Click for next page ( 218


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 217
6 The Effectiveness of Police Activities in Reducing Crime, Disorder, and Fear T he public expects the police to accomplish many things, and the re- duction of violent crime is often ranked among the most important. But public disorder, as characterized by public alcohol consumption, prostitution, vagrancy, and drug dealing, may loom larger in any given neighborhood than a concern for serious violence (Schneider, 1980). More recently, both researchers and police have come to view disorder and crime as linked to a community's level of fear about crime and its ability to pre- vent it (Skogan, 1990). In large part because of prominent public concern over these issues, considerable research has been conducted on the effective- ness of the police at reducing three public concerns: crime, disorder, and fear of crime. This body of research is the main concern of this chapter. It is important to note at the outset that the committee recognizes that other mechanisms, both within and outside law enforcement, play a role in addressing these concerns. For instance, an offender might be successfully deterred from committing crime by criminal penalties; a community might diminish its own fear of crime by forming neighborhood patrols; an indi- vidual might simultaneously impose an effective criminal deterrent and re- duce his or her fear by installing a home security system; and so on. Other social and economic forces may also have impact on crime and disorder. Criminologists and economists, for example, have long recognized that de- mographic changes, have important impacts on crime rates. These include shifts in age, family composition, household organization, poverty, and employment patterns (Becker and Landes, 1974; Yamada, 1985; Corman et al., 1987; Glaeser et al., 1996). Crime is importantly affected by drug use 217

OCR for page 217
218 FAIRNESS AND EFFECTIVENESS IN POLICING and availability, access to guns, and alcohol consumption. Private citizens can increase or decrease their risk of victimization by the precautions they choose to adopt and their daily routines and lifestyles. The police are but one part of the formal and informal mechanisms in place to reduce crime, disorder, and fear of crime, and there is no definitive evidence of the relative size of the role they play. Moreover there are significant measures of police work that are not treated fully in this chapter. A vitally important concern since the inception of modern policing has been public acceptance of police use of authority as legitimate (Miller, 1977). We examine what is known about the specific relationship between legitimacy and outcome measures, but the broader problem of the legitimacy of police practices is examined in Chapters 7 and 8. Many other measures of police effectiveness have yet to receive sustained research attention. For example, there has been little or no research on the quality of police as first responders to emergency calls. Do the police pro- vide appropriate medical attention? Do the police calm victims and provide solace? Do the police provide relevant, accurate, and useful information to victims and witnesses? Similarly, there has been little research on the ser- vices provided in police-initiated stops of citizens. Are citizens treated po- litely? Are the citizens informed as to the nature of the stop? How intrusive and inconvenient are the stops? And surprisingly, there is also little research on how the police contribute to criminal justice; are police actions just in both outcome and process? These and other understudied police activities warrant serious and sustained research. In this chapter we review a large of body of research on police using a very specific criterion: How effective are police strategies at reducing crime, disorder, and fear of crime? We begin by discussing how the evidence was evaluated and assessed by the committee. What criteria did we use for dis- tinguishing the value of studies for coming to conclusions about the effec- tiveness of police practices? How did we decide when the evidence was persuasive enough to draw more general statements about specific programs or strategies? We then turn to a series of propositions concerning specific police practices that can be drawn directly from the research literature. Our approach here is to identify what existing studies say about the effects of core police practices. Having summarized the research literature in this way, we conclude the chapter with a more general synthesis of the evidence re- viewed. Are there more general conclusions that can be reached regarding what works in policing from the specific research we review? Are there suggestive patterns that can lead us to identify new and promising direc- tions, even if the present research does not directly test these practices?

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 219 STRENGTH OF THE EVIDENCE There is no hard rule for determining when studies provide more reli- able or valid results, or any clear line to indicate when there is enough evidence to come to an unambiguous conclusion. Nonetheless, social scien- tists generally agree on some basic guidelines for assessing the strength of the evidence available. Perhaps the most widely agreed-on criterion relates to what is often referred to as internal validity (Sherman et al., 2002; Weisburd et al., 2001). Research designs that allow the researcher to make a stronger link between the interventions or programs examined and the outcomes observed are generally considered to provide more valid evidence than designs that provide for a more ambiguous connection between cause and effect. In formal terms, the former designs are considered to have higher internal validity. In reviewing studies, we used internal validity as a primary criterion for assessing the strength of the evidence provided. It is generally agreed that randomized experiments, if successfully imple- mented and sustained, generally provide a higher level of internal validity than do nonexperimental studies (see, e.g., Boruch et al., 2000; Campbell and Boruch, 1975; Cook and Campbell, 1979; Farrington, 1983; Feder et al., 2000; Shadish et al., 2002; Weisburd, 2003). In randomized experi- ments, people or places are randomly assigned to treatment and control or comparison groups (Pocock, 1983). This means that all causes, except treat- ment, can be assumed to be randomly distributed among the groups. Ac- cordingly, if an effect for an intervention is found, the researcher can con- clude with confidence that the cause was the intervention itself and not some other confounding factor. In general in policing, seldom are there groups that can be called true control groups in a randomized study, since they generally receive some type of intervention. Policing experiments mostly compare a new or innovative program or strategy with traditional police practices. Another class of studies, referred to here as quasi-experiments, typi- cally allow for less confidence in making a link between the programs or strategies examined and the outcomes observed (Cook and Campbell, 1979). Quasi-experiments generally fall into three classes. In the first class, the experiment compares an "experimental" group with a control or com- parison group, but the subjects of the study were not randomly assigned to the categories. This results in less confidence than true experimental ap- proaches that the treatment was the actual cause if there is a difference in outcomes. It may have been, but the difference in outcome may be due to some preexisting difference between the treatment and the control groups that has not been taken into account by the researchers. In the second class of quasi-experiments, a long series of observations is made before the treatment, and another long series of observations is made

OCR for page 217
220 FAIRNESS AND EFFECTIVENESS IN POLICING after the treatment. The established before-treatment trend allows research- ers to predict what might have happened without intervention. The differ- ence between what really happened after the treatment and the predicted outcome demonstrates the treatment effect. But these approaches are still subject to the criticism that other confounding factors not accounted for in the design may have caused the observed differences. The third class of quasi-experiments combines the use of a control group with time-series data: this approach provides the strongest conclusions in quasi-experiment research. In theory, these designs are still likely to have a lower level of internal validity than randomized experimental studies. How- ever, some scholars argue that such quasi-experiments can sometimes pro- duce results of the same quality as randomized experiments (see, e.g., Shadish et al., 1993; Lipsey and Wilson, 1993). Others have found that even strongly designed quasi-experiments are still likely to produce less valid outcomes than well executed randomized experiments. In practice it is nec- essary to judge the persuasiveness of quasi-experiments on a case-by-case basis. (Weisburd, Lum and Petrosino, 2001). Finally, studies that rely only on statistical controls--generally termed nonexperimental or observational designs--are often seen to represent the weakest level of confidence (Cook and Campbell, 1979; Sherman et al., 1997). In nonexperimental research, neither researchers nor policy makers intentionally vary treatments to test for outcomes. Rather, researchers ob- serve natural variation in outcomes and examine the relationships between that variation and police practices. For example, when trying to determine if police staffing levels influence crime, researchers might examine the rela- tionship between staffing levels and crime rates across cities. The difficulty with this approach is apparent: there can easily be other factors that influ- ence crime that are not accounted for. To address this concern, researchers attempt to control for these other factors statistically. One issue is that variation in the policy variable of interest (for example, police staffing lev- els) may be correlated with unknown or unmeasured variables that also influence the outcome (crime rates). The estimated effect of the policy vari- able will then be biased because it will be confounded by the effects of the unobserved variables (Kunz and Oxman, 1998). Under some circumstances this problem can be overcome by the use of instrumental-variables estima- tion technique, which exploits the availability of a variable (the instrument) that has a known effect on the policy variable, but no direct effect on the outcome variable (Angrist and Krueger, 2001). For example, Levitt's (1997) study of the effects of police hiring on crime rates takes advantage of the fact that municipal police hiring tends to follow the electoral cycle. Increases in the size of the police forces are disproportionately concentrated in may- oral and gubernatorial election years, but, he asserts, there is little reason to believe that the election-cycle has any direct effect on crime rates. The elec-

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 221 toral cycle then serves as an "instrument" for police hiring. However, it is generally agreed that causes unknown or unmeasured by the researcher are likely to be a serious threat to the internal validity of these observational studies (Pedhazur, 1982; Kunz and Oxman, 1998; Feder and Boruch, 2000). Importantly, despite the limitations of such research, sometimes nonexperi- mental studies are the only investigation method possible in sensitive areas of crime and justice. Some scholars have argued that when the theory un- derlying a research problem is well developed and the quality of data are very high, it is possible to develop statistical models that provide highly valid results (Heckman and Smith, 1995). However, in the view of the com- mittee, theories of police effectiveness and the data generally available for research on policing do not warrant such confidence. In our evaluation, we rely strongly on these general assessments of the ability of research to make statements of high internal validity regarding the practices evaluated. However, we also recognize that other criteria are im- portant in assessing the strength of research. While it is generally recog- nized that randomized experiments have higher internal validity than nonrandomized studies, a number of scholars have suggested that the re- sults of randomized field experiments can be compromised by the difficulty of implementing such designs (Clarke and Cornish, 1972; Pawson and Tilly, 1997). In practice it can be difficult to sustain randomized trials in the field, and the integrity of randomized experiments must also be assessed on a case-by-case basis. Accordingly, in assessing the evidence, we also took into account the quality of both the measures identified and the implementation of the research design. The external validity of a study refers to our ability to generalize from its findings to other settings, times and even conditions. Even if a researcher can make a very strong link between the practices examined in a study and their influence on crime, disorder, or fear, if one cannot generalize from that study to the more general problems and practices of policing, then the findings will not be very useful. Randomized field experiments may be prob- lematic in this respect, since they are conducted under artificial circum- stances, with the orders to the police and the data collection procedure dictated to some extent by the experimental design rather than by standard operating procedures. On the other hand, the process of instituting new procedures and management routines may mimic the reform measures that are being evaluated, if they are in turn to be implemented in practice. An experiment evaluating a particular intervention is often limited with respect to time and place. For these and other reasons our confidence with which an impact assessment's findings can be extrapolated to similar programs or from the program as tested to the program as implemented must be exam- ined. Moreover, it is generally agreed by social scientists that caution should be used in drawing strong policy conclusions from a single study, no matter

OCR for page 217
222 FAIRNESS AND EFFECTIVENESS IN POLICING how well designed (Weisburd and Taxman, 2000; Manski, 2003). For these reasons we took into account additional factors related to our ability to generalize from study findings in drawing our conclusions. For example, when the committee found multiple studies with similar conclusions, we call them a "consistent" body of evidence. The strength of this consistent evidence was judged primarily on the quality of the majority of the studies. When there were contradictory findings among studies, we also assessed the quality of studies involved in order to be able to characterize the re- search. When we found a body of weakly designed studies with contradic- tory results, or strongly designed studies with contradictory results, we call that body of research conclusions "ambiguous." That is, when the strength of studies supporting opposite conclusions was judged comparable, no defi- nite conclusion could be drawn. And when we found a group of strong studies contradicted by a group of weak studies, the committee favored the conclusions of the better quality studies, with appropriate qualifications. It is also important to note that although the committee makes judg- ments about police effectiveness, the strength and scope of these judgments were limited by the extent and rigor of policing research more generally. For example, it is common in research in the social sciences today to com- pute standardized effect coefficients that allow assessment of the absolute size of the impacts of interventions and comparison of impact of interven- tions across studies (Wilson and Lipsey, 2001). Very few of the studies we reviewed included standardized effect sizes. Moreover, it was beyond the scope of the committee's work to independently construct these estimates, especially since the reporting of results was many times not detailed enough to allow independent and accurate computation of such statistics. The com- mittee sought to gain a general assessment of the strength of the findings reported in each of these studies based on the evidence presented by the original investigators and the methods used. It is also important to point out that evaluations of police tactics to reduce crime and disorder often involve comparisons of areas receiving vary- ing levels of the tactic in question. Evaluations based on spatial compari- sons must account for complex spatial dependencies. Two well-recognized spatial dependencies are displacement (Cornish and Clarke 1986) and dif- fusion of crime prevention benefits (Clarke and Weisburd 1994). Substan- tial progress has been made in consistently measuring these phenomena over the past two decades. However, many of the evaluations reviewed either did not describe findings regarding spatial displacement or diffusion, or it appeared that measurements of displacement and diffusion were not central to the overall evaluation. Displacement and diffusion relate to spatial dependencies introduced by the police tactic. Of equal concern should be spatial dependencies that confound presumed causal relationships between tactics and outcomes. One

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 223 set of confounders comes from the nested nature of crime: crime events are clustered at specific locations--for example, addresses or street blocks (Eck and Weisburd, 1995), and such locations tend to cluster in larger areas (Sherman, Gartin, and Buerger, 1989). While such nesting of events sug- gests the importance of taking into account the hierarchical nature of the effects of policing, police research has seldom taken such concerns into account (see Kelling and Sousa, 2001 for a recent attempt to develop hierar- chical models for evaluating a police strategy). A related phenomena is the tendency for nearby locations to have correlated characteristics. Just as fail- ure to account for temporal autocorrelation in time-series analysis can re- sult in biased results, failure to account for spatial autocorrelation in nonexperimental and quasi-experimental evaluations of police tactics may also bias the results (Odland, 1988). Again, police research today has gen- erally not addressed such potential biases. Once again, we present our evaluation in the form of propositions. These propositions summarize what is known regarding police effectiveness in the areas that we have reviewed. They represent both recurring themes in policing and relatively recent but nevertheless influential changes in police practice or philosophy. STANDARD MODEL OF POLICE PRACTICES Proposition 1: The standard model of policing has relied on the uni- form provision of police resources intended to prevent crime and disor- der across a wide array of crimes and across all parts of the jurisdic- tions that police serve. Despite the continued reliance of many police agencies on these standard practices, the evidence the committee re- viewed suggests that such approaches are generally not the most effec- tive strategy for controlling crime and disorder or reducing fear of crime. Over the past two decades, there has been increasing criticism of what has come to be considered the standard model of police practices (Goldstein, 1990; Visher and Weisburd, 1998). This model relies generally on a "one size fits all" application of reactive strategies to suppress crime, in contrast to more customized and proactive strategies. The standard model also em- phasizes the role of arrests and the threat of punishment in achieving this objective, with less emphasis on other capabilities of the police. The stan- dard model of policing has assumed that generic strategies for crime reduc- tion can be applied throughout a jurisdiction, regardless of the level of crime, the nature of crime, or other possible variations. Because the model is fo- cused on providing a generalized model of police service, it has often been criticized as focused more on the means of policing or the resources that

OCR for page 217
224 FAIRNESS AND EFFECTIVENESS IN POLICING police bring to bear (i.e., providing the generalized strategy throughout a jurisdiction) than on the effectiveness of policing in reducing crime, disor- der, or fear (Goldstein, 1979). This model has also been criticized because of its almost sole reliance on the traditional law enforcement powers of police in preventing crime (Moore et al., 1992). We identified five general types of strategies that have been prominent in the standard model of policing and have been the focus of systematic research over the last three decades: 1. Increasing the size of police agencies. 2. Random patrol across all parts of the community. 3. Rapid response to calls for service. 4. Generally applied follow-up investigations. 5. Generally applied intensive enforcement and arrest policies. Increasing the Size of Police Agencies There has long been a presumption among politicians and the public that more police will mean more safety in the community. However, re- search in this area is ambiguous and the evidence is often drawn from stud- ies that use quasi-experimental or observational methods. Moreover, it is generally very difficult to identify what the added police actually did in such studies once they were brought into the community. This is particularly the case when a long time series is employed, because police organization, tech- nology, and even their core missions may have changed dramatically over time and these may be confounded with re-sizings of police forces. In this sense, it is difficult to distinguish the circumstances under which the addi- tion of police is likely to provide benefit. Research on the number of police employed shows two different results with regard to crime (disorder has not been studied). Evidence from case studies in which police have suddenly left duty (e.g., police strikes) shows that the absence of police is likely to lead to an increase in crime (Sherman, 1997; Sherman and Eck, 2002). While these studies are generally not very strong in their design, their conclusions are consistent and appear persua- sive. To completely abandon police service in the community would be likely to result in more crime. One can logically conclude from these case studies that there is an "absolute benefit" to the creation of police agencies or the introduction of police to areas that have not been policed previously. Aside from this indirect evidence that the introduction of police in situ- ations in which no police are found can reduce crime, does research support the idea that adding officers to an existing police force is effective? This might be called the marginal effectiveness of policing that is produced by

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 225 increasing the relative number of police in an agency. The results here are ambiguous, and it is difficult to reach an overall conclusion. A series of studies involving comparisons of police agencies with more officers than others as well as studies of the effects of variation in police strength over time suggest that normal variation in police strength has no influence on crime (Van Tulder, 1992; Niskanen, 1994; Chamlin and Langworthy, 1996; Eck and Maguire, 2000). Nonetheless, two recent studies using sophisti- cated econometric methods conclude that marginal increases in the number of police are related to decreases in crime rates (Marvell and Moody, 1996; Levitt, 1997). The Levitt study is of particular interest. It attempts to over- come the statistical problem of reverse causation (that crime rates influence police hiring as well as the reverse) by employing an instrumental-variables estimation technique. Levitt documents the fact that police hiring is concen- trated in mayoral and gubernatorial election years, and asserts that the elec- tion cycle is unlikely to affect crime rates directly. Using a large sample of municipalities over a number of years, he estimates the effect on crime rates of that portion of variability in police staffing that is associated with the election cycles. His estimates indicate that the effect is negative, suggesting that at the margin the police are effective in controlling crime. The commit- tee views these latter studies as representing important efforts to advance statistical methods of analyzing the relationship between police strength and crime rates. However, while recognizing the elegance of the statistical designs employed, in our view these nonexperimental designs are still likely to be confounded with many unmeasured or indeed unknown causes of variability in crime rates. For example, these studies generally do not take into account what the police do. In policing it is common for police agencies to increase and rede- ploy officer strength as they also change policing strategies. New York City in the 1990s is a recent example of this (McDonald, 2001). Changes in mayoral administrations presaged by the electoral cycles employed in some econometric studies can bring new police chiefs, new management policies, changes in deployment patterns, the adoption of new tactics and programs, and new commitments regarding levels of police service, all of which almost inevitably seem to require more police officers. As a result, shifts in their numbers can covary with unmeasured changes in what they actually do to attack crime. As we show later, there is also evidence that some policing strategies are more effective than others. Accordingly, it may be that the effects of police strength observed in such studies are confounded with the unmeasured impacts of the effects of changes in the strategies of policing. Such potential confounding makes it difficult to draw strong policy conclu- sions from these and other studies of the relationship between police strength and crime rates.

OCR for page 217
226 FAIRNESS AND EFFECTIVENESS IN POLICING Random Patrol Across All Parts of the Community Random preventive patrol across police jurisdictions has continued to be one of the most enduring of standard police practices. Despite the con- tinued use of random preventive patrol by many police agencies, the evi- dence supporting this practice is very weak and the studies reviewed by the committee are more than a quarter-century old. Two studies, both using weaker quasi-experimental designs, suggest that random preventive patrol can have an impact on crime (Dahmann, 1975; Press, 1971). However, a much larger scale and more persuasive evaluation of preventive patrol in Kansas City found that the standard practice of preventive patrol does not reduce crime, disorder, or fear of crime (Kelling et al., 1974). The Kansas City Preventive Patrol Experiment increased preventive patrol in five beat areas in the city, reduced it in five others, and left patrol at standard levels in still five others. The study found no significant differ- ences among the three types of beats with regard to crime, disorder, or fear, leading the investigators to conclude that there is no evidence that preven- tive patrol significantly influences any of these core measures of the out- comes of policing. However, while this is a landmark study, the validity of its conclusions has been challenged because of methodological flaws that relate to the allocation of study beats, the measurement of crime and disor- der, and the actual number of police cars on patrol (Minneapolis Medical Research Foundation, 1976; Larson and Cohn, 1985; Sherman and Weisburd, 1995). These problems illustrate the general difficulty of sustain- ing an experimental intervention in the field, especially as in this case where some police were required to change their routines in ways that they may have thought inappropriate. Rapid Response to Calls for Service A third component of the standard model of policing, rapid response to emergency calls for service, has also not been shown to reduce crime or even lead to increased chances of arrest in most situations. The crime reduc- tion assumption behind rapid response is that if the police get to crime scenes rapidly they will apprehend offenders, thus providing a general de- terrent against crime; there are no studies of the effects of this strategy on disorder or fear of crime. In principle, one can distinguish between absolute and marginal effects of rapid response, as is the case with adding police. There are no documented studies of the absolute effect of rapid response: that is, the shifting from a situation with rapid response to one of no re- sponse or long-delayed response. The best evidence about the marginal effectiveness of rapid response comes from two studies conducted in the late 1970s (Kansas City Police Department, 1977; Spelman and Brown, 1981). Evidence from five cities

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 227 examined in these two studies consistently shows that most crimes (about 75 percent at the time of the studies) are discovered some time after they have been committed. Accordingly, offenders in such cases have had plenty of time to escape. For the minority of crimes in which the offender and the victim have some type of contact, citizen delay in calling the police blunts whatever effect a marginal improvement in response time might provide. These studies show that, on average, citizens delayed calling the police for about five minutes, so a slight improvement in how fast the police arrive at the crime scene is unlikely to make much difference in the probability of apprehending the suspect. Contrary to the view that rapid response on the part of the police can increase police effectiveness, both of these studies point to the importance of creating change in the activities of citizens. The greatest benefit in terms of effectiveness is seen to come from increasing the speed at which citizens call the police. There have been no direct studies of the effects of response time on crime rates. It is possible that a slight decline in police response time might be detectable by the average offender, and the average offender might, as a consequence, perceive an increase in apprehension risk sufficient to serve as a deterrent. This seems to be a remote possibility, however, given the fact that offenders often have limited or sometimes even invalid information on police activity. It is important to distinguish between the effectiveness of rapid response as a general policy to reduce crime and the effectiveness of rapid response to clear emergencies in which offenders are present or citizens are injured. In the former, the question is whether crime for the jurisdiction is reduced, and the evidence suggests that this is unlikely. In the latter case, the ques- tion is whether improving rapid responses to crimes in which an offender is present or there are injuries results in more arrests and improved assistance to citizens in crisis. The committee found no studies shedding light on the effectiveness of focused rapid response. Generally Applied Follow-Up Investigations Studies consistently show that most property crimes and many violent crimes are unsolved (Greenwood et al., 1975; Eck, 1983; Brandl and Frank, 1994; Horvath et al., 2001). Another example of the standard model of policing assumes that general improvements in police investigations to im- prove the rate at which crimes are solved will prevent future crime. Once a serious crime has been reported to the police, and if officers who first re- spond to the incident have arrested no one, there is often a follow-up inves- tigation. This model suggests that such investigations increase the risk of apprehension for offenders, thus providing a general deterrent (Eck, 1983). Alternatively, the capture and sanctioning of offenders stemming from in- vestigations might reduce crime by imprisoning active offenders.

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 241 not target specific individuals simply because of their racial or ethnic back- grounds (Harris, 2002, 1999, 1997). Of course, police have always profiled individuals who they believe are suspicious, and such profiling is a well- known part of police culture. Nonetheless, only recently has profiling be- come an important focus of police research. One area in which the targeting of specific offenders has been studied by police scholars is in what are often termed "repeat offender" programs. Here, the criteria for focused police intervention are not related to the social or demographic characteristics of offenders, but rather to their criminal records. Police have demonstrated substantial ability to apprehend known repeat offenders. Randomized controlled trials in Washington, DC (Martin and Sherman, 1986), and in Phoenix, AZ (Abrahamse et al., 1991), of re- peat offender units found that covert investigation of high-risk, previously convicted offenders has a high yield in arrests and incarceration per officer hour, relative to other investments of police resources. Thus, it appears that investigation effectiveness is enhanced to the degree that it focuses on a relatively few high-rate offenders rather than spreading resources over a large number of crimes with few clues. It is important to note, however, that these evaluations have examined the apprehension effectiveness of re- peat offender programs. These are only indirect examinations of their effect on reducing crime, and conclusions about their crime reduction effective- ness rely on ancillary assumptions about the effectiveness of selective incar- ceration and incapacitation. A recent and promising application of the repeat offender approach was developed in the Boston Ceasefire project. In this instance, the problem was the high level of firearm related killings of black youth. Extensive analy- sis led to a multiagency and community project designed to communicate directly to youth gangs that firearm violence would not be tolerated (Ken- nedy, Braga, and Piehl, 1997). This is sometimes referred to as the "pulling levers" strategy (Kennedy 1997). A time-series analysis of young adult and youth homicide indicated a 63 percent reduction in these killings as well as declines in other gun related events (Kennedy et al., 2001). Statistical con- trols were used to control for citywide violence trends, employment, and the number of youth in the population. Comparisons with other large cities during the same time were used to control for national trends. Although other cities have implemented similar pulling-levers projects, the results are descriptive rather than evaluative. Another method for identifying and apprehending repeat offenders are "anti-fencing" or property sting operations. These typically involved un- dercover police posing as receivers of stolen goods (Pennell, 1979; Criminal Conspiracies Division, 1979; Weiner, Chelst and Hart, 1984). Did these programs apprehend repeat offenders? And did this result in reductions in property crime? Although a number of evaluations were conducted of these

OCR for page 217
242 FAIRNESS AND EFFECTIVENESS IN POLICING operations, most employed weak research designs, thus making it difficult to provide definitive answers to these questions. There seems to be a con- sensus among the evaluations that antifencing operations of this type do apprehend more older and criminally active offenders than are arrested in more traditional law enforcement practices (Criminal Conspiracies Divi- sion, 1979; Pennell, 1979; Weiner et al., 1983). However, as noted, the empirical support for this consensus is weak. Evidence that the effect of arrest policies varies according to the social and demographic characteristics of offenders is drawn from nonexperi- mental reanalyses of the domestic violence experiments discussed earlier under Proposition 1. In Milwaukee, WI, and Omaha, NE (Sherman and Smith, 1992), as well as Miami-Dade County, FL (Pate, Hamilton, and Annan, 1991) and Colorado Springs, CO (Berk et al., 1992), arrest effects varied with the employment status of the suspect. In all four correlational tests for interaction in the randomized experiments, suspects who were em- ployed at baseline (the time of random assignment) were more deterred by arrest than unemployed suspects. In three of the cities (Milwaukee, Omaha, and Miami-Dade), the unemployed suspects became substantially more likely to reoffend during the follow-up period if they were arrested than if they were merely warned. Marciniak (1994) found the same interaction effect across neighborhoods in Milwaukee, where arrest had deterrent ef- fects in areas of low unemployment but increased repeat domestic violence in areas of medium to high unemployment. However, Garner and Maxwell (2000), in an examination of victimiza- tion interviews, found that combining either employment or marriage with arrest did not reduce significantly the quantity of victimization any more than did arrest by itself. Therefore, these results call into question the valid- ity of the stakes in conformity theses. They suggest that the results reported earlier using the official data may in large part reflect victims' willingness to report the incidents and the police officer's willingness to record the inci- dent. In other words, victims may be less willing to call the police and, if they do call, the police may also be less willing to record the incident when the suspect is employed. They both may independently choose to do this because they both understand the negative consequences that arrest may have on continued employment. Overall, these studies suggest that there may be an important interac- tion between the nature of offenders and the effects of police practices. Nonetheless, the committee notes that even if this relationship were to be confirmed directly through experimental research, it is unlikely that it would lead to significant policy changes in policing since larger constitutional ques- tions remain. It is unlikely that any jurisdiction would be able to vary police arrest practices according to the social and demographic characteristics of individual citizens. Although it may be more acceptable to vary such prac-

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 243 tices across jurisdictions with differing types of populations, important ques- tions regarding fairness and constitutionality are likely to arise. With these caveats in mind, findings in these studies do provide a more textured under- standing of the effects of arrest policies. PROBLEM-ORIENTED POLICING Proposition 4: Problem-oriented policing emerged in the 1990s as a central police strategy for solving crime and disorder problems. There is a growing body of research evidence that problem-oriented policing is an effective approach. As we have already noted, the standard model of policing came under increasing criticism in the 1970s and 1980s (see, e.g., Goldstein, 1979; Visher and Weisburd, 1998). Herman Goldstein proposed one response to the standard model in 1979, calling it "problem-oriented policing." Gold- stein stated that police were focusing so much attention on internal man- agement concerns and standard law enforcement that they had largely ig- nored the objectives, or ends, of policing. His proposal challenged the one size fits all approach of standard models of policing, replacing it with an approach focused on specific problems, and looked to the development of tailor-made police practices to address such problems. Importantly, his ap- proach also demanded that the police focus more attention on the ends of policing. In the standard model, the necessity of providing a common type of service across broad areas or problems naturally led police to a concern with monitoring whether that service was actually being delivered. In this process, the goals of policing were often given a secondary place in police management. In the problem-oriented model, the results of policing are placed very much at the center of police efforts. In the problem-oriented policing model, police are expected to under- take systematic analysis of community problems, engage in broad searches for effective solutions, and evaluate the results of their efforts. Problems are patterns of events that members of the public expect the police to address. Problems have both a behavioral and an environmental component (Eck and Clarke, 2003). The behavioral component describes how the motives of the parties involved (offenders, victims, and third parties) and the parties interact. The environmental component defines the type of place in which the problem is located. Consequently, problems are highly specific. In a problem-oriented approach, law enforcement is one of many possible means for reducing problems, and this tactic will vary in appropriateness, depend- ing on the specific characteristics of each problem. Examples of problems include thefts from vehicles in downtown parking lots, aggressive panhan-

OCR for page 217
244 FAIRNESS AND EFFECTIVENESS IN POLICING dling in commercial areas, robberies of gas stations, speeding in residential neighborhoods, and club drug use at teen-oriented entertainment businesses. We have already noted in Proposition 2 that there is often overlap be- tween community-oriented and problem-oriented policing, in the sense that both approaches expand the police function beyond the traditional law en- forcement powers of the police. In addition, we have also noted the overlap in Proposition 3 between focused policing and problem solving. Problem- oriented policing extends focused policing beyond geography or specific offenders to include the analysis of community problems. In this sense, prob- lem-oriented policing includes elements of both trends in policing innova- tion that we reviewed above. Problem-oriented policing was originally directed at department-wide analysis of persistent problems. Police agencies expanded the scope of prob- lem-oriented policing to include beat-level problem-solving efforts by pa- trol officers and detectives as part of their normal duties (Cordner, 1986; Eck and Spelman, 1987). This aspect of problem-oriented policing has re- ceived the most attention in police research. Furthermore, various forms of beat-level problem solving have been incorporated into community policing programs. Problem-oriented policing refers to the overall organizational direction of a police agency that focuses its efforts on addressing problems (Goldstein, 1990). In this context, a problem is a recurring set of related harmful events in a community that members of the public expect the police to address (Eck and Spelman, 1987; Goldstein, 1990). Problem solving per se refers to the actual handling of a specific problem. Problem solving can be employed by any agency on an ad hoc basis, whether that agency adopts a compre- hensive organization-wide problem-oriented policing approach or not. Con- sequently, far more agencies employ problem solving than might be defined as problem-oriented. In some ways, problem-oriented policing is untestable. The claim that policing should direct its attention to problems is a normative proposition. One could argue that policing should focus solely on emergency response, or should focus solely on law enforcement. No amount of evidence could test either of these assertions. If one accepts that the police should focus on problems, it is hard to argue against systematic and empirical analysis of problems before selecting the intervention that best fits the data. In essence, a problem-oriented approach calls for the application of the scientific method to policing. Nevertheless, it is important to answer two questions: Does the application of problem solving actually reduce problems? Is this application more effective than other police activities? Research is consistently supportive of the capability of problem solving to reduce crime and disorder. However, as in other areas of our review, we raise the caution that many studies to date have used relatively weak re-

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 245 search designs. There are a number of quasi-experiments going back to the mid-1980s that consistently demonstrate that problem solving can reduce violent and property crime (Eck and Spelman, 1987), firearm-related youth homicide (Kennedy et al., 2001), fear of crime (Cordner, 1986), and vari- ous forms of disorder, including prostitution and drug dealing (Eck and Spelman, 1987; Hope, 1994; Capowich and Roehl, 1994). For example, a quasi-experiment in Jersey City, NJ, public housing complexes (Mazerolle et al., 2000) found that police problem-solving activities caused measurable declines in reported violence and property crime, although the results var- ied across the six housing complexes studied. In another example, Clarke and Goldstein (2002) report a reduction in thefts of appliances from new home construction sites following careful analysis of this problem by the Charlotte-Mecklenburg, NC, Police Department and the implementation of changes in building practices by construction firms. The two experimental evaluations of applications of problem solving in hot spots, reviewed above, also suggest its effectiveness in reducing crime and disorder. In a randomized trial with Jersey City violent crime hot spots, Braga et al. (1999) report reductions in property and violent crime in the treatment locations. While this study tested problem-oriented approaches, it is important to note that focused police attention was brought only to the experimental locations. Accordingly, it is difficult to distinguish between the effects of bringing focused attention to hot spots and that of such fo- cused efforts being developed using a problem-oriented approach. In the Jersey City Drug Market Analysis Experiment (Weisburd and Green, 1995a), more direct support can be found for the application of problem- solving approaches as opposed to standard models of policing. In that study, a similar number of narcotics detectives was assigned to treatment and con- trol hot spots. However, Weisburd and Green compared the effectiveness of unsystematic, arrest-oriented enforcement based on ad hoc target selection (the control group) with a treatment strategy involving analysis of assigned drug hot spots, followed by site-specific enforcement and collaboration with landlords and local government regulatory agencies, and concluding with monitoring and maintenance for up to a week following the intervention. Compared with the control drug hot spots, the treatment drug hot spots fared better with regard to disorder, but there were no significant differ- ences between the two groups with regard to violent crime. Evidence of the effectiveness of situational and opportunity-blocking strategies, while not necessarily police based, provides indirect evidence sup- porting the effectiveness of problem solving in reducing crime and disorder. Problem-oriented policing has been linked to routine activity, rational choice perspectives, and situational crime prevention (Eck and Spelman, 1987; Clarke, 1992b). Recent review of prevention programs designed to block crime and disorder opportunities in small places noted that most of the

OCR for page 217
246 FAIRNESS AND EFFECTIVENESS IN POLICING studies report reductions in target crime and disorder events (Eck, 1997; Poyner, 1981; Weisburd, 1996). Furthermore, many of these efforts were the results of police problem-solving efforts. We note that many of the stud- ies reviewed employed relatively weak designs (Clarke, 1997; Eck, 1997; Weisburd, 1997). One example of a methodologically strong investigation into the way in which the actions of the public can influence the effectiveness of police is research on the deterrent effects of Lojack, a hidden radio-transmitter de- vice used for retrieving stolen vehicles (Ayers and Levitt, 1998). The au- thors found an association between increases in the fraction of Lojack- equipped vehicles--an unobservable victim precaution--and a substantial decline in auto theft, without any displacement of crime to other crime categories. They estimate the marginal social benefit of Lojack installation at 15 times greater than its social cost, concluding that Lojack affects auto theft even at low market penetration rates (p. 74). This is a quasi-experi- mental study with very high internal validity, used in a situation in which a randomized experiment would be impossible because the crime reduction benefits accrue to the metropolitan area, not the individual users of Lojack. Furthermore, it has a very high level of generalizability. CONCLUSION We have reviewed the literature of research on the effectiveness of po- lice practices in controlling crime, disorder and fear in the context of four broad propositions: Proposition 1: The standard model of policing has relied on the uniform provision of police resources intended to prevent crime and disorder across a wide array of crimes and across all parts of the jurisdictions that police serve. Despite the continued reliance of many police agencies on these stan- dard practices, the evidence the committee reviewed suggests that such ap- proaches are generally not the most effective strategy for controlling crime and disorder or reducing fear of crime. Proposition 2: Over the past two decades there has been a major investment on the part of the police and the public in community policing. Because community policing involves so many different tactics, its effect as a general strategy cannot be directly evaluated. Some community policing strategies appear to reduce crime, disorder, or fear of crime. Many others have not been found to be effective when evaluated. Proposition 3: There has been increasing interest over the past two decades in police practices that target very specific types of crimes, criminals, and

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 247 crime places. In particular, policing crime hot spots has become a common police strategy for reducing crime and disorder problems. A strong body of evidence suggests that taking a focused geographic approach to crime prob- lems can increase the effectiveness of policing. There is no consistent evi- dence regarding the targeting of specific types of offenders. Proposition 4: Problem-oriented policing emerged in the 1990s as a central police strategy for solving crime and disorder problems. There is a growing body of research evidence that problem-oriented policing is an effective approach. It is important to note that there is much about police practice and its effectiveness that is still unknown. While our review documents a substan- tial amount of research over the past four decades, it also illustrates the fact that many established police practices have not been carefully evalu- ated. Even in the case of programs that have received broad national atten- tion, such as community-oriented policing, there is often little research evidence about what works and under what circumstances. One reason for the lack of evidence is the complexity and ambiguity of police strategies. But even in the case of approaches that are more clearly defined, such as problem-oriented policing, our review suggests a need for more carefully designed studies. It is also important to state again, as noted at the beginning of this chapter, that a century of criminological research has documented the pow- erful impact of a long list of social and economic factors on crime. These include some of the most fundamental aspects of our society. Their influ- ence extends far beyond crime, and they are mainly beyond the reach of the police. A thoroughgoing research agenda on crime and policing would en- deavor to identify the role of the police among these factors, but that ques- tion lies beyond the scope of this report. To draw useful lessons from 20 years of police effectiveness research the committee had to organize these strategies in a meaningful way. Figure 6-1 depicts the relationships of four somewhat overlapping policing strate- gies. The figure has two dimensions. The first dimension represents the content of the practices employed. Strategies that rely primarily on tradi- tional law enforcement are low on this dimension. Strategies that expand the "toolbox" of policing (see Taylor, 2001), whether in partnerships with the community or through the use such strategies as civil remedies, are high on this dimension. The second dimension represents the extent of focus or targeting of police activities. Strategies that are focused on specific places or that are tailor made to respond to specific types of problems, are high on

OCR for page 217
248 FAIRNESS AND EFFECTIVENESS IN POLICING array Wide Problem-oriented Community Policing Policing Approaches of Diversity Standard Focused Policing enforcement Model law Mostly low high Level of Focus FIGURE 6-1 Dimensions of policing strategies. this dimension. Strategies that are generalized and applied uniformly across places or offenders score low on this dimension. The four strategies are shown in their relative position. Although one can describe these strategies as clearly separable ideal types, in practice there are no clear boundaries among them. The committee has noted, for example, that community-ori- ented policing often incorporates elements of problem-oriented policing, and that problem-oriented policing and focused policing share important features. The standard model provides a common set of services throughout a jurisdiction. Police agencies using the standard model employ a limited range of approaches, overwhelmingly oriented toward enforcement, and make relatively little use of institutions outside policing (with the notable excep- tion of other parts of the criminal justice system). Community policing makes greater use of institutions outside policing, particularly community groups and private organizations. Although law enforcement powers still are an important tool of community-oriented policing agencies, such agen- cies can employ a greater range of tools to address crime and disorder. When community policing is employed without problem solving, it too pro- vides a common set of services throughout a jurisdiction. The services pro-

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 249 vided by focused policing--generally incorporating hot spots and repeat offender strategies--vary considerably by geography or type of offender. Like the standard model, focused policing tends to rely almost exclusively on law enforcement. Unlike both the standard model and community polic- ing, the application of police powers is tightly focused. Problem-oriented policing incorporates the precise resource targeting of focused policing with the diversity of approaches in community policing. So a problem-oriented policing agency selects from a wide variety of approaches to address very specific problems. The committee used the framework depicted in Figure 6-1 to synthesize its findings. This synthesis is found in Table 6-1. It is the committee's judg- ment, based on our review, that the standard model of policing, represented in the lower left corner of the table, which draws on generally applied tac- tics and uses primarily the law enforcement powers of the police, has gener- ally not been found to be effective either in reducing crime or disorder or TABLE 6-1 Synthesis of the Committee's Findings on Police Effectiveness Research Police Strategies That... Are Unfocused Are Focused Apply a diverse array of Little or no evidence of Moderate evidence of approaches, including law effectiveness effectiveness enforcement sanctions. --Impersonal community --Problem-oriented policing policing, e.g., newsletters Strong evidence of Weak to moderate evidence effectiveness --Personal contacts in --Problem solving in hot spots community policing --Respectful police-citizen contacts --Improving legitimacy of police --Foot patrols (fear reduction) Rely almost exclusively on Little or no evidence of Inconsistent or weak evidence law-enforcement sanctions. effectiveness of effectiveness --Adding more police --Repeat offender --General patrol investigations --Rapid response Moderate to strong evidence --Follow-up investigations of effectiveness --Undifferentiated arrest --Focused intensive for domestic violence enforcement --Hot-spots patrols

OCR for page 217
250 FAIRNESS AND EFFECTIVENESS IN POLICING decreasing citizen fear. Whether the strategy examined was generalized pre- ventive patrol, efforts to reduce response time to citizen calls, increases in numbers of police officers, or the introduction of generalized follow-up investigations or undifferentiated intensive enforcement activities, studies fail to show consistent or meaningful crime or disorder prevention benefits or evidence of reductions in citizen fear of crime. Given the widespread and continued use of the standard model of police practice in policing in the United States, these findings have significant policy implications. In contrast to these findings, there is evidence of police effectiveness in each of the remaining cells of the table. In the cell that represents focused policing, we found promising evidence regarding the effects of arrest target- ing specific types of people committing specific types of offenses. We found even stronger evidence regarding the use of traditional enforcement strate- gies that are targeted at specific places. Indeed, studies that focused police resources on crime hot spots provide the strongest collective evidence of police effectiveness that is now available. On the basis of a series of ran- domized experimental studies, we conclude that the practice described as hot-spots policing is effective in reducing crime and disorder and can achieve these reductions without significant displacement of crime control benefits. Indeed, the research evidence suggests that the diffusion of crime control benefits to areas surrounding treated hot spots is stronger than any dis- placement outcome. The two remaining cells of the table indicate the promise of new direc- tions for policing in the United States. The choice of police tactics is of course a normative concern that must be considered not only in relation- ship to what is effective but also to the effects it has on communities. The choice of police tactics must also be assessed in light of their economic cost to the community. In this context, it is often argued that the law enforce- ment powers of the police, found mostly in their power to arrest, are costly both in human and economic terms and should be limited when possible (Goldstein, 1979). While substantive research on the costs of police strate- gies is not presently available (for an exception, see Levitt, 1997), we dis- cuss in the next chapter some of the normative concerns surrounding tradi- tional law enforcement practices. Our findings suggest that other strategies beyond law enforcement can be effective in controlling crime and disorder and reducing fear. It is particularly instructive that specific community policing strategies may be effective. Importantly, these strategies, while consistent with the standard model of policing in their application to a general class of prob- lems, contrast strongly with that model in their reliance on nonenforce- ment-related approaches and community crime control. The evidence base here does not allow for definitive conclusions. Nonetheless, the research available suggests that when the police partner more generally with the

OCR for page 217
THE EFFECTIVENESS OF POLICE ACTIVITIES 251 public, levels of citizen fear will decline. Moreover, when the police are able to gain wider legitimacy among citizens and offenders, nonexperimental evidence suggests that the likelihood of offending will be reduced. These promising themes form a primary focus of discussion in Chapter 8. There is greater and more consistent evidence, although it is based pri- marily on nonexperimental studies, that focused strategies drawing on a wide array of nonlaw enforcement tactics can also be effective in reducing crime and disorder. These strategies, found in the upper right of the table, may be classed more generally within the model of problem-oriented polic- ing. While many problem-oriented policing programs employ traditional law enforcement practices, many also draw on a wider group of strategies and approaches. The research available suggests that such tools can be ef- fective when they are combined with a tactical philosophy that emphasizes the tailoring of policing practices to the specific characteristics of the prob- lems or places that are the focus of intervention. Taken together, these research findings allow for strong policy recom- mendations for policing in the 21st century. Several decades of research have found weak or, at best, mixed evidence regarding the effectiveness of what we have defined as the "standard model" of policing. A large body of carefully conducted research has found much evidence of the effectiveness of what we have defined as the focused model of policing. Many police innovations over the past two decades have sought to re- duce crime and disorder and reduce fear by focusing on specific problems. Such approaches have great promise and should be the subject of more systematic investigation. There is very strong research evidence that the more focused and specific the strategies of the police, the more they are tailored to the problems they seek to address, the more effective police will be in controlling crime and disorder. This should be a guiding principle of developing strategies to reduce crime and disorder in the 21st century.