investment: The Importance of Implementation Strategies and ProgramIncentives. Washington, DC: Division of Research and Statistics, Boardof Governors of the Federal Reserve System.

Bourgois, Phillipe. 1995. In Search of Respect: Selling Crack inEl Barrio. New York: Cambridge University Press.

Bowman, William. 1993. Evaluating JTPA Programs for EconomicallyDisadvantaged Adults: A Case Study of Utah and General Findings. Annapolis,MD: Annapolis Economic Research.

Bushway, Shawn. 1996. The Impact of a Criminla History Record onAccess to Legitimate Employment. Unpublished Dissertation, CarnegieMellon University.

Case, Anne and Lawrence Katz. 1991. The Company You Keep: The Effectsof Family and Neighborhood on Disadvantaged Youth. National Bureauof Economic Research.

Cave, George, et al. 1993. JOBSTART: Final Report on a Program forHigh School Dropouts. New York: Manpower Demonstration Research Corp.

Cave, George and Janet Quint. 1990. Career Beginnings Impact Evaluation:Findings from a Program for Disadvantaged High School Students. NewYork: Manpower Demonstration Research Corp.

Chiricos, Ted. 1986. "Rates of Crime and Unemployment: An Analysisof Aggregate Research Evidence." Social Problems. 34:187-212.

Clark, P.M. 1990. Crime, Employment and Recidivism: A Review of EmploymentRelated Interventions with Criminal Offenders. College Station, TX:Technology and Society Division, Texas Engineering Experiment Station,Texas A&M University.

Clasby, Richard. 1996. Recidivism Rates and UCi Participation.Draper, UT: Department of Corrections, Correctional Industries Division.

Coffey, Osa and Maia Gemignani. 1994. Effective Practices in JuvenileCorrectional Education: A Study of the Literature and Research 1980-1992.Washington, DC: Office of Juvenile Justice Prevention.

Cook, Philip and Gary Zarkin. 1985. "Crime and the Business Cycle."Journal of Legal Studies. 14: 115-28.

Correctional Educational School Authority. 1990. Academic and VocationalProgram Completers Released from Prison During Fiscal Year 1986-1988: Employment,Recidivism and Corson, Walter, and Joshua Haimson. 1994. The NewJersey Unemployment Insurance Re-Employment Demonstration Project: SixYear Follow-Up Report. Princeton, NJ: Mathematica Policy Research,Inc.

Corson, Walter and Marsha Silverberg. 1994. The School-to-Work/YouthApprenticeship Demonstration: Preliminary Findings. Research and EvaluationReport Series 94-E. Washington, DC: U.S. Department of Labor.

Cost Avoidance. Tallahassee, FL: Correctional Education SchoolAuthority, Planning Research and Evaluation.

District of Columbia Pretrial Services Agency quarterly Drug UseAmong Juvenile Arrestees

Donohue III, John J. and Peter Siegelman. 1996. Is the United Statesat the Optimal Rate of Crime? Allocating Resources Among Prisons, Police,and Social Programs. Working paper for the American Bar Foundation.

Dreier, Peter and David Moberg. 1996. "Moving From the 'Hood: TheMixed Success of Integrating Suburbia." The American Prospect.Winter: 75-9.

Dynarski, Mark and Linda Rosenberg. 1995. Evaluation of Youth FairChance: Summary of Program Characteristics After Year One. Princeton,NJ: Mathematica Policy Research.

Engberg, John and Robert Greenbaum. 1996. "State Enterprise Zonesand Local Housing Markets." Ongoing research at Carnegie Mellon University.

Erickson, Rodney A. and Susan W. Friedman. 1991. "Comparative Dimensionsof State Enterprise Zone Policies." In R.E. Green ed., EnterpriseZones: New Dimensions in Economic Development. Newbury Park, CA: SageFocus Edition.

Fabelo, Tony. 1992. Evaluation of the Reading to Reduce RecidivismProgram. Austin, TX: Texas Criminal Justice Policy Council.

Fagan, Jeffrey. 1995. "Legal Work and Illegal Work: Crime, Work,and Unemployment." In Burton Weisbrod and James Worthy, eds., Dealingwith Urban Crisis: Linking Research to Action. Evanston, IL: NorthwesternUniversity Press.

Farkas, George, et al. 1982. Impacts from the Youth Incentive EntitlementPilot Projects: Participation, Work and Schooling Over the Full ProgramPeriod. New York: Manpower Demonstration Research Corp.

Farrington, David P., et al. 1986. "Unemployment, School Leavingand Crime." British Journal of Criminology. 26: 335-56.

Faupel, Charles E. 1988. "Heroin Use, Crime and Unemployment."The Journal of Drug Issues. 18: 467-79.

Ferguson, Ronald and Jason Snipes. 1995. The Process of Youth Developmentin the Youthbuild Demonstration Project. Cambridge, MA: John F. KennedySchool of Government, Harvard University.

Finn, Mary A. And Katherine G. Willoughby. 1996. "Employment Outcomesof Ex-Offender Job Training Partnership Act (JTPA) Trainees." EvaluationReview. 20: 67-83.

Flanagan, T.J., T.P. Thornberry, K. Maguire, and E. McGarrell. 1988.The Effect of Prison Industry Employment on Offender Behavior: Reportof the Prison Industry Research Project. Albany, NY: Hindelang CriminalJustice Research Center.

Freeman, Richard B. 1995."The Labor Market." In J.Q. Wilsonand Joan Petersilia, eds., Crime. San Francisco, CA: Institute forContemporary Studies.

Gerber, Jurg and Eric J. Fritsch. 1993. Prison Education and OffenderBehavior: A Review of the Scientific Literature. Huntsville, TX: TexasDepartment of Criminal Justice, Institutional Division.

Gleason, Philip and Mark Dynarski. 1995. Falling Behind: Characteristicsof Students in Federally Funded Dropout Prevention Programs. Princeton,NJ: Mathematica Policy Research.

Good, David H. and Maureen A. Pirog-Good. 1987. "Employment, Crimeand Race." Contemporary Policy Issues. 5: 91-104.

Gottfredson, Denise C. 1985. "Youth Employment, Crime, and Schooling:A Longitudinal Study of a National Sample." Developmental Psychology.21: 419-32.

Gottfredson, Michael and Travis Hirschi. 1990. A General Theory ofCrime. Stanford: Stanford University Press.

Gray, Tara and Jon'a F. Meyer. 1995. "Expanding Prison IndustriesThrough Privatization." In G. Larry Mays and Tara Gray, eds., Privatizationand the Provision of Correctional Services. Anderson Press.

Green, T., J. Richmond, and J. Taira. 1993. POWER Program (ProvidingOpportunities for Work, Education and Readiness): Final Evaluation Report.Minneapolis, MN: Hennepin County Administrative Support Division, Bureauof Community Corrections.

Greenwood, Peter and Susan Turner. 1993. "Evaluation of the PaintCreek Youth Center: A Residential Program for Serious Delinquents."Criminology. 31: 263-280.

Grey, Doug and Greg Wren. 1992. Pre-release at South Idaho CorrectionalInstitution: Something Works. Boise, ID: South Idaho Correctional Institution.

Grossman, Jean Baldwin. 1992. The Evolution of Youth Employment andTraining Programs. Philadelphia: Public/Private Ventures.

Gruber, David and Frazierita Davidson. 1986. State Employment Initiativesfor Youth: A Assessment of the Demonstration. Philadelphia: Public/PrivateVentures.

Hale, Chris and Dima Sabbagh. 1991. "Testing the Relationship BetweenUnemployment and Crime: A Methodological Comment and Empirical AnalysisUsing Time Series Data from England and Wales." Journal of Researchin Crime and Delinquency. 28: 400-17.

Hartmann, David, P. Friday, and Kevin Minor. 1994. "ResidentialProbation: A Seven Year Follow-Up Study." Journal of Criminal Justice.22: 503-15.

Hayward, Becky and G. Talmadge. 1993. Evaluation of Dropout Preventionand Re-entry Projects in Vocational Education: Draft Final Report.Research Triangle Park, NC: Research Triangle Institute.

Heckman, James J. 1994. "Is Job Training Oversold?" ThePublic Interest. Spring: 91-115.

________. 1996. "What Should Our Human Capital Investment PolicyBe?" Jobs & Capital. 5: 3-10.

_________ , Rebecca Roselius and Jeffrey Smith. 1993. "U.S. Educationaland Training Policy: A Re-evaluatiion of the Underlying Assumption Behindthe `New Consensus." University of Chicago Graduate School of PublicPolicy.

Henry, Patrick. 1988. Effects of Private Industry Participation onInmates Adjustment Experiences Before and After Release. St. Petersburg,FL: Department of Sociology, Eckerd College.

Hershey, Alan and Linda Resenberg. 1996. School-to-Work ImplementationProgress: The State Perspective in Early 1996. Princeton, NJ: MathematicaPolicy Research, Inc.

Hillsman, Sally. 1982. " Pretrial Diversion of Youthful Adults:A Decade of Reform and Research." The Justice System Journal,7/3:361-387.

Hoffman, P. And B. Meierhoefer. 1979. "Post-release Experiencesof Federal Prisoners: A Six-Year Follow-Up." Journal of Researchin Crime and Delinquency. 28: 400-17.

Holzer, Harry J., Richard N. Block, Marcus Cheatham, and Jack H. Knott.1993. "Are Training Subsidies for Firms Effective? The Michigan Experience."Industrial and Labor Relations Review. 46: 625-636.

Home Builders Institute. 1996. Project TRADE (Training, Restitution,Apprenticeship, Development and Education) Report. Washington, DC:Home Builders Institute.

Hughes, Mark Alan. 1993. Over the Horizon: Jobs in the Suburbs ofMajor Metropolitan Areas. Philadelphia, PA: Public/Private Ventures.

Jacobs, James B., Richard McGahey, and Robert Minion. 1984. "Ex-OffenderEmployment, Recidivism, and Manpower Policy: CETA, TJTC, and Future Initiatives."Crime and Delinquency. 30: 486-506.

James, Thomas S. and Jeanne M. Granville. 1984. "Practical Issuesin Vocational Education for Serious Juvenile Offenders." In R. Mathias,P. DeMuro, and R. Allinson, eds., Violent Juvenile Offenders: An Anthology.San Francisco, CA: National Council on Crime and Delinquency.

Johnson, Candace. 1984. The Effects of Prison Labor Programs on Post-ReleaseEmployment and Recidivism. Doctoral Dissertation, Florida State University.

Johnson, Terry, et al. 1991. Evaluation of the Impacts of the WashingtonAlternative Work Search Experiment. Unemployment Insurance OccasionalPaper 91-4. Washington, DC: Department of Labor.

Katz, Jack. 1989. The Seductions of Crime. New York: Basic Books.

Kiss, Dr. Mary Ellen. 1996. Youth Pilot Projects: Multisite Competition.Bowie, MD: Resource Development Group, Inc.

Lattimore, Pamela K., Ann Dryden Witte, and Joanna R. Baker. 1990. "ExperimentalAssessment of the Effect of Vocational Training on Youthful Property Offenders."Evaluation Review. 14: 115-33.

Leiber, Michael J. And Tina L. Mawhorr. 1995. "Evaluating the Useof Social Skills Training and Employment with Delinquent Youth." Journalof Criminal Justice. 23: 127-41.

Lemert. Edwin. 1951. Social Pathology. New York: McGraw-Hill.

Levin, Laurie, Bernard McMullan, and Paul Rynders. 1987. Venturesin Community Improvement: Findings from a Four Site Replication Initiative.Philadelphia: Public/Private Ventures.

Maguire, Kathleen E., Timothy J. Flanagan, and Terence P. Thornberry.1988. "Prison Labor and Recidivism." Journal of QuantitativeCriminology. 4: 3-18.

Mallar, Charles and Craig V.D. Thornton. 1978. "Traditional Aidfor Released Prisoners: Evidence from the LIFE Experiment." HumanResources. 13: 208-36.

Mallar, Charles, et al. 1982. Third Follow-Up Report of the Evaluationof the Economic Impact of the Job Corps. Princeton, NJ: MathematicaPolicy Research, Inc.

Markley, H., K. Flynn, and S. Bercaw-Dooen. 1983. "Offender SkillsTraining and Employment Success: An Evaluation of Outcomes." Correctionaland Social Psychiatry and Journal of Behavior Technology Therapy. 29:1-11.

Martinson, Robert. 1974. "What Works? Questions and Answers aboutPrison Reform." Public Interest. 35: 22-54.

Mathematica Policy Research. 1993. Fifth Year Employment Impactsof CET. Princeton, NJ: Mathematica Policy Research, Inc.

Mayer, Susan. 1991. "How Much Does a High School's Racial and SocioeconomicMix Affect Graduation and Tennage Fertility Rates?" In C. Jencks andP. Peterson, eds., The Urban Underclass. Washington, DC: The BrookingsInstitution.

Menon, Ramdas, Craig Blakely, Dottie Carmichael, and Laurie Silver.1992. An Evaluation of Project RIO Outcomes: An Evaluative Report.College Station, TX: Public Policy Resources Laboratory.

Moody, Mark, et al. 1994. Evaluation of Maryland's Tomorrow Program:Cohort I and II. Maryland State Department of Education.

Moses, Marilyn C. 1996. Project Re-Enterprise: A Texas Program.Washington,DC: U.S. Department of Justice, National Institute of Justice.

Myers, Samuel. 1982 . "Racial Differences in Postprison Employment."Social Science Quarterly. 655-669.

Myers, Samuel L. 1983. "Estimating the Economic Model of Crime:Employment Versus Punishment Effects." The Quarterly Journal ofEconomics. February: 157-66.

Nagin, Daniel and Joel Waldfogel. 1995. "The Effects of Criminalityand Conviction on the Labor Market Status of Young British Offenders."International Reveiw of Law and Economics, 15:109-126.

________. 1994. "The Effect of Conviction on Income Over the LifeCourse." Working Paper, Yale University.

National Institute of Justice. 1991. Post-release Employment Project:Summary of Preliminary Findings. Washington, DC: U.S. Department ofJustice, Federal Bureau of Prisons.

Needels, Karen. 1996. "Go Directly to Jail and Do Not Collect?A Long-term Study of Recidivism, Employment and Earnings Patterns AmongPrison Releasees." Journal of Research on Crime and Delinquency.33:471-496.

Office of Juvenile Justice Delinquency Prevention. 1995. Guide forImplementing the Comprehensive Strategy for Serious, Violent and ChronicJuvenile Offenders. Washington, DC: U.S. Department of Justice.

Office of Juvenile Justice Delinquency Prevention. 1994. EffectivePractices in Juvenile Correctional Education: A Study of the Literatureand Research 1980 - 1992. Washington, D.C., U.S. Department of Justice.

Papke, Leslie E. 1994. "Tax Policy and Urban Development: Evidencefrom the Indiana Enterprise Zone Program." Journal of Public Economics.54: 37-49.

________. 1993. "What Do We Know About Enterprise Zones?."In James M. Poterba ed., Tax Policy and the Economy. Cambridge,MA: MIT Press.

Pezzin, Lillian E. 1995. "Earnings Prospects, Matching Effects,and the Decision to Terminate a Criminal Career." Journal of QuantitativeCriminology. 11: 29-50.

Phillips, Llad and Harold L. Votey, Jr. 1987. "The Choice BetweenLegitimate and Illegitimate Work: Micro Study of Individual Behavior."Contemporary Policy Issues. 5: 59-72.

Piehl, Anne M. 1995. Learning While Doing Time. Cambridge, MA:John F. Kennedy School of Government, Harvard University.

Piliavin, Irving and Rosemary Gartner. 1981. The Impact of SupportedWork on Ex-Offenders. Madison, WI: Institute for Research on Povertyand Mathematical Policy Research, Inc.

Report prepared for Manpower Demonstration Research Corp.

Piliavin, Irving and Stanley Masters. 1981. The Impact of EmploymentPrograms on Offenders, Addicts, and Problem Youth: Implications from SupportedWork. Madison, WI: University of Wisconsin, Institute for Researchand Poverty Discussion.

Public/Private Ventures. 1994. Youth in JTPA: Findings from a 1990Field Study. Philadelphia: Public/Private Ventures.

Reuter, Peter; Robert J. MacCoun and Patrick J. Murphy. 1990. Moneyfrom Crime: The Economics of Drug Selling in Washington, DC. SantaMonica: RAND.

Rosenbaum, James E. 1992."Black Pioneers--Do Their Moves to theSuburbs Increase Economic Opportunity for Mothers and Children." HousingPolicy Debate. 2: 1179-1213.

Rovner-Pieczenik, Roberta. 1973. A Review of Manpower R&D Projectsin the Correctional Field (1963-1973). Manpower Research Monograph,no. 28. U.S. Department of Labor: Washington, DC.

Rubin, M. 1990. "Urban Enterprise Zones: Do They Work? Evidencefrom New Jersey." Public Budgeting and Finance. 10: 3-17.

Rubinstein, Michael. 1995. Giving Youth a Second Chance: The Evaluationof the Alternative Schools Demonstration Program. Washington, DC: PolicyStudies Associates.

Sadd, S., et al. 1983. Alternative Youth Employment Strategies Project:Final Report. New York: Vera Institute.

Saner, Hilary; Robert J.MacCoun and Peter Reuter. 1995. "On theUbiquity of Drug Selling Among Youthful Offenders in Washington, D.C.,1985-1991." Journal of Quantitative Criminology. 11: 337-362.

Saylor, William G. and Gerald G. Gaes. 1996. PREP: Training Inmatesthrough Industrial Work Participation, and Vocational and ApprenticeshipInstruction. Washington, DC: U.S. Federal Bureau of Prisons.

Schwartz, Richard D. and Jerome H. Skolnick. 1964. "Two Studiesof Legal Stigma." In Howard Becker, ed., The Other Side: Perspectivesof Deviance. New York: Free Press.

Silverberg, Marsha. 1995. Facing the Challenge of Change: Experiencesand Lessons of the School-to-Work/Youth Apprenticeship Demonstration.Princeton, NJ: Mathematica Policy Research.

Smith, Michael E. and James W. Thompson. 1983. "Employment, Youth,and Violent Crime." In Kenneth Feinberg, ed., Violent Crime inAmerica. Washington, DC: National Policy Exchange.

Smith, Thomas J., Gary Walker, and Rachel Baker. 1987. Youth andthe Workplace: Second Chance Programs and the Hard-to-Employ. Philadelphia:Public/Private Ventures.

Spencer, F. 1980. "The Effects of an Experimental Vocational InterventionModel upon Hard-Core Unemployed Ex-Offenders." Journal of OffenderCounseling, Services and Rehabilitation. 4: 343-356.

Stewart, James K. 1986. "The Urban Strangler." Policy Review.summer: 6-10.

Taylor, Ronald L. 1990. "Improving the Status of Black Youth."Youth & Society. 22: 85-107.

Tennebaum, Judith. 1996. Southwest Metro Reverse Commute Project:Interim Reports and Preliminary Findings. Minneapolis, MN: Loring Nicollet-BethlehemCommunity Centers, Inc.

Thornberry, Terence P., Melanie Moore, and R.L. Christenson. 1985. "TheEffect of Dropping Out of High School on Subsequent Criminal Behavior."Criminology. 23: 3-18.

Uggen, Christopher. 1996. "Estimating the `True Effect' of Workon Crime: A Dynamic Analysis of Supported Employment and Desistance."Paper delivered at the American Society of Criminologists meetings.

_______. 1994. "Innovators, Retreatists, and the Conformist Alternative:A Job Quality Model of Work and Crime." unpublished paper, Universityof Wisconsin-Madison.

U.S. Department of Housing and Urban Development. 1996. ExpandingHousing Choices for HUD-Assisted Families. Washington, DC: U.S. Departmentof Housing and Urban Development, Office of Policy Development and Research.

U.S. Department of Labor. 1995. What's Working (and What's Not).Washington, DC: Office of the Chief Economist, U.S. Department of Labor.

________. 1990. Guidebook for the Operation of the Federal BondingProgram. Washington, DC: U.S. Department of Labor, Employment and TrainingAdministration.

________. 1992. Training and Employment Report of the Secretary ofLabor. Washington, DC: U.S. Department of Labor, Employment and TrainingAdministration.

Van Stelle, Kit, Julie R. Lidbury, and D. Paul Moberg. 1995. FinalEvaluation Report Specialized Training and Employment Project (STEP).Madison, WI: University of Wisconsin Medical School, Center for HealthPolicy and Program Evaluation.

Vera Institute of Justice. 1970. Pre-Trial Intervention: The ManhattanCourt Employment Project. New York: Vera Institute of Justice.

Walker, Gary and Frances Viella-Velez. 1992. Anatomy of a Demonstration.Philadelphia: Public/Private Ventures.

Walker, R. and D. Greenstreet. 1991. "The Effect of GovernmentIncentives and Assistance on Location and Job Growth in Manufacturing."Regional Studies. 25: 13-30.

White, Thomas, et al. 1993. Maryland's Tomorrow Evaluation: YearFour. Washington, DC: Pelavin Associates.

Wilder, Margaret G. and Barry M. Rubin. 1996. "Rhetoric vs. Reality:A Review of Studies on State Enterprise Zone Programs." Journalof the American Planning Association. Fall: .

Williams, Nicolas, Francis Cullen, John Paul Wright. 1996. "LaborMarket Participation and Youth Crime: The Neglect of 'Working' in DelinquencyResearch." Forthcoming, Social Pathology.

Wilson, Paul R. 1994. "Recidivism and Vocational Education."Journal of Correctional Education. 45: 158-63.

Wilson, William Julius. 1987. The Truly Disadvantaged: The InnerCity, The Underclass and Public Policy. Chicago,IL: University of ChicagoPress.

__________________ . 1996. When Work Disappears: The World of theNew Urban Poor . New York: Alfred A. Knopf.

Urban Institute. 1995. Federal Funds, Local Choices: An Evaluationof the Community Development Block Grant Program. Washington, DC: UrbanInstitute, Center for Public Finance and Housing.

Wolf, Wendy, Sally Leiderman, and Richard Voith. 1987. The CaliforniaConservation Corps: An Analysis of Short-Term Impacts on Participants.Philadelphia: Public/Private Ventures.

Wolf, Wendy, John M. Kelley, Jerene Good, and Richard Silkman. 1982.The Impact of Pre-Employment Services on the Employment and Earningsof Disadvantaged Youth. Philadelphia, PA: Public/Private Ventures.

Wolfgang, Marvin, Robert Figlio and Thorstin Sellin. 1972. Delinquencyin a Birth Cohort. Chicago, IL: University of Chicago Press.

Wren, C. and M. Waterson. 1991. "The Direct Employment Effectsof Financial Assistance to Industry." Oxford Economic Papers.43: 116-138.

Chapter 7

PREVENTING CRIME AT PLACES

John E. Eck

WHY PLACES ARE IMPORTANT

Most places have no crimes and most crime is highly concentrated inand around a relatively small number of places. If we can prevent crimeat these high crime places, then we might be able to reduce total crime.Do we have evidence that this is feasible?

Places have received relatively little attention in crime policy soit is important to define "place." A place is a very smallarea reserved for a narrow range of functions, often controlled by a singleowner, and separated from the surrounding area. By small we mean thata location is smaller than a neighborhood or community. Examples of placesinclude stores, homes, apartment buildings, street corners, subway stations,and airports. We will also include mobile places, such as buses, in ourdiscussions.

Concentration of crime at places is predicted by routine activity theory(Cohen and Felson 1979; Felson 1994) and offender search theory (Brantinghamand Brantingham 1981). Some of the original evidence for clustering ofcrime at places was found in Boston (Pierce, Spaar and Briggs 1986) andMinneapolis (Sherman, Gartin and Buerger 1989). Additional evidence forcrime concentration at places has been found for specific types of crime.Crow and Bull (1975) noted over 20 years ago that most convenience storeshave no or few robberies, but a few have many robberies. In England andCanada a growing body of research has revealed that in high burglary neighborhoodsmost residences have no burglaries, but a few residences suffer repeatedburglaries (Forrester et. al. 1988; Forrester et. al. 1990; Polvi et. al.1990; Farrell 1995). Among drinking establishments, a few bars have mosttavern-related violence (Sherman, Schmidt, and Velke 1992). Ten percentof the fast food restaurants in San Antonio, Texas account for one thirdof the property crimes at such restaurants (Spelman 1995b). In Kansas Cityand Indianapolis, gun crimes were found to be highly concentrated at afew places (Sherman and Rogan 1995b). Drug dealing is highly concentratedin a few locations, even in areas with a high volume of drug dealing (Weisburd,Green and Ross 1994; Eck 1994; Sherman and Rogan 1995a). This clusteringis most apparent when compared to repeat offending and repeat victimizations.Combining the results from several studies, Spelman estimated that 10 percentof the victims in the United States are involved in about 40 percent ofthe victimizations, that 10 percent of the offenders are involved in over50 percent of the crimes, and that 10 percent of the places are sites forabout 60 percent of the crimes (Spelman and Eck 1989). Further, the concentrationof crimes at a few places is relatively stable over time (Spelman 1995a,1995b). These findings suggest that something about a few places facilitatescrimes and something about most places prevents crimes.

Blocking Criminal Opportunities

The oldest forms of crime prevention were undertaken with the knowledgethat making changes to places might prevent criminal events. These changesinvolve making crime more difficult, risky, less rewarding, or less excusable.This approach is known as opportunity blocking (Clarke 1992; 1995; Clarkeand Homel, forthcoming). Opportunity blocking does not have to be doneat places. It can also be built into targets (for example, designing anti-theftdevices into automobiles [Clarke 1995] or printing holograms and photoson credit cards to curtail forgery and fraud).

Designing methods for blocking crime opportunities is the domain ofSituational Crime Prevention (Clarke 1992; 1995). In this chapter we examineopportunity blocking at places, a subset of Situational Crime Prevention.It not only has a much longer history than offender-based prevention measures,it is used much more widely and in more settings than any other form ofcrime prevention. The vast majority of efforts to block crime opportunitiesat places are carried out and paid for by businesses, individuals, andlocal governments. Because places themselves have only recently becomea subject for study by criminologists (Eck and Weisburd 1995), the Officeof Justice Programs has funded very few explicit place-focused programsor tests of place-focused prevention. We will see that this lack of attentionhas limited our knowledge about this approach to prevention.

Opportunity blocking at places may have a greater direct effect on offendersthan other crime prevention strategies. This is because place-focused tacticsmight influence offenders when they are deciding to commit a specific crime.Most offender based strategies try to sway offenders weeks, months, oryears before they confront a tempting criminal opportunity. If offenderspay closer attention to the situation immediately before them than to theuncertain long term risks of their behavior, then it is quite possiblethat prevention at places may have a greater impact on offending than increasesin penalties or less tangible increases in risks (e.g., decreases in policeresponse time, increased police presence, or greater numbers of arrestsand convictions). Because hotspots of crime are themselves clustered, ifcrime at these few places can be substantially reduced, communities canbe made safer.

Although opportunity blocking takes a different approach than programsdesigned to change the life-course of potential and existing offenders,these two approaches can work together. Keeping cookies out of sight oftoddlers is not only different from instructing them not to take the cookies-- and sanctioning them when they yield to temptation -- it reinforcesinstructions and sanctions by eliminating the temptation. For people withlow self-control and low ability to see long term consequences of behavior(Gottfredson and Hirschi 1990) addressing the immediate circumstances surroundingcrime opportunities may amplify the effectiveness of other strategies designedto address the prevalence of such offenders.

The evaluations selected for review in this chapter were required tomeet three criteria. First, they must describe crime opportunity blockingat places. Second, they had to examine the manipulation of places, usuallyintentional changes in which the changes clearly precede any change incrime. Third, each evaluation must report outcome data, typically a measureof crime. We did not examine studies of implementation and management thatdid not measure an impact on crime. In short, we looked at evaluationsof the impact on crime of intentional changes at places.

Over the last decade, police have paid attention to places, or "hotspots,"of crime (Eck and Spelman 1989; Sherman and Weisburd 1995). This chapterdoes not review police efforts at places that relied solely on patrolling,investigations, or other enforcement. These are reviewed in Chapter 8 ofthis report. We did review evaluations of interventions involving policeagencies when the intervention was a tactic that could also have been implementedby other agencies or institutions. Nuisance abatement, for example, hasbeen implemented by police agencies, but it has also been implemented byprosecutors' offices, city attorneys, and citizen groups. In short, whoimplemented the tactic was of less importance than the fact that the tacticwas applied at places.

Table 7-1 summarizes the evaluations examined. When a report describedseveral separate quasi-experiments we treated them as distinct interventions.Two-thirds of the evaluations were conducted outside the United States,particularly in the United Kingdom and Australia. Only six studies werefunded by an OJP-related agency. Although the OJP funded evaluations compriseonly 15 percent of U.S. interventions, recent efforts by the National Instituteof Justice (NIJ) are improving our understanding of places. NIJ's DrugMarket Analysis Project helped introduce computer mapping of crime anddrug places to police agencies and funded one of the randomized experimentsdescribed in this chapter (as well as several evaluations described inChapter 8, on police prevention). The recent establishment of the CrimeMapping Research Center at NIJ has the potential to increase our knowledgeof what works at places.

Violent crimes -- homicide, robbery and assault (no rape preventionevaluations were found) -- were the focus of 31 percent of the evaluations.Three of the six drug dealing prevention evaluations were OJP funded. Overhalf the evaluations examined serious crime (either a mixture of violentand non-violent crimes, or just non-violent crimes). Thus, 90 percent ofthe evaluations focused on serious property, personal, or drug crimes.Only 20 percent of the evaluations examined minor offenses, such as propertydamage, vandalism, minor thefts or incivilities.

As in earlier chapters, evaluations were graded using the scientificmethods score (1=correlations between tactics and crime and studies withoutpre-intervention measures; 2=pre-post designs without control places; 3=pre-postdesigns with controls or time-series designs with at least five time periodsprior to the intervention; 4=studies of interventions in a large sampleof places compared to similar places without interventions; and 5=randomizedcontrolled experiments). The modal score was 3, but a substantial numberof evaluations only scored 2. There were few studies at either extreme(1 or 5).

Table 7-1: SUMMARY OF EVALUATIONS Reports Examined 78 Percent of Interventions Interventions Examined 99 100% Funded by OJP agency* 6 6% Setting & Crimes Study U.S. Setting 40 40% Violent Crime 31 31% Drug Dealing 6 6% Serious Crime 52 53% Minor Crime 20 20% Scientific Methods Score 1 -- correlation/cross 7 7% section 2 -- before/after (b/a) 36 36% 3 -- b/a with control & 50 51% time series 4 -- large sample 4 4% quasi-exper. 5 -- randomized experiments 2 2% Evidence of Crime Change Down 89 90% Up 2 2% * Including predecessor agencies within the Department of Justice. 

GENERAL FINDINGS

These evaluations are consistent with the hypothesis that opportunityblocking at places can prevent crime, at least under some circumstances.Ninety percent of the evaluated interventions displayed evidence of crimereduction effects. Often these reductions were large. As we will see, thesefindings are consistent across a variety of evaluation designs, settings,and interventions. Although few of them have been replicated at a stronglevel of scientific evidence, there is good reason to invest in furthertesting of these tactics. Do these tactics displace crime? We will delveinto this issue at the end of this chapter, but for now we will state thatdisplacement seldom overwhelms prevention effects.

How much can we conclude about specific types of intervention, at specificplaces, against specific crimes? The answer is, we usually cannot be confidentabout what works where. We will discuss this finding in greater detaillater in this chapter. We looked at nine types of places in four broadcategories: In the following sections we describe the results of evaluationsat residential places; money spending places (retail stores, banks andmoney handling businesses, and bars and drinking establishments); transportationplaces (public transportation facilities, parking lots, and airports);and other public places (open urban spaces and public coin machines). Thenine types of places examined were not selected on theoretical grounds.They were selected because these were the places for which evaluationsexisted. Clearly, our knowledge about place-focused tactics is limitedto a relatively few place types. Within each category we examine look ata variety of crime prevention tactics.

APARTMENTS AND RESIDENCES

Places where people live are the subject of this section. We will examinesix types of interventions at residential properties, many of which arein public housing in Great Britain and the United States. Public housingcomplexes have become notorious for high crime rates in the United States.Dunworth and Saiger (1994) found that public housing complexes in threecities had higher rates of violent crimes and drug arrests than nearbyneighborhoods or surrounding cities, but there was a great deal of variationamong housing projects within each of the cities. We will see that crimein British public housing estates can also be a problem. First we willlook at efforts to reduce crime by restricting movement through apartmentcomplexes. Next we will look at improving security by improving locks andbarriers on windows and doors. Third, we will examine property marking.Improving watching of residences is the subject of the fourth section.In the fifth section we will look at the effectiveness of multiple tacticinterventions to prevent burglaries at dwellings with a history of burglary.Finally, we will turn our attention to methods to compel place managersto reduce drug dealing on their rental property. Table 7-2 summarizes theevaluations of crime prevention in residential settings.

Restricting Pedestrian Access and Movement

Oscar Newman's Defensible Space (1972) stimulated interest inthe link between the built environment and crime in residential areas.Newman compared two public housing complexes and asserted that the differencesin design were the principal reasons for the differences in crime. Thelimited number of places observed and the failure to take into accountother differences (most notably the age distribution of tenants) suggeststhat his conclusions may have been overstated (Mawby 1977; Mayhew 1979;Merry 1981; Taylor, Gottfredson, and Brower 1980). Newman expanded on hisideas in a later book (1980). Other studies of the influence of designhave compared more sites (Coleman 1985; Poyner 1983; Poyner and Webb 1991).All pointed to the association of design features and crime, particularlyfeatures that allow unfettered movement through residential complexes.Two of these evaluations examine changes in residential sites that breakup large residential complexes into smaller components.

Newman (1980; 1996) reports on the effect of changes to the ClawsonPoint public housing complex in the Bronx. The complex was changed by reducingthe number of pedestrian routes through the project, creating separateareas within the complex, improving lighting, and enhancing the surfaceappearance of the buildings. Newman (1996) reports a 54 percent declinein the crime rate and a 62 percent decline in the rate of serious crime(burglary, robbery and assault). No control group was used.

Poyner (1994) describes a retrospective evaluation of the effect ofthe removal of elevated walkways connecting buildings in a British publichousing complex. The walkways were thought to facilitate robberies of residents.He reports a reduction in purse snatching, but no reduction in burglaries.An entry phone was installed at one entrance and this too may have contributedto the decline in purse snatches. Although auto thefts declined, Poynerwas unable to determine if this was due to the removal of the walkwaysor the presence of construction workers while the removal was underway.There was no comparison to control places.

Restricting the movements of pedestrians was also part of a 1991 effortto reduce crime in several of Chicago's worst public housing buildings(Popkin et. al. 1995b). The approach included door-to-door police inspectionsof all units within the buildings. Ground floor entrances were enclosedin new lobbies and guard stations were installed along with metal detectors.Residents were issued identification cards and asked to present them whenentering the buildings. In addition to housing authority and private securityguards, the Chicago Public Housing Authority organized tenant patrols.Finally, a set of drug prevention services were provided tenants.

Popkin and her colleagues (1995b) attempted to evaluate this program.They interviewed a sample of residents in two complexes and asked themif conditions had improved, remained the same, or became worse followingthe interventions. The surveys found that 74 percent and 88 percent ofrespondents (depending on the complex) said shootings and fighting in theirbuilding had declined. It also found that 40 percent and 64 percent ofthe residents interviewed said drug dealing in their building had declined.These retrospective assessments by residents were a substitute for pre-treatmentmeasures of crime and drug problems. The lack of control groups and truepre-treatment measures of crime, along with the implementation of multiplesimultaneous interventions means that we cannot determine if the restrictionson pedestrian access contributed to improvements.

Collectively, these evaluations are suggestive of possible beneficialeffects of reducing pedestrian movement through large public housing complexes.The weak designs used to evaluate these interventions temper our confidencein these types of interventions.

Target Hardening

Providing locks and improved security to access points is a commonlyused burglary prevention tactic. The installation of improved locks anddoors at two English public housing complexes was evaluated by Tilly andWebb (1994). Both studies used a pre-post design compared to a controlarea. In one complex burglaries declined 59 percent. In the other, burglariesdecline over 90 percent relative to the control area.

The displacement of burglars to less protected locations is commonlyraised as a threat to the effectiveness of place-focused interventions.Patricia Allatt (1984) has been one of the few evaluators to explicitlytest for displacement effects. In addition to identifying the target residenceswhich received improved ground floor entrance security, she examined theresidences in the area immediately adjacent to the target area. And sheused a control area that was far enough from the treatment area that itwould not be contaminated by displacement. She found that burglaries inthe target area increased by 9 percent one year after implementation, butin the control area burglaries had increased 77 percent. This suggeststhe program may have reduced potential burglaries, compared to what theywould have been in the absence of the program. Burglaries increased 86percent in the displacement area, but relative to the control area thiswas only a 9 percent increase over what could have been expected withoutthe program. Thus, she was able to determine that displacement may haveoccurred, but was small relative to the overall program effect on the targetarea.

Target hardening appears to reduce burglaries without major displacementeffects. However, with only two studies, more rigorous evaluations wouldmake valuable contributions to our knowledge of what works in place-focusedcrime prevention.

Property Marking

A third approach to controlling burglaries is to make burglary targetsunattractive to offenders. Laycock (1985; 1991) reports on the evaluationof a property marking campaign in two isolated Welsh communities. She reportsa 40 percent decline in burglaries at residences where people said theyengaged in property marking compared to the control group of non-participatingresidences. These results might be due to property marking, but the resultscould also occur if less vulnerable residents participated in the programand more vulnerable residents did not participate. Gabor (1981) also evaluatedproperty marking in a Canadian neighborhood. He found a 75 percent increasein seasonally adjusted burglaries per dwelling unit by comparing the 24months before the program to 18 months after the property marking. Clearly,with two contradictory studies we cannot be confident that property markingis an effective method for reducing burglaries to residences.

Closed-Circuit Television (CCTV)

CCTV was used in fifteen housing complexes for elderly residents inManchester, England. Chatterton and Frenz (1994) report a decline in burglaryand burglary attempts of 79 percent across all complexes. Again, naturaltrends in burglary were not reported due to the absence of control places.This single weak study is insufficient as a basis for crime preventionpolicy. We will return to the use of CCTV in other settings.

Multi-tactic Interventions and Repeat Victimizations

Crime prevention in residential settings often involves the implementationof a variety of measures. Evaluations of such interventions usually cannotestimate the relative effectiveness of the component parts, but they canshow whether prevention is possible. Meredith and Paquette (1992) examineda multiple tactic approach to controlling burglaries in a Canadian apartmentbuilding. The program included apartment watch (like neighborhood watchbut for apartment dwellers), target hardening, property marking, lightingimprovements, and an assortment of other interventions. Reported burglariesdropped 82 percent from the year before to the year after the preventionmeasures were put in place. No control group was used, so again this dropmay have been due to a general trend toward fewer burglaries in the surroundingarea.

A growing body of evidence suggests that a few victims are involvedin a large proportion of victimizations (Farrell 1995). Most of the researchon this topic has been conducted in Great Britain, where programs to reduceburglaries of dwellings have been based these findings. The Kirkholt publichousing complex has received considerable attention in England becauseevaluations indicated that focusing on residences with previous burglariesis effective (Forrester, Chatterton, and Pease 1988; Forrester, et. al.1990; Pease 1991; Tilly 1993a). A number of interventions were used ateach targeted residence, including target hardening and organizing residentsin surrounding homes to watch the burgled house. However, two tactics deservesspecial mention. Like many low income publicly subsidized projects in England,the residences in Kirkholt had coin-operated gas meters. Residents putcoins in the meter to get a preset amount of gas for heating and cooking.Officials periodically empty these meters, but for weeks the meters cancontain a great deal of cash. These meters were the target of many of theburglaries in Kirkholt and removing them was an important tactic in theproject. Another part of the Kirkholt repeat victimization project wasorganizing the residents surrounding burgled dwellings to watch the victimizedhome. This was referred to as "cocoon neighborhood watch" becauseinstead of organizing the entire neighborhood, the police focused onlyon the people living around at-risk places (Forrester, Chatterton, andPease 1988; Forrester, et. al. 1990).

The 40 percent decline in burglaries in the first year following thestart of the program, and subsequent decline over the next three years(controlling for seasonality and surrounding area burglary trends) cannotbe attributed to any single tactic (Forrester, Chatterton, and Pease 1988;Forrester, et. al. 1990). Thus, we do not know which tactics worked.

Another repeat victimization program in Great Britain used a gradedresponse to repeat victimization (Anderson, Chenery, and Pease 1995a; 1995b).Residents that reported a single burglary received a "bronze"response. This included crime prevention advice from the police, cocoonneighborhood watch, and improvement in dwelling security. If a residentwas a victim of a second burglary within a year the police stepped up patrollingof the location, and put warning stickers on the dwelling. This was the"silver" response. If a third burglary was reported within ayear then the "gold" response was put into place. This includedthe use of video surveillance of the location and even more intense policepatrols. Anderson, Chenery and Pease (1995b) report a 19 percent reductionin burglaries relative to changes in burglary in the surrounding area.

Repeat victimization and crime prevention programs based on repeat victimizationsare interesting. Because housing projects in Great Britain and the UnitedStates have important differences (the presence of coin-operated gas metersis just one example), research in the United States should be undertakento determine if repeat burglaries are a problem in the United States, andif repeat victimization responses are effective. The National Instituteof Justice is currently sponsoring studies examining repeat victimization.

Reducing Drug Dealing and Crime in Private Rental Places

Despite the fact that the management of private rental housing has onlyrecently been examined as a crime risk factor, we have strong evidencethat improving management of rental properties can reduce drug relatedcrime. A study of retail drug dealing locations in San Diego found thatsmaller apartment buildings were more likely to be selected by drug dealersthan the larger buildings, primarily because owners of the smaller buildingshad less management resources to control the behaviors of place users (Eck1994; 1995). Spelman (1993) studied residential locations that had beenabandoned by their owners and found that they were magnets for crime. Theeffectiveness of compelling place managers to control the behaviors ofpeople that use their properties has been the subject of a number of evaluations.

The civil law has been the primary tool used to make owners of privaterental property evict drug dealers or make physical changes to their property.Hope (1994) describes three case studies from St. Louis where police officersinfluenced the changing of ownership of drug houses. Calls for servicefrom blocks with the houses declined 54 percent to 94 percent relativeto nearby blocks, suggesting a decline in drug selling.

Most efforts to influence landlords threaten civil action, but do nottypically result in the transfer of property ownership or the seizure ofproperty. Nuisance abatement programs threaten court action to seize propertyunless owners take action to curtail drug dealing. Three evaluations ofnuisance abatement programs were found.

Lurigio and colleagues (1993) evaluated an abatement program run bythe State's Attorney Office in Cook County, Illinois. They compared theperceptions of residents living near 30 abated properties to the perceptionsof residents on nearby untreated blocks. They found no difference in perceptions.If the abatement program did reduce drug dealing or related crime, nearbyresidents did not notice it. The weakness of this design is that it doesnot have a true pre-treatment measures of crime, but only perceptions ofchange.

Green (1993; 1995; 1996) examined changes in drug arrests, police fieldcontacts, and citizen calls around 275 abated drug dealing sites in Oakland,California. Relative to citywide changes in these measures, Green founda 15 percent decline in arrests, a 38 percent decline in field contacts,and a 14 percent decrease in citizen calls.

Finally, Eck and Wartell (1996) report on the results of a randomizedcontrolled experiment using threatened property seizure in San Diego, California.No landlords were taken to court and no properties were seized. Instead,following police drug enforcement, owners of properties in one randomlyselected group received a letter from the police ("letter" group).Owners of properties in another randomly selected group met with a narcoticsdetective and a city codes inspector ("meeting" group). Ownersof properties in a third (control) group received no follow-up contactfrom the police or the city. Drug offenders who were lease holders weremore likely to be evicted from the properties in meeting group. Further,for the six month period following treatment, the properties in the meetinggroup had a significantly lower number of reported crimes. The letter groupalso had a decline in crimes, but it was not significantly different fromthe control group or the meeting group.

Three of the four studies report some reduction in crime or calls forservice at treated drug properties or the block around the properties.The three studies that reported the positive findings were more rigorousthan the single study showing no results. Thus we can be reasonably confidentthat holding owners responsible for drug dealing on their property mayreduce drug related crime.

Conclusions About Residences

Collectively, there is reason to be optimistic about the efficacy ofopportunity blocking tactics in residential settings. As a group, theseevaluations -- from the weakest to the strongest -- suggest that improvementsin crime reduction can be achieved. Nevertheless, it is difficult to beprecise about what works, at which types of residential sites, and againstwhich crimes. One set of tactics, however, does have a limited number ofrigorous evaluations. Nuisance abatement is a place-focused tactic that"works." With the evidence available we are relatively certainthat holding private landlords accountable for drug dealing on their propertyby threatening abatement reduces drug related crimes. A weaker body ofevidence suggests that reducing the ability of people to move freely aboutlarge public housing complexes can reduce crime.

Addressing repeat victimization deserves more attention in the UnitedStates but there is insufficient evidence to recommend that this tacticbe applied wholesale at this time. Nevertheless, research on repeat victimizationprevention in housing and other settings will be useful for public housingauthorities, police agencies, and private landlords. Finally, by that standardsused in this report, the evidence for target hardening is weak so it isof unknown effectiveness. Of particular concern is the lack of significancetests in target hardening evaluations that could provide evidence thatobserved crime reductions were not due to chance. More rigorous evaluationsneed to be conducted to improve our confidence in this tactic.

Table 7-2: RESIDENCES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Allatt 1984 3 target British 52% reduction hardening public relative to housing controls in burglary Anderson, 3 graded British 19% reduction in Chenery, response public burglary relative Pease depending on housing to control 1995a; number of (Huttersfield) 1995b prior burglaries Chatterton 2 cctv elderly 79% decline in & Frenz including housing burglary and 1994 dummy cameras complexes, attempt burglary Manchester, Great Britain Gabor 1981 3 property residential 75% increase in marking dwellings, burglary Canada Laycock 3 property public 40% reduction in 1985, 1991. marking housing, burglary Great Britain Tilly & 3 improving Birmingham 59% reduction in Webb. 1994 security of pubic housing burglary doors and , Great windows Britain 3 improved door Bradford 91% reduction in locks and public burglary removal of housing, prepayment Great Britain meters Forrester, 3 removal of Public 40% reduction in Chatterton, pay gas housing, burglary in one & Pease meters; Great Britain year; continued 1988; Cocoon drop over next Forrester, neighborhood three years Frenz, watch; O'Connell & security Pease survey and 1990. hardware Pease installation 1991. Tilly 1993a. Meredith & 2 crime watch apartment 82% drop in Paquette (and target building burglary, little 1992 hardening) drop in other crimes Popkin, et. 1 guards, Two high rise 40% to 64% drop in al. 1995a; design public drug dealing; 74% 1995b changes, housing to 88% drop in enforcement, buildings, shootings and identification Chicago fighting cards, and other changes Newman 1996 2 restricting Bronx public 54% drop in pedestrian housing reported crime. movement and 62% drop in other design burglary, robbery, changes & assault Poyner 1994 3 closing London public reported reduction walkways housing in purse snatches connecting buildings and installation of entry phone Eck & 5 nuisance private 59% drop relative Wartell abatement residential to controls in 1996 rental reported crime for property, San most stringent Diego CA intervention, 51% drop for less stringent intervention not significant Green 1993; 4 nuisance private 15% decline in 1995; 1996 abatement residential arrests, 38% properties decline in field with drug contacts, & 14% dealing in decline in calls Oakland, CA Hope 1994 3 closing or 3 addresses 54%, 67% and 95% selling of used for drug reduction in calls property dealing in for service St. Louis, MO Lurigio, 2 nuisance residential no difference et. al. abatement properties in between treated and 1993 Cook County, untreated blocks IL relative to drug dealing 

RETAIL STORES

Places that sell goods to the public are frequent crime sites. The theftof goods represents a large proportion of these crimes. Some of these theftsare committed by patrons and some by employees. In addition to thefts,robberies of store clerks and burglaries after store hours can also beproblems. In this section we examine all of these crime types. First wewill look at convenience store robberies. Much has been written on thistopic, but most of it describes correlational studies with very small samples,comparing stores with and without robberies. As we will see, the numberof evaluations of interventions is limited. We will then turn to burglariesand robberies in other retail settings. Third, we will examine credit cardfraud. The largest group of studies involves shoplifting prevention. Finally,we will look at thefts by employees.

Convenience Store Robberies

Although convenience stores have received considerable attention inthe crime prevention literature, robberies of these retail establishmentspeaked around 1980-81, declined through 1983 and remained stable for thenext 10 years at around 16,000 per year. Over the same period, the numberof such stores has increased and gas station robberies have trended upward(Bellamy 1996). Comparisons of convenience stores with and without robberieshave been carried out for over two decades. These studies attempted tofind store features that are associated with few or no robberies. The studiesgenerally suffer from three major scientific problems. First, they usuallyexamine a variety of store features using a small sample of stores. Sincethese features are often correlated with each other, it is difficult todetermine which features are related to robberies. Second, since the storefeatures and robberies are measured at about the same time, it is unclearif the features preceded the robberies (and could possibly have influencedthe chances of the crime) or whether the robberies cause store managersto change the store's features. Finally, most convenience stores have norobberies, but a few have many robberies. Crime prevention measures maywork in the few stores with repeated robberies but have no influence onthe other stores (Crow and Bull 1975). It is not surprising, therefore,that these studies can arrive at contradictory findings.

One of the most debated questions is whether two clerks reduces therisk of robberies. Hunter and Jeffrey (1992) cite a number of studies showingthat stores with fewer robberies are associated with two clerks being onduty. LaVigne (1991) provides evidence that the number of clerks is unrelatedto robberies. Another study, conducted by Robert Figlio, compared 230 conveniencestores with two or more clerks on duty at night, to 346 stores with onlyone clerk on duty, and examined a subsample of one-clerk stores beforeand after they shifted to two clerks. The evaluation found no impact onrobberies by the switch to two clerks, compared to similar stores thatdid not increase the number of clerks from one to two. However, for storeswith robberies prior to the switch, two clerks did reduce the chances ofa robbery (National Association of Convenience Stores 1991).

The Gainesville (Florida) Police Department evaluated a city ordinancerequiring two clerks to be on duty. The police department found that conveniencestore robberies declined immediately after the ordinance took place (Clifton1987). Wilson (1990) reviewed the initial evidence and found that a plausiblerival explanation for the decline in robberies was the arrest of activeoffenders responsible for a rash of convenience store robberies just beforethe ordinance took place. Although the short term reduction may have beendue to these arrests, robberies of these stores in Gainesville continuedto decline for seven years following the ordinance and the arrests of therepeat offenders (Bellamy 1996). The controversy surrounding this ordinance,and Florida-wide efforts to increase the number of clerks, may have sensitizedthe convenience store industry and the police to this problem. Thus, manyother changes could have created the long term reduction. Changes in stores'operations may also have been responsible for the reduction in robberies.Thus we cannot be certain the decline was due to the two clerk rule.

One of the first randomized experiments in crime prevention was undertakenover 20 years ago to determine if prevention measures in convenience storesreduced robberies. Crow and Bull (1975) matched 120 stores according toprevious robberies and other characteristics. These stores were randomlyassigned to either a control group or a prevention group. The type of preventionwas selected based on site visits, so it was not possible to determinewhat type of prevention had what effects. The treated stores with two ormore previous robberies had 30 percent fewer robberies after treatmentthan the untreated stores with two or more previous robberies.

In a later convenience store study, cameras and silent alarms did notappear to prevent robberies when 55 convenience stores in Columbus, Ohioand New Orleans, Louisiana receiving these devices were compared to 53stores in Dayton, Ohio and Baton Rouge, Louisiana not receiving them (Crowand Erickson 1984). In the treated stores signs announcing the equipmentwere posted. These changes were accompanied by publicity in the treatmentareas. No significant changes in robberies were found.

The National Association of Convenience Stores (1991) reported on twoother interventions evaluated by Robert Figlio. The installation of interactiveCCTV (allowing communication between the clerk and the personnel watchingthe TV monitor in a remote location) reduced robberies in 189 stores bya statistically significant 31 percent in the first year following theinstallations. By the second year, the reduction had shrunk to 15 percent,which was not statistically significant. No control stores were used inthe analysis. One chain of 81 stores installed color video monitors thatwere visible to patrons and staff. Robbery rates were reported to havedeclined by 53 percent a year after installation. Again, no control storeswere used.

The convenience store industry has conducted some of the most sophisticatedcrime prevention experiments available. These studies suggest that thereare two types of stores, those with few or no robberies where crime preventionefforts are unlikely to influence future robberies, and a fewer numberof stores with several robberies where prevention efforts may be more productive.

Burglary and Purse Snatching in Other Retail Places

Burrows and Speed (1996) report on an effort to curb "wire-cut"burglaries of electronics stores. Since alarm systems in these stores areconnected to a remote monitoring station, burglars cut the telephone linesbefore entering. Electronically monitoring the integrity of the phone linesappears to have reduced losses from these types of burglaries. Unfortunatelythe authors only show a graph of the data without reporting the figuresfor burglaries or losses. Trends in wire-cut burglaries were compared toother types of burglaries and indicated that the decline was unlikely tobe due to a general decline in burglaries, independent of the preventivetactic studied.

"Ram-raiding" involves crashing a vehicle (often stolen) intothe front of a retail establishment and then removing valuable products.The costs of the damage to the store are considerable and often exceedthe costs of the stolen merchandise (Jacques 1994). This is a problem inGreat Britain, but its extent in the United States is unknown. Jacques(1994) reports that the installation of metal shutters in six large retailestablishments cut burglary costs 53 percent (from an average of 20,892pounds sterling to 9613 pounds sterling). In one store, burglars shiftedto a roof entry thus providing evidence of limited displacement in burglarytactics. No control stores were examined.

Thefts from shoppers at retail places can also be a problem. In shoppingmarkets in one British city, women's purses were being taken from theirshopping bags. The aisles of the markets were widened to reduce the bumpingof patrons that facilitated the thefts. Poyner and Webb (1992) report thata comparison of reported thefts for the three years prior to the changesto the two years after, showed a 44 percent decline in these offenses.Simultaneous changes in nearby markets makes them unsuitable as controlplaces, so we have no evidence about background trends.

Credit Card Fraud

Three evaluations examined attempts to prevent credit card fraud atthe point of sales. All three involved staff training and increased attentionto customers. Two studies describe providing clerks with more informationabout potential offenders, either through liaison with law enforcementauthorities (Masuda 1993) or by providing computer-aided identificationof shoppers wishing to use credit cards to pay for purchases (Masuda 1996).Both evaluations compared pre-program losses to post-program losses, butdid not use control stores. Losses declined 82 to 90 percent.

A British experiment in lowering the limit for unauthorized credit cardpurchases along with improved information exchange about possible offendersmay have reduced fraud losses by 25 to 41 percent nationwide, dependingon the length of the pre-treatment period used (Webb 1996).

Although these studies did not use strong evaluation designs, they consistentlyreport that tightening restrictions on credit card use and use of informationabout people with a history of credit card fraud can reduce this crime.Such findings underscore the point that many losses by retailers are dueto choices about how to conduct their business. Challinger's (1996) evaluationof refund fraud reduction reinforces this point. Refund fraud involvesthe return of stolen goods for a refund. The store ends up paying for themerchandise twice, the first time at the wholesale price and the secondtime at the retail price. Challinger (1996) reports that requiring proofof purchase may reduce the losses from this form of theft. For confidentialityreasons, he does not report the amount of losses for stores involved inthe evaluation.

Shoplifting

Here we will look at several methods for preventing shoplifting. Twointerventions, electronic article surveillance and ink tags, have receivedmultiple evaluations. Electronic article surveillance (EAS) involves placingtags on merchandise that only clerks can remove at time of payment. Ifa clerk does not remove the tag and the shopper leaves the store, the tagcauses an alarm to sound. EAS technology improves employee surveillanceof goods. Ink tags deface the merchandise if it is removed from the storewithout paying. This destroys the value of the goods to thieves.

Five evaluations of EAS were reviewed and each reported reductions incrime events or shrinkage. All compared crime or shrinkage (unaccountedfor declines in inventory) before the installation of EAS to the same measuresafter, and all used a control store to measure background trends. The reductionin shrinkage varied from 32 percent (Bamfield 1994) to 80 percent (DiLonardo1996). Farrington and colleagues (1993) report even greater reductionsin shoplifting in the two stores they examined (76 to 93 percent). Furthermore,EAS was found to be more effective than security guards (no improvement)or store redesign (50 to 80 percent improvement) (Farrington et. al. 1993).Unfortunately, with one exception (Farrington et. al. 1993) significancetests were not reported so we cannot determine the probability that thereported reductions were due to chance.

Ink tags may also reduce shoplifting, but we have fewer studies andthey used weaker evaluation designs. DiLonardo and Clarke (1996) reporton two quasi-experiments involving ink tags. Both used repeated inventorycounts to measure inventory reduction before and after the installationof the tags. In the first study, 14 new stores were compared to the chain-wideaverage. Shrinkage was reduced 14 percent in the new stores. In the secondstudy, ink tags were installed in four stores, but no control stores wereused. Shrinkage declined by 47 percent. As we will see below, repeatedinventory counts have been linked to reduced employee theft, so we cannotbe certain that the changes reported in these two ink tag studies are dueto the ink tags or the method of measuring shrinkage.

The final shoplifting evaluation is a case study of a single store wherethe problem was minor thefts by elementary school children. A combinationof individual and collective rewards were offered the children for refrainingfrom stealing small items. The period before the program, program period,and a period after the program ended were compared. Shoplifting of targeteditems declined by 58 percent and profits increased 42 percent during theprogram period compared to the periods before and after the program.

Shoplifting appears to be controllable by the use of EAS technology,and possibly ink tags. If more evaluations had used significance testswe could have classified EAS as "works." In the absence of thisinformation EAS must be placed in the "do not know" category.Limited evaluations of other approaches suggest that there may be alternativeapproaches as well. The single study that examined the value of guardsfound that they were of no assistance in reducing shoplifting, but as Farringtonand colleagues (1993) point out, this may be due to an implementation failure.

Employee Theft

Masuda (1992) examined the effectiveness of increasing the frequencywith which articles at great risk of theft are counted. Since the increasedinventory counts were unknown to shoppers but were known to store employees,it is reasonable to assume that the 100 percent reduction in shrinkagehe found was due to the deterrence of employees. The 85 percent reductionin non-target item shrinkage may be attributable to a diffusion of benefitseffect. However, the absence of an uncontaminated control makes it difficultto determine if this reduction was an unexpected program effect or evidenceof declining shrinkage independent of the intervention.

Table 7-3: RETAIL STORES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Crow & Bull 5 variety convenience stores with 2 prior 1975 stores robberies had 30% fewer robberies relative to controls Crow & 4 surveillance convenience No significant Erickson cameras stores change in robberies 1984 National 4 two clerks convenience 15% reduction in Association stores robberies over 2 of year period in high Convenience robbery stores Stores 1991 2 cctv convenience 15% reduction in stores robberies over 2 year period 2 video convenience 53% reduction in monitors for stores robberies patrons and staff Poyner & 2 widening public market 44% reduction in Webb 1992 aisles in in thefts from purses open market Birmingham, Great Britain Burrows & 3 electronic electronic noticeable decline Speed 1996 monitoring of retail stores in wire cut phone lines burglaries but amount difficult to determine from chart provided Jacques 2 metal electronic 53% drop in losses 1994 shutters retail stores due to ram-raiding burglaries Masuda 1993 2 profiling retail store 82% decline in offenders, chain credit card fraud training, losses liaison with law enforcement Masuda 1996 3 computer retail stores 90% reduction in aided credit card fraud positive losses identification at point of sales Webb 1996 2 lowering point of 25% to 41% decrease limits for sales in in credit card use of credit retail fraud losses cards, establish-ment nationwide improved in Great information Britain exchange, and other tactics Challinger 3 requiring retail stores decline in losses 1996 proof of and reports. purchase for refund, and related procedures to prevent refund fraud Bamfield 3 EAS to retail stores 32% reduction in 1994 prevent shrinkage shoplifting DiLonardo 3 EAS to retail stores 47% decline in 1996 prevent shrinkage over 5 shoplifting years 3 EAS to retail stores 80% decrease when prevent installed. When shoplifting reinstalled over 80% decline repeated. 3 EAS to retail stores 52% decrease in prevent shrinkage shoplifting DiLonardo & 3 ink tags to retail stores 14% reduction in Clarke prevent inventory shrinkage 1996 shoplifting 3 ink tags retail stores 47% decline in replace EAS inventory shrinkage to prevent shoplifting Farrington 3 uniformed retail stores No measurable et.al. guards in Great impact on 1993 Britain shoplifting 3 store retail stores 58% drop in redesign in Great shoplifting at one Britain store and 80% decline in another in target items stolen. 3 tagging retail stores 76% reduction in in Great shoplifting at one Britain store and 93% reduction in another in target items stolen. McNees, 3 awards for single 58% decline in Schnelle, compliance to convenience shoplifting of Kirchner, & prevent store targeted items. Thomas 1980 shoplifting Estimated increase by elementary in profits of 42% school during program children Masuda 1992 2 increased retail stores Elimination of frequency of shrinkage for inventory targeted products, counts to 85% decline in prevent shrinkage of employee non-targeted theft products 

BANKS AND MONEY HANDLING PLACES

The robbery of banks and other places that provide money handling servicesis a serious problem in many countries. In this section we will examineevaluations of security measures in U. S. and Swiss banks, British postoffices, and Australian betting shops.

Guards may prevent bank robberies. A study of 236 banks in the Philadelphiaarea found one less robbery per year at banks with guards compared to bankswithout them, controlling for the surrounding area, police response time,proximity to major streets, and other prevention measures used. Screensprotecting tellers and cameras were not associated with fewer robberies(Hannan 1982). Since these tactics are often found together, the evidenceabout the effectiveness of any specific measure is weak. Though this isa correlational study, the evaluator made special efforts to control fortemporal order. Information about security measures came from surveys administeredby the Federal Reserve and only crimes reported after the survey were usedin the analysis. Because we can be sure that the interventions were installedprior to the crimes, this evaluation was given a scientific methods scoreof 2.

Two other studies provide better evidence that screens protect clerksfrom robberies. A study of over 300 Swiss banks found that banks with screenshad a 52 percent lower robbery rate than banks without them (Grandjean1990). Ekblom (1987, 1988) examined the installation of bullet proof barriersto protect post office clerks. He estimated that the barriers reduced robberiesfrom 55 percent to 65 percent, net of changes in control group robberies.Both studies found evidence for displacement, but even accounting for displacement,robberies declined substantially.

Clarke and McGrath (1990) examined the effects of time-lock cash boxesand safes on Australian betting shop robberies. Relative to control places,robberies may have been reduced by 52 to 139 percent. The results may behighly unstable given that there were three interventions throughout a10-year period.

An examination of a drop in the number of bank robberies in Victoria,Australia asserts that this was due to the installation of screens protectingclerks, guards, cameras, and other security devices (Clarke, Field, andMcGrath 1991). After increasing from 1979 through 1987, the number of bankrobberies dropped to levels similar to those found in earlier years. Similarpatterns of growth and rapid decline were found in bank robberies in anadjacent state and in robberies of other businesses. It is unclear whetherthe protective measures were installed only in Victoria's banks and whenthey were installed.

We do not know what works to prevent crimes at banks and other moneyhandling places because the scientific methods scores for the interventionsare either below 3 or significance tests were not reported. These evaluationssuggest the possibility that guards, bullet proof screens, and secure cashcontainers might reduce crimes, but more rigorous evaluations are neededto draw firm conclusions.

Table 7-4: BANKS AND MONEY HANDLING PLACES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Clarke & 3 time-lock Betting shops robberies declined McGrath. cash boxes in Australia 52% to 139% 1990 and safes Clarke, 1 security Banks in drop in bank Field & screens, and Victoria, robberies McGrath other Australia 1991 measures Ekblom 3 counter Post offices 55% to 65% 1987 1988 screen in London reduction in barriers in robberies front of clerks Grandjean 2 bulletproof Banks in 52% reduction in 1990 screens for Switzerland robberies tellers Hannan 2 security Banks in reduction of one 1982 guards, Philadelphia, robbery per year screens and PA area for most robbery cameras prone banks due to guards 

BARS, TAVERNS AND DRINKING ESTABLISHMENTS

There is a consistent research literature that points to a relationshipbetween the presence of bars and crime in the surrounding area (Roncekand Bell 1981; Roncek and Pravatiner 1989; Roncek and Meier 1991; Blockand Block 1995). Despite this reputation, most bars may be relatively crimefree while a few may be hotspots of crime (Engstad 1975; Sherman, Schmidt,and Velke 1992; Homel and Clark 1994). The behavior of bartenders and bouncersmay be contribute to violence in these places (Homel and Clark 1994) andchanges in bar management practices (from server training and changes inlegal liability of bartenders) may reduce assaults (Putnam et. al. 1993),drunk driving (Saltz 1987), and traffic accidents (Wagenaar and Holder1991).

Two Australian programs to reduce violence created agreements amongpub managers to improve the training of bouncers, reduce crowds of youths,and improve relationships with police, along with other tactics (Homelet. al. 1997). In one evaluation observers reported a 53 percent reductionin assaults per 100 hours of observation in the first year of the program.The prevention effects decayed over time. Three years after implementationthe reduction had declined to 15 percent. No control pubs were observed(Homel et. al. 1997). The other evaluation examined serious assaults atdowntown pubs for the year before and four years after the management accord,and compared these changes to the same period for six other cities in thesame state. Serious assaults declined 40.5 percent in the target city butincreased 14.3 percent in the control cities (Felson, et. al. 1997).

The consistent results from Australia and the United States summarizedin Table 7-5, suggest that changing the management of drinking places isa promising method for prevention of drinking-related offenses.

Table 7-5: BARS AND TAVERNS STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Felson, 3 code of bars and 60% decline in et. al. practice for drinking serious assaults, 1997 pubs establishments net controls in Geelong Australia Homel, et. 2 training for bars and 53% decline in al. 1997 bouncers, drinking assaults/100 hours code of establishments of observation 1st practice in year after Australian implemented, but town only 15% decline compared to 3 years after Putnam, 3 training of alcohol sales decline in alcohol Rockett, & alcohol outlets in related assaults & Campbell servers and one Rhode vehicle crash 1993 police Island injuries, relative enforcement community to control communities Saltz, 3 changing Navy enlisted Over 50% reduction 1987 serving club in in driving when policies and California drunk training 

PUBLIC TRANSPORTATION

Two types of public transportation have been the subject of evaluations:buses and subways. Evaluations investigated prevention measures directedat four types of crime: crimes against riders; attacks on staff; fare evasion;and vandalism. The types of interventions have been quite varied, rangingfrom complete system design to volunteer citizen patrols.

Incivilities and Crimes Against the Public

The Washington, D.C. Metro System has been singled out in crime preventionliterature as having been designed to prevent crime (LaVigne 1997) andis sometimes contrasted with the New York City subway system which gaineda reputation for crime in the 1970's (Sloan-Howitt and Kelling 1990; Dwyer1991). "Designing in" crime prevention may be effective, butit is difficult to determine if a design is effective. LaVigne (1997) comparedthe Washington, D.C. Metro to three other urban rail transit systems andfound that it had less crime than the other systems. She also comparedsubway station crime to crime in the areas above-ground. If the systemhad no influence on crime then the above-ground crime levels and stationcrime levels should be correlated. If the system design prevented crime,then there should be no relationship between station and above ground crime.LaVigne (1997) found that, except for assaults, ground level and stationcrime were not correlated. Although this is not a strong research design,it is the best evidence available that system design influences crime patterns.

To improve passenger confidence in the safety of the New York subwaysystem, an intensive cleanup program was undertaken to remove graffitifrom all train cars and stations. Rapid cleanup would deprive vandals ofthe benefit of seeing their graffiti (Sloan-Howitt and Kelling 1990). Bytreating the physical appearance of the system, it was hoped that thiswould make the public feel safe and bring more people into the system.More riders would increase the number of people watching out for each other,and this could drive down crime. This chain of events is expected accordingto the "broken-windows" hypothesis (Wilson and Kelling 1982).Sloan-Howitt and Kelling (1990) show that graffiti was virtually eliminated,and despite increased police attention to graffiti, arrests for this offensealso declined.

A similar effort was carried out by the Victoria (Australia) transitsystem which includes trains, trams and buses. The Victoria program involvedrapid repair and cleaning of vandalized equipment, along with stepped uppolice enforcement. Carr and Spring (1993) show that train availabilityincreased 45 percent and reported crimes against persons declined 42 percent.

Another comprehensive program to clean up a problematic transit facilityhas been described by Felson and colleagues (1997). The title of theirpaper, "Redesigning Hell," suggests the state of disrepair intowhich the New York Port Authority Bus Terminal had fallen. Sixty-threeinterventions were made at the terminal, at about the same time. Theseincluded closing off spaces, improving shopping, cleaning, increased enforcement,and other measures to remove situations that facilitated offending or increasethe number of patrons and their ability to watch each other. Although robberiesand assaults declined in the station, they also declined in the surroundingarea. Outside crime control efforts or diffusion of crime control benefitsto the surrounding area may account for these parallel trends. Annual surveysof patrons that began with the cleanup in 1991 show declines in incivilitiesand disorder.

Vandalism against buses is another problem in transit systems. Poyner(1988) describes how the installation of CCTV on a portion of a bus fleetwas followed by reduced vandalism throughout the fleet. There was alsoa public information campaign directed at the group of people most likelyto be responsible for the damage, school children. Poyner (1988) attributesthe diffusion of benefits from the targeted buses to the entire set ofbuses, to offenders' confusion over which buses had the CCTV. Unfortunately,this evaluation only describes trends in vandalism after CCTV was installed.

Kenney (1986) evaluated the effectiveness of Guardian Angel patrolsat stations by comparing crime changes to control stations without thesepatrols. He found that these citizen patrols had no discernible impacton crime in the patrolled stations. This may be because the base ratesof crime in the stations were too low to detect an effect (Kenney 1986).

Webb and Laycock (1992) also found no evidence that the Guardian Angelsreduced crime in the London Underground. They did find that the installationof CCTV in London Underground stations reduced robberies 11 to 28 percent,relative to control stations without CCTV. Twenty-two months of data beforeCCTV installation and 26 months after installation at selected stationswere compared.

On the whole, we have limited information about how to prevent incivilitiesand crime against transit. In part this is due to the difficulty in assessingsystem-wide designs and comprehensive changes. Selecting a control systemand disentangling the effects of multiple interventions is very difficult.Rapid cleanup and repair to deprive offenders of the pleasure of seeingtheir graffiti appears to be effective, but the evidence to date is weak.

Attacks On Bus Drivers

The two evaluations of attacks on bus drivers provide evidence thatthese crimes can be reduced. The rise in robberies of bus drivers in thelate 1960's and early 1970's prompted New York City officials, along withtransportation officials in other U.S. cities, to remove accessible cashthat was the target of the robbers. They required passengers to give exactfares and prohibited bus drivers from giving change. Fares were put insecure boxes. Chaiken, Lawless, and Stevenson (1974) reported a 90 percentreduction in bus driver robberies following these changes. The StanfordResearch Institute (1970) reported similar results in its review of theeffect of exact fare systems in 18 other cities (Clarke 1992, page 216).

If the target of the attack cannot be removed, then maybe it can beprotected. A bus company in northern England used two approaches to protectits drivers from assaults by riders (Poyner and Warne 1988). The firstwas to simplify the fare system so it would be less aggravating. They alsoinstalled protective screens around bus drivers. Assaults on drivers declined90 percent following the installation of screens. Assaults on all employeesfell during this period, but not as much as it fell for drivers (37 percent).

Fare Evasion

Transit systems suffer from people who try to enter without paying thecorrect fare. Fare evasion can simply mean jumping gates or moving throughentries without paying, or it can involve the use of slugs in gates orticket machines. Three evaluations examined the redesign of gates or ticketmachines to curtail fare evasion. All three report evidence suggestingdeclines in this form of theft. Clarke (1993) reports an increase in ticketsales of 10 percent, relative to control stations where new automatic gateswere not installed. Clarke, Cody and Matarajan (1991) show that one formof slug use was totally eliminated by modifying ticket machines so theywould not accept a type of coin for which a slug could be substituted.This was a system-wide change so no control stations were available. Finally,Weidner (1997) gives results of the effect on fare evaders of the installationof new gates in the New York City subway. While arrests declined in thetarget station, they increased in adjacent control stations. Whether thiswas due to changes in police enforcement, displacement, or background trendscannot be determined from the evidence provided.

Two evaluations examined personnel changes to reduce fare evasion. Increasesin ticket takers at a Canadian ferry terminal may have reduced fare evasionby 20 percent, although there were no control sites to assess backgroundtrends (DesChamps, Brantingham, and Brantingham 1992). A Dutch effort toreduce fare evasion in three cities decreased fare dodging by 18 to 78percent. Authorities recruited over 1100 unemployed young people to monitorticket use on the buses, trains and trams in the three cities. This report(vanAndel 1986) claims that there was also a 60 percent decline in assaulton and harassment of patrons. Like the Canadian study, there was no controlgroup.

Conclusions about Transportation System Prevention

Although there are several evaluations of crime prevention in transportationsettings, we know relatively little about the effectiveness of these interventions.This is in part due to the variety of crime types that are applicable totransportation systems. It is also due to the number of settings (buses,trains, and stations) within the system, as well as the variety of victims(patrons, staff, and facilities). Thus a large number of studies are neededto learn what works to prevent crime in transit systems. However, thereare methodological complications that make learning about crime preventioneffectiveness quite difficult. Many of the systems are large and thereare few, if any, plausible control settings are available to measure backgroundtrends. Places within systems are linked, so internal changes to part ofa system can influence crime in other parts of the system. If untreatedparts of the system are used as controls, diffusion of benefits or displacementeffects can confound the findings. We cannot, therefore, identify, withreasonable certainty, any specific tactic against specific crimes, thatcan be said to "work" across similar settings in other cities.

Table 7-6: PUBLIC TRANSPORTATION FACILITIES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE LaVigne 1 subway system Washington, system design may 1997 design DC metro prevent crime subway Carr & 2 improved public 45% improvement in Spring cleaning and transportation train availability. 1993 vandalism system, 42% reduction in repair; Victoria, crimes against patrolling Australia persons Felson et. 3 63 different Port reduction in al. 1997 tactics Authority Bus robberies & imple-mented Terminal, New assaults but not about the York City compared to same time surrounding area; reductions in incivilities Kenney 3 Guardian subways no detectable 1986 Angels impact on crime Poyner 1 cctv buses steady decline in 1988 vandalism Webb & 3 cctv (and stations on 11% to 28% Laycock other London reduction in 1992 tactics) Underground robberies Chaiken, 2 exact fare buses in New 90% decline in Lawless, & requirement York City robberies of bus Stevenson drivers 1974 Poyner & 2 protective buses in 90% reduction in Warne 1988 screens for Cleveland, assaults on drivers drivers Great Britain Clarke 3 automatic London 10% increase in 1993 gates to Underground ticket sales prevent fare evasion Clarke, 2 modification London elimination of Cody & of ticket Underground problem of slug use Matarajan vending within 4 months of 1991 machines modification DesChamps, 2 increase in ferry 20% reduction in Brantingham rush hour terminal fare evasion rate , & attendants to Brantingham check 1992 tickets, training in fraud detection vanAndel 2 recruiting buses, metro 18% to 72% decrease 1989 over 1100 trains and in fare dodging young trams in 3 depending on city unemployed large cities and mode of people as in the transport, 60% public Netherlands decline in attack transit or harassment monitors victimizations Weidner 3 installation stations on fare evasions 1997 of new fare New York City declined in target gates subway station 

PARKING LOTS AND GARAGES

Evaluations of crime prevention in parking lots and garages examinedchanges in people who watch cars. These people were often security guards,although one evaluation looked at placing a taxi business near the entranceto a parking garage to increase informal guardianship (Poyner 1991). Anotherset of interventions used close-circuit television to centralize watching.

Guards and Security Attendants

Four evaluations are available reporting on the effectiveness of addingsecurity guards to parking lots. Four showed reductions in car-relatedcrimes (Barclay et. al. 1996; Laycock and Austin 1992; Poyner 1994; andPoyner 1991) and one found no improvement (Hesseling 1995). Although thesestudies suggest auto thefts and thefts from automobiles might be preventedby increasing people who watch lots, there are two important caveats. Poyner(1991) notes that parking lot strategies that control access may curb theftsof vehicles, but may be ineffective at controlling thefts fromvehicles. The failure of Hesseling (1995) to find a reduction in theftsfrom vehicles may be due to the way the guards were deployed. Thus, whatthe guards do may be as important as their deployment. Second, none ofthese studies examined personal violence against people using parking facilities.In conclusion, because of the mixed results of the evaluations, we do notknow if guards or security attendants prevent crimes in parking lots.

Closed-Circuit Television

There are seven evaluations from Great Britain of the effects of CCTVon vehicle crimes (thefts of vehicles, thefts from vehicles, and damageto vehicles), but no evaluations of its effect on other crimes in parkingfacilities (Poyner 1992; Tilly 1993c). The weakest of the evaluations foundno effect (Coventry lots, in Tilly 1993c). The other six evaluations foundvarying levels of decline in vehicle crimes. In the CCTV parking lots evaluated,thefts from vehicles declined 46 to 94 percent, and thefts ofvehicles dropped 18 to 89 percent, depending on the evaluation. We do notknow if these results can be replicated in the United States. There isno empirical basis for recommending CCTV to prevent parking lot violence.The results suggest that CCTV should be tested in high vehicle crime parkinglots within the United States. Because of the lack of significance testswe must classify CCTV in parking facilities as having "unknown"prevention effectiveness.

Conclusions About Parking Facilities

Evaluations in parking lots and garages outside the United States consistentlysupport the hypotheses that guards and CCTV reduce vehicle-related propertycrime. Though several CCTV studies had scientific methods scores of 3,they lacked of significance tests. Therefore, CCTV's effectiveness in parkinglots is "unknown." These studies do not report on violent crimesin parking lots, including robberies and car-jacking. The highly crime-specificnature of intervention effectiveness suggests that we must be careful drawinginferences about the effectiveness of interventions to places and settingwhere they have not been tested.

Table 7-7: PARKING FACILITIES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Barclay 3 security commuter 53% reduction in et. al. guards on parking lot car thefts/ month 1996 bikes Hesseling 3 guards parking area 2% increase in 1995 in Rotterdam thefts from automobiles relative to control Laycock & 3 security parking area 52% to 60% in auto Austin attendant theft reduction 1992 Poyner 2 guard parking area Reduction in auto 1994 thefts. Amount cannot be estimated Poyner 3 restricting parking 29% increase in 1991 foot access, garage thefts from improved vehicles, 35% lighting, reduction in thefts increased of vehicles guardianship 3 cctv parking lots 71% & 94% reduction in thefts from cars Tilly 3 cctv parking lots, 75% reduction in 1993c Hartlepool, theft of autos, 60% Great Britain reduction in theft from autos 2 cctv one parking 45% reduction in lot, Hull, damage to autos, Great Britain 89% reduction in theft of autos, and 76% reduction in theft from autos 2 cctv one parking 75% reduction in lot, auto crimes Lewisham, Great Britain 3 cctv one parking 73% to 78% lot, reduction in theft Bradford, from autos, 49% to Great Britain 75% reduction in thefts of autos 2 cctv one parking 18% reduction in lot, thefts of autos, Wolverhampton, 46% reduction in Great thefts from autos Britain 2 cctv 5 parking no discernible lots, pattern in auto Coventry, crimes Great Britain 

AIRPORTS

Aircraft hijacking by armed passengers has been a problem since WorldWar I. Wilkinson (1977) has documented the worldwide trends in this problem.From 1948 (when records were first kept) through 1957 there were 15 attemptsworldwide and none involved aircraft originating in the United States.In the next decade there were 48 hijackings worldwide (23 of them NorthAmerican originating flights). In 1968, the number of world-wide aircrafthijackings began an explosive climb. There were 38 that year, and 82 thenext. In response, policy makers implemented a number of strategies, includingtreaties to ensure the return of hijackers and aircraft. By 1973, hijackingattempts had dropped to 22 worldwide and 2 in the United States (Wilkinson1977).

Since several interventions were put into place over a short time periodduring the early 1970s, it is difficult to determine which tactics madethe greatest contribution to the decline. Sky marshals (armed nonuniformedsecurity guards) were assigned to selected flights beginning in 1970. Tothwart parachuting from aircraft, modifications were made to the rear doorsof Boeing 727's and DC 9's to prevent them from being opened in flight(Landes 1978). In early 1973, the U.S. and Cuba signed a treaty that requiredeach country to extradite or punish hijackers (Landes 1978).

Landes (1978) attempted to determine the effectiveness of sky marshalsand passenger screening. He used a time series analysis of 64 quarter yearsand 143 incidents. He also controlled for hijacking of aircraft originatingfrom foreign airports to remove world-wide trends in skyjacking and attemptedto remove the effects of the Cuba treaty. He provides evidence for an 82percent decline in U.S. hijacking due to the combined effects of the Cubatreaty, sky marshals, and passenger screening. He then estimated the contributionof the three policies: screening was the cause of a decline of 45 percent,sky marshals created a 28 percent decline, and the remainder (9 percent)was probably attributed to the Cuba treaty.

Two other studies, using annual data for different time periods andweaker evaluation designs, also found large declines in aircraft hijackingin the United States following passenger baggage screening (Wilkinson 1977;Easteal and Wilson 1991). These studies did not attempt to estimate theeffects of different hijacking programs.

The variation in aircraft hijacking from year to year and the virtuallysimultaneous implementation of multiple prevention methods at airportsaround the world make it difficult to come to definitive conclusions regardingany particular intervention. Nevertheless, the weight of the evidence supportsthe effectiveness of passenger screening.

These findings are important. First, they demonstrate the potentialutility of opportunity blocking against highly determined offenders. Second,they illustrate some of the difficulties of evaluating place-focused prevention(multiple simultaneous interventions, detecting reductions in rare events,and the difficulty of finding control places). And third, they may haveimplications for other places.

What do these findings about the use of metal detectors to screen forweapons at airports tell us about their deployment at other places? Thesedevices have been used to enhance the security of court buildings, schools,government offices, and public housing. Are they effective? From an empiricalperspective, we can only say we do not know. Evaluations are scant andweak. A New York City study of the use of metal detectors found that weaponcarrying in schools with metal detectors (n=19) was lower than in schoolswithout the devices (n=96), but there were no differences in assaults withinor outside these schools (Centers for Disease Control and Prevention 1993).This evaluation has a scientific methods score of four, and although therewas a decline in risk-factors for violence, there was no significant declinein violence. In the residential places section we noted an evaluation ofa multi-tactic intervention in a particularly troubled set of public housingbuildings (Popkin, et. al. 1996). Metal detectors were a part of this program,but it is impossible to determine what, if any, influence they had becauseso many other things were implemented at the same time. We cannot, therefore,be confident about the transferability of this tactic to other, very differentsettings.

Table 7-8: AIRPORTS  STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Easteal & 2 passenger US Airports 64% reduction in Wilson. screening and hijacking of 1991 with metal originating passenger aircraft detectors flights Landes 3 passenger US Airports 45% reduction in 1978 screening and hijacking of with metal originating passenger aircraft detectors flights 3 sky marshals 28% reduction in hijacking of passenger aircraft Wilkinson 3 passenger airports 41% reduction in 1977 screening hijacking of with metal passenger aircraft detectors in US, 3% drop world-wide 

OPEN PUBLIC SPACES

The places considered in this section are open spaces in cities, includingstreet corners and segments. Four types of interventions will be examined.The first is the control of problem offenders. The second is improved lighting.The fourth is the use of closed-circuit television (CCTV). Finally, weexamine street closures and rerouting.

Controlling Problem Offenders

Two efforts to control public drinking as a means to reduce assaultsand incivilities in downtown areas provide evidence that controlling problemoffenders may be effective. Ramsay (1990; 1991) reports on the banningof public drinking in one English town. Comparing the year before and theyear after the ban (with no control group) he found no changes in assaults,but surveys of people using the area suggest that there may have been areduction in incivilities. A Swedish effort to reduce disorder at an annualfestival reported a decline in drunkenness and disorderly conduct arrestsfollowing the prohibition of public drinking, banning high risk offenders,and the closing of a popular camping site (Bjor, Knutsson and Kuhlhorn1992). This study compared arrests at the previous year's festival to arrestsat the festival with the restrictions, without control area comparisons.

Lighting

Lighting campaigns seek to enhance the ability of people to provideprotection for each other. In 1979, the predecessor agency of NIJ, theNational Institute of Law Enforcement and Criminal Justice, reported ona review of 60 lighting evaluations. The authors of this review concluded:

"Is street lighting an effective approach in the reduction anddeterrence of crime? The answer is inconclusive. The paucity of reliableand uniform data and the inadequacy of available evaluation studies precludea definitive statement regarding the relationship between street lightingand crime." (Tien, et. al. 1979, page 93, emphasis in the original)

Almost twenty years later, we know little more about the effectivenessof lighting.

In the 1980's, a borough in London upgraded all of its street lighting.Atkins, Husain and Storey (1991) compared reported crimes the year beforethe relighting to the year following for 39 sections of the borough. Nocontrol areas were used, so background trends in crime cannot be assessed.No systematic changes in crime were detected. Surveys of residents of onearea found no changes in perceptions of security.

A Scottish study of relighting in a Glasgow neighborhood and a smalltown near Glasgow found that there was a short term reduction in victimizationsthat varied from 32 percent to 68 percent, depending on how victimizationwas measured (respondent victimizations, victimization of respondents'children, victimization of other family members, victimization of friends,or car victimization). Reported crime dropped 14 percent. The evaluatorscompared a three-month period prior to relighting to a three-month periodfollowing (Ditton and Nair 1994). No control group was used and the resultsfor the two neighborhoods were combined.

Finally, we need to consider three separate evaluations, with similardesigns, undertaken by Painter (1994). She examined lighting improvementson two separate street segments and a footpath, all located in "crimeprone" areas within London. Pedestrians were interviewed before andafter the lighting improvement. All interviews were conducted after darkand were completed within 6 weeks of the relighting. No interviews wereconducted in control areas. Substantial reductions in robberies, auto crimes,and threats were reported in two sites (86 percent, 79 percent). Thesecrimes were eliminated in the third site, but the number of crimes beforerelighting was small so this could have been the result of other factors.

Not much has changed since Tien and his colleagues (1979) gave theircritical assessment of the impact of lighting on crime. In part this isdue to the lack of research on lighting, particularly in the United States.However, the limited research on lighting continues to use weak designs(typically without control areas) which fail to substantially reduce ouruncertainty about the effect of lighting on crime. We may speculate thatlighting is effective in some places, ineffective in others, and counterproductive in still other circumstances. The problematic relationship betweenlighting and crime increases when one considers that offenders need lightingto detect potential targets and low-risk situations (Fleming and Burrows1986). Consider lighting at outside ATM machines, for example. An ATM usermight feel safer when the ATM and its immediate surrounding area are welllit. However, this same lighting makes the patron more visible to passingoffenders. Who the lighting serves is unclear.

Closed-Circuit Television

Closed-circuit television (CCTV) enhances the ability of a designatedguardian to watch people in an area and to call for police interventionif potential trouble is detected. This is supposed to increase the risksof offending, but only if the CCTV surveillance is well known to the peoplewho use the area. This project was unable to locate any published scientificevaluations of the use of CCTV in urban areas of the United States.

Three CCTV evaluations have been reported in Great Britain (Brown 1995).As deployed, a set of video cameras are posted in center city areas andmonitored at a central station. The cameras cover many, but not all locationsin the target area. Finding locations with clear unobstructed views, yearround, can be difficult. CCTV cameras were installed around the town centerof Newcastle-upon-Tyne in late 1992 and early 1993. Using a time seriesof 23 months prior to the installation of cameras, four months during,and 14 months after, and comparing CCTV covered areas to uncovered areasin the same period, Brown (1995) found that burglaries declined by 18 percent,auto thefts dropped 9 percent, thefts from autos went down 11 percent,and other thefts declined 7 percent. No effect was found for robberies.

Brown (1995) used a similar design to assess the impact of CCTV in Birmingham.He compared reported crime 12 months before, two months during, and 30months after installation to control areas. Unfortunately, no figures wereprovided with the reported charts, but visual inspection of the time-seriescharts provided suggests reductions in robbery, burglary, and thefts. Similarresults were reported for another town center in Great Britain, King'sLynn. Four quarters of reported crime before installation were comparedto seven quarters after. A control area was used. Again, the data was notgiven, but visual inspection of the charts suggests reductions in burglary,assaults, thefts from vehicles, and thefts of vehicles. Significance testswere not reported in any of these case studies.

The effectiveness of CCTV in open spaces is unknown due to the lackof significance tests. Given recent interest in the use of CCTV in theUnited States, this tactic should be given a high priority for rigorousevaluations. Absent evaluation results from installations in the UnitedStates, the level of uncertainty about CCTV effectiveness is too high toadvocate its use except to test its effectiveness.

Street Closures

Research has suggested that areas with easy access have more crime thanareas with street layouts that restrict access (White 1990; Beavon, Brantinghamand Brantingham 1994). Oscar Newman (1982) reported on crime and its associationwith privately owned streets with limited access. He compared these streetsin a St. Louis neighborhood to nearby publicly owned, free access streetsand found that the private streets had less crime. In this section we willexamine five evaluations that support the hypothesis that closing and reroutingautomobile traffic can reduce crime.

In 1986 the citizens of Miami Shores, Florida (just outside Miami, inDade County) voted to increase taxes to fund closing off 67 streets (Atlasand LeBlanc 1994). The closings took place between July 1988 and March1991. The evaluation compared changes in reported crime within the townto the changes in the same crimes in the surrounding county and Miami.Mean 1986 and 1987 crimes (before installation) were compared to the meannumber of reported crimes in 1991 and 1992 (Atlas and LeBlanc 1994). Therewere no significant changes in reported robberies and aggravated assaultswithin Miami Shores compared to the two control jurisdictions. Relativeto changes in Dade County, reported burglaries significantly declined atleast 8 percent. Larcenies and auto theft in Miami Shores also declinedsignificantly, relative to changes in Miami and Dade County.

Newman (1996) reports the results of a street closure program in a Dayton,Ohio neighborhood. The Five Oaks neighborhood is a half-mile square areacontaining 2,000 homes on a grid street layout. Streets were closed offso that the area was subdivided into small areas and so one could not drivedirectly through the area. Newman (1996) summarized the City of Daytonevaluation results. Police-reported crime statistics showed that crimein the city rose one percent, but that total crime in the target neighborhooddeclined 26 percent, and violent crime declined 50 percent. Significancetests were not reported. Citizen surveys reported that over half of theresidents felt crime had declined. Newman also reports that housing valuesincreased after having declined prior to the street closures.

Two efforts to curb prostitution activity in London neighborhoods usedroad closures and rerouting coupled with increased police enforcement.In the Finsbury Park area police had steadily increased enforcement fortwo years prior to changes in the street closures. However, with the changesin the streets, "Soliciting and curb-crawling virtually disappearedand the area was transformed from a noisy and hazardous 'red-light' districtinto a relatively tranquil residential area." (Matthews 1992, page94). Reported crime declined 50 percent for the 12-month period after thestreet closures compared to the previous 12 months. Observations of thearea suggest that most of the prostitutes left the area but did not displaceto adjacent neighborhoods (Matthews 1992).

In the Streatham neighborhood of London, street closures were also usedin conjunction with increased police enforcement. Matthews (1993) reportsa decline in traffic flow along key streets. Although police enforcementwas maintained, arrests of "curb-crawlers" seeking sexual servicesdeclined by two-thirds (comparing the first quarter of 1990, after theprogram, to the first quarter of 1988, before the program began). Interviewsof residents suggests a decline in noticeable prostitution activity, althoughsome of this activity may have shifted to the periphery of that area.

The final evaluation of street closures was a retrospective analysisof the Los Angeles Police Department's Operation Cul-De-Sac. In 1990 theLos Angeles Police Department installed traffic barriers on 14 streetsin a South Central Los Angeles neighborhood with a high level of drug activity,shootings and homicides. Much of the violence was created by disputes overdrug sales locations by local gang members. The barriers were designedto make the driveup purchase of drugs more difficult and prevent drive-byshootings. This effort was part of a larger law enforcement effort to suppressthese crimes. Two years following the installation of the barriers, thebarriers were abandoned and then removed as the police became embroiledin the controversy surrounding the Rodney King beating.

The evaluation of the Los Angeles Police Department project comparedreported crimes in the neighborhood for four quarters before the barrierswere installed, the eight quarters while they were being maintained, and16 quarters after the program was abandoned (Lasley 1996). Reported crimefor the four adjacent areas was also examined. If one uses the surroundingbeats as control areas, the net effect of the installation of the barriers(before, compared to during) was that homicides decreased 65 percent. Infact, during the two years when the barriers were installed there was onlya single killing in the target area. Once the barriers were no longer maintainedand were removed (comparing the installed period, to the after period)homicides rose 800 percent, relative to the surrounding area killings.Total violent crimes (homicide, rape, street robbery, aggravated assaultand purse snatching) declined from the pre-program period to the two yearsduring the program (8 percent for the first year and 37 percent for thesecond year) and then rose again after the program fell into disuse. Atthe same time the surrounding areas remained relatively stable. Lasleyattributes most of the decline in violent crime to changes in aggravatedassaults. Significance tests were not reported for any of these comparisons.

Closing streets makes offenders' escapes more problematic. In the caseof prostitution cruising and drive-by shootings, the offenders are likelyto follow a circular driving pattern in their search for targets. By makingcircular driving patterns more difficult and increasing the chances offenderswill find themselves at the end of a dead end street, criminal behaviormay be thwarted.

The street closure evaluations used moderately strong designs and theirconclusions are consistent with theory and prior research. This gives usconfidence that this approach to curbing crime should be classified as"promising." In at least three of the programs (the two Londonprostitution cases and the Los Angeles drive-by shooting case), the streetclosures were undertaken along with police crackdowns. Matthews (1992)hypothesizes that street closures and enforcement may be more effectivewhen used together than when used separately and enforcement should beused prior to street changes. This opportunity-blocking tactic for controllingcrime in open urban areas deserves more attention by, particularly sinceit might reduce violence under some circumstances.

Conclusions for Open Urban Places

Four types of tactics were considered in this section. There is limitedevidence that controlling offenders, particularly public drinking, mightbe useful. However, the evaluations are small in number and weak in design,leaving its effectiveness unknown.

Lighting has received considerable attention. Yet, evaluation designsare weak and the results are mixed. We can have very little confidencethat improved lighting prevents crime, particularly since we do not knowif offenders use lighting to their advantage. In the absence of bettertheories about when and where lighting can be effective, and rigorous evaluationsof plausible lighting interventions, we cannot make any scientific assertionsregarding the effectiveness of lighting. In short, the effectiveness oflighting is unknown.

The installation of CCTV in urban areas might be a fruitful area forresearch, but its effectiveness is unknown. Though several evaluationshad scientific methods scores of 3, the absence of significance tests limitswhat we can claim for the effectiveness of this tactic. We cannot recommendthe adoption of this tactic, except for purposes of testing.

Finally, compared to the other tactics examined, street closure evaluationshave been conducted with greater rigor. We also have evaluation evidencethat is consistent with theory and research. This tactic appears to bepromising and deserves greater attention, particularly in high crimeareas.

Table 7-9: OPEN PUBLIC PLACES  STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Bjor, 2 ban on public Open spaces 8% reduction in Knutsson, drinking & of downtown drunkenness arrests & Kuhlhorn high risk area, Sweden 64% reduction in 1992 offenders & disorderly conduct closing of a arrests parking site Ramsay 2 ban on public Open spaces No change on 1991 drinking of a British assaults Ramsay downtown area 33% reduction in 1990 insults from strangers Atkins, 2 lighting 39 sections no systematic Husain, of London effect of lighting and Storey. 1991 Ditton & 2 lighting Glasgow 32% to 68% Nair 1994 neighborhood reduction in victimizations. 14% reduction in reported crime Painter 2 lighting London 86% reduction in 1994 street robberies, auto crimes, and threats 2 lighting London 78% reduction in street robberies, auto crimes, and threats 2 lighting London 100% reduction in street robberies, auto crimes, and threats (base rates too small to be meaningful) Brown 1995 3 cctv town center, decline in burglary Newcastle (18%), criminal upon Tyne damage (9%), auto Great Britain theft (7%), theft from auto (11%), other theft up (7%). 3 cctv town center, charts suggest Birmingham, reductions in Great Britain robbery, burglary, and theft but do not allow calculation of reductions 3 cctv town center, charts suggest King's Lynn, reductions in Great Britain burglary, assaults, thefts from autos, and thefts of autos but do not allow calculation of reductions Atlas, and 3 street Florida town 8% decline in LeBlanc closures burglary, drops in 1994 larceny and autotheft. No change in robbery or aggravated assault Lasley 3 street Los Angeles, 65% reduction in 1996 barricades CA homicides Matthews 2 street Finsbury reduction in 1992 closures & Park, London prostitution rerouting activity Matthews 2 street Streatham, reduction in 1993 closures & London prostitution rerouting activity Newman 3 street Dayton, OH 26% reduction in 1996 closures reported crime and 50% reduction in violent crime 

PUBLIC COIN MACHINES

Parking meters and public telephones are the principal subject of thissection. These devices occupy small but important places in cities andare subject to fraud and vandalism. The six studies we will examine hereshow reductions in property offenses due to changes in the physical structure(target hardening) or operations of these devices.

Two evaluations examined the effectiveness of strengthening the materialused in public telephone cash boxes. Target hardening was supplementedby other prevention measures in both instances. In Britain, electronicmonitoring of phone booths helped identify attacks quickly and act as adeterrent (Barker and Bridgeman 1994). The evaluators reported a 49 percentreduction in attacks on cash compartment attacks as a result of these changes.Australian evaluators claimed a comparable reduction in vandalism incidentsfollowing a combined target hardening and rapid repair program (Challinger1991). Both studies have weak designs due to their absence of control places.

Fraudulent use of public telephones has been addressed in two studies.In both, new systems were installed that prohibited calls that prior analysissuggested were likely to be fraudulent. At the New York Port AuthorityBus Terminal, international calls were blocked, keypads were disabled toprevent routing calls through outside automated systems, and the numberof available phones were reduced and relocated (Bichler and Clarke 1997).Calls and number of minutes of phone use declined from the pre-interventionperiod to the post-intervention period. This is indirect evidence of adrop in fraudulent phone use because one cannot distinguish between reducedlegitimate phone use due to increased inconvenience to users and reducedillegitimate phone use.

LaVigne (1994) evaluated the effects of restricting inmate access tophones at Rikers Island, a New York City jail facility. The Departmentof Corrections restricted inmate phone use to control the costs of fraudulentcalls. Not only did phone costs go down, but phone related fights amonginmates declined, controlling for overall trends in fights and changesin inmate population.

Finally, Decker (1972) examined the effectiveness of a target hardeningmethod to prevent slug use in parking meters (i.e., installation of metersthat reject certain types of slugs and display the last coin inserted).Rates of slug use were measured in 10 areas of New York. Slug use declinedin all areas. In another study, Decker (1972) looked at the effectivenessof warning labels on parking meters. He found short-term reductions inslug use for some labels, but overall the labels were less effective thanmeters that reject slugs.

These evaluations imply that target hardening is a promising methodfor reducing theft and vandalism. When evaluators looked for displacementeffects, they were not found. LaVigne's (1994) evaluation suggests thatillegal use of some facilities might stimulate other more serious criminalbehavior and blocking minor offenses might reduce other more serious crimes.The Rikers Island evaluation is an illustration of the possible diffusionof crime prevention benefits (Clarke and Weisburd 1994).

Table 7-10: PUBLIC COIN MACHINES STUDY SCIENTIFIC TACTIC SETTING RESULTS METHODS SCORE Barker & 2 publicity, public 49% reduction in Bridgeman target telephones in vandalism/ 1994 hardening, Great Britain theft electronic monitoring Wilson 2 hardened coin Australian 48% reduction in 1988; boxes, and public vandalism Challinger other telephones 1991 changes, and rapid repair Bichler & 3 removing Port 37% reduction in Clarke international Authority Bus calls and 72% 1996 dialing Terminal, reduction in capacity and Manhattan minutes of phone disabling use. No telephone displacement found keypads to prevent pay phone toll fraud LaVigne 3 restrictions Rikers 46% reduction in 1994 on inmate Island, New telephone related phone use and York fights. phone system 49% reduction in phone costs Decker 4 installation parking reduction in slug 1972 of slug meters in New use due to changes rejecting York in meters. Short parking term reduction with meters & two labels, but no warning signs long term effect of on parking any labels meters 

SCIENTIFIC CONCLUSIONS

Blocking crime opportunities at places can reduce crime, under somecircumstances. Over 90 percent of the interventions reported evidence ofcrime reduction following the installation of an opportunity blocking tactic.This evidence is encouraging but it must be tempered by three considerations.

First, we know little about the place- and crime-specific effects ofthese tactics. That there is a great deal of uncertainty about what works,at which places, against which crimes, should not distract us from thebroader finding that opportunity blocking tactics at places can be productive.We will address specific tactics below.

Second, 94 percent of these evaluations are case studies of a very fewsites, typically a single site. We cannot treat the 99 interventions asa random sample of all interventions of this type. These may have beenevaluations of programs that were far more likely to succeed than is typical.Nevertheless, authors of many of the evaluations asserted that their placeswere hotspots of crime and had resisted other interventions, such as policeenforcement. Thus, the interventions may have tackled tougher problemsthan would be found at the average place.

Third, many of the evaluations studied the effect of multiple interventionsimplemented at about the same time. Even when the effects of a single tacticwere identified, it was sometimes reported that other changes had occurredthat could confound the evaluation results. Thus we might learn that crimewas prevented, but we do not know what caused the prevention. The largenumber of multiple interventions deserves some explanation. Many of theefforts evaluated were the result of some form of problem-solving processin which a specific crime problem was analyzed and a set of appropriatesolutions were implemented. This must be contrasted with efforts undertakento test the efficacy of a particular prevention measure in a particularsetting. Problem-specific interventions may have a greater likelihood ofsuccess than generic interventions, but we may have more difficulty learningfrom them. Later we will return to the subject of problem-solving and situationalcrime prevention.

Fourth, the scientific rigor (as shown by the scientific methods score)supporting the conclusions is usually moderate at best, and is frequentlyweak. Forty-three percent of the studies did not use control places ormeasure crime for a minimal number of pre-intervention time periods. Only6 percent of the evaluations compared the same intervention in at least20 places and used control places. There were only two randomized controlledexperiments among the studies examined. Often evaluators did not reportsignificance levels for crime reductions, so we cannot determine the chancesthat the results were due to random changes in crime. In summary, a typicalevaluation of a place-focused intervention involves a before-after comparisonof a prevention tactic at a single location, compared to a roughly similarlocation or the surrounding area.

The Effects of Displacement are Limited

There is little reason to believe that side effects from place-focusedefforts are greater than the intentional effects. Further, some of theseside effects enhance prevention, rather than undermine it. There are twoside effects: displacement of crime and diffusion of prevention benefits.

One reason for community resistance to place-focused prevention (orany area specific tactic) is the fear of the displacement of crime fromthe target places to other, presumably safer, locations nearby. Displacementcan take on a number of forms. Offenders can change locations. They canchange the times of offending . They can change the target of their criminalbehavior. They can adopt new behaviors to attack the same targets. Andthey can switch the type of crime they commit. Fear of displacement isoften based on the assumption that offenders are like predatory animals(they will do what ever it takes to commit crimes just as a rat will dowhatever it takes to steal food from the cupboard).

In the last 10 years there have been four reviews of the empirical evidenceand theoretical underpinnings for displacement. Theoretical explorationsbased on a rational choice perspective (Cornish and Clark 1986) find nobasis for believing offenders always completely displace if they cannotattack their favorite targets (Cornish and Clark 1987; Barr and Pease 1990;Eck 1993; Barnes 1995; Bouloukos and Farrell 1997). Reviews of empiricalstudies examining place-focused prevention, police enforcement, and otherpreventive tactics in the United States, Canada, Great Britain, continentalEurope, and Australia, find that there is often no displacement, but whendisplacement occurs it does not overwhelm other gains from blocking crimeopportunities (Cornish and Clark 1987; Barr and Pease 1990; Eck 1993; Hesseling1995). There is no evidence to suggest that these interventions increasescrime by displacing it. There have been only a very few examples wheresomething close to 100 percent displacement has been observed (for example,100 crimes are prevented at one set of targets, but there is an increaseof 100 crimes at similar targets). Displacement far less than 100 percentis not uncommon (for example, 100 crimes are prevented against one setof targets but there is an increase of 30 crimes against other targets,yielding a net reduction of 70 crimes). But usually, evaluators who havelooked for empirical evidence of displacement have found little evidenceof displacement. Concern about displacement is usually based more on pessimismthan empirical fact.

It is possible that more displacement would be found if evaluators weremore diligent in their search for it. Most prevention evaluations do notreport on possible displacement effects and when they do, the evidenceused is almost always weaker than the evidence used to support the mainfindings. Still, if the evidence for limited displacement is weak, theevidence for large amounts of displacement is even weaker.

Prevention Benefits Can Spread

Overlooked in our concern about displacement is the possibility of justthe opposite effect, diffusion of crime prevention benefits (Clarke andWeisburd 1994). For example, Scherdin (1992) reported that when magnetictags were put in books in a university library, book theft declined. Butso did the theft of audio and video tapes which were not tagged. Thievesapparently were unaware of which items were protected. We have noted severalother examples of possible diffusion of benefits effects in the evaluationsexamined in this chapter (Felson et. al. 1997; LaVigne 1994; Masuda 1992;Poyner 1988). Evidence for diffusion of benefits is weaker than evidenceagainst displacement, largely because few people have looked for it. Nevertheless,this possibility cannot be rejected on empirical or theoretical grounds.In fact, there are good theoretical reasons to believe diffusion of benefitsmight be common. Diffusion is the flip side of the coin of crime contagion.Contagion suggests that when offenders notice one criminal opportunitythey often detect similar opportunities they have previously overlooked.Crime then spreads. The broken-windows theory (Wilson and Kelling 1982)is an example of a contagion theory. Thus under some circumstances offendersmay be uncertain about the scope of prevention efforts and avoid both theblocked opportunities and similar unblocked opportunities. When this occurs,prevention may spread.

There is Much Uncertainty About Place- and Crime-Specific Tactics

Table 7-11 summarizes the place-specific findings described in detailin the body of this chapter. Each evaluated intervention was put into oneof four categories. Tactics that "work" had to have two or morepositive studies with a scientific methods score of 3 or more and had toreport the statistical significance of the findings. Only one tactic, nuisanceabatement to control drug dealing and related crime at private rental places,received this classification.

To be classified as "does not work" an intervention had tomeet the same qualifications as "works" but the findings reportedno relationship between the intervention and crime. The scientific methodsused were insufficient to detect tactics that did not work, so we haveno tactics in this category. With improved knowledge from more rigorousevaluations some of the tactics in the "unknown" category mightmove into this category. Most tactics may be effective at some type ofplace and against particular crimes, but it is unlikely that all tacticsare effective at all places against all crimes. The absence of tacticsin the "does not work" category reveals our ignorance.

"Promising" tactics had to have at least one evaluation witha scientific methods score of 3 that used significance tests, and showedthat crime declined. If significant tests had been reported some tacticsof "unknown" effectiveness might have been classified as "promising."Seven interventions had sufficient scientific evidence to be classifiedas "promising." Putting metal detectors in this category revealsthe limits of the application of standard social science research methods.Few would question the efficacy of metal detectors and passenger screeningto prevent aircraft hijacking, but because this tactic has not been widelystudied and many of the studies use weak research methods, we cannot putthis tactic in the "works" category. We can be far less certainabout its effectiveness in other settings. Street closures may be anothertactic that is underrated because of a lack of rigorous evaluations, particularlythe absence of significance tests.

The "unknown" category contains the majority of interventions.Many of these interventions had multiple studies showing positive effects,but the evaluations had scientific methods scores less than 3, or did notreport significance test results. Examples of these tactics include CCTVin open spaces and parking lots, and EAS in retail stores. Other tacticshad several weak studies reporting conflicting results. Lighting in openareas is an example of this type of tactic. Finally, some tactics may notprevent crime. Cameras were found to be ineffective at preventing robberiesof convenience stores in a single rigorous test. In a less rigorous analysis,cameras were found to be unrelated to bank robberies.

Clearly there is much to learn if we are to develop a set of well-testedinterventions that can be applied to specific problems. Most cells in Table7-11 empty and the places listed are only a small set of places with crimeproblems. Even when we have tactics that work or look promising, they haveonly been tested against a limited set of crimes.

Table 7-11: SUMMARY OF PLACE SPECIFIC-FINDINGS Works Does Not Work Promising Unknown Residential nuisance target abatement hardening restricting movement guards CCTV cocoon watch property marking Commercial stores multiple EAS clerks CCTV store design target hardening frequent inventory counts prohibiting offenders electronic monitoring ink tags guards cameras restricting movement banking & cameras money target handling hardening guards bars & server taverns training Transport public removing transportation targets rapid cleanup design informal watching parking lots CCTV guards restricting movement airports metal detectors guards Public Setting open spaces street CCTV closures prohibiting offenders controlling drinking lighting public target removing facilities hardening targets signs 

Situational Crime Prevention and Problem-Solving are Promising

This chapter has described what we have learned about the effectivenessof specific tactics to prevent crimes at specific types of places. It isbased on the assumption that if know the type of place and the type ofcrime, we should be able to recommend a specific tactic that can preventcrimes of this type and this place. We have seen limited evidence thatthis assumption is valid.

There is another approach to addressing crime problems, however, thatmay also be valid. Rather than look for a generic solution to a specificcrime problem at a place, one could undertake a thorough examination ofthe problem and then craft a unique set of interventions to address thisproblem. Such an approach is advocated by both situational crime prevention(Clarke 1992) and problem-oriented policing (Goldstein 1990). Many of theevaluations examined multiple simultaneous interventions that addressedspecific problems at places. In these projects the selection of tacticswas preceded by some form of crime analysis. Their evaluations are examinationsof the effectiveness of problem-solving and situational crime prevention.Additionally, one of the two randomized experiments was a study of problem-solving(scientific methods score=5). Stores in the treatment groups did not geta standard intervention, but had an on-site diagnosis and a recommendationof a set of tactics that fit the circumstances (Crow and Bull 1975). Repeatvictimization evaluations (Anderson, Cheery and Pease 1995; Forrester,Chatterton and Pease 1988) are also a form of problem-solving because thecomplex interventions were based on site-specific analysis (both with scientificmethods scores=3). It is difficult to determine how many of the studiesreviewed in this chapter should be considered as problem-solving or situationalcrime prevention efforts, but almost half provide evidence they can beinterpreted in this light. This implies that we have relatively strongevidence for the effectiveness of problem-solving and situational crimeprevention. At minimum, these complementary strategies are a promisingapproach to crime prevention.

EFFECTIVENESS OF DOJ PROGRAMS

There is no single program within the Department of Justice that fundsplace-focused prevention. Instead, place-focused prevention tactics maybescattered throughout a variety of program areas. Within the Byrne FormulaGrant Program, place-focused tactics may be funded under the domestic drugcontrol, community crime prevention, property crime prevention, law enforcementeffectiveness, and public housing purpose areas. These areas comprises$151.8, or about 8 percent of all Byrne Funds for fiscal years 1989 through1994 (Dunworth, Haynes and Saiger 1997). We do not know what proportionof these funds actually went to place-focused tactics, but it is probablyvery small. Within the Local Law Enforcement Block Grant Program for 1996through 1997, $50 million has been allocated to security measures and crimeprevention. This comprises about 14 percent of the program total. Onceagain, we cannot determine how much of these funds go to place-focusedcrime prevention.

Programs to foster problem-solving and situational crime preventionefforts at places may be effective. The NIJ sponsored the earliest researchon problem-solving in Madison, Wisconsin (Goldstein 1990) and Newport News,Virginia (Eck and Spelman 1987). The Bureau of Justice Assistance has fundeda number of programs that applied problem-solving, including the Problem-OrientedApproach to Drug Enforcement, the Systems Approach to Crime and Drug Prevention,and the Comprehensive Gang Initiative. The COPS program, with its focuson the police problem-solving with communities, could make good use ofplace-focused crime prevention. Improving the ability of police and communitiesto identify and analyze problems and then craft effective prevention methodsto alleviate these problems may improve police effectiveness.

Though police problem-solving has received much attention in the UnitedStates, the police are not the only social institution that uses problem-solvingto prevent crime problems. Improving the ability of small businesses, socialorganizations, community groups, and non-criminal justice public agenciesto craft problem-specific solutions to crime problems would have the effectof democratizing crime prevention. Two types of knowledge are requiredfor such efforts. First, people addressing crime problems at places mustknow how to go about identifying problems, analyzing the causes of problems,crafting feasible solutions, and determining if the problems have declined.Second, these people need knowledge about what place-focused preventionhave been tried and which have been found to be effective. Congressionalsupport for developing both sets of knowledge might improve the abilityof private and public institutions to prevent crime.

To the extent that Department of Justice program funds are used to supportnuisance abatement to prevent drug dealing and related crime, these fundsare probably being used in an effective manner. The Bureau of Justice Assistancehas singled out three programs (Boston's Safe Neighborhood Initiative,Lansing's Neighborhood Reclamation program, and Los Angeles's FALCON NarcoticsAbatement Unit) involving nuisance abatement as particularly innovative(Bureau of Justice Assistance 1995).

Nuisance abatement points out a very important fact about place-focusedprevention. Most place-focused prevention takes place at privately ownedlocations. If these owners do not employ prevention measures at their places,then mechanisms are required to induce them to undertake relevant preventionmeasures. Nuisance abatement provides a threat in order to compel the installationof prevention. A positive alternative is landlord training. Landlord trainingprograms provide information to landlords to help them manage their propertiesand prevent drug dealing. Unfortunately, this positive approach has notbeen evaluated so we do not know how effective it is, either in absoluteterms or relative to nuisance abatement.

IMPROVING EFFECTIVENESS THROUGH EVALUATION AND RESEARCH

Providing citizens and businesses, as well as local governments, withscientifically based information on crime prevention may be more productivethan directly funding such programs. Such information can only be providedby a program of rigorous research.

What should a research program look like? First, it must enlist theactive participation of the people and organizations that own and controlplaces. Some basic research can be undertaken using police records, otherpublic data bases, and surveys. Most systematic evaluation and experimentationinvolving changes to the characteristics of places will require the cooperationof the businesses and property owners.

Second, a place-focused research and evaluation program should builda body of theoretically sound and rigorously tested interventions. Theprogram should address six questions:

  1. Where is each type of crime most likely to occur?
  2. What place characteristics protect places from crime or facilitatecrime?
  3. What innovative prevention tactics come from problem-solving and situationalcrime prevention efforts?
  4. What methods for analyzing problems and developing prevention tacticsare particularly useful for local decision makers?
  5. Which tactics are found effective, based on impact evaluations withscientific methods scores not less than 3?
  6. Of those tactics that appear promising based on impact evaluationsin single sites, which survive multi-site evaluations with scientific methodsscores of 4 and 5?

The Drug Market Analysis Project (DMAP) is a useful example of how demonstration,research and evaluation can work together. In five cities (Jersey City,Hartford, Pittsburgh, Kansas City, and San Diego) NIJ funded the developmentof advanced computer mapping. These efforts improved police ability toanalyze their crime and drug problems and they supported basic researchinto drug market places and rigorous evaluations of interventions to controldrug dealing.

DMAP also addressed another research priority. A place-focused researchprogram should foster improvements in scientific methods used in evaluations.All evaluations should employ control groups or interrupted time-seriesdesigns, unless there are overwhelming reasons why such controls cannotbe employed. Further, significance tests and effect sizes should to bereported. NIJ's new Crime Mapping Research Center has the potential toexpand on what was learned through DMAP and to extend our knowledge ofeffective place-focused tactics.

Special efforts need to be made to address side effects: displacementand diffusion of benefit. These side effects can contaminate control groupsand confound evaluation results. If crime displaces into control placesthen program effects can be overestimated. If crime prevention diffusesinto control places then program effects will be underestimated. In neithercase can diffusion or displacement effects be estimated. Evaluation protocolsfor separating control places and displacement/diffusion places need tobe required for all federally funded research. Additionally, these sideeffects should be the subject of research to determine the conditions underwhich they are most likely to occur and what can be done to reduce displacementand facilitate diffusion.

Several place-focused interventions should be given priority for testingto determine if they are effective at controlling violence. These includestreet closures around retail drug markets, CCTV at locations that arehotspots for robberies and assaults, landlord training programs to curbdrug related violence in apartment buildings, and metal detectors in schoolsand public housing with high violent crime rates. Research into the relationshipbetween lighting and violent crime needs to be conducted. Such researchshould examine how offenders use lighting, the circumstances under whichlighting facilitates crime, and the conditions under which lighting isassociated with low crime rates. Evaluations could then be undertaken atplaces where this earlier research suggested that lighting improvementsmight be effective. Finally, studies should examine how repeat victimization,repeat crime places, and repeat offenders are related.

REFERENCES

Allatt, Pat

1984 Residential security: Containment and displacement of burglary.Howard Journal of Criminal Justice. 23:99-116.

Anderson, David, Sylvia Chenery, and Ken Pease

1995a Biting Back: Tackling Repeat Burglary and Car Crime. Crime Detectionand Prevention Series Paper 58. London: Home Office.

Anderson, David, Sylvia Chenery, and Ken Pease

1995b Preventing repeat victimizations: A report on progress in Huddersfield.Home Office Police Research Group Briefing Note 4/95. London: Home Office.

Atkins, Stephen, Sohail Husain, and Angele Storey

1991 The Influence of Street Lighting on Crime And Fear of Crime. CrimePrevention Unit Paper 28. London: Home Office.

Atlas, Randall and William G. LeBlanc

1994 The impact on crime of street closures and barricades: A Floridacase study. Security Journal. 5:140-45.

Bamfield, Joshua

1994 Electronic article surveillance: Management learning in curbingtheft. In Martin Gill (ed.) Crime at Work: Studies in Security and CrimePrevention. Leiscester: Perpetuity Press.

Barclay, Paul, Jennifer Buchley, Paul J. Brantingham, Patricia Brantingham,and Terry Whinn-Yates

1996 Preventing auto theft in suburban Vancouver commuter Lots: Effectsof a bike patrol. In Ronald V. Clarke (ed.) Preventing Mass Transit Crime.Crime Prevention Studies. Vol. 6. Monsey, NY: Criminal Justice Press.

Barker, Mary and Cressida Bridgeman

1994 Preventing Vandalism: What Works? Crime Detection and PreventionSeries paper 56. London: Home Office.

Barnes, Geoffrey C.

1995 Defining and optimizing displacement. In John E. Eck and DavidWeisburd (eds.) Crime and Place. Monsey, NY: Criminal Justice Press andPolice Executive Research Forum.

Barr, Robert and Ken Pease

1990 Crime placement, displacement, and deflection. In Michael Tonryand Norval Morris (eds.), Crime and Justice: A Review of Research, Vol.12. Chicago, IL: University of Chicago Press.

Beavon, Daniel J.K., P.L. Brantingham and P.J. Brantingham

1994 The influence of street networks on the patterning of propertyoffenses. In Ronald V. Clarke (ed.). Crime Prevention Studies. Vol. 2.Monsey, NY: Criminal Justice Press.

Bellamy, Lisa C.

1996 Situational crime prevention and convenience store robbery. SecurityJournal. 7:41-52.

Bichler, Gisela and Ronald V. Clarke

1996 Eliminating pay phone toll fraud at the Port Authority buss terminalin Manhattan. In Ronald V. Clarke (ed.) Preventing Mass Transit Crime.Crime Prevention Studies. Vol. 6. Monsey, NY: Criminal Justice Press.

Bjor, Jill, Johannes Knutsson, and Eckhart Kuhlhorn

1992 The celebration of midsummer eve in Sweden: A study in the artof preventing collective disorder. Security Journal. 3:169-74.

Block, Richard and Carolyn Block

1995 Space, place and crime: Hot spot areas and hot place of liquor-relatedcrime. In John E. Eck and David Weisburd (eds.) Crime and Place. Monsey,NY: Criminal Justice Press and Police Executive Research Forum.

Bouloukos, Adam C. and Graham Farrell

1997 On the displacement of repeat victimization. In Graeme Newman,Ronald V. Clarke, and S. Giora Shoham (eds.), Rational Choice and SituationalCrime Prevention: Theoretical Foundations. Dartmouth, NH: Dartmouth UniversityPress.

Brantingham, Patricia L. and Paul J. Brantingham

1981 Notes on the geometry of crime. In Paul J. Brantingham and PatriciaL. Brantingham (eds.), Environmental Criminology. Beverly Hills, CA: Sage.

Brown, Ben

1995 CCTV in Town Centres: Three Case Studies. Crime Detection and PreventionSeries Paper 68. London: Home Office.

Bureau of Justice Assistance

1995 Ten Years of Innovation: Pioneering BJA Programs. Washington, DC:Bureau of Justice Assistance.

Burrows, John and Martin Speed

1996 Crime analysis: Lessons from the retail sector. Security Journal.7:53-60.

Carr, Kerri, and Geoff Spring

1993 Public transport safety: A community right and a communal responsibility.In Ronald V. Clarke (ed.) Crime Prevention Studies. Vol. 1. Monsey, NY:Criminal Justice Press.

Centers for Disease Control and Prevention

1993 Violence-related attitudes and behaviors of high school students-- New York City, 1992. Morbidity and Mortality Weekly Report. 42: 773-77.

Chaiken, Jan M., Michael W. Lawless, and Keith A. Stevenson

1974 Impact of Police on Crime: Robberies on The New York City SubwaySystem. R-1424-NYC. New York: The New York City Rand Institute.

Challinger, Dennis

1991 Less telephone vandalism: How did it happen? Security Journal.2:111-19.

Challinger, Dennis

1996 Refund fraud in retail stores. Security Journal. 7:27-35.

Chatterton, Michael R. and Samantha J. Frenz.

1994 Closed-circuit television: Its role in reducing burglaries andthe fear of crime in sheltered accommodation for the elderly. SecurityJournal. 5:133-39.

Clarke, Ronald V. ed.

1992 Situational Crime Prevention: Successful Case Studies. Albany,NY: Harrow and Heston.

Clarke, Ronald V.

1993 Fare evasion and automatic ticket collection in the London underground.In Ronald V. Clarke (ed.) Crime Prevention Studies. Vol. 1. Monsey, NY:Criminal Justice Press.

Clarke, Ronald V.

1995 Situational crime prevention. In Michael Tonry and David P. Farrington(eds.) Building a Safer Society: Strategic Approaches to Crime Prevention.Crime and Justice. Vol. 19. Chicago, IL: University of Chicago Press.

Clarke, Ronald V., Ronald Cody, and Mangai Natarajan

1994 Subway slugs: Tracking displacement on the London underground.British Journal of Criminology. 34:122-38.

Clarke, Ronald V., Simon Field, and Gerry McGrath

1991 Target hardening of banks in Australia and displacement of robberies.Security Journal. 2:84-90.

Clarke, Ronald V. and Ross Homel

in press "A revised classification of situational crime preventiontechniques." In Steven P. Lab (ed.), Crime Prevention at the Crossroads.Cincinnati, OH: Anderson.

Clarke, Ronald V. and Gerry McGrath

1990 Cash reduction and robbery prevention in Australian betting shops.Security Journal. 1:160-63.

Clarke, Ronald V. and David Weisburd

1994 Diffusion of crime control benefits: Observations on the reverseof displacement. In Ronald V. Clarke (ed.), Crime Prevention Studies. Vol.2. Monsey, NY: Willow Tree Press.

Clifton, Wayland, Jr.

1987 Convenience store robberies in Gainesville, Florida: An interventionstrategy by the Gainesville Police Department. Gainesville, FL: GainesvillePolice Department. Photocopy.

Cohen, Lawrence E. and Marcus Felson

1979 Social change and crime rate trends: A routine activity approach.American Sociological Review. 44:588-605.

Coleman, Alice

1985 Utopia on Trial. London: Hilary Shipman, Ltd.

Cornish, D. and R. V. Clarke

1986 The Reasoning Criminal. New York: Springer-Verlag.

Cornish, D. and R. V. Clarke

1987 Understanding crime displacement: An application of rational choicetheory. Criminology. 25:933-947.

Crow, W.J., and J.L. Bull

1975 Robbery Deterrence: An Applied Behavioral Science Demonstration--FinalReport. La Jolla, CA: Western Behavior Sciences Institute.

Crow, W.J., and Rosemary J. Erickson

1984 Cameras and Silent Alarms: A study of Their Effectiveness as aRobbery Deterrent. Jackson Hole, WY: Athena Research Corporation.

DesChamps, Scott, Patricia Brantingham, Paul Brantingham

1992 The British Columbia transit fare evasion audit. In Ronald V. Clarke(ed.) Situational Crime Prevention: Successful Case Studies. New York:Harrow and Heston.

Decker, John F

1972 Curbside deterrence: An analysis of the effect of a slug rejectiondevice, coin view window and warning labels on slug usage in New York Cityparking meters. Criminology. 10:127-42.

DiLonardo, Robert L.

1996 Defining and measuring the economic benefit of electronic articlesurveillance. Security Journal. 7:3-9.

DiLonardo, Robert L. and Ronald V. Clarke

1996 Reducing the rewards of shoplifting: An evaluation of ink tags.Security Journal. 7:11-14.

Ditton, Jason and Gwyneth Nair

1994 Throwing light on crime: A case study of the relationship betweenstreet lighting and crime prevention. Security Journal. 5:125-32.

Dunworth, Terence and Aaron J. Saiger

1994 Drugs and Crime in Public Housing: A Three City Analysis. Washington,DC: National Institute of Justice.

Dunworth, Terence, Peter Haynes, and Aaron J. Saiger

1997 National Assessment of the Byrne Formula Grant Program. Researchin Brief. Washington, DC: National Institute of Justice.

Dwyer, Jim

1991 Subway Lives: 24 Hours in the Life of the New York City Subway.New York: Crown.

Easteal, Patricia Weiser and Paul R. Wilson

1991 Preventing Crime on Transport. Canberra: Australian Institute ofCriminology.

Eck, John E.

1993 The threat of crime displacement. Criminal Justice Abstracts. 25:527-46.

Eck, John E.

1994 Drug markets and drug places: A case-control study of the spatialstructure of illicit drug dealing. Unpublished Ph.D. dissertation, Universityof Maryland, College Park.

Eck, John E.

1995 A general model of the geography of illicit retail marketplaces.In John E. Eck and David Weisburd (eds.) Crime and Place. Monsey, NY: CriminalJustice Press and Police Executive Research Forum.

Eck, John E. and William Spelman

1987 Problem-solving: Problem Oriented Policing in Newport News. Washington,DC: Police Executive Research Forum.

Eck, John E. and Julie Wartell

1996 Reducing Crime and Drug Dealing by Improving Place Management:A Randomized Experiment. Report to the San Diego Police Department. Washington,DC: Crime Control Institute.

Eck, John E. and David Weisburd

1995 Crime and Place. Monsey, NY: Criminal Justice Press and PoliceExecutive Research Forum.

Ekblom, Paul

1987 Preventing Robberies at Sub-Post offices: An Evaluation of a SecurityInitiative. Crime Prevention Unit Paper no. 9. London: Home Office.

Ekblom, Paul

1988 Preventing post office robberies in London: Effects and side effects.Journal of Security Administration. 11:36-43.

Engstad, Peter A.

1975 Environmental opportunities and the ecology of crime. In RobertA. Silverman and James J. Teevan (eds.), Crime in Canadian Society. Toronto:Butterworth.

Farrell, Graham

1995 Preventing repeat victimization. In Michael Tonry and David P.Farrington (ed.) Building a Safer Society: Strategic Approaches to crimePrevention. Crime and Justice. Vol. 19. Chicago, IL: University of ChicagoPress.

Farrington, David P., Sean Bowen, Abigail Buckle, Tony Burns- Howell,John Burrows, and Martin Speed

1993 An experiment on the prevention of shoplifting. In Ronald V. Clarke(ed.) Crime Prevention Studies. Vol. 1. Monsey, NY: Criminal JusticePress.

Felson, Marcus

1994 Crime and Everyday Life: Insight and Implications for Society.Thousand Oaks, CA: Pine Forge Press.

Felson, Marcus, Mathieu E. Belanger, Gisela M. Bichler, Chris D. Bruzinski,Glenna S. Campbell, Cheryl L. Fried, Kathleen C. Grofik, Irene S. Mazur,Amy B. O'Regan, Patricia J. Sweeney, Andrew L. Ullman, and LaQuanda M.Williams

1996 Redesigning hell: Preventing crime and disorder at the Port Authoritybus terminal. In Ronald V. Clarke (ed.) Preventing Mass Transit Crime.Crime Prevention Studies. Vol. 6. Monsey, NY: Criminal Justice Press.

Felson, Marcus, Robyn Berends, Barry Richardson, and Arthur Veno

1997 A community policing initiative to discourage abuse of alcohol.In Ross Homel (ed.) Crime Prevention Studies. Vol. 7. Monsey, NY: CriminalJustice Press.

Fleming, R. and J. Burrows

1986 The case for lighting as a means of preventing crime. (ResearchBulletin, Research and Planning Unit, No. 22, pp. 14-17). London: HomeOffice.

Forrester, David H., Michael R. Chatterton, and Ken Pease

1988 The Kirkholt Burglary Prevention Demonstration Project, Rochdale.Crime Prevention Unit Paper no. 13. London: Home Office.

Forrester, David H., Samantha Frenz, Martin O'Connell, and Ken Pease

1990 The Kirkholt Burglary Prevention Project: Phase II. Crime PreventionUnit Paper no. 23. London: Home Office

Gabor, Thomas

1981 The crime displacement hypothesis: An empirical examination. Crimeand Delinquency. 26:390-404.

Goldstein, Herman

1990 Problem-oriented Policing. New York: McGraw-Hill.

Gottfredson, Michael R. and Travis Hirschi

1990 A General Theory of Crime. Stanford, CA: Stanford University Press.

Grandjean, Christian

1990 Bank robberies and physical security in Switzerland: A case studyof The escalation and displacement phenomena. Security Journal. 1:155-59.

Green, Lorraine

1995 Cleaning Up drug Hotspots in Oakland, California: The displacementand diffusion effects. Justice Quarterly. 12:737-54.

Green, Lorraine

1996 Policing Places with Drug Problems. Thousand Oaks, CA: Sage.

Green, Lorraine

1993 Treating deviant places: A case study examination of the beat healthprogram in Oakland, California. Ph.D. dissertation, Rutgers, The StateUniversity of New Jersey.

Hannan, Timothy H.

1982 Bank robberies and bank security precautions. Journal of LegalStudies. 11:83-92.

Hemingway, Richard C.

1989 Environmental design, access control, and surveillance as deterrentsto thefts in hospitals. Security Journal. 1:47-57.

Hesseling, Rene B. P.

1995 Displacement: A review of the empirical literature. In Ronald V.Clarke (ed.), Crime Prevention Studies. Vol. 3. Monsey, NY: Criminal JusticePress.

Hesseling, Rene'

1995 Theft from cars: Reduced or displaced? European Journal of CriminalPolicy and Research. 3:79-92.

Homel, Ross and Jeff Clark

1994 The prediction and prevention of violence in pubs and clubs. InRonald V. Clarke (ed.) Crime Prevention Studies. Vol. 3. Monsey, NY: CriminalJustice Press.

Homel, Ross, Marg Hauritz, Richard Wortley, Gillian McIlwain, and RussellCarvolth

1997 Preventing alcohol-related crime Through community action: TheSurfers Paradise safety action project. In Ronald V. Clarke (ed.) CrimePrevention Studies. Vol. 7. Monsey, NY: Criminal Justice Press (forthcoming).

Hope, Tim

1994 Problem-oriented policing and drug market Locations: Three casestudies. In Ronald V. Clarke (ed.). Crime Prevention Studies. Vol. 2. Monsey,NY: Criminal Justice Press.

Hunter, Ronald D. and C. Ray Jeffrey

1992 Preventing convenience store robbery through environmental design.In Ronald V. Clarke (ed.) Situational Crime Prevention: Successful CaseStudies. Albany, NY: Harrow and Heston.

Jacques, Christopher

1994 Ram raiding: The history, incidence and scope for prevention. InMartin Gill (ed.) Crime at Work: Studies in Security and Crime Prevention.Leiscester: Perpetuity Press.

Kenney, Dennis J

1986 Crime on the subways: Measuring the effectiveness of the GuardianAngels. Justice Quarterly. 3:481-496.

Landes, William M.

1978 An economic study of U.S. aircraft hijacking, 1961-1976. Journalof Law and Economics. 21:1-32.

Lasley, James R.

1996 Using Traffic Barriers to 'Design out' crime: A Program Evaluationof LAPD's Operation Cul-De-Sac. Report to the National Institute of Justice.California Sate University, Fullerton (November).

LaVigne, Nancy G.

1991 Crimes of convenience: An analysis of criminal decision-makingand convenience store crime in Austin, Texas. Master's thesis. Universityof Texas at Austin.

La Vigne, Nancy G.

1994 Rational choice and inmate disputes over phone use on Rikers Island.Ronald V. Clarke (ed.) Crime Prevention Studies. Vol. 3. Monsey, NY: CriminalJustice Press.

La Vigne, Nancy G.

1997 Safe Transport: Security by design on the Washington Metro. InRonald V. Clarke (ed.) Preventing Mass Transit Crime. Crime PreventionStudies. Vol. 6. Monsey, NY: Criminal Justice Press.

Laycock, Gloria

1985 Property Marking: a Deterrent To Domestic Burglary? Crime PreventionUnit Paper no. 3. London: Home Office.

Laycock, Gloria

1991 Operation identification, or the power of publicity? Security Journal2:67-72.

Laycock, Gloria and Claire Austin

1992 Crime prevention in parking facilities. Security Journal. 3:154-60.

Lurigio, Arthur J., Robert C. Davis, Thomas A. Regulus, Victoria E.Gwiasda, Susan J. Popkin, Mark L. Dantzker, Barbara Smith, and LawrenceOuellet

1993 An Evaluation of the Cook County State's Attorney's Office NarcoticsNuisance Abatement Unit. Chicago, IL: Illinois Criminal Justice InformationAuthority.

Masuda, Barry

1996 An alternative approach to the credit card fraud problem. SecurityJournal. 7:15-21.

Masuda, Barry

1992 Displacement vs. diffusion of benefits and the reduction of inventorylosses in a retail environment. Security Journal. 3:131-36.

Masuda, Barry

1993 Credit card fraud prevention: A successful retail strategy. InRonald V. Clarke (ed.) Crime Prevention Studies. Vol. 1. Monsey, NY: CriminalJustice Press.

Matthews, Roger

1993 Kerb-Crawling, Prostitution and Multi-Agency Policing. Police ResearchGroup. Crime Prevention Unit Series Paper 43. London: Home Office.

Matthews, Roger

1992 Developing more effective strategies for curbing prostitution.In Ronald V. Clarke (ed.) Situational Crime Prevention: Successful CaseStudies. New York: Harrow and Heston.

Mawby, R. I.

1977 Defensible space: A theoretical and empirical appraisal. UrbanStudies. 14:169-79.

Mayhew, Pat

1979 Defensible space: The current status of a crime prevention theory.Howard Journal of Penology and crime Prevention. 18:150-59.

McNees, M.P., J.F. Schnelle, R.E. Kirchner, and M.M. Thomas

1980 An experimental analysis if a program to reduce retail theft. AmericanJournal of Community Psychology. 8:379-385.

Meredith, Colin and Chantal Paquette

1992 Crime prevention in high-rise rental apartments: Findings of ademonstration project. Security Journal. 3:161-69.

Merry, Sally F.

1981 Defensible space undefended: social factors in crime preventionthrough environmental design. Urban Affairs Quarterly. 16:397-422.

National Association of Convenience Stores

1991 Convenience store security Report and Recommendations: November1991. Alexandria, VA: National Association of Convenience Stores.

Newman, Oscar

1972 Defensible Space. New York: Macmillan.

Newman, Oscar

1980 Community of Interest. Garden City, NY: Anchor Press.

Newman, Oscar

1996 Creating Defensible Space. Washington, DC: U.S. Department of Housingand Urban Development.

Painter, Kate

1994 The impact of street lighting on crime, fear and pedestrian use.Security Journal. 5:116-24.

Pease, Ken

1991 The Kirkholt project: Preventing burglary on a British public housingestate. Security Journal. 2:73-77.

Pierce, Glenn L., Susan Spaar, and LeBaron R. Briggs

1986 The Character of Police Work: Strategic and Tactical Implications.Center for Applied Social Research. Northeastern University. Photocopy.

Polvi, Natalie, Terah Looman, Charlie Humphries and Ken Pease

1990 Repeat break-and-enter victimizations: Time-course and crime preventionopportunity. Journal of Police Science and Administration. 17:8-11.

Popkin, Susan J., Victoria E. Gwiasda, Dennis P. Rosenbaum, Andrea A.Anderson, Lynn M. Olson, Arthur J. Lurigio, and Nina Taluc

1995a An Evaluation of the Chicago Housing Authority's Anti-Drug initiative:A Model of comprehensive crime Prevention in Public Housing. Cambridge,MA: Abt Associates Inc.

Popkin, Susan J., Lynn M. Olson, Arthur J. Lurigio, Victoria E. Gwiasda,and Ruth G. Carter

1995b Sweeping out drugs and crime: Residents' views of the ChicagoHousing Authority's public housing drug elimination program. Crime andDelinquency. 41: 73-99.

Poyner, Barry

1983 Design Against Crime: Beyond Defensible Space. London: Butterworths.

Poyner, Barry

1994 Lessons from Lisson Green: An evaluation of walkway demolitionon a British Housing estate. In Ronald V. Clarke (ed.) Crime PreventionStudies. Vol. 3. Monsey, NY: Criminal Justice Press.

Poyner, Barry

1991 Situational crime prevention in two parking facilities. SecurityJournal. 2:96-101.

Poyner, Barry

1988 Video cameras and bus vandalism. Journal of Security Administration.11:44-51.

Poyner, Barry and Caroline Warne

1986 Violence to Staff: A Basis for Assessment and Prevention. London:Tavistock.

Poyner, Barry and Barry Webb

1991 Crime Free Housing. London: Butterworths-Architecture.

Poyner, Barry and Barry Webb

1992 Reducing theft from shopping bags in city center markets. In RonaldV. Clarke, (ed.) Situational Crime Prevention: Successful Case Studies.New York: Harrow and Heston.

Putnam, Sandra L., Ian R. Rockett and Miriam K. Campbell

1993 Methodological issues in community-based program effects. In ThomasK. Greenfield and Robert Zimmerman, (ed.) Center for Substance Abuse PreventionMonograph 14. Rockville, MD: U.S. Department of Health and Human Services.

Ramsay, Malcolm

1991 A British experiment in curbing incivilities and fear of crime.Security Journal. 2:120-125.

Ramsay, Malcolm

1990 Lagerland Lost: An Experiment in Keeping Drinkers off the Streetsin Central Conventry and Elsewhere. Crime Prevention Unit Paper 22. London:Home Office.

Roncek, Dennis W. and Ralph Bell

1981 Bars, blocks and crime. Journal of Environmental Systems. 11:35-47.

Roncek, Dennis and Pamela A. Meier

1991 Bars blocks and crimes revisited: Linking the theory of routineactivities to the empiricism of 'hot spots'. Criminology. 29:725-55.

Roncek, Dennis W. and Mitchell A. Pravatiner

1989 Additional evidence that taverns enhance nearby crime. Sociologyand Social Research. 73:185-188.

Saltz, R.F.

1987 The role of bars and restaurants in preventing alcohol-impaireddriving: An evaluation of server education. Evaluation in Health Professions.10:5-27.

Scherdin, Mary Jane

1992 The halo effect: Psychological deterrence of electronic securitysystems. In Ronald V. Clarke (ed.) Situational Crime Prevention: SuccessfulCase Studies. Albany, NY: Harrow and Heston.

Sherman, Lawrence W., Patrick R. Gartin, and Michael E. Buerger

1989 Hot spots of predatory crime: Routine activities and the criminologyof place. Criminology. 27:27-55.

Sherman, Lawrence and Dennis P. Rogan

1995a Deterrent effects of police raids on crack houses: A randomized,controlled experiment. Justice Quarterly. 12:755-781.

Sherman, Lawrence and Dennis P. Rogan

1995b Effects of gun seizure on gun violence: "Hot spots"patrol in Kansas City. Justice Quarterly. 12:755-781.

Sherman, Lawrence, Janell D. Schmidt, and Robert J. Velke

1992 High Crime Taverns: A RECAP Project in Problem-Oriented Policing.Final report to the National Institute of Justice. Washington, DC: CrimeControl Institute.

Sloan-Howitt, Maryalice and George L. Kelling

1990 Subway graffiti in New York City: 'Gettin up' vs. 'meanin it andcleaning it.' Security Journal. 1:131-36.

Spelman, William

1993 Abandoned buildings: Magnets for crime? Journal of Criminal Justice.21:481-495.

Spelman, William

1995a Criminal careers of public places. In John E. Eck and David Weisburd(eds.) Crime and Place. Monsey, NY: Criminal Justice Press and Police ExecutiveResearch Forum.

Spelman, William

1995b Once bitten, then what? Cross-sectional and time-course explanationsof repeat victimization. British Journal of Criminology. 35:366-383.

Spelman, William and John E. Eck

1989 Sitting ducks, ravenous Wolves, and Helping Hands: New approachesto Urban policing. Public Affairs Comment. 35(2):1-9.

Stanford Research Institute

1970 Reduction of Robbery and Assault of Bus Drivers. Vol. 3. Technologicaland Operational Methods. Stanford, CA: Stanford Research Institute.

Taylor, Ralph B., Stephen D. Gottfredson, and Sidney Brower

1980 The defensibility of defensible space. In Travis Hirschi and MichaelGottfredson (eds.) Understanding Crime. Beverly Hills: Sage.

Tien, J.M., V.F. O'Donnell, A. Barnett, and P.B. Mirchandani

1979 Phase I Report: Street Lighting Projects. Washington, DC: U.S.Government Printing Office.

Tilly, Nicholas

1993a After Kirkholt--Theory, Method and Results of Replication Evaluations.Crime Prevention Unit Paper 47. London: Home Office.

Tilly, Nicholas

1993b The Prevention of Crime Against Small Businesses: The Safer CitiesExperience. Crime Prevention Unit Paper 45. London: Home Office.

Tilly, Nicholas

1993c Understanding Car Parks, Crime and CCTV: Evaluation Lessons fromSafer cities. Crime Prevention Unit Paper 42. London: Home Office.

Tilly, Nicholas, and Janice Webb

1994 Burglary Reduction: Findings From Safer Cities Schemes. Crime PreventionUnit Paper 51. London: Home Office.

vanAndel, Henk

1989 Crime prevention that works: The case of public transport in theNetherlands. British Journal of Criminology. 29:47-56.

Veno, Arthur and Elizabeth Veno

1993 Situational prevention of public disorder at the Australian motorcycleGrand Prix. In Ronald V. Clarke (ed.) Crime Prevention Studies. Vol. 1.Monsey, NY: Criminal Justice Press.

Wagenaar, Alexander C. and Harold D. Holder

1991 Effects of alcoholic beverage server liability on traffic crashinjuries." Alcoholism: Clinical and Experimental Research. 15:942-947.

Webb, Barry

1996 Preventing plastic credit card fraud in the UK. Security Journal.7:23-25.

Webb, Barry and Gloria Laycock

1992 Reducing Crime on The London Underground: An Evaluation of ThreePilot Projects. Crime Prevention Unit Paper 30. London: Home Office.

Weidner, Robert R

1997 Target hardening at a New York City subway station: Decreased fareevasion -- At what price? In Ronald V. Clarke (ed.) Preventing Mass TransitCrime. Crime Prevention Studies. Vol. 6. Monsey, NY: Criminal Justice Press.

Weisburd, David, Lorraine Green and D. Ross

1994 Crime in street level drug markets: A spatial analysis. Criminologie.27:49-67.

Weisel, Deborah L.

1990 Tackling Drug Problems in Public Housing: A Guide for Police. Washington,DC: Police Executive Research Forum.

White, Garland F.

1990 Neighborhood permeability and burglary rates. Justice Quarterly.7:57-67.

Wilkinson, Paul

1977 Terrorism and the Liberal State. New York: John Wiley and Sons.

Wilson, James Q. and George L. Kelling

1982 Broken windows. Atlantic Monthly. (March):29-38.

Wilson, Jerry V.

1990 Gainesville convenience store ordinance: Findings of fact. conclusionsand recommendations. Crime Control Research Corporation.

Wilson, Paul

1988 Reduction of telephone vandalism: An Australian case study. SecurityJournal. 1:149-154.

Chapter Eight

POLICING FOR CRIME PREVENTION

by Lawrence W. Sherman

The more police we have, the less crime there will be. While citizensand public officials often espouse that view, social scientists often claimthe opposite extreme: that police make only minimal contributions to crimeprevention in the context of far more powerful social institutions, likethe family and labor markets. The truth appears to lie in between. Whetheradditional police prevent crime may depend on how well they are focusedon specific objectives, tasks, places, times and people. Most of all, itmay depend upon putting police where serious crime is concentrated, atthe times it is most likely to occur: policing focused on risk factors.

The connection of policing to risk factors is the most powerful conclusionreached from three decades of research. Hiring more police to provide rapid911 responses, unfocused random patrol, and reactive arrests does not preventserious crime. Community policing without a clear focus on crime risk factorsgenerally shows no effect on crime. But directed patrols, proactive arrestsand problem-solving at high-crime "hot spots" has shown substantialevidence of crime prevention. Police can prevent robbery, disorder, gunviolence, drunk driving and domestic violence, but only by using certainmethods under certain conditions.

These conclusions are based largely on research supported by the NationalInstitute of Justice, the research arm of the Office of Justice Programsin the U.S. Department of Justice. In recent years, increasing numbersof police executives have incorporated these findings into their crimeprevention strategies. University of Wisconsin law professor Herman Goldstein's(1979) paradigm of "problem-oriented policing" directed researchattention to the specific things police do, and how they can focus theirresources to attack the proximate causes of public safety problems. TheJustice Department's adoption of this perspective has yielded an increasinglycomplex but useful body of knowledge about how policing affects crime.

One of the most striking recent findings is the extent to which thepolice themselves create a risk factor for crime simply by using bad manners.Modest but consistent scientific evidence supports the hypothesis thatthe less respectful police are towards suspects and citizens generally,the less people will comply with the law. Changing police "style"may thus be as important as focusing police "substance." Makingboth the style and substance of police practices more "legitimate"in the eyes of the public, particularly high-risk juveniles, may be oneof the most effective long-term police strategies for crime prevention.

This chapter begins with a review of the eight major hypotheses abouthow the police can prevent crime (Figure 1). It then describes the varyingstrength of the scientific evidence on those hypotheses, in relation tothe "rigor" of the scientific methods used to test them. Theavailable studies are summarized for both their conclusions and their scientificrigor. The chapter then attempts to simplify these results by answeringthe questions about what works, what doesn't, and what's promising. Majorgaps in our knowledge are also examined. The chapter concludes with recommendationsderived from these findings for future federal investment in both evaluationresearch and police methods to be developed for evaluation.

Figure 8-1

Eight Major Hypotheses About Policing and Crime

Other things being equal,

1. Numbers of Police. The more police a city employs,the less crime it will have.

2. Rapid Response to 911. The shorter the police travel timefrom assignment to arrival at a crime scene, the less crime there willbe.

3. Random Patrols. The more random patrol a city receives, themore a perceived "omnipresence" of the police will deter crimein public places.

4. Directed Patrols. The more precisely patrol presence is concentratedat the "hot spots" and "hot times" of criminal activity,the less crime there will be in those places and times.

5. Reactive Arrests. The more arrests police make in responseto reported or observed offenses of any kind, the less crime there willbe.

6. Proactive Arrests. The higher the police-initiated arrestrate for high-risk offenders and offenses, the lower the rates of seriousviolent crime.

7. Community Policing. The more quantity and better quality ofcontacts between police and citizens, the less crime.

8. Problem-Oriented Policing. The more police can identify andminimize proximate causes of specific patterns of crime, the less crimethere will be.

VARIETIES OF POLICE CRIME PREVENTION

1. Numbers of Police. Like the deterrence hypothesis itself (Gibbs,1975), the claim that police prevent crime is not a "theory"in a truly scientific sense. The idea was developed not as a mathematicalequation but as a general "doctrine" of public policy in theheat of democratic debate. The doctrine was based not just on speculation,but also on the apparent results of several "demonstration projects"with some empirical results. These included the court supervised "BowStreet Runners" (Lee, 1901 [1971]; Pringle, 1955) and the privatelyoperated but publicly chartered Thames River "Marine Police"(Stead, 1977). As the level of violence throughout the 19th century declinedwhile the number of police increased (Gurr, et al, 1977: 93-96; 140), manyobservers concluded that the more police, the less crime.

2. Rapid Response to 911. The general form of this claim is thatthe shorter the police travel time from assignment to arrival at a crimescene, the more likely it is that police can arrest offenders before theyflee. This claim is then extended to rapid response producing three crimeprevention effects. One is a reduction in harm from crimes interruptedin progress by police intervention. Another, more general benefit of rapidresponse time is a greater deterrent effect from the threat of punishmentreinforced by response-related arrests. The third hypothesized preventioneffect comes from the incapacitation through imprisonment of offendersprosecuted more effectively with evidence from response-related arrests.All of these claims presume, of course, that police are notified duringor immediately after the occurrence of a crime. This premise, like thehypotheses themselves, is empirically testable, and it's falsificationcould logically falsify the hypotheses built upon the assumption of itsvalidity.

3. Random Patrols. Early beat officers were directed to checkin at specific places at specific times, with rigid supervision of theprescribed patrol patterns (Reiss, 1992). The increasing emphasis on rapid911 response in automobiles gradually put an end to directed patrols, allowingofficers to patrol at random far beyond their assigned beats. This policywas justified by the theory that unpredictability in patrol patterns wouldcreate a perceived "omnipresence" of the police that deters crimein public places. Chicago Police Chief and Berkeley Criminology Dean OrlandoW. Wilson (1963: 232) was a widely cited proponent of this view. Althoughhe favored the use of police workload analysis to determine how many officersshould be assigned to different beats and shifts, modern police practiceshows little variation in patrol presence by time and place. Nonetheless,many police chiefs and mayors claim that hiring more officers to patrolin this fashion would reduce crime.

4. Directed Patrols. Since the advent of computerized crime analysis,however, a far greater precision in the identification of crime patternshas become possible. Police have used that precision to focus patrol resourceson the times and places with the highest risks of serious crime. Thehypothesis is that the more patrol presence is concentrated at the "hotspots" and "hot times" of criminal activity, the less crimethere will be in those places and times. The epidemiological underpinningfor this claim is NIJ-funded research showing that the risk of crime isextremely localized, even within high crime neighborhoods, varying widelyfrom one address to another (Pierce, Spaar and Briggs, 1988; Sherman, Gartinand Buerger, 1989).

5. Reactive Arrests. Like police patrol, arrest practices canbe either unfocused or focused on crime-risk factors. Reactive arrests(in response to specific citizen complaints) are like random patrol inthat they cast a wide net, warning all citizens that they can be arrestedfor all law violations at all times. This net is necessarily quite thin.Observations of thousands of police encounters with criminal suspects showsthat police choose not to arrest suspects in the majority of thecases in which there was legal basis to do so (Black, 1980: 90; Smith andVisher, 1981: 170). The frequent decision not to arrest has been noticedby crime victims' advocacy groups, who argue that more arrests will produceless crime. This hypothesis, like deterrence generally, is expressed attwo levels of analysis: the "general" or community-wide, andthe "specific" or individual-level. The individual-level hypothesishas been questioned for decades by social scientists, and even some police,who suggest exactly the opposite: that arrest, especially for minor offenses(which are by far the most common), provokes a response by offenders makingthem more likely to commit future crime than if they had not been arrested.

6. Proactive Arrests. Like directed patrol, proactive (police-initiated)arrests concentrate police resources on a narrow set of high-risk targets.The hypothesis is that a high certainty of arrest for a narrowly definedset of offenses or offenders will accomplish more than low arrest certaintyfor a broad range of targets. In recent years the theory has been testedwith investigations of four primary high risk targets: chronic seriousoffenders, potential robbery suspects, drug market places and areas, andhigh-risk places and times for drunk driving. All but the first can betested by examining the crime rate. The hypothesis about chronic seriousoffenders is tested by examining the rate at which such offenders are incapacitatedby imprisonment from further offending.

Another version of the proactive arrest hypothesis is called "zerotolerance," based on the "broken windows" theory (Wilsonand Kelling, 1982). The theory is that areas appearing disorderly and out-of-controlprovide an attractive climate for violent crime--just as a window withone broken pane attracts more stones than a completely unbroken window.The crime prevention hypothesis is that the more arrests police make foreven petty disorder), the less serious crime there will be (Skogan, 1990).

Community vs. Problem-Oriented Policing

The hypotheses about community- and problem-oriented-policing are lessfocused than the others, so much so that some observers have even advisedagainst trying to test them (Moore 1992: 128). They both involve far morevariations and possible combinations of police activities than the narrowdeterrence hypotheses. As in the community- and school-based programs reviewedin chapters 2 and 4, community and problem-oriented policing are put intopractice more like a "stew" of different elements than a singletype of "food." Yet it is just this flexibility that proponentshypothesize to give them their power. Crime problems vary so widely innature and cause that effective policing for prevention must vary accordingly,and arguably require many elements to succeed.

While community and problem-oriented policing are often said to be overlappingstrategies (Skogan, 1990; Moore, 1992), they actually have very differenthistorical and theoretical roots. Community policing arises from the crisisof legitimacy after the urban race riots of the 1960s, the proximate causesof which several blue-ribbon reports blamed on police (President's Commissionon Law Enforcement and Administration of Justice, 1967; National AdvisoryCommission on Civil Disorders, 1968). The reports claimed police had lostcontact with minority group residents, both by changing from foot patrolsto radio cars and by taking a more legalistic approach to law enforcement.In various ways, most notably "team policing" (Task Force Report;Sherman, et al 1973), the police were urged to increase their contact withcitizens in more positive settings than just responding to emergencies.Thus for almost three decades the Community Policing hypothesishas been that increasing the quantity and quality of police-citizen contact(Kelling, 1988) reduces crime.

Problem-oriented policing, in contrast, arose from the crisis of policeeffectiveness at crime prevention provoked in the 1970s by some of thevery studies reviewed in this chapter. As one of its early sponsors, GaryHayes (1979), put it, the studies told police chiefs that nothing theywere doing--putting more police on the street into random patrols, rapidresponses--was working to fight crime. The strategy of problem-orientedpolicing conceived by Professor Goldstein (1979) provided a new paradigmin which to focus innovation, regardless of any contact with the citizenry.Where the core concept of community policing was community involvementfor its own sake, the core concept for problem-oriented policing was results:the effect of police activity on public safety, including (but not limitedto) crime prevention. Nonetheless, community policing has also been justifiedby its hypothesized effects on crime, not the least of which has been therationale for the 100,000 federally funded police officers.

7. Community Policing. The crime prevention effects of communitypolicing are hypothesized to occur in four major ways.

7.a. Neighborhood Watch. This hypothesis justifies one of themost widespread community policing programs, "block watch": increasingvolunteer surveillance of residential neighborhoods by residents, whichshould deter crime because offenders know the neighbors are watching.

7.b. Community-Based Intelligence. This hypothesis justifiesthe many community meetings (Sherman et al, 1973) and informal contactspolice sought through storefront offices, foot patrol (Trojanowicz, 1986)and other methods: increasing the flow of intelligence from citizens topolice about offenses and offenders, which then increases the probabilityof arrest for crime and the deterrent incapacitative effects of arrest.The increased flow of citizen intelligence can also increase police effectivenessat crime prevention through problem-solving strategies.

7.c. Public Information About Crime. This hypothesis is justthe reverse of the last one: increased flow of police intelligence aboutcrime back to citizens improves citizen ability to protect themselves,especially in light of recent changes in crime patterns and risks. Thelatest version of this idea is "reverse 911," under which policefax out warnings of criminal activity to a list of residential and businessfax numbers requesting the service.

7.d. Police Legitimacy. Given the historical roots of communitypolicing, perhaps the most theoretically compelling version of its crimeprevention hypothesis addresses police legitimacy, or public confidencein the police as fair and equitable (Eck and Rosenbaum, 1994). Recent theoreticaland basic research work in "procedural justice" (Tyler, 1990)provides a more scientifically elaborate version of this hypothesis thanits proponents in the 1960s intended. The claim is not just that policemust be viewed as legitimate in order to win public cooperation with lawenforcement. The claim is that a legitimate police institution fostersmore widespread obedience of the law itself. Gorer (1955: 296) even attributesthe low levels of violent crime in England to the example of law-abidingmasculinity set by 19th Century police, a role model that became incorporatedinto the "English character." There is even evidence that thepolice themselves become less likely to obey the law after they have becomedisillusioned with its apparent lack of procedural justice (Sherman, 1974).

8. Problem-Oriented Policing offers infinite specific hypothesesabout crime prevention, all under this umbrella claim: the more accuratelypolice can identify and minimize proximate causes of specific patternsof crime, the less crime there will be. In recent years this claim hastaken two major forms:

8.a. Criminogenic Commodities. The more police can removecriminogenic substances from the micro-environments of criminal events,the fewer crimes there will be. This claim arises from the growing emphasison the causation of criminal events as partly independent of the causationof individual criminality (Hirschi, 1986). Like the theories about preventingcrime in places (Chapter 7), the premise is that many crimes require certainpreconditions, such as guns or cash or moveable property (Cohen and Felson,1979).

8.b. Converging Offenders and Victims. Another precondition ofviolent criminal events is that victims and offenders must intersect intime and space. A major problem-solving theory of crime prevention is tokeep the more basic elements of criminal events from combining: the morepolice can reduce the intersection of motivated offenders in time and spacewith suitable targets of crime, the less crime there will be.

TESTING THE HYPOTHESES

All of these hypotheses pose formidable challenges to scientific testing.The measurement of crime is difficult under any circumstances, let alonein relation to experiments or natural variation in police practices. Controlover police practices is difficult for police administrators under normalconditions, let alone under experimental protocols. Measuring the manydimensions of police activity, from effort to manners, is expensive andoften inaccurate. Only a handful of studies have managed to produce strongscientific evidence about any of these hypotheses. But the accumulatedevidence of the more numerous weaker studies can also provide some insightson policing for crime prevention.

As noted in chapter 1, this report employs a scale of 1 to 5 to summarizeseveral different dimensions of scientific "rigor": the strengthof scientific evidence. A score of 5 = strongest evidence for inferringcause and effect, while 1 = the weakest. These dimensions vary somewhatby institutional setting, with different issues inherent in the kinds ofprograms being evaluated. problems. In the police evaluation literature,crime is almost always measured by either official crime reports (withall their flaws) or by victimization surveys of the public (with all theircosts). Police practices are measured either not at all, through citizenperceptions of those practices, through police records, or (in one instance)through direct observation of police patrol activity. It is not clear thatany of these methods except the last is superior to any others in drawingvalid inferences about the actual practices of the police. Thus the greatestdifference across police evaluations lies not in their methods of measurement,but in their basic research designs: the logical structure for drawingconclusions about cause and effect.

Evaluations of police crime prevention generally follow five basic researchdesigns, which can be ranked for overall strength of the inferences theycan suggest about cause and effect. These designs are 1) correlations atthe same point in time (e.g., in 1995 the cities with the most police hadthe most crime) 2) before-and-after differences in crime without a comparisongroup (e.g., doubling drunk driving arrests was followed by a 50% reductionin fatal accidents, 3) before-after differences with comparison (e.g.,the 50% reduction in fatal crashes compared to a 10% increase in fatalcrashes in three cities of comparable size in the same state), 4) beforeand after large sample comparisons of treated and untreated groups (e.g.,30 neighborhoods organized for neighborhood watch compared to 30 that werenot), and 5) randomized controlled experiments (300 offenders selectedby a computerized equal probability program to be arrested had higher repeatoffending rates than 300 offenders selected to be given warnings only).

SCIENTIFIC EVALUATIONS

This section reviews and interprets the reported tests of each of thehypotheses. The discussion attempts to integrate both the scientific scoreof the various studies and the number of studies converging on the sameconclusion. More detailed discussion is offered for some of the major findings,in order to connect the evidence more clearly to the hypotheses. The mainconcern throughout this section is the cumulative success or failure ofthe studies in ruling out competing theories in the attempt to providea conclusive test of each hypothesis.

1. Numbers of Police

As Table 1 shows, most of the studies of the effects of police numberson crime are scientifically weak. They consist of two basic research designs.One is evidence from police strikes1about a sudden and drastic reduction in police numbers. The other is evidencefrom correlational studies of police strength and crime rates.

The police strike evidence, while weak in both measurement and design,is fairly consistent in showing the effect of this natural experiment:crime rates skyrocket instantly. The strongest design is the Makinen andTakala (1980) study of crime in Helsinki before and during a police strike.The Helsinki measures included systematic observation counts of fightsin public places, as well as emergency room admissions for assault-relatedinjuries. Both measures rose substantially during the strike despite severewinter weather. The only purportedly negative evidence on this conclusionis the Pfuhl (1983) study of police-recorded crime in 11 American policestrikes, in which 89% of the "strike" period in the analysisconsisted of non-strike days. Both the measure and the definition of thestrike period hopelessly confound cause and effect, rendering the studyirrelevant to the conclusion reached from the stronger evidence.

None of the strike findings have comparison groups, so in theory itis possible that crimes would have risen dramatically during the strikeperiod even without the strike. The substantial magnitudes of some of theincreases, however, greatly exceed typical daily variations in crime inbig cities. In the Montreal police strike of 1969, for example, there were50 times more bank robberies and 14 times more commercial burglaries thanaverage (Clark, 1969). Thus despite the weak research design, the largeeffect size suggests that abolishing a police force can cause crime toincrease.

Table 8-1: Numbers of Police

The more police a city employs, the less crime it will have.

Studies Scientific Methods Score Findings Andenaes 1974 2 1944 No Danish Police, large robbery & larceny increase Clark 1969 2 Police Strike, major increase in violent & property crime Russell 1975 2 Same as preceding (Boston) Sellwood 1978 2 Same as preceding (Liverpool) Makinen and Takala 1980 2 Same as preceding (Helsinki) Pfuhl 1983 1 No crime increase during quarters with police strike Marvell & Moody 1996 3 Higher police numbers in cities reduce most types of crime 36 study review 2 Little evidence that more police reduce crime; weak methods 

Whether adding more officers to an already large police force causescrime to decrease, however, is somewhat less clear. A recent review of36 correlational studies, most of them weak in research design, found littleevidence that more police reduce crime (Marvell and Moody, 1996). The sameauthors, however, offer a twenty-year analysis of 56 cities of over 250,000people each and of 49 states. Using a complex technique called the Grangertest, Marvell and Moody (1996) find consistent evidence that increasesin the numbers of police cause reductions in crime in the following year.This study rates a level 4 because it employs multiple comparison groupsand uses appropriate controls for well-specified differences across units.While it lacks random assignment, it is the best evidence available aboutthe effect of modest increases in police numbers. While it runs againstthe conclusion of the preponderance of the other studies, the differencein scientific rigor tips the preponderance of the evidence in the directionof the conclusion that police numbers alone do help to reduce crime ina big city or a state. What the causal mechanism for that effect may beor how it may be enhanced, however, is not clear.

The Marvell and Moody (1996: 632) analysis also allows a test of thehypothesis that the prevention benefits of hiring more police officersare greater in higher-crime cities than across the country in general.The analysis estimates that for each additional officer added to a policeforce in a big city, 24 Part I crimes are prevented annually. For eachofficer hired anywhere in a state, only 4 Part I crimes are prevented.States, on average, have much lower crime rates than the big cities (over250,000 population); in 1995 the rate of Part I crimes was 8,563 per 100,000in the big cities, compared to 5,624 per 100,000 across all police agencies.Yet the ratio of crime prevention benefit is far greater than the ratioof reported crime risks. The Marvell and Moody estimate shows that sixtimes as much crime is prevented for each officer added in cities thanadded in all places on average. Why the benefit ratio exceeds the riskratio is unknown, but one likely candidate is the greater population densityin cities which lets additional police officers have greater effects onpatrol visibility per resident.

2. Rapid Response to 911

One major theory about the crime prevention benefits of hiring moreofficers is that it reduces police response time. The research on thistheory is an excellent example of how different conclusions can resultfrom research results with very different levels of scientific strength.The initial studies of the response time hypothesis produced strong support,suggesting that shaving minutes off response time could lead to the arrestof many more offenders. The extension of this hypothesis into a strategyof policing included the development of 911 systems to speed victim contactwith police radio dispatchers, and the hiring of more police nationwidein the early 1970s in order to reduce average response times and detercrime through greater certainty of arrest. Only the 1977 NIJ response timeanalysis in Kansas City study, and the NIJ replications in four other cities,were able to call that strategy into question, and open the door to morefocused alternatives (Goldstein, 1979).

The original test of the hypothesis was based on a scientifically weakresearch design, a non-random sample of 265 police responses to citizencalls by the Los Angeles Police Department (Isaacs, 1967). Its resultswere confirmed by a later study in Seattle (Clawson and Chang, 1977): theprobability of arrest per police response increased as police time in travelto the scene decreased. Two other studies (Brown, 1974; Holliday, 1974,as cited in Chaiken, 1978) failed to find that pattern, perhaps because,as Chaiken (1978: 130) observes, "the curves are essentially flatfor response times larger than three minutes, and therefore a substantialamount of data for responses under three minutes is needed to observe anyeffect."

The Kansas City (1977) response time analysis took a far more systematicapproach to the issue. Its first step was to divide crimes into victim-offender"involvement" (e.g., robbery, assault, rape) and after-the-crime"discovery" categories (e.g., burglary, car theft). It then focusedresponse time analysis on involvement crimes, since the offender wouldnot be present at the discovery crimes. The analysis then divided the involvementcrime "response time" into three time periods: crime initiationto calling the police ("reporting time"), police receipt of callto dispatch ("dispatch time"), and "travel time" ofpolice from receipt of dispatch to arrival at the scene. Using systematicobservation methods and interviews of victims, the Kansas City study (1977,Vol. 2: 39) found that

Table 8-2: Rapid Response

The shorter the police travel time from assignment to arrival at a crimescene, the less crime there will be.

Studies Scientific Methods Score Findings Isaacs 1967 1 Shorter police travel time, more arrests Clawson and Chang 1977 1 Same as preceding Pate et al 1976 1 Police travel time unrelated to arrest Kansas City (MO) Police 2 Same as preceding, most 1977 crimes Spelman and Brown 1981 2 Same as preceding 

there was no correlation between response-related arrest probabilityand reporting time once the time exceeded 9 minutes. The average reportingtime for involvement crimes is 41 minutes (K.C.P.D. 1977, Vol. 2: 23).Cutting police travel time for such crime from 5 to 2.5 minutes could requirea doubling of the police force, but it would have almost no impact on theodds of making an arrest.

Police chiefs in the Police Executive Research Forum (PERF) told NIJthat they did not think citizens in their own communities would take solong to call the police. NIJ responded by commissioning PERF to replicatethe citizen reporting component of the response time analysis in four othercities. Over 4,000 interviews about 3,300 "involvement" crimesproduced unequivocal support for the findings of the Kansas City responsetime analysis (Spelman and Brown, 1981). The probability of arrest in thoseserious crimes was only 29 per 1,000 reports, with 75% of serious crimesbeing discovered by victims long after the crimes occurred. Of the 25%that directly involved the victims, almost half were reported five minutesor more after the crime was completed. The findings were consistent acrosscities, including one that had a 911 system and three that did not.

The conclusion that reduced response time will not reduce crime is basedon strong but indirect evidence. The evidence is strong because it is basedon large samples, careful measurement, and a replicated research designin five diverse cities showing little variation in arrest rates by policetravel time, the main factor that tax dollars can affect. It isindirect because an experimental test of the effects of reduced policetravel time on city-wide arrest and crime rates has never been conducted.Yet there is neither empirical nor theoretical justification for such anexpensive test. Given the strong evidence of citizen delays in reportinginvolvement crimes, and the small proportion of serious crimes that featuredirect victim-offender involvement, further tests of this theory seem tobe a waste of tax dollars. Those dollars might be better spent on communicatingthe findings to the general public, which still puts great priority onpolice travel time for public safety (Sherman, 1995).

3. Random Patrols

Another major theory about the benefits of more police is that theycan conduct more random patrols. Table 3 summarizes the evidence for thepolice numbers hypothesis tested at the level of uniformed patrols withincities, in non-directed or random patrol patterns. The Table shows weakevidence of no effect of moderate variations in numbers or method of patrols.The most famous test of the random preventive patrol hypothesis,the Kansas City Preventive Patrol Experiment (Kelling et al, 1974), revealssome of the difficulty in testing this claim. This experiment claimed tohave varied the dosage of patrol presence for one year across three groupsof five randomly assigned beats each, preceded and followed by extensivemeasures of crime from both household surveys and official records. Theresults of the experiment showed no statistically significant differencesin

Table 8-3: Random Patrol

The more random patrol a city receives, the more a perceived "omnipresence"of the police deters crime in public places.

Studies Scientific Methods Score Findings Kelling et al 1974 3 No difference in crime by N of police cars assigned Police Foundation 1981 3 No difference in crime by N of foot patrol assigned Trojanowicz 1986 3 Foot patrol areas had fewer crimes than controls, but no significance tests reported. 

crime across the three groups.

Many criminologists conclude from this experiment that there is no crimeprevention effect of adding patrol presence in a big city, where low densityof crime makes the extra patrol a mere drop in the bucket (Felson, 1994).Yet the experiment has been criticized for its failure to measure the actualdifferences in patrol dosage and the possible lack of them (Larson, 1975),its inadequate statistical power to detect large percentage differencesin crime as not due to chance (Fienberg, et al, 1976), and its failureto assign patrol dosage at random (Farrington, 1982). Similar limitationsare found in the Newark Foot Patrol Experiment (Police Foundation, 1981),where despite large victimization surveys no crime prevention effects weredetected in association with adding or eliminating daytime and early eveningfoot patrols from selected patrol beats.

The weakness of the evidence is even greater for the one study claimingto find a crime prevention effect from random patrols not focused on crimerisks (Trojanowicz, 1986). The design of this study was limited to recordedcrime and calls for service, with no victimization surveys. After daytimefoot patrols were added to 14 beats in Flint Michigan for three years,the official crime counts in those beats were down by 9% in the foot patrolbeats and up 10% in the other beats city-wide. Large increases in burglaryand robbery in the foot patrol areas were matched by reportedly greaterincreases in the rest of the city. No significance tests were reported,nor were there any controls for the demographic characteristics of theareas selected for foot patrol compared to the rest of the city. Sincethe foot patrol areas were not selected at random, it is possible thatthose areas might have experienced different crime trends even withoutthe foot patrols. The fact that the increase in burglary and robbery occurredlargely at night when the foot patrols were not working is perhaps themost interesting fact in the study, supporting the conclusion reached fromevaluations of directed patrols focused on high crime-risk times and places.

4. Directed Patrols

The evidence from the directed preventive patrol hypothesis ismore voluminous, scientifically stronger (in two tests), and consistentlyin the opposite direction from the weight of the (weak) evidence on therandom patrol hypothesis. In order to be assigned to this category, thestudies had to indicate that they were somehow focused on high-crime places,times or areas. In the New York City study (Press, 1971: 94), for example,the test precinct was known as a high robbery area, and had over threetimes as many robberies per week as each group of five areas in the KansasCity experiment. All eight of the reported tests of this hypothesis showcrime reductions in response to increased patrol presence.

Table 8-4 Directed Patrol

The more precisely patrol presence is concentrated at the "hotspots" and "hot times" of criminal activity, the less crimethere will be in those places and times.

Studies Scientific Methods Score Findings Press 1971 3 40% more police, reductions of outdoors crime Chaiken et al 1975; Chaiken 3 Police on subways at night 1978 reduced crime Dahman 1975 2 More police, reductions of outdoors crime Schnelle et al 1977 2 400% more patrol, less Part I crime Sherman and Weisburd 1995 5 100% more patrol, less observed hot spot crime Koper 1995 4 Longer patrol visits, longer post-visit crime-free time Reiss 1995 Review: Barker et al 1993 2 Squad focused on hot spots, where street crime dropped Burney 1990 2 Saturation patrols, reduced street crime 

The crime prevention effects of extra uniformed patrol in marked policecars at high crime "peaks" are especially evident in two verydifferent research designs imposed on one large NIJ study designed to improveupon the Kansas City Preventive Patrol Experiment. Based on the NIJ-fundedresearch showing extreme concentrations in spatial and temporal distributionsof crime, the Minneapolis Police Department (MPD) reorganized its entirepatrol force in 1988-89 to test a pattern of directed patrols at hot spotsduring hot times. With the unanimous consent of the City Council, the MPDsubstantially reduced patrols from low-crime areas in order to provide2 to 3 hours of extra patrol each day during high crime hours at 55 streetcorner hot spots. The corners were randomly selected for extra patrolsfrom a carefully compiled list of 110 high crime locations that were visuallyseparate from each other (Buerger, Petrosino and Cohn, 1995). Under a milliondollar NIJ grant, both the patrolled and unpatrolled hot spots were subjectedto over 7,000 randomly selected hours of observations by independent researchersover the course of a year. The observers recorded every minute of 24,813instances of police presence in the hot spots, and 4,014 observed actsof crime and disorder (Koper, 1995: 656).

Koper's (1995) analysis of the Minneapolis Hot Spots Patrol data founda very strong relationship between the length of each police patrol presence(which averaged 14 minutes) and the amount of time the hot spot was freeof crime after the police left the scene. The longer the policestayed before they left, the longer the time until the first crime (ordisorderly act) after they left. This relationship held for each additionalminute of police presence from one to fifteen minutes, after which therelationship began to reverse. Thus the "Koper curve" in theMinneapolis data suggests the optimum length of a police patrol visit toa hot spot for the purpose of deterring crime is about 15 minutes.

Koper's correlational analysis of all police presences observed in boththe extra-patrol and no-extra-patrol hot spots combined is consistent withthe results of Sherman and Weisburd's (1995) comparisons of the two groups.The experimental analysis found that there was an average of twice as muchpatrol presence and up to half as much crime in the extra-patrol hot spotsas in the no-extra-patrol group. The observational data showed crime ordisorder in 4 percent of all observed minutes in the control group comparedto 2 percent in the experimental group (Sherman and Weisburd, 1995: 64).Most of the difference in the observed crime was found when police werenot present in the hot spots. Crime-related calls for service increasedfor both groups of hot spots over the one-year experiment as well as city-wide,but the average growth per hot spot was up to three times as great in theno-extra-patrol group (17%) as in the extra patrol group (5%) (Shermanand Weisburd, 1995: 644).

These findings can be questioned, like most place-linked crime preventioneffects, with the possible side effect that the crime simply moved elsewhere(but see the discussion of displacement in Chapter Seven). So, too, cana reduction of crime in one city be questioned on the grounds that offendersmay have focused on other jurisdictions. The theoretical perspective of"routine activities" (Cohen and Felson, 1979; Felson, 1994),under which crimes are only likely to happen in certain places and times,makes the displacement hypothesis less plausible. It suggests that if crimeis displaced, it would have to be displaced to other hot spots. That argumentis still consistent with the experimental-comparison group analysis, giventhe rising numbers of calls in the experimental year relative to the baselineyear. But it does not explain away Koper's cross-sectional analysis ofthe effects of longer patrol presence on post-patrol crime rates.

5. Reactive Arrests

The evidence in support of the reactive arrest hypothesis is remarkablyunencouraging at both the community and individual levels of analysis.As a matter of general deterrence, the tests are all fairly weak and generallynegative. As a matter of individual deterrence, the results are consistentlynegative for juveniles, and contradictory for two different groups of domesticassailants, employed an unemployed. The scientific evidence for the latteris among the strongest available in the police literature, while the evidenceabout juveniles is much weaker. Taken as a whole, these results make avivid demonstration of the complexity of police effects on crime.

The evidence on the general deterrent effects of reactive (Reiss,1971) arrests is based on correlational analyses, with and without temporalorder. There is some weak evidence that there is a threshold beyond whichthe effect of increased arrest rates becomes evident, while no such effectis apparent below the "tipping point" of minimum dosage level(Tittle and Rowe, 1974). This evidence is complicated by the suggestionthat the arrest effects are only evident among cities of less than 10,000people, even with the "tipping point." The finding by Greenbergand his colleagues (1979, 1982) of no arrest rate deterrent effects ina temporal sequence design in big cities throws great doubt on a simpleclaim of general deterrence. Here again, without focusing arrests on highrisk persons or places, the effects of higher arrest levels may get lostin the many factors causing crime.

Table 8-5 Reactive Arrests

The more arrests police make in response to reported or observed offensesof any kind, the less crime there will be.

5.a. General Deterrence. The higher the arrest rate per crimefor each type of crime in a city, the lower the city's rate of that typeof crime.

Studies Scientific Methods Score Findings Tittle and Rowe 1974 2 Cities with higher arrest rates beyond a "tipping point" have less crime, but under tip point no arrest effects Logan 1975 2 No correlation of arrest rates and crime across cities Brown 1978 2 Tipping effect limited to cities under 10,000 people Greenberg et al 1979 2 No effect of arrest rates on crime across cities Greenberg and Kessler 1982 3 No arrest rate effect even when other factors controlled Chamlin 1988 3 More arrests reduce robberies, not 4 property crimes Chamlin 1991 3 No arrest rate effect for cities over 10,000 

5.b. Specific Deterrence. Individual offenders arrested for anoffense are less likely to repeat that offense in the future than offenderswho are not arrested.

Table 8-5b: Specific Deterrence

Studies Scientific Methods Score Findings Juvenile Offenses: Gold and Williams 1970 2 Arrested juveniles offend more post-arrest than those not arrested Klein 1986 4 More formal arrest processing increased recidivism Huizinga and Esbensen 1992 2 Same as preceding Smith and Gartin 1989 2 Arrested juveniles offend less post-crime than those not arrested, if they are first offenders; others more Farrington 1977 2 Arrested juveniles offend more post-arrest than those not arrested Domestic Violence Sherman and Berk 1984 4 Arrest reduced recidivism 50% Dunford et al 1990 5 Arrest had no effect on recidivism at 6 mos Dunford 1992 5 Arrest increased offense frequency at 12 mos Dunford 1990 5 Arrest warrant reduced absent offender recidivism 50% Sherman et al 1991 5 Arrest had no effect on recidivism at 6 mos; short arrest increased recidivism after 12 mos Sherman et al 1992 (Milwaukee) 4 Arrest deters employed, criminogenic for unemployed (Omaha) 4 Same as preceding Berk et al 1992a 5 Arrest reduced recidivism Berk et al 1992b 4 Arrest deters employed, not unemployed Pate et al 1991 5 Arrest reduced recidivism Pate and Hamilton 1992 5 Arrest deters employed, criminogenic for unemployed Hirschel et al 1992 5 Arrest increases official recidivism Marciniak 1994 4 Arrest deters in areas of high employment & marriage; increases recidivism in areas of low employment & marriage 

The consistent individual level evidence of the criminogeniceffects of arrests for juveniles is all longitudinal, but only one of thestudies is a randomized experiment (Klein, 1986). The other studies arenatural observations of the difference in self-reported offending beforeand after juvenile offenders were arrested. These studies cannot adequatelycontrol for the rival hypothesis that the same factors that led to theyouth being arrested also caused a higher level of repeat offending. Apattern of "defiance" (Sherman, 1993), for example, would accountfor both variables and their correlation. The Klein (1986) experiment reportedsome difficulties in maintaining random assignment, but still managed tomake the formal charging of juveniles in police custody a matter of equallikelihood across cases. Holding juvenile characteristics relatively constant,then, Klein found that the more legalistic the processing of a juvenilesuspect, the higher the official recidivism rate.2In interpreting these results, it is necessary to recall that most juvenileoffenses are for fairly minor offenses, and that most juveniles with onepolice contact never have another (Wolfgang, Figlio and Sellin, 1972).Thus to a certain degree, arresting some juveniles and not others for suchoffenses may be perceived as arbitrary or procedurally unfair.

The evidence on the effects of arrest for misdemeanor domestic violenceis contradictory across cities but consistent within arrestee characteristics.While three experiments have found some evidence of deterrent effects ofarrest (Sherman and Berk, 1984; Pate, Annan and Hamilton, 1991; Berk etal, 1992), three other experiments have found some evidence that arrestincreases the frequency of officially detected offending (Sherman, et al,1991; Hirschel et al, 1992; Dunford, 1992). All four of these six experimentsfor which the data have been analyzed separately by employment status ofthe offender show consistent results. Arrest increases repeat offendingamong unemployed suspects while reducing it among employed suspects. Marciniak(1994) has shown that this difference operates even more powerfully atthe census tract level than at the individual level, with arrest backfiringirrespective of individual employment status in neighborhoods of concentratedunemployment and single parent households. There is a literature raisingconcerns about measurement issues in these data (Garner and Fagan, 1995;Fagan, 1996) that are not generally raised about other studies in the policeliterature. Yet there is no other example in the police literature of sixsimilar randomized experiments all testing similar hypotheses with similar(though not identical) designs, and these studies feature a scientificrigor score that is twice the mean of all studies classified for this chapter.The consistency of the effects of arrest on crime for employed and unemployedoffenders even extends to similarity in effect sizes.

6. Proactive Arrests

Like the evidence on focused patrol, the evidence on the focused proactivearrest hypothesis is generally supportive across a wide range of studiesand research designs. While most of the studies are relatively modest inscientific strength, there are some randomized controlled experiments.With the exception of arrests targeted on drug problems, there appear tobe substantial results from focusing scarce arrest resources on high riskpeople, places, offenses and times.

The evidence on high-risk people comes from two strong ( level 4) evaluationsof police units aimed at repeat offenders. The Washington, D.C. unit employedpre-arrest investigations, designed to catch offenders in the act of crimeto enhance the strength of evidence. The Phoenix police unit employed post-arrestinvestigations, designed to enhance the evidence in the offenders latestcase based upon the length and nature of the offender's prior record. Bothprojects aimed at increasing the incarceration rate of the targeted offenders,and both succeeded. Just how serious or active the offenders were is animportant issue in these studies, one which could illuminate future analysesof dollars invested per crime prevented.

Two weaker studies use national samples of cities to test the effectsof police arrest rates for minor offenses on robbery. Both employ multivariatemodels to control for the effects of some of the other factors that couldinfluence the city's robbery rate. Both find that the higher the per capitarates of traffic arrest, the lower the rates of robbery. One uncontrolledfactor in these analyses is the number of pedestrian robbery opportunities.This may be much higher in cities where there is less use of automobiles,such as New York City, in which under 3% of the US population suffers 12%of the reported robberies. Since that is the only crime type for whichNew York is so disproportionate, and since other dense, pedestrian citieslike Baltimore and Boston also have high robbery rates, there may be aspurious relationship between traffic enforcement and robbery. That is,the more cars per capita, the fewer robbery opportunities and the moretraffic enforcement opportunities.

Table 8-6: Proactive Arrests.

The higher the arrest rate for high-risk offenders and offenses, thelower the rates of serious violent crime.3

8-6.a. Repeat Offenders.

Studies Scientific Methods Score Findings Martin and Sherman 1986 4 Targeted offenders more likely to be arrested and incarcerated (Washington) Abrahamse et al 1991 4 Post-arrest case enhancement increases odds of arrestees being incarcerated (Phoenix) 

Table 8-6.b. Traffic and disorderly conduct arrests

Studies Scientific Methods Score Findings Wilson and Boland 1978 2 Cities with higher arrest rates have less crime Sampson and Cohen 1988 2 Same as preceding Weiss and McGarrell 1996 3 Increased traffic tickets, reduced robbery 

Table 8-6.c. Drug market areas

Studies Scientific Methods Score Findings Kleiman 1988 (Lynn) 2 Crackdown on heroin market, violence down Kleiman 1988 (Lawrence) 2 Crackdown on heroin market, violence up Zimmer 1990 and Kleiman 2 Crackdown on heroin market, 1988 (NYC) violence down Sviridoff et al 1992 3 Crackdown on crack market, violence flat Uchida et al 1992 3 Inconsistent changes in (Birmingham) violence after arrests up Uchida et al 1992 (Oakland) 3 Buy & bust plus door-to-door, robbery down Each strategy alone, no effect Sherman and Rogan 1995 5 Raids of crack houses reduced crime for 12 days Weisburd and Green 1995 4 Crackdowns on hot spots reduced disorder; no effects on violence or property crime Annan and Skogan 1993 3 Drug crackdown in public housing, no effect on crime 

Table 8-6.d. Drunk driving

Studies Scientific Methods Score Findings Ross 1981 review: Arrests up sharply, drop in Ross 1973 (U.K.) 2 crashes decays over time Ross 1975 2 Same as preceding (Scandinavia) 2 Same as preceding Ross 1977 (Chesire 1) 2 Same as preceding Ross 1977 (Chesire 2) 2 Same as preceding Hurst-Wright 1980 (NZ1) 2 Same as preceding Hurst-Wright 1980 2 Same as preceding (NZ2) Ross et al 1982 (France) Homel 1993 3 Increased state arrest rate reduced deaths over 10 years, but not in comparable states 

Table 8-6.e. Zero Tolerance Arrests

Studies Scientific Methods Score Findings Boydstun 1975; Sherman 1990 3 More field interrogations, fewer outdoors crimes Reiss 1985 3 More police regulation of conduct, fewer "soft" crimes Pate et al 1985; Skogan 3 Same as preceding 1990 Sherman 1990 2 Disorder crackdown, no robbery reduction Kelling and Coles 1996 2 Fare-beating, crackdown, robbery reduction in subways 

That is just the kind of limitation in causal inference that experimentscan address. Quasi-experimental evidence on this hypothesis was recentlyreported by the Hudson Institute study of the Indianapolis Police Department,in which substantial increases in traffic enforcement in a high robberyarea were followed by a sharp reduction in robbery (Weiss and McGarrell,1996).

The evidence on drug crackdowns shows no consistent reductions in violentcrime during or after the crackdown is in effect. The strongest evidenceis the randomized experiment in raids of crack houses (Sherman and Rogan,1995), in which crime on the block dropped sharply after a raid. The rapiddecay of the deterrent effect in only seven days, however, greatly reducesthe cost-effectiveness of the labor-intensive raid strategy. Only the highyield of guns seized per officer-hour invested (Shaw, 1994) and its possibleconnection to community gun violence over a longer time period (Sherman,Shaw and Rogan, 1995) showed great cost-effectiveness. Other drug enforcementstrategies in open-air markets have even less encouraging results, withthe exception of the Jersey City experiment in which the principal outcomemeasure was disorder, not violence.

The evidence on drunk driving, in contrast, is one of the great successstories of world policing. Despite relatively low rigor scores, the sheernumbers of consistent results from quasi-experimental evaluations of proactivedrunk driving arrest crackdowns suggest a clear cause and effect. The abilityof the police to control drunk driving appears to be a direct and linearfunction of the amount of effort they put into it (Homel, 1990). Sincemore deaths are caused annually by drunk driving than by homicide, thecost effectiveness of saving lives through DUI enforcement may well befar greater than for homicide prevention. The evidence on drunk drivingprevention sees far clearer than anything we know how to do to have policeprevent murders.

The evidence for the broken windows-zero tolerance arrests hypothesis(Wilson and Kelling, 1982) is also consistently supportive. The researchdesigns are only moderately strong, but they all suggest that a policefocus on street activity can help reduce serious crime. The specific tacticsby which this is accomplished can be controversial, and some methods usedin the 1982 Newark test have been described in the literature as "unconstitutional"(Skolnick and Bayley, 1985:199), including the ordering of loitering teenagemales off of street corners on the grounds of obstructing traffic. Fieldinterrogations have often been a flash point of poor police-community relations,yet they have also been a favorite crime prevention tactic for police inboth the US and Europe. The evidence from both the San Diego field interrogationexperiment (Boydstun, 1975) and the NIJ Oakland city center study (Reiss,1985) suggest that it is possible to regulate public behavior in a politemanner that fosters rather than hinders police legitimacy. That possibility,however, is by no means guaranteed, and generally takes substantial managerialinvestment in order to bring about.

The larger concern about zero tolerance is its long-term effect on peoplearrested for minor offenses. Even while massive arrest increases, suchas those in New York City, may reduce violence in the short run--especiallygun violence--they may also increase serious crime in the long run. Thenegative effects of an arrest record on labor market participation aresubstantial (Schwartz and Skolnick, 1963; Bushway, 1996). The effects ofan arrest experience over a minor offense may permanently lower policelegitimacy, both for the arrested person and their social network of familyand friends. The criminogenic effect of arrest may make arrestees moredefiant (Sherman, 1993) and more prone to anger in domestic violence andchild abuse. The data suggest that zero tolerance programs should be evaluatedin relation to long-term effects on those arrested, as well as short-termeffects on community crime rates. Program development to foster greaterlegitimacy in the course of making the arrests is also advisable, basedon findings from procedural justice research (see hypothesis 7.d below).This could include, for example, a program to give arrested minor offendersan opportunity to meet with a police supervisor who would explain the programto them, answer questions about why they are being arrested, and give thema chance to express their views about the program while listening respectfullyto them. Such innovations would not be expensive, but would also pose manytestable hypotheses.

7. Community Policing

The results of available tests of the community policing hypothesesare mixed. The evidence against the effectiveness of police organizingcommunities into neighborhood watches is consistent and relatively strong.The evidence about the crime prevention benefits of more information flowingfrom citizens to police is at best only promising. The two tests of policesending more information to citizens are both very strong, but clearlyfalsify the hypothesis. The tests of increasing police legitimacy are themost promising, especially since they draw on a powerful theoretical perspectivethat is gaining growing empirical support.

One of the most consistent findings in the literature is also the leastwell-known to policymakers and the public. The oldest and best-known communitypolicing program, Neighborhood Watch, is ineffective at preventingcrime. That conclusion is supported by moderately strong evidence, includinga randomized experiment in Minneapolis that tried to organize block watchprograms with and without police participation in areas that had not requestedassistance (Pate et al, 1987). The primary problem found by the evaluationsis that the areas with highest crime rates are the most reluctant to organize(Hope, 1995). Many people refuse to host or attend community meetings,in part because they distrust their neighbors. Middle class areas, in whichtrust is higher, generally have little crime to begin with, making measurableeffects on crime almost impossible to achieve. The program cannot evenbe justified on the basis of reducing middle class fear of crime and flightfrom the city, since no such effects have been found. Rather, Skogan (1990)finds evidence that Neighborhood Watch increases fear of crime.

Another popular program for increasing contact between police and publicis community meetings. The careful NIJ evaluation of the Madison,Wisconsin community policing project in which meetings played a centralrole found no reduction in crime (Wycoff and Skogan, 1993). A differentapproach to the meetings in Chicago shows more promise, with the meetingsfocused much more precisely on specific crime patterns in the area andideas for what the police should do to attack those problems. While thecrime reduction evidence for "community policing, Chicago style"is mixed, it is striking that Chicago has mobilized high crime communitiesto participate in these meetings (Skogan, 1996). Unlike neighborhood watchmeetings, the Chicago meetings are held in public places rather than localresidences. The best attendance at these meetings for almost two yearshas been found in the police districts with the highest crime rates.

A less popular but often effective community policing practice is doorto door visits by police to residences during the daytime. These visitsmay be used to seek information, such as who is carrying guns on the street(Sherman, Shaw and Rogan, 1995). The visits may be used to give out information,such as burglary reduction tips (Laycock, 1991). The visits may be usedsimply to introduce local police officers to local residents, to make policingmore personal (Wycoff et al, 1985). Four out of six available tests ofthe door to door visits show modestly strong (rigor = 3) evidence of substantialcrime prevention. In the NIJ-funded Houston test, for example, the overallprevalence of household victimization dropped in the target area substantially,with no reduction in the comparison area. The prevention effects were primarilyfor car-breakins and other minor property crime. Here again, however, therewas a substantial "Matthew effect" (see Chapter 1): the benefitsof the program were highly concentrated among white middle class homeowners,with virtually no benefit for the Asian, Hispanic and African-Americanminorities living in rental housing in the target area (Skogan, 1990).

A far more popular program is far less effective. Police storefrontsare often requested by communities, often staffed during business hoursby a mix of sworn police, paid civilians and unpaid volunteers. The evidencefrom tests of substations in Houston, Newark and Birmingham

(continued after Tables)

Table 8-7: Community Policing

Increasing the quantity and quality of police-citizen contact reducescrime. Tests of this basic hypothesis omitting measurement of an interveningcausal mechanism have been done:

Table 8-7.a. Neighborhood Watch

Studies Scientific Methods Score Findings Lindsay and McGillis 1986 3 Burglary reduced for 18 mos Pate et al 1987; 4 No effect of block watch on crime Skogan 1990 Poorer areas had less surveillance Rosenbaum 1986 3 Same as preceding Bennett 1990 3 Same as preceding 

Table 8-7.b. Intelligence from citizens to police.

Studies Scientific Methods Score Findings Community Meetings Wycoff and Skogan 1993 3 No drop in victimization after increase in police-community meetings in target district Skogan et al 1995 3 After 18 monthly police-community meetings in each beat in 5 districts, reductions in some crimes and victimization measures but not others Door-to-Door Contacts Wycoff et al 1985; Skogan 3 Door-to-door police visits; 1990 victimization dropped Pate et al 1985; Skogan 3 Door-to-door visits & 1990 storefront, crime dropped Laycock 1991 3 Door-to-door visits, burglary down by ___% Sherman et al 1995 3 Door-to-door visits, no drop in crime Uchida et al 1992 3 Visits plus Buy and Bust, crime down Uchida et al 1992 3 Visits alone, no crime reduction Storefronts Wycoff and Skogan 1986 3 Storefront open, no drop in victimization Uchida et al 1992 3 Storefront open, no difference in crime Pate et al 1985; Skogan 3 (See above under 1990 "door-to-door") 

Table 8-7.c. Increasing the flow of information from police to citizens.

Studies Scientific Methods Score Findings Pate et al (Newark) 5 Monthly newsletter with crime data failed to reduce victimizations of recipients Pate et al (Houston) 5 Same as preceding 

Table 8-7.d. Legitimacy

Studies Scientific Methods Score Findings Skogan 1990 (Houston) 3 Doorknock visits reduced fear of police, reduced crime Tyler 1990 1 Definition of past police treatment as fair increases expected obedience to law in the future Paternoster et al 1996 2 Definition of treatment at arrest as fair, lower recidivism in domestic violence Skogan et al 1995 3 Perceived increased responsiveness of police to community in 4 districts, perceived reduction in serious crime in 3 of those 4 

(AL) consistently shows no impact on crime. While there are some positivecitizen evaluations associated with storefronts, the problems of staffingthe offices once they are open may counterbalance any non-crime benefits.

Increasing the flow of information from police to public has been testedin the form of police newsletters. In two randomized NIJ-fundedexperiments, the Newark and Houston police departments found no effectof newsletters on the victimization rates of the households receiving them.The finding was true for both newsletters with and without specific dataon recent crimes in the community.

The most promising approach to community policing is also the most theoreticallycoherent. Based on two decades of laboratory and field studies on the socialpsychology of "procedural justice," a growing body of researchsuggests that police legitimacy prevents crime. Tyler (1990) findsa strong correlation across a large sample of Chicago citizens betweenperceived legitimacy of police and willingness to obey the law. The legitimacywas measured by citizen evaluations of how police treated them in previousencounters. This finding is consistent with the Houston door-to-door experiment,in which citizen fear of police after a major scandal over police beatingto death a Mexican immigrant was reduced by the door-to-door visits. Communitypolicing Chicago style (Skogan, et al 1996) also find the greatest perceivedreduction in serious crime in the districts where surveys showed policewere "most responsive" to citizen concerns. The most powerfultest of this hypothesis is the Paternoster et al (1996) reanalysis of theMilwaukee Domestic Violence Experiment, which found that repeat domesticviolence was lowest among arrestees who thought police had treated themrespectfully; a powerful effect on recidivism was associated with policesimply taking the time to listen to the offender's side of the story. Thecapacity of police legitimacy to prevent crime is something community policingmay well be effective at creating; Skogan's (1994: 176) review of six communitypolicing evaluations (SM scores = 2 or 3) found every one showed positiveor improved perceptions of police in the treated areas.

Still in progress, but with encouraging preliminary results, is theAustralian test of community accountability conferences. The AustralianFederal Police in the Australian capital, Canberra, use this procedureas an alternative to prosecuting juveniles. Only cases in which the offender(s)admit(s) guilt and the victim(s) are willing to attend the conference areeligible. The conference of offenders and victims with their respectivefamilies and friends is led by a trained police officer, who focuses thediscussion on what happened, what harm it caused, and how the harm canbe repaired. The officer tries to insure that everyone, especially victims,is allowed to have their say. Sometimes offenders apologize, but alwaysan agreement for repaying the cost of the crime to the victim is reached;failure to do so results in the case being prosecuted. Preliminary findingsfrom subsequent interviews with victims and offenders in a randomized experimentshow that the procedure greatly increases respect for police and a perceptionof justice, regardless of the outcomes (Strang, 1996; Barnes, 1996). TheNational Institute of Justice has funded a similar ongoing project in Bethlehem,Pennsylvania. This method may turn out to have long-term effects on policelegitimacy in the eyes of both juvenile offenders and their families, whichcould in turn reduce crime.

The interesting point about the Australian model of community policing,as noted in Chapter 2, is that it builds on actual community ties ratherthan anonymous geographic areas. Moreover, the attendees form a communityof concern about the criminal act bringing them together, holding the offenderaccountable for over an hour to a "village-like" community ratherthan for a few minutes to a distant and anonymous judge. Of all the approachesto community policing yet tried, this one may have the most focused empowermentof "community" to prevent future crimes.

8. Problem-Oriented Policing

The tests of this hypothesis are generally more positive than the testsof community policing. As Moore (1992) suggests, however, this may be dueto a process of selective reporting, in which failures are not included.The most basic problem with testing this very rich and complex hypothesisis that it is essentially about insight, imagination and creativity. Theessence of problem-oriented policing as Goldstein (1979) defined it isscience itself (Sherman and Strang, 1996): classification, prediction,and causation. Evaluations of the scientific method, paradoxically, arenot readily susceptible to the scientific method--except in gross comparisonto unscientific methods. From this perspective, problem-orientedpolicing embraces all of the other strategies described in this chapter,with the problem to be solved that of crime prevention.

This section reviews some evidence on police efforts to prevent crimethat do not fall into the preceding seven hypotheses, and that self-consciouslyadopted a scientific process that involved police officers in analyzingcrime patterns, imagining and creating an intervention, and testing itin the field. The two basic categories of interventions reported in theliterature to date are "removing criminogenic substances" and"separating potential victims and offenders." These two categoriessimply reflect a convergence of police and criminological thinking aboutthe proximate causes of criminal events. There is nothing in the basicproblem-oriented policing (POP) strategy (Goldstein, 1979) that requiresthe use of these two approaches. Many others are possible, and may evenbe more effective. If POP succeeds at making scientific research and developmentthe core technology of police work (Reiss, 1992), we may expect that itsapproaches to crime prevention will evolve with the evolution of knowledgeabout crime causation.

8.a. Criminogenic Substances. The evidence on cash controlis weak but suggestive. As part of a multiple intervention strategy toreduce crime in an English public housing project, the coin-operated gasheaters were removed from residences. Rather than having the cash in thehouse as an attraction to burglars, the gas charges were switched to monthlybilling. Burglary went down substantially. It is uncertain, however, whetherother efforts, such as the "cocoon" neighborhood watch aroundrecently burglarized residences, might account for the crime reduction.

The evidence on gun carrying is stronger. In the NIJ Kansas CityGun Experiment, police focused traffic enforcement and field interrogationson gun crime hot spots during hot times (Sherman, Shaw and Rogan, 1995).With special training in the detection of carrying concealed weapons, policefocused on seizing illegally carried weapons. Gun seizures in the targetarea rose by 60%, and gun crimes dropped by 49%. A similar area in a differentpart of town showed no change in either guns seized or gun crimes. In Boston,police have used a mix of strategies to discourage gun carrying in publicplaces among juveniles, especially gang members and probationers. Qualitativeevidence from an NIJ project suggests gun carrying by the high-risk groupshas been substantially reduced, while early quantitative evidence showsan elimination of juvenile gun homicide (Kennedy et al 1996).

The evidence on alcohol and prostitution is also encouraging,and was presented in Chapter Seven in the discussions of taverns, bars,traffic barriers and street closures.

In the Minneapolis RECAP (REpeat Call Address Policing) experiment,however, four police officers were unable to implement a broad mixof efforts to separate potential victims and offenders across a sampleof 250 target addresses. The YMCA refused to limit access to its lobbyduring evening hours, the Public Library refused to bar intoxicated persons,public housing officials were unable to segregate young "disabled"but predatory alcoholics from elderly co-residents, and private landlordsresisted efforts to evict drug dealers (Sherman, 1990; Buerger, 1994).While a randomized experimental design gave the test strong science, policeinexperience at persuading property managers gave the strategy a weak technology.Given the theoretical power of the idea, further development of the methodsof persuasion might be justified, and only then followed by further research.

One of the most popular practices for separating victims and offendersis evening curfews for juveniles. While such curfews give police additionalpowers to search for guns, they have not been used consistently in thatfashion. The primary objective is to get kids, not guns, off the streets.Some cities, such as San Antonio, have reported reductions in reportedcrimes against juveniles. But in preliminary results of an NIJ evaluation,Adams (1996) finds no consistent crime reduction effects across citiesadopting curfews. The scientific rigor of these studies is quite low giventheir complete absence of control groups, and there may also be difficultiesin police willingness to follow curfew policies. Thus the question of theeffectiveness of curfews at preventing youth violence is still quite opento further research and development.

Table 8-8: Problem-Oriented Policing

The more accurately police can identify and minimize proximate causesof specific patterns of crime, the less crime there will be.

Table 8-8.a. Reducing Gun Carrying in Public. The more policecan remove guns from public places or deter people from carrying them inthe micro-environments of criminal events, the fewer crimes there willbe.

Studies Scientific Methods Score Findings Sherman et al 1995 3 Increased gun seizures, reduced gun crimes Kennedy et al 1996 2 Reduced gun carrying, fewer gun crimes 

Table 8-8.b. Separating Potential Victims and Offenders. Themore police can reduce the intersection of motivated offenders in timeand space with suitable targets of crime, the less crime there will be.

Studies Scientific Methods Score Findings Sherman 1990; Buerger 1994 5 Unable to get landlords to restrict offender access Adams 1996 2 Youth curfews, no consistent reduction in crime. 

CONCLUSIONS

For all of its scientific limitations, the evidence shows substantialconsistency on a number of the hypotheses, and some tentative conclusionson others. All science, of course, is provisional, with better researchdesigns or theories revealing previously undiscovered patterns. It is nosmall achievement that police crime prevention research has developed tothe point of having reached some conclusions to discard.

The available evidence supports two major conclusions about policingfor crime prevention. One is that the effects of police on crime are complex,and often surprising. The other is that the more focused the police strategy,the more likely it is to prevent crime. The first conclusion follows fromthe findings that arrests can sometimes increase crime, that traffic enforcementmay reduce robbery and gun crime, that the optimal deterrent effect ofa police patrol may be produced by 15 minutes of presence in a hot spot,and that prevention effects generally fade over time without modificationand renewal of police practices. The second conclusion follows from thelikely failure to achieve crime prevention merely by adding more policeor shortening response time across the board.

The substantial array of police strategies and tactics for crime prevention(Reiss, 1995) has a small but growing evaluation literature. Using thestandard of at least two consistent findings from level 3 scientific methodsscore (well-measured, before-after studies with a comparison group) anda preponderance of the other evidence in support of the same conclusion,the research shows several practices to be supported by strong evidenceof effectiveness, and several with strong evidence of ineffectiveness.

What works:

o increased directed patrols in street-corner hot spots of crime

o proactive arrests of serious repeat offenders

o proactive drunk driving arrests

o arrests of employed suspects for domestic assault

What doesn't:

o neighborhood block watch

o arrests of some juveniles for minor offenses

o arrests of unemployed suspects for domestic assault

o drug market arrests

o community policing with no clear crime-risk factor focus

Several other strategies fail to meet the test of strong evidence forgeneralizable effectiveness, but merit much more research and developmentbecause of encouraging findings in the initial research.

What's Promising:

o police traffic enforcement patrols against illegally carried handguns

o community policing with community participation in priority setting

o community policing focused on improving police legitimacy

o zero tolerance of disorder, if legitimacy issues can be addressed

o problem-oriented policing generally

o adding extra police to cities, regardless of assignments

o warrants for arrest of suspect absent when police respond to domesticviolence

What is notably absent from these findings, however, are many topicsof great concern to police. Gang prevention, for example, is a matter aboutwhich we could not find a single impact evaluation of police practices.Police curfews and truancy programs lack rigorous tests. Police recreationactivities with juveniles, such as Police Athletic Leagues, also remainunevaluated. Automated identification systems, in-car computer terminals,and a host of other new technologies costing billions of dollars remainunevaluated for their impact on crime prevention. There is clearly a greatdeal of room for further testing of hypotheses not listed here due to theabsence of available scientific evidence.

These conclusions suggest important implications for both DOJ crimeprevention funding of police agencies, and improving that effectivenessthrough stronger evaluations.

The Effectiveness of DOJ Programs

Local police agencies receive crime prevention funding from a wide rangeof DOJ programs (see Chapter One). The evidence cited in this chapter indicatesthat most of the funding supports programs shown to be effective. Thereis also evidence that Congress could increase the effectiveness of thefunding with modifications to several formula grant allocation criteria.The two largest components (multijurisdictional task forces and policeequipment) of the two largest OJP programs (estimated $361 million totalin FY 1996 funding) are of unknown effectiveness, suggesting a high priorityfor evaluation research. Also of unknown effectiveness are the ViolenceAgainst Women grants for police. Byrne grants in the drug enforcement purposearea supporting unfocused proactive arrest programs in drug market areasappear from the available evidence to be ineffective at preventing crime.

How Police Funds Are Allocated. The largest single funding sourceis the Office of Community-Oriented Policing Services (COPS), which distributesfunding for the 100,000 planned extra police officers irrespective of crimerate and partially on the basis of population served by each police agency;the major constraint is that half of all funds must go evenly to policeagencies serving over 150,000 people. Other DOJ funds for police are distributedthrough the Office of Justice Programs (OJP) and its constituent offices.

Figure 8-2

Federal Funding Programs for Local Police

DOJ Office and Program Purpose Areas Total Funding (in bold) COPS Office Cops on the beat $1.4 Billion Office of Justice Programs Bureau of Justice Law Enforcement Equipment $171 million Assistance Law Enforcement Hiring 65 million Local Law Enforcement Law Enforcement Overtime 51 million Block Grants Total $287 million Bureau of Justice Multijurisdictional Task Assistance Forces $190 million Byrne Memorial Grants4 against drugs 26 million Urban Enforcement vs. drugs Law Enforcement 15 million Effectiveness 3 million Organized Crime $234 million Total Violence Against Women Grant Office Encourage Arrests Program Total $46 Million Violence Against Women Grant Office Law Enforcement STOP Block Grants Total $30 Million OJJDP Community Policing for Juveniles Total $16 million Other Programs, amounts n.a.: Weed and Seed, OJJDP Serious Chronic Violent and Anti-Gang, BJA Comprehensive Communities Program, others Total of Major Funding Programs $2.013 Billion 

Because each OJP grant award may be allocated among a variety of localagencies including police, there is no exact count of how much federalfunding goes to police agencies. Purpose areas within the major fundingprograms, however, provide a good approximation (See Figure 8-1). Whilesimply summing the purpose area allocations may overestimate police agencyfunding as distinct from other "enforcement" agencies, such asprosecutors, the difference is probably more than made up by other programsfor which we have no precise estimates.

The largest OJP source of local police funding is apparently the LocalLaw Enforcement Block Grants Program, which distributes formula grantsto units of local government on the basis of both state and local PartI violent crimes for the preceding three years; 71 percent of the $405million ($287 million) in 1996 formula funds were allocated to PurposeAreas specifically directed to law enforcement, and more may have beenawarded through other purpose areas. At similar levels of funding are the$475 million in 1996 formula funds provided as Byrne Grants on the basisof population, of which 50% ($237 million at 1996 funding levels) wereallocated to Purpose areas specifically directed to law enforcement in1989-94 (Dunworth, et al, 1997). The Violence Against Women Act includestwo major funding mechanisms for local policing, the $120 STOP ViolenceAgainst Women Formula Grants (of which 25%, or $30 million in FY 1996,must be allocated to improving law enforcement) and the competitive GrantsTo Encourage Arrest Policies ($46 million in 1996). There are also fundsfor community policing components appropriated through Weed and Seed, BJA'sComprehensive Communities Program (CCP), and these OJJDP Programs: Kidsand Guns, the Comprehensive Community-Wide Gang Prevention, Interventionand Suppression Program, the Comprehensive Strategy for Serious, Violentand Chronic Offenders, and Juvenile and Child-Centered Community-OrientedPolicing. Many other smaller funding programs support local police crimeprevention programs. The current estimated total is in excess of $2 billionper year.

Implications of Available Science. The DOJ funding programs supporta wide range of local police activity. Some types of police activity DOJsupports have no impact evaluations, while others can be evaluated directlyor indirectly with the evidence reviewed in this chapter. Much of the fundingsimply supports additional police presence regardless of the activitiespolice undertake. Given the promising evidence on the effectiveness ofsimply adding police officers to police agencies, the scientific reviewsuggests that these funding programs may be effective. It also suggests,however, that the funding programs could be even more effective if thestatutory formula were changed.

In general, the evidence suggests that federal appropriations toprevent crime through additional policing is most effective when allocatedon the basis of serious crime rather than on the basis of population size.This implication is drawn from several scientific conclusions. One is the"promising" finding that across all large cities, more policeproduced less serious crime. A second is the finding that each additionalpolice officer assigned to a big city prevents six times as many seriouscrimes each year as an officer assigned nationally by population (Marvelland Moody, 1996). A third conclusion is the finding that directed patrolin crime hot spots "works" to prevent crime in those hot spots,the greatest micro-level concentrations of crime. A fourth conclusion isthe "promising" finding that police can reduce gun crime by intensifiedenforcement of the laws against carrying concealed weapons. This findingsuggests that federally funded police work in hot spots of gun crime couldhave a substantial impact on the national homicide rate, just as policemay have done in New York City (Reppetto, 1996). Taken together, thesefindings suggest that the Congress could consider revising the statutoryallocation formula based not only on city-level violent crime, but beat-leveland block level crime as well. Such a revision would be more effectivein directing federal funds as precisely as possible for maximum crime prevention.

Refining a Crime-Based Grant Formula. If the Congress did decideto move towards more crime-based grant formulas for allocating police funding,it would be worth considering more precise criteria. The LLEBG formulabased on total Part I violent crimes is problematic for several reasons.One is that police agencies vary in how they report the largest singlecategory of Part I violence, aggravated assault. The boundary between aggravatedand simple assaults is marked very differently in different cities. InMilwaukee in the early 1990s, for example, when someone pointed a gun atanother person and threatens to shoot, the offense is classified as anaggravated assault. In many other police agencies, that conduct might noteven result in an offense report being taken, or at most a simple assaultreport would be filed; this merely reflects different local traditionsin defining "attacks" and "attempts" (the latter ofwhich the FBI asks police to count as completed crimes) for Uniform CrimeReporting Purposes. Differences in aggravated assault rates thus do notreflect the level of serious violence as reliably as differences in homiciderates. But aggravated assault counts clearly determine the allocation ofLLEBG money; they constituted sixty-one percent of all Part I violencein 1995, while homicides constituted only one percent.

Taking aggravated assaults out of a crime-based formula raises otherissues. Homicide alone is a more consistently reported but more unstableindicator, vary widely in many cities from year to year, which would createinstability in funding levels if used to allocate funding. Robberies aremuch more numerous, and more consistently reported than aggravated assaultand rape. On balance, the Congress may find a combination of robbery andhomicide counts to be the most reliable indicator of the greatest needfor supplementary police presence. The same is true for possible statutoryrequirements on how federal funds should be spent on policing within cities,with hot spots of robbery and homicide receiving top priority. The concentrationsof those crimes in the "hot times" of 7 pm to 3 am is a furtherelement a refined crime-based formula for allocating police funding couldconsider.

COPS Program. The procedure for distributing COPS funds by jurisdictionis the major implication of the scientific review for the COPS Program.Another important issue, however, is the purposes for which COPS officersare funded. While there is promising evidence that any increase in policeofficers is helpful, there is even stronger evidence of crime preventioneffects of specific activities. While COPS Program language has stresseda community policing approach, there is no evidence that community policingper se reduces crime without a clear focus on a crime risk factorobjective. There is strong evidence, however, that directed patrols andprograms targeted on criminogenic substances like guns and alcohol canbe effective in attacking crime hot spots. The evidence on crime preventionin places reviewed in Chapter Seven also finds promising support for problem-orientedpolicing, which could be another more tightly defined purpose area forsupplementary police. Thus while the scientific evidence indicates theCOPS program is effective, it also suggests it could be more effectiveif its funding was more focused upon police programs of proven effectiveness.

Local Law Enforcement Block Grants (LLEBG). The scientific evidencealso suggests that most of the wide range of police activities supportedby the LLEBG program are effective in preventing crime. The major exceptionis for law enforcement equipment and technology, which received 60 percentof 1996 appropriations directed to specifically to police (see bar graphs).As noted above, there are no published impact evaluations of the effectsof equipment and technology on crime. Thus the effectiveness of this fundingis unknown. Impact evaluations of this activity are certainly feasible,and could result in substantial improvements in the uses of such technologiesas firearms identification, automated fingerprint identification systems,and in-car computer search capacity for stolen cars and arrest warrants.While the common sense value of such systems may appear substantial, theprior history of other equipment items suggests that there is much to belearned from careful analysis of its ultimate effects upon crime, and notjust intermediate indicators like arrests.

The Congress could also consider refining the crime-based formula forLLEBG as described above, especially for the usage of police overtime.Many police agencies are now using such overtime to mount directed patrolsof the kind found effective in this Chapter. The statutory plan could betterinsure that overtime is used in the most effective ways possible by incorporatingthe "hot times, hot spots" criteria, or other programs of proveneffectiveness, for overtime work. It could offer additional special purposeareas, such as repeat offender units, which have also been found effectivein apprehending and incarcerating serious violent felons.

What Works, What Doesn't, What's Promising (5)

Byrne Grants. The many uses of Byrne grants almost certainlyinclude the programs of proven effectiveness identified in this Chapter.The most heavily funded Purpose Area, however, is of unknown effectiveness.Multijurisdictional Task Forces against drugs received 40 percent of ByrneFormula funding in the years 1989-94 (Dunworth, et al, 1997), but theyhave never been subjected to a published impact evaluation. To the extentthat the Byrne Program was intended to apprehend drug dealers, it may beinappropriate to consider these task forces a prevention program. It doesnot seem inappropriate, however, to specify measurable goals for the program,and to design an impact evaluation to test the effectiveness of the TaskForces in accomplishing those goals.

A Purpose area for Byrne Grants in which evaluation research indicatesineffective use of funding is "urban enforcement" against drugs,estimated at $26 million in FY 1996. To the extent that these grants supportstreet-level drug enforcement with an emphasis on arrests or drug raids,the money is unlikely to prevent crime. The conclusions of multiple evaluationsshow that such practices do not reduce violent crime or disorder in theabsence of constant police presence, and sometimes not even then.

New purpose areas under the Byrne Grants include both drunk drivingand gang enforcement and prevention. The scientific evidence strongly supportsthe use of Byrne grants for drunk driving enforcement as likely to preventmany deaths and serious injuries. It may also have the prevention effectof reducing gun crime, since so many illegally carried guns and gun criminalswanted on warrants can be removed from the streets through traffic enforcement.There is also a preponderance of available evidence that traffic enforcementthat can help reduce robbery. There are no impact evaluations availableon the effectiveness of police strategies against gangs.

STOP Violence Against Women Block Grants. A review of the detailedlisting of FY 1995 STOP grants for law enforcement shows that they generallysupported activities of unknown effectiveness. Programs such as trainingpolice about domestic violence, hiring domestic violence specialists inpolice agencies, and computer software for domestic violence records allappear to be useful at face value, but have not been subject to publishedevaluations. While the individual grant awards are small, there are manyin the same program categories. An evaluation program addressing the effectivenessof the major funding categories could enhance the currently unknown effectivenessof most of these grants.

Grants to Encourage Arrest Policies. These grants apparentlysupport similar activities as the STOP grants, although with a more narrowfocus on domestic violence against women. To the extent that these grantsresult in more arrests in areas of high employment, the scientific evidencesuggests they will be effective in reducing domestic violence against women.There is also strong scientific evidence, however, that under certain conditionsarrests substantially increase future domestic violence against women.This research raises a critical need for further rigorous research, developmentand program evaluation, which would attempt to discover means to overcomethe apparent criminogenic effects of arrest on certain batterers. Thisresearch program, much of which has already been suggested by a NationalAcademy of Sciences panel report (Crowell and Burgess, 1996), could testcombinations of arrest with greater use of supplementary measures suchas battered women's shelters, detoxification centers for batterers, prosecution,counseling, and other strategies.

Juvenile Crime. Substantial federal funds are spent on policingjuvenile crime, for which scientific evidence also shows that policingcan increase crime under certain conditions. The effectiveness or harmresulting from federal support of juvenile policing cannot be determinedfrom the present review, since the kinds of activities and kinds of offendersare too diverse. The available evidence, however, suggests that there isa substantial need for randomized controlled tests of federally fundedjuvenile policing strategies, in order to provide the greatest possiblecertainty that these programs at least do no harm. Federal support of juvenilecurfew enforcement is of unknown effectiveness (and quantity), but theapparent growth of the idea suggests a need for rigorous program evaluationbeyond the current NIJ-funded survey.

Other Programs. Federal support of policing in high-crime Weedand Seed target areas is strongly supported by the scientific evidence,as described above. Federal support of policing in the Comprehensive CommunitiesProgram is also supported by the evidence that extra police prevent crimemore effectively in big cities.

Improving Effectiveness Through Stronger Evaluations

This analysis of how DOJ funding effectiveness shows many critical knowledgegaps. While the scientific evidence does suggest that the majority of DOJfunding for police is indeed effective at preventing crime, there is noevidence available on a large percentage of the other funding. A conservativeestimate is that we lack even indirect scientific evidence on the effectivenessof some $500 million in Congressionally directed federal funding for localpolice in 1996. The record also suggests that evaluation results couldhelp to revise and channel these funds in ways that would prevent crimemore effectively. Moreover, the past two decades have seen police becomemuch more sensitive to the significance of crime prevention evaluationresults, and actively put them to good use (Blumstein and Petersilia, 1995).

Evaluation needs for each specific funding program have been noted aboveas appropriate. A basic statutory plan for accomplishing these evaluationsmore effectively is offered in Chapter Ten. The remainder of this chaptersummarizes the evaluation needs of the current DOJ funding, and then addressesthe highest priority areas for police effectiveness research implied bythis review of the available evidence.

LLEBG Police Equipment and Technology. Historically, DOJ supportof police technologies has focused on the engineering issues in accomplishingtechnological goals, rather than the human factors in using technologyeffectively. A major Congressional investment in human factors evaluationscould provide the Congress with far better guidance on the effectivenessof its substantial appropriations in this area.

A prime example is NIJ's support of lighter-weight bullet-proof vests,which has apparently saved hundreds of police officers' lives. Even morelives might be saved, however, if evaluation research examined police officercompliance in wearing the vests, factors affecting that compliance, andstrategies for increasing the compliance. Similar questions can be answeredfor the role of automated fingerprint identification systems (AFIS). Thepercentages of cases in which fingerprints are respectively sought, detected,submitted to an AFIS, resulting in a suspect's identification, leadingto an arrest, conviction and incarceration can all be evaluated in a varietyof police agencies. The results could encourage greater use of AFIS, ifwarranted, or if not, a redirection of federal funding into police expendituresthat may be more cost-effective in preventing crime.

Another major example is police use of in-car computer terminals. Intheory, this equipment can help patrol officers make far more productiveus of the time they spend patrolling hot spots, or otherwise awaiting thenext dispatch to a call for service. Whether the officers will actuallyuse the terminals to look for stolen cars or check suspicious persons,however, is a key question for an equipment program evaluation. The NationalInstitute of Justice can help design controlled tests (Scientific MethodsScore = 5) randomly assigning new in-car computer systems to some officersbut not others, with observations of how the officers spend their patroltime both before and after the new equipment is installed. This in turncould inform analyses of the number of arrests made per patrol hour, thenumber of guns seized, stolen cars recovered, and so on. Similar experimentscould be done at the patrol beat level over longer periods of time, testingthe hypothesis that beats patrolled by computer-equipped cars will haveless crime than beats patrolled without them. If these hypotheses cannotbe supported by rigorous scientific testing, additional research couldidentify the reasons the technology does not prevent crime as expectedand possible ways to solve those problems.

Other possible examples of technology evaluations are limited only bythe diverse array of police equipment already on the market and currentlyin development, from hand-held gun detectors revealing weapons concealedunder clothing to electrical devices for police to shut off the ignitionof pursued vehicles. A Congressional plan setting aside ten percent ofprogram funding for controlled testing, and another ten percent for researchcosts, would allow evaluations to identify police technology and equipmentof proven effectiveness.

Byrne Grant Multijurisdictional Task Forces Against Drugs. Thisprogram may be defined as serving purposes other than crime prevention.Other goals might be measured in amounts of drugs seized or the numberof mid-level drug dealers arrested and incarcerated. Testing the effectivenessof these programs in accomplishing the goals might be done through randomassignment of a large sample of cases to single jurisdiction versus taskforce investigation. Alternatively, before-and-after comparisons of drugabuse problems could be made in metropolitan statistical areas where thetask forces operate, with further comparisons to areas not creating thesetask forces. Comparisons might also be made across task forces of differentsizes. Basic productivity indicators could also be computed and comparedacross all Byrne-funded task forces, with an analysis of the reasons forvariation in productivity. Further funding might then be conditional uponachieving specific productivity levels. Task force leadership might collaboratewith NIJ researchers in framing a set of questions to be answered by suchan evaluation, and agree upon scientifically and operationally appropriatemeans of designing an evaluation of this $190 million annual program.

Violence Against Women Grants. Both the STOP and Encourage Arrestgrants have two critical areas in which program evaluation can help. Oneis discovering programs of proven effectiveness in preventing almost everykind of crime against women. The other is identifying the most effectivemeans of delivering a wide array of support services, from police trainingto data banks. Both tasks are hindered by the fact that many of the grantsawarded under these funding programs are under $20,000, and are too smallin scope to warrant separate evaluations. This issue, which also appliesto the Byrne Grants and is addressed in Chapter Ten, is one that a Congressionalplan for evaluation can resolve. It is arguably inefficient for each granteeto confront similar issues separately, such as classroom instructionalmaterials for police training. A national evaluation program to identifyViolence Against Women programs of proven effectiveness would provide muchbetter guidance for how to focus the thousands of small grants scheduledto be awarded by these programs in future years.

The methods of testing program effectiveness in crime prevention arediscussed generally in Chapter Four. The most important police researchissues concern the prediction and prevention of serious domestic violence,for which no scientifically validated risk assessment tools are currentlyavailable (Sherman and Strang, 1996). The effectiveness of police-monitoredpersonal radio alarm necklaces for women given court orders of protectionis a high priority for a randomized controlled trial. So is a comparisonof the crime prevention effectiveness of misdemeanor domestic assault arrestswith and without prosecution, which could indicate a need for Congressionalearmarking of funding for the specific purpose of prosecution of such cases.Issuance of arrest warrants for absent misdemeanor assault offenders isa promising practice (Dunford, 1990) that needs replication. Various policeresponses to non-violent domestic disputes (which are more numerous thanviolent ones) can be compared and tested for their effectiveness in preventingsubsequent violence.

Program effectiveness at accomplishing goals other than crime preventioncan also benefit from evaluations. Improved gender equality and victimservices in police actions can also be measured scientifically as programoutcomes. Regardless of the effectiveness of mandatory arrest, for example,the literature reveals substantial difficulty in obtaining patrol officercompliance with arrest policies for misdemeanor assaults--of which themajority require no medical treatment and one-third have no visible signsof injury. The tendency of officers to trivialize these crimes, to respondslowly to domestic calls, and to refuse to make arrests are all behaviorsthat DOJ-funded training and technical assistance programs may seek tochange. Whatever methods are used to pursue those goals, randomized controlledtests can reveal which methods are most effective. Followup observationsof police treatment of women victims in the field would be a criticallyimportant--although expensive--component of evaluating training programs.Absent such careful scrutiny by a "big science" national evaluationeffort, however, the effectiveness of programs for changing police behaviorwill remain unknown. Here again, a Congressional plan for developing programsof proven effectiveness could make a major difference.

Getting Guns Off the Streets--With Legitimacy. One major hypothesisabout the declining homicide rate in the US is that police have becomemore effective at deterring illegal gun carrying in public places (Moore,1980; Wilson, 1994). Further testing of the gun carrying hypothesis seemsto warrant the highest priority for federal research, given the clear connectionof guns to serious juvenile and gang violence. At the same time, the issueof police legitimacy and perceived harassment of young black males is acrucial aspect of gun enforcement. A research agenda developing both policeeffectiveness at detecting illegal guns, while enhancing police legitimacyin the eyes of all citizens including offenders, could address both issuessimultaneously. On these issues, research could help reduce both homicidesand riots, and increase general compliance with the law through greaterrespect for the moral authority of police.

Patrol Location and Timing Strategies. Since gun violence isheavily concentrated in less than 100 of the 10,000 police agencies reportingto the FBI, research is also needed on more general approaches to directedpatrols in hot spots and hot times. One example is the apparently mundaneis of police schedules, which may be vital to crime prevention. Policechiefs face enormous resistance from police unions in changing work assignmentsand schedules to concentrate police in high crime areas between 7 pm and3 am, with the most officers assigned on weekends. Many must use overtimepay to even move in that direction. If experiments comparing crime-focusedstaffing patterns with conventional procedures found a reduction in crime,that could support police chiefs trying to make better use of taxpayerdollars.

Juvenile Shaming and Restorative Justice. Every police agencymust deal with juvenile offenders. The Australian community accountabilityconferences can be tested in police agencies large and small. Given thenegative findings about the effects of arrest on juvenile offending, thereis much to be gained and perhaps little to lose by developing alternativesto arrest. The growing concern over serious juvenile violence, especiallygun offenses in big cities, should not distort the truth that most juvenilesare still arrested for shoplifting and other minor offenses. A programfor first-offenders that works better to nip criminal careers in the budmay well prevent more serious property crime, such as auto theft, and violentcrime. It may also increase police legitimacy in the eyes of the participatingadults, far more effectively than conventional approaches to communitypolicing.

Multi-Agency Experiments. The proposed Congressional restructuringof evaluations in Chapter Ten would make possible a major breakthroughin police research: comparing strategies across large sample of policedepartments. Random assignment of enhanced federal funding for specificstrategies to half of the hundred largest cities could go a long way towardslearning what works of agency-wide policies. A prime example is trafficenforcement. Proactive police arrests for drunk driving are generally sporadic(Ross, 1994), in part because there is no direct evidence that trafficdeaths will rise if drunk driving arrests decline. Moreover, the evidencethat traffic enforcement reduces robbery is suggestive but not conclusive.Taken together, the twin objectives of reducing traffic deaths and robberieswould justify investment in a 100-agency randomized experiment in trafficenforcement. An experiment in which 50 police agencies selected at randomfrom 100 volunteering agencies received substantial federal funding forgreatly increased traffic enforcement--by 300 or 400%-- would be an idealtest of the hypothesis now weakly supported by merely correlational studies.

Another approach would go right to the core of the 1994 Crime Act--the100,000 police. An experiment in which 20% more officers (over currentlevels including COPS grants) were randomly funded in half of a sampleof police agencies would provide a far more definitive test of the crimeprevention effectiveness of the $1.4 billion annual expenditure. The popularsupport for this program may render the question moot for the moment, butthe question remains of just how effective the program is. Experimentsusing this design could also test other theories, such as problem-solvingor community policing uses of extra officers.

Evaluation Funding Priorities. Over half of all DOJ funding forlocal crime prevention is directed to the police. The same cannot be said,however, for the allocation of program evaluation funding. The Congresshas not addressed the question of evaluation funding priorities with thesame clarity as it has identified program funding priorities. This is onemore reason for the Congress to consider the restructuring of DOJ crimeprevention evaluations as discussed in Chapter Ten.

NOTES

1And in one case,the arrest of the entire Copenhagen police force by the Nazis in 1944,which was equivalent to a strike because the occupying German army didnothing to enforce civilian criminal laws before or after arresting thepolice (Andenaes, 1974).

2There was no differencein the self-reported offending data, but only 60% of the offenders gavefollowup interviews.

3Given the potentialfor vehicular homicide attached to drunk driving, that offense is includedhere in the definition of violent crime. It would not, however, be classifiedthat way for most other purposes.

4These amounts areextrapolated from the Dunworth, et al (1997) analysis of the award of grantsin 1989-94, proportionately applied to the FY 1996 allocation of $475 million.

REFERENCES

Abrahamse, Allan F., Patricia A. Ebener, Peter W. Greenwood, Nora Fitzgerald,and Thomas E. Kosin

1991 An Experimental Evaluation of the Phoenix Repeat Offender Program.Justice Quarterly 8:141-168.

Adams, Kenneth

1996 Paper Presented to the National Evaluation Conference, NationalInstitute of Justice, Washington, D.C.

Annan, Sampson and Wesley Skogan

1993 Drug Enforcement in Public Housing: Signs of Success in Denver.Washington, DC: Police Foundation.

Andenaes, Johannes

1974 Punishment and Deterrence. Ann Arbor, MI: University of MichiganPress.

Barker, M., Geraghty, J., Webb, B. & Key, T.

1993 The Prevention of Street Robbery. Police Research Group Crime PreventionSeries Paper No. 44. London: Police Department, Home Office.

Barnes, Geoffrey C.

1996 Paper presented to the American Society of Criminology, Chicago.

Bennett, Trevor

1990 Evaluating Neighborhood Watch. Basingstoke: Gower.

Berk, Richard A., Alec Campbell, Ruth Klap, and Bruce Western

1992a A Bayesian Analysis of the Colorado Spouse Abuse Experiment. Journalof Criminal Law and Criminology 83: 170-200.

1992b The Deterrent Effect of Arrest in Incidents of Domestic Violence:A Bayesian Analysis of Four Field Experiments. American Sociological Review57: 698-708.

Black Donald

1980 The Manners and Customs of the Police. NY: Academic Press.

Boydstun, John

1975 The San Diego Field Interrogation Experiment. Washington, DC: PoliceFoundation.

Brown, W.

1974 Evaluation of Police Patrol Operations. Unpublished M.A. Thesis,University of Ottowa.

Brown, Don W.

1978 Arrests and Crime Rates: When Does a Tipping Effect Occur?. SocialForces 57: 671-82.

Buerger, Michael E. (ed)

1994 The Crime Prevention Casebook: Securing High Crime Locations. Washington,DC: Crime Control Institute.

Buerger, Michael E., Anthony Petrosino and Ellen G. Cohn

1995 In John Eck and David Weisburd, eds. Crime and Place. Monsey, N.Y.:Criminal Justice Press and Police Executive Research Forum.

Burney, E.

1990 Putting Street Crime in Its Place: A Report to the Community/PoliceConsultative Group for Lambeth. London: Goldsmith's College.

Bushway, Shawn

1996 The Impact of a Criminal History Record on Access to LegitimateEmployment. PhD Dissertation, H. John Heinz School of Public Policy andManagement, Carnegie Mellon University.

Chaiken, Jan M.

1978 What Is Known About Deterrent Effects of Police Activities. InJames A. Cramer, ed., Preventing Crime. Beverly Hills, CA: Sage Publications.

Chaiken, Jan M., M. Lawless and K. Stevenson

1975 The Impact of Police Activity on Crime: Robberies on the New YorkCity Subway System. Urban Analysis 3: 173-205.

Chamlin, Mitchell

1988 Crimes and Arrest: An Autoregressive Integrated Moving Average(ARIMA) Approach. Journal of Quantitative Criminology.

1991 A Longitudinal Analysis of the Arrest-Crime Relationship: A FurtherExamination of the Tipping Effect. Justice Quarterly 8: 187-199.

Clark, Gerald

1969 What Happens When the Police Go On Strike. New York Times Magazine,Nov. 16, sec. 6, pp.45, 176-85, 187, 194-95.

Clawson, C. and S.K. Chang

1977 The Relationship of Response Delays and Arrest Rates. Journal ofPolice Science and Administration. 5: 53-68.

Cohen, Lawrence and Marcus Felson

1979 Social Change and Crime Rate Trends: A Routine Activities Approach.American Sociological Review 44: 588-607.

Crowell, Nancy A. and Ann W. Burgess

1996 Understanding Violence Against Women. Washington, DC: NationalAcademy of Sciences.

Dahmann, J.S.

1975 Examination of Police Patrol Effectiveness. McLean Va.: Mitre Corporation.

Dunford, Franklyn W.

1990 System-Initiated Warrants for Suspects of Misdemeanor DomesticAssault: A Pilot Study. Justice Quarterly 7: 631-653.

1992 The Measurement of Recidivism in Cases of Spouse Assault. Journalof Criminal Law and Criminology 83: 120-136.

Dunford, Franklyn W., David Huizinga, and Delbert S. Elliott

1990 The Role of Arrest in Domestic Assault: The Omaha Police Experiment.Criminology 28: 183-206.

Dunworth, Terence, Peter Haynes, and Aaron J. Saiger

1997 National Assessment of the Byrne Formula Grant Program. NationalInstitute of Justice Research in Brief. Washington, DC: US Department ofJustice.

Eck, John E. and William Spelman

1987 Problem Solving: Problem-Oriented Policing in Newport News. Washington,DC: Police Executive Research Forum.

Eck, John E. and Dennis Rosenbaum

1994 The New Police Order: Effectiveness, Equity and Efficiency in CommunityPolicing. In Dennis Rosenbaum, Ed., The Challenge of Community Policing:Testing the Promises. Thousand Oaks: Sage.

Epperlein, T.

1985 The Use of Sobriety Checkpoints as a Deterrent: An Impact Assessment.Phoenix: Arizona Department of Public Safety.

Fagan, Jeffrey

1996 The Criminalization of Domestic Violence: Promises and Limits.Research Report. Washington, DC: National Institute of Justice.

Farrington, David P.

1977 The Effects of Public Labeling. British Journal of Criminology17: 112-25.

1982 Randomized Experiments on Crime and Justice. In Michael Tonry andNorval Morris, Eds., Crime and Justice: An Annual Review of Research. Chicago:University of Chicago Press.

Felson, Marcus

1994 Crime and Everyday Life. Thousand Oaks, CA: Pine Forge Press.

Fienberg, Stephen E., Kinley Larntz, and Albert J. Reiss, Jr.

1976 Redesigning the Kansas City Preventive Patrol Experiment. Evaluation3: 124-131.

Garner, Joel, Jeffrey Fagan and Christopher D. Maxwell

1995 Published Findings From the NIJ Spouse Assault Replication Program:A Critical Review. Journal of Quantitative Criminology 8: 1-29.

Gibbs, Jack

1975 Crime, Punishment and Deterrence. New York: Elsevier.

Gold, Martin and Jay Williams

1970 National Study of the Aftermath of Apprehension. Prospectus 3:3-12.

Goldstein, Herman

1979 Improving Policing: A Problem-Oriented Approach. Crime and Delinquency25: 236-258.

Gorer, Geoffrey

1955 Exploring English Character. London: Cresset.Gorer, Geoffrey

Greenberg, David F., Ronald C. Kessler, and Charles H. Logan

1979 A Panel Model of Crime Rates and Arrest Rates. American SociologicalReview 44: 843-50.

Greenberg, David and Ronald C. Kessler

1982 The Effects of Arrests on Crime: A Multivariate Panel Analysis.Social Forces 60: 771-90.

Gurr, Ted Robert, Peter N. Grabosky, and Richard C. Hula

1977 The Politics of Crime and Conflict: A Comparative History of FourCities. Beverly Hills, CA: Sage.

Hayes, Gary

1979 Personal Communication

Hirschi, Travis

1986 On The Compatibility of Rational Choice and Social Control Theoriesof Crime. In Derek B. Cornish and Ronald V. Clarke, Eds., The ReasoningCriminal. NY: Springer-Verlag.

Hirschel, David, Ira W. Hutchison III, Charles W. Dean, Joseph J. Kelley,and Carolyn E. Pesackis

1990 Charlotte Spouse Assault Replication Project: Final Report. Charlotte,NC: University of North Carolina at Charlotte.

Holliday, L.P.

1974 A Methodology For Radio Car Planning. Unpublished Ms., New YorkCity RAND Institute.

Homel, Ross

1990 Random Breath Testing and Random Stopping Programs in Australia.In R.J. Wilson and R. Mann, Eds., Drinking and Driving: Advances in Researchand Prevention. NY: Guilford Press.

Homel, Ross

1993 Reference supplied upon request

Hope, Tim

1995 Community Crime Prevention. In Michael Tonry and David P. Farrington,Eds., Building a Safer Society. Crime and Justice, Vol. 19. Chicago: Universityof Chicago Press.

Huizinga, David and Finn Esbensen

1992 School Safety 15: 15-17.

Hurst, P. and P. Wright

1980 Deterrence at Last: The Ministry of Transport's Alcohol Blitzes.Paper Presented to the Eighth International Conference on Alcohol, Drugsand Traffic Safety, Stockholm.

Isaacs, H.

1967 A Study of Communications, Crimes and Arrests in A MetropolitanPolice Department. Task Force Report: Science and Technology, A Reportto the President's Commission on Law Enforcement and Administration ofJustice. Washington, DC: USGPO.

Kansas City, MO, Missouri Police Department

1977 Response Time Analysis. Kansas City, MO: Author.

Kelling, George L.

1988 Community Policing. Presentation to the Executive Sessions on thePolice, Kennedy School of Government, Harvard University.

Kelling, George L., Antony M. Pate, Duane Dieckman and Charles Brown

1974 The Kansas City Preventive Patrol Experiment: Technical Report.Washington, DC: Police Foundation.

Kelling, George L. and Catherine M. Coles

1996 Fixing Broken Windows: Restoring Order and Reducing Crime in OurCommunities. NY: Free Press.

Kennedy, David M., Anne M. Piehl, and Anthony A. Braga

1996 Youth Gun Violence in Boston: Gun Markets, Serious Youth Offenders,and a Use Reduction Strategy.

Kleiman, Mark

1988 Crackdowns: The Effects of Intensive Enforcement on Retail HeroinDealing. In Marcia Chaiken, Ed., Street-Level Drug Enforcement: Examiningthe Issues. Washington, DC: National Institute of Justice.

Klein, Malcolm

1986 Labeling Theory and Delinquency Policy: An Empirical Test. CriminalJustice and Behavior 13: 47-79.

Koper, Christopher

1995 Just Enough Police Presence: Reducing Crime and Disorderly BehaviorBy Optimizing Patrol Time in Crime Hot Spots.

Larson, Richard

1975 What Happened to Patrol Operations in Kansas City. Journal of CriminalJustice. 3: 299-330.

Laycock, Gloria

1991 Operation Identification, or the Power of Publicity? Security Journal2: 67-72.

Lee, W.L. Melville

1901

[1971] A History of Police in England. Montclair, NJ: Patterson-Smith.

Lindsay, Betsy and Daniel McGillis

1986 Citywide Community Crime Prevention: An Assessment of the SeattleProgram. In Dennis Rosenbaum, Ed., Community Crime Prevention: Does ItWork? Beverly Hills, CA: Sage.

Logan, Charles H.

1975 Arrest Rates and Deterrence. Social Science Quarterly 56: 376-89.

Makinen, Tuija, and Hannu Takala

1980 The 1976 Police Strike in Finland. Scandinavian Studies in Criminology7: 87-106.

Marciniak, Elizabeth

1994 Community Policing of Domestic Violence: Neighborhood Differencesin the Effect of Arrest. PhD Dissertation, University of Maryland.

Martin, Susan and Lawrence Sherman

1986 Selective Apprehension: A Police Strategy for Repeat Offenders.Criminology 24: 55-72.

Marvell, Thomas B. and Carlisle E. Moody

1996 Specification Problems, Police Levels and Crime Rates. Criminology34: 609-646.

Moore, Mark

1980 The Police and Weapons Offenses. Annals of the American Academyof Political and Social Sciences 452: 22-32.

1992 In Michael Tonry and Norval Morris, Eds., Modern Policing. Crimeand Justice, Vol. 15. Chicago: University of Chicago Press.

National Advisory Commission on Civil Disorders

1968 Report. NY: Bantam Books.

Pate, Tony, Amy Ferrara, Robert A. Bowers, and Jon Lorence

1976 Police Response Time: Its Determinants and Effects. Washington,DC: Police Foundation.

Pate, Antony M. and Wesley Skogan

1985 Coordinated Community Policing: The Newark Experience. TechnicalReport. Washington, DC: Police Foundation.

Pate, Antony M., Marlys McPherson, and Glenn Silloway

1987 The Minneapolis Community Crime Prevention Experiment: Draft EvaluationReport. Washington, DC: Police Foundation.

Pate, Antony, Edwin E. Hamilton and Sampson Annan

1991 Metro-Dade Spouse Abuse Replication Project: Draft Final Report.Washington, D.C.: Police Foundation.

Pate, Antony M. and Edwin E. Hamilton

1992 Formal and Informal Deterrents to Domestic Violence: The Dade CountySpouse Assault Experiment. American Sociological Review. 57: 691-698.

Paternoster, Raymond, Bobby Brame, Ronet Bachman and Lawrence W. Sherman

1996 Do Fair Procedures Matter? The Effect of Procedural Justice onSpouse Assault.

Paper accepted for publication in Law and Society Review.

Pfuhl, Erdwin H., Jr.

1983 Police Strikes and Conventional Crime: A Look At the Data. Criminology21: 489- 503.

Pierce, Glenn, Susan A. Spaar, and LeBaron Briggs, IV.

1988 The Character of Police Work: Strategic and Tactical Implications.Unpublished Ms., Northeastern University, Center for Applied Social Research.

Police Foundation

1981 The Newark Foot Patrol Experiment. Washington, D.C.: Police Foundation.

President's Commission on Law Enforcement and Administration of Justice

1967 The Challenge of Crime in a Free Society. Washington, DC: USGPO.

Press, S.J.

1971 Some Effects of an Increase in Police Manpower in the 20th Precinctof New York City. NY: New York City Rand Institute.

Pringle, Patrick

1955 Hue & Cry: The Birth of the British Police. London: MuseumPress.

Reppetto, Thomas

1996 Reducing Gun Crime in New York City: A Research and Policy Report.NY: Citizens Crime Commission of New York City

Reiss, Albert J. Jr.

1971 The Police and the Public. New Haven: Yale University Press.

1985 Policing a City's Central District: The Oakland Story. Washington,DC: National Institute of Justice.

1992 Police Organization in the Twentieth Century. In Michael Tonryand Norval Morris, eds., Modern Policing. Crime and Justice, Vol. 15. Chicago:University of Chicago Press.

1995 The Role of the Police in Crime Prevention. In Per-Olof Wikstrom,Ed., Integrating Crime Prevention Strategies: Propensity and Opportunity.Stockholm: National Council for Crime Prevention.

Reiss, Albert J. Jr., and Jeffrey A. Roth, eds.

1993 Understanding and Preventing Violence. Washington, DC: NationalAcademy of Sciences.

Rosenbaum, Dennis, Dan A. Lewis and Jane A. Grant

1986 Neighborhood-Based Crime Prevention: Assessing the Efficacy ofCommunity Organizing in Chicago. In Dennis Rosenbaum, editor, CommunityCrime Prevention: Does It Work? Beverly Hills, CA: Sage.

Ross, H. Laurence

1973 Law, Science and Accidents: The British Road Safety Act of 1967.Journal of Legal Studies 2: 1-78.

1975 The Scandinavian Myth: The Effectiveness of Drinking-and-DrivingLegislation in Sweden and Norway. Journal of Legal Studies 4: 285-310.

1977 Deterrence Regained: The Chesire Constabulary's "BreathalyserBlitz." Journal of Legal Studies 6: 241-249.

1981 Deterring the Drinking Driver: Legal Policy and Social Control.Lexington, Mass: Lexington Books.

1992 Confronting Drunk Driving: Social Policy for Saving Lives. NewHaven: Yale University Press.

1994 Personal Communication

Ross, H. Laurence, R. McCleary and T. Epperlein

1982 Deterrence of Drinking and Driving in France: An Evaluation ofthe Law of July 12, 1978. Law and Society Review.

Russell, Francis

1975 A City In Terror: 1919--The Boston Police Strike. New York: Viking.

Sampson, Robert and Jacqueline Cohen

1988 Deterrent Effects of Police on Crime: A Replication and TheoreticalExtension. Law and Society Review 22:163-189.

Schnelle, J.F., R.E. Kirchner, Jr., J.D. Casey, P.H. Uselton Jr., andM.P. McNees

1977 Patrol Evaluation Research: A Multiple-Baseline Analysis of SaturationPolice Patrolling During Day and Night Hours. Journal of Applied BehavioralAnalysis 10: 33-40.

Schwartz, Richard and Jerome Skolnick

1962 Two Studies of Legal Stigma. Social Problems 10: 133-138.

Sellwood, A.V.

1978 Police Strike--1919. London: W.H. Allen.

Shaw, James W.

1994 Community Policing Against Crime: Violence and Firearms. PhD dissertation,University of Maryland at College Park.

Sherman, Lawrence W.

1974 Becoming Bent. In Lawrence W. Sherman, Ed., Police Corruption:A Sociological Perspective. NY: Anchor-Doubleday.

1990 Police Crackdowns: Initial and Residual Deterrence. In MichaelTonry and Norval Morris, Eds., Crime and Justice: A Review of Research,Vol. 12. Chicago: University of Chicago Press.

1992 Policing Domestic Violence: Experiments and Dilemmas. NY: FreePress.

1993 Defiance, Deterrence, and Irrelevance: A Theory of the CriminalSanction. Journal of Research in Crime and Delinquency 30: 445-473.

1995 The Police. In James Q. Wilson and Joan Petersilia, eds. Crime.San Francisco: ICS Press.

Sherman, Lawrence W., Catherine H. Milton and Thomas Kelly

1973 Team Policing: Seven Case Studies. Washington, DC: Police Foundation.

Sherman, Lawrence W., Patrick R. Gartin and Michael E. Buerger

1989 Hot Spots of Predatory Crime: Routine Activities and the Criminologyof Place. Criminology 27: 27-55.

Sherman, Lawrence W. and Richard A. Berk

1984 The Specific Deterrent Effects of Arrest for Domestic Assault.American Sociological Review 49: 261-272.

Sherman, Lawrence W., Janell D. Schmidt, Dennis P. Rogan, Patrick R.Gartin, Ellen G. Cohn, Dean J. Collins and Anthony R. B