Preventing Crime: What Works, What

¶ … Preventing Crime: What Works, What Doesn't, What's Promising" by Lawrence W. Sherman, Denise C. Gottfredson, Doris L. MacKenzie, John Eck, Peter Reuter, and Shawn D. Bushway

According to Sherman, one of the major problems with modern crime prevention programs is that many of them simply do not work. However, society continues to put resources into these programs, largely because they have not been appropriately evaluated, making it possible to determine which ones are successful and which ones are unsuccessful. Not only is this lack of effective evaluation a waste of resources, it also creates the potential for a public backlash against crime prevention programs, because it increases the likelihood that public funds will be used on ineffective programs. While Sherman believes that there is not enough evidence available to conclusively determine the effectiveness of the various programs, he does find that there is sufficient evidence to determine which programs are successful, which are clearly unsuccessful, and which programs seem to be promising. Concentrating funding efforts on programs that have been prove to work, as well as on additional research into the effectiveness of promising strategies, is a necessary step in the implementation of an effective crime-prevention regime.

The article began by explaining that there is a problem with how crime prevention programs are evaluated. By 1996, federal law mandated an independent, scientific review of state and local crime prevention programs, geared at measuring the reduction in juvenile delinquency and gang related crime, reductions in risk factors for criminal behavior, and increases in protective factors shown to reduce the likelihood of criminal behavior. (Sherman, 1998, p.75). The University of Maryland's Department of Criminology and Criminal Justice performed the review, which looked at any program that resulted in less crime than without the practice. Furthermore, the review examined programs that claimed to reduce crime or drug abuse rates. The review classified programs into seven settings: community, family, school, labor markets, places, police, and post-arrest criminal justice agencies. The focus of the review was whether or not there was scientific evidence favoring the types of programs that are eligible for funding, by examining whether they could accomplish their goals.

Sherman acknowledged that a scientific approach to evaluating these programs had strengths and weaknesses. The major weakness was that scientific knowledge could not necessarily be generalized from one population to another, making the results of each study provisional. (Sherman, 1998, p.76). Furthermore, no causal-relationship can ever be conclusively proved in science, merely disproved. However, the scientific method has a major strength, which is that scientific evaluation can provide consistent and objective ways to determine cause and effect. (Sherman, 1998, p.76). There are widely agreed-upon rules for determining the statistics involved in scientific studies, which help researchers determine the validity and reliability of those studies.

To determine whether or not a program was effective, the review examined whether the programs had a demonstrated impact on reducing crime and delinquency. First, the reviewers identified and reviewed reports evaluating the effectiveness of crime prevention programs, by examining impact reports rather than process evaluation reports. Furthermore, the review concentrated on the impact the programs had on crime reduction and did not consider other social issues. The reviewers limited their search mainly to published reports and reports available through the National Institute of Justice, acknowledging that there was a large body of literature about crime prevention that was not included in the report. To rank the effectiveness of the program, the reviewers developed and used the Maryland Scale of Scientific Methods. This scale permitted them to rank studies from 1 to 5 on overall internal validity. The researchers defined the Maryland Scale of Scientific Methods' five levels as follows: level one measured correlation between a program and a single measurement of crime in a point in time; level two reflected an observation of temporal sequence between the program and the crime or risk outcome observed, or, in the alternative, the presence of a control group lacking demonstrated comparability; level three involved a comparison between two comparable groups, one exposed to the program and one unexposed; level four involved comparisons between multiple comparable units with and without the program; level five involved randomization of program subjects, and an analysis of the study with control groups. (Sherman, 1998, p.78) They also used an overall rating system, which considered: control of other variables in the study, measurement error, and statistical issues such as study size. The reviewers recognized four main threats to validity: causal direction; history or other external factors that could have impacted the outcome; chance factors; and selection bias. (Sherman, 1998, p.78).

Based on the above scale, the researchers divided the studied programs into four categories: what works, what doesn't work, what's promising, and what's unknown. The what works category consisted of programs that had "at least two level 3 evaluations with statistical significance tests and the preponderance of all available evidence showing effectiveness." (Sherman, 1998, p. 79). The what doesn't work category consisted of programs that had "at least two level 3 evaluations with statistical significance tests showing ineffectiveness and the preponderance of all available evidence supporting the same conclusion." (Sherman, 1998, p.79). The what's promising category reflected programs that did not have a substantial amount of available evidence, but that had "at least one level 3 evaluation and the preponderance of the remaining evidence." (Sherman, 1998, p. 79). The unknown category was reserved for any program not fitting into the three other categories.

The program found a number of programs that work, though it found no community programs that fit into that category. In families, it found that home visits for infants reduced child abuse and injury to those infants. Preschool and home visits for preschool-aged children also reduced the likelihood those children would be arrested while a juvenile and into early adulthood. Family therapy and parent training were both found to be successful in reducing risk factors for delinquency. In schools, the review found that using school teams or similar strategies could reduce crime and delinquency. Moreover, clarifying expectations and behavioral norms reduces crime, delinquency, and substance abuse. Life skill training can reduce delinquency and substance abuse. Coaching in thinking skills can help high-risk students reduce substance abuse. In labor markets, ex-offender training helps reduce recidivism. In places, nuisance abatement proceedings can help reduce drug dealing and crime. By police, increased patrol of high crime areas can reduce crime in those areas. Monitoring of repeat offenders can reduce the number of crimes recidivist repeat between incarcerations. Arresting domestic abusers reduces domestic violence in some communities. In the post-arrest scenario, incarceration does prevent additional crimes, at least during the time of incarceration and with civilian victims. However, rehabilitation programs that use treatments appropriate to the risk posed by the offender can reduce recidivism rates. Finally, drug treatment during incarceration can reduce the risk of incarceration. (Sherman, 1998, p.80).

While the reviewers found that many programs were effective, they also discovered that many programs are ineffective. In communities, both gun buyback programs and community mobilization of residents in high-crime areas have failed to reduce crime. In families, they discovered that follow-up visits by police officers do not reduce domestic violence, and that arrests of unemployed offenders actually increase the rate of domestic violence in those scenarios. Raids or arrests in high-drug-traffic areas only have a very temporary positive impact on the crime rate in those areas, if they have one at all. Storefront police offices do not prevent crime in the surrounding areas, nor do police newsletters discussing local crime reduce local victimization rates. There are some post-arrest programs that are ineffective as well, including: military-style boot camps, scared straight programs for juvenile offenders, shock probation or parole programs, electronic monitoring of home detainees, intensive supervision while on parole or probation, rehabilitation programs using non-specific counseling, and rural residential programs for juvenile offenders. (Sherman, 1998, p.81-82).

There were several programs that were promising. In communities, gang offender monitoring, community-based mentoring, and community based afterschool recreation programs all showed promise. In families, battered women's shelters reduced the short-term rate of repeat victimization. In schools, grouping students into smaller units for specific programs, training or coaching in thinking skills, using school teams to sustain innovation, and improved classroom management and instructional techniques all showed promise for reducing delinquency or substance abuse. In labor markets, job corps, prison-based vocational educational programs, dispersing inner-city public-housing residents to suburban public housing, and the establishment of enterprise zones all showed promise. In places, adding additional clerks to previously targeted convenience stores, redesigning the layout of retail stores, improved training of bar and tavern staff, the use of metal detectors in schools and airports, the use of sky marshals on airplanes, street closures, target hardening, and site-and-crime specific target analysis seemed to suggest a reduction in crime rates. By the police, proactive arrests of people carrying concealed, proactive breath tests for drunk drivers, community policing, policing with greater respect for offenders, field interrogations of suspicious persons, mailing arrest warrants to domestic violence offenders who flee the scene before the police arrive, and increasing the number of police in an area all show promise for reducing crimes. Post-arrest, the promising programs included: drug courts, drug treatment in jails, intensive supervision and aftercare of juvenile offenders, and the use of fines and other penalties in lieu of incarceration for technical violations. (Sherman, 1998, p. 82-85).

From the results, the reviewers were able to come to several different conclusions. First, they concluded that there is widespread interest in the use of scientific evidence in criminological policy-making. However, they also concluded that "the current development of scientific evidence is inadequate to the task of policymaking." (Sherman, 1998, p.86). They suggest the use of control groups to help increase external validity. They also suggest looking at the geography of crime. Finally, they believe that there is simply a need for a greater number of impact evaluations. All of these recommendations make sense and help explain how the scientific process can help policymakers identify appropriate crime-prevention strategies.

One of the most encouraging things that the research revealed is that family-intervention programs can help reduce the rate of child abuse. At least five studies substantiated the idea that "frequent home visits to infants aged 0-2 by trained nurses and other helpers reduce child abuse and other injuries to infants." (Sherman, 1998, p.79). While the researchers do not go into an explanation of why these visits are promising, it easy to hypothesize several different reasons for their success. First, parents know that their babies are being monitored, and are, therefore, probably going to be less likely to engage in any type of behavior that would cause injury to the infant. However, it is also likely that these new parents receive support and training from the nurses and helpers who are visiting their homes. It is impossible to know with any real certainty how much child abuse stems from poor-parenting skills, but the cyclical nature of child abuse certainly suggests that parenting skills are related to family violence. Child abuse has been definitely linked to stress, therefore having a trained visitor who may be able to provide stress-reduction tips may help decrease abusive behavior. For example, new parents may not realize that a fussy-seeming baby might be suffering from a treatable condition, such as an ear infection or teething, because young babies often do not manifest signs of illness. Moreover, accidental injuries in the household likely stem from parents not knowing how to adequately baby-proof their homes or supervise their infants. One of the things that the researchers neglected to mention was that it would be interesting to follow-up on these studies and see if early intervention in homes reduces the lifelong likelihood of child abuse, or whether the positive impact of the early-childhood visits ends when the visits end.

Another interesting program that works is the mandatory arrest of some domestic violence offenders. Domestic violence advocates have long advocated for the mandatory arrest of domestic violence offenders, because police departments had a long history of treating domestic violence as if it were not a crime. What the research reveals is that arresting domestic violence offenders who are employed or live in neighborhoods where most homes have an employed adult reduces domestic violence. (Sherman, 1998, p.80). This makes sense, because employed individuals face greater consequences from arrest than unemployed individuals face. Therefore, it is not surprising that those results are not the same when the offender is unemployed. In fact, the most interesting aspect about mandatory arrest policies is that arresting unemployed domestic violence offenders actually increases the rate of long-term recidivism. (Sherman, 1998, p.82). These two studies, when viewed together, demonstrate that it is impossible to generalize from one population to another population. They also pose a problem for policymakers. People with unemployed spouses or in neighborhoods with high unemployment rates, while not more likely to be victimized by domestic violence, are frequently more vulnerable to the impact of domestic violence, because they lack the resources to leave the abuser. Moreover, unemployment and poverty are correlated with race in the United States. Therefore, the proper application of this information, if the goal is to reduce overall repeat domestic violence offenses, might mandate a different approach to domestic violence, depending on the neighborhood where the offense occurred and the employment status of the offender. Such an approach would almost certainly violate the 14th Amendment's guarantee of Equal Protection, because it would have a disparate impact based on race, and the science supporting the difference is simply not strong enough to provide the necessary rational basis for such a discriminatory impact. Therefore, it is impossible to make a policy recommendation based on the evidence of even multiple valid studies, unless those studies discuss all of the population types contained within a state or local governmental unit.

Sticking with the issue of domestic violence, what makes it even more difficult for policy makers is the fact that it has been discovered that "home visits by police to couples after domestic violence incidents to provide counseling and to merely warn them or use other alternatives to formal charging" do not reduce recidivism. (Sherman, 1998, p.82). Combined with the evidence from the other two studies, one can see a difficult scenario for lawmakers. Not only can they not rely on a blanket policy of mandatory arrest to drive down repeat occurrences of domestic violence, but they also cannot trust in the fact that simply increasing monitoring of known spousal abusers will result in a reduction of violence. Because the arrest has a definite impact on the abuser, it is clear that arrest sends a message that monitoring fails to send. Unfortunately, that message can either help or hurt the victim, depending on the circumstances of the abuser. However, what those studies make clear is that treating the incident like something less than a serious criminal offense can be very detrimental to the victim. It also suggests that more creative methods, such as using the employment status of the offender as a condition for bail, perhaps by linking it to flight risk, might need to be employed to help reduce future incidents of violence.

One of the other interesting programs that failed to work is the "Scared Straight" programs. These programs involve "bringing minor juvenile offenders to visit maximum security prisons to see the severity of prison conditions." (Sherman, 1998, p.82). What is even more surprising is that these programs may actually increase crime. While the authors do not provide an explanation for this phenomenon, and it seems very counterintuitive, the assumption is based upon the idea that delinquent kids will be frightened by prison life. What the "Scared Straight" programs ignore is that prison life is celebrated by many of the same children who engage in delinquent behavior. Rather than being frightened by prison life, they may feel excited and invigorated to be around cons, who may make up a large percentage of their peer groups. Moreover, the true horrors of prison life, such as sexual abuse, physical assaults, and the demoralizing truth of having no personal freedom are not things that these offenders directly experience in a "Scared Straight" type of program. Instead, these programs must insulate juvenile offenders from some of the harsher aspects of prison life. Furthermore, the reality is that prison is not a real deterrent for people who are actually incarcerated. Instead, incarcerated offenders have very high recidivism rates, which may have been a good preliminary argument against the idea that such programs would be successful.