Evidence-based crime prevention: conclusions and directions for a safer society.

• Author: Welsh, Brandon C.
• Position: Includes text in French - Canada

Dans une societe qui mise sur les resultats, les politiques gouvernementales et les pratiques locales en matiere de prevention du crime privilegient les interventions qui ont fait leurs preuves--ce qui marche le mieux. Dans une telle societe, la meilleure facon d'evaluer l'efficacite des mesures de prevention du crime consiste a effectuer des etudes systematiques qui guident les gouvernements dans l'elaboration de leur programme d'action en la matiere. Cet article resume les principaux resultats d'un projet du groupe de coordination Campbell pour la justice et la prevention du crime, projet de recherche sur les moyens efficaces de lutte contre le crime, retenus parmi une grande variete d'interventions ciblees sur les enfants a risque, les delinquants, les victimes et les lieux mal fames. Les conclusions completes en sont publiees dans le livre a paraitre intitule Preventing Crime: What Works for Children, Offenders, Victims, and Places. Cette premiere serie d'analyses donne des resultats encourageants en ce sens que la plupart des interventions sont efficaces pour la prevention du crime et donnent, dans bien des cas, des resultats significatifs. Ces interventions sont diverses: cours d'apprentissage social pour les enfants, therapie cognitivo-comportementale et traitement des delinquants toxicomanes en etablissement, face a face entre victimes et delinquants au nom de la justice reparatrice, prevention des cambriolages a repetition, maintien de l'ordre dans les lieux mal fames, surveillance par television en circuit ferme, amelioration de l'eclairage public. On pourrait contribuer a une societe plus sure en donnant suite aux conclusions de ces etudes methodiques, tant dans l'immediat qu'a plus long terme. Outre la preparation et la mise a jour des etudes methodiques entreprises par le groupe Campbell au profit des decideurs, des praticiens et du grand public, il faudrait lancer un programme de recherche sur les nouvelles experiences faites en matiere de prevention du crime et de lutte contre la criminalite.

Introduction

Crime prevention should be rational and should be based on the best possible evidence. One would expect that decision makers would weigh heavily any available evidence on what works. How can a program that has produced no discernible evidence of effectiveness, as shown through numerous evaluations, be considered for implementation? Unfortunately, this happens all the time. Consider the short-lived revival of the prison deterrence program known as Scared Straight, for which past evaluations showed that it had failed to deter juvenile delinquents from future criminal activity (Finckenauer and Gavin 1999; Petrosino, Turpin-Petrosino, and Finckenauer 2000). Consider, also, the long-standing school-based substance abuse prevention program known as DARE (Drug Abuse Resistance Education), for which the accumulated evidence shows that it has only a trivial effect on substance use and crime (Gottfredson, Wilson, and Skroban Najaka 2002; U.S. General Accountability Office 2003). Many other examples exist in the U.S. and elsewhere.

There are many considerations involved in selecting and implementing new crime prevention programs (as well as in expanding effective programs or putting an end to ineffective or harmful ones). For example, there may be different government priorities, such as military defence spending, environmental protection, or prescription drug benefits for seniors, which are competing for scarce public resources. National polls may show that the public is more concerned with matters of public policy other than preventing crime. Other political considerations that are germane to crime prevention include politicians' apprehension that they may be perceived as being soft on crime if they support social crime prevention efforts (see Gest 2001), and politicians' short time horizons, which make programs that show results only in the longer run less appealing to those who are trying to get elected every few years. Regrettably, it seems that evidence of what works best is rarely a factor in implementing new crime prevention programs. Political and policy considerations seemingly dominate.

Evidence-based crime prevention attempts to overcome these obstacles by ensuring that the best available evidence is considered in any decision to implement a program designed to prevent crime. As noted by Petrosino (2000), "An evidence-based approach requires that the results of rigorous evaluation be rationally integrated into decisions about interventions by policymakers and practitioners alike" (635). This is an approach that has garnered much support in medicine (Halladay and Bero 2000; Millenson 1997). But even in medicine, a discipline noted for its high educational requirements and its adherence to scientific principles, most practice is "shaped by local custom, opinions, theories, and subjective impressions" (Sherman 1998: 6). Of course, making available to policy makers and practitioners scientific evidence on what works best and having them put it into practice are two entirely different things.

Support for evidence-based crime prevention is growing (see Sherman 2003; Welsh and Farrington 2001). This growth has been fostered by a number of recent developments, including a movement toward an evidence-based approach in other disciplines, such as medicine (Millenson 1997) and education (Mosteller and Boruch 2002); large-scale, government- and foundation-sponsored reviews of "what works" in crime prevention (Goldblatt and Lewis 1998; Sherman, Gottfredson, MacKenzie, Eck, Reuter, and Bushway 1997; Sherman, Farrington, Welsh, and MacKenzie 2002; Tonry and Farrington 1995); and, most recently, the establishment of the Campbell Collaboration and its Crime and Justice Group (Farrington and Petrosino 2001).

This article summarizes the main findings of leading scientific evidence on what works best in preventing crime. These findings are based on a project of the Campbell Collaboration Crime and Justice Group (CCJG). The main aim of the project is to advance knowledge on what works to prevent crime for a wide range of interventions, organized around four important domains in criminology: at-risk children, offenders, victims, and high-crime places. Another important aim of the project is to make the field of evidence-based crime prevention, along with the method of systematic review, known to a much wider audience. This includes policy makers, scholars, and researchers, all of whom share an interest in the study of crime prevention and come from a wide range of disciplines, including criminology and criminal justice, sociology, psychology, psychiatry, public health, mental health, social work, economics, and education. The full conclusions of this project are published in Preventing Crime: What Works for Children, Offenders, Victims, and Places (Welsh and Farrington forthcoming).

Evaluation research

When can we have confidence that the reported conclusions of an evaluation of a crime prevention program are valid, whether they suggest that it is effective, ineffective, or, worse yet, harmful? This is a central question for an evidence-based approach to preventing crime.

The randomized controlled experiment is considered the gold standard in evaluation research designs. It is the most convincing method of evaluating crime prevention programs (Farrington 1983; Farrington and Welsh forthcoming). The key feature of randomized experiments is that the experimental and control groups are equated, before the experimental intervention, on all possible extraneous variables. Hence, any subsequent differences between the groups must be attributable to the intervention. Randomization is the only method of assignment that controls for unknown and unmeasured confounders as well as for those that are known and measured. However, the randomized experiment is only the most convincing method of evaluation if it is implemented with full integrity. To the extent that there are implementation problems (e.g., problems of maintaining random assignment, differential attrition, or cross-over between control and experimental conditions), internal validity could be undermined.

Another important feature of the randomized experiment is that a sufficiently large number of units (e.g., people, areas) need to be randomly assigned to ensure that the treatment group is equivalent to the comparison group on all extraneous variables (within the...