Displacement of crime and diffusion of crime control benefits in large-scale geographic areas

Displacement of crime and diffusion of crime control benefits in large-scale geographic areas: a systematic review

Cody W. Telep & David Weisburd & Charlotte E. Gill & Zoe Vitter & Doron Teichman

Published online: 28 June 2014 # Springer Science+Business Media Dordrecht 2014

Abstract Objectives To conduct a systematic review examining the extent to which there is crime displacement or a diffusion of crime control benefits in social control interven- tions implemented in medium sized or large geographic areas. Methods A number of search strategies were used to identify and code eligible experimental or quasi-experimental studies that measured displacement in areas larger than crime hot spots. A total of 33 publications covering 43 quasi-experimental studies were identified as eligible. Nineteen of these publications covering 20 studies were included in a meta-analysis. Results The narrative results overall suggest that displacement is not a common occurrence in interventions implemented at larger units of geography and a diffusion of crime control benefits is somewhat more likely to occur. The effect sizes from the meta-analyses suggest that, while the interventions, on average, were associated with a significant decline in crime, displacement was not likely to occur. The meta-analyses found no significant overall evidence of displacement or a diffusion of benefits. Conclusions These findings are in line with previous reviews that have focused on displacement at smaller geographic units. When examining larger geographic scales

J Exp Criminol (2014) 10:515–548 DOI 10.1007/s11292-014-9208-5

C. W. Telep (*) School of Criminology and Criminal Justice, Arizona State University, 411 N. Central Ave. Mail Code 4420, Phoenix, AZ 85004, USA e-mail: [email protected]

D. Weisburd : C. E. Gill : Z. Vitter Center for Evidence-Based Crime Policy, Department of Criminology, Law and Society, George Mason University, Fairfax, VA, USA

D. Weisburd Institute of Criminology, Faculty of Law, Hebrew University, Jerusalem, Israel

D. Teichman Faculty of Law, Hebrew University, Jerusalem, Israel

and a broader array of interventions, spatial displacement is still a fairly unlikely occurrence.

Keywords Diffusionofbenefits .Displacement .Largeareas .Meta-analysis .Systematic review

Introduction

A series of reviews have found that spatial displacement is an uncommon outcome in place-based interventions (e.g., see Barr and Pease 1990; Bowers et al. 2011a; Guerette and Bowers 2009; Hesseling 1994; Johnson et al. 2012). When there is evidence of displacement, the amount of crime displaced tends to be far less than the amount of crime prevented by the initiative. Research also suggests that a “diffusion of crime control benefits” (Clarke and Weisburd 1994) to surrounding areas is a more common occurrence (Bowers et al. 2011a). Much of the primary research on displacement and most of these reviews have focused on interventions implemented at “micro-places” such as crime hot spots. Because formal social control interventions are often imple- mented at larger geographic units (e.g., police beats and districts, cities, jurisdictions), it is also important to examine displacement and diffusion of crime control benefit outcomes in broadly targeted place-based interventions. We conducted a systematic review in order to synthesize evidence on crime displacement and diffusion that results from formal social control interventions in larger areas.

Our main question is to what extent do formal social control interventions targeted at meso- or macro-places lead to spatial displacement of crime or diffusion of crime prevention benefits? Our results overall suggest that displacement of crime is not very common as a result of policing and other governmental interventions at a range of larger geographic scales. While there has only been limited research to date on interventions in very large macro-geographic areas, successful police interventions at meso-units larger than hot spots (e.g., neighborhoods or police beats) follow the pattern of studies at micro-geographic units. Crime displacement is not inevitable and it appears that a diffusion of crime prevention benefits is just as likely or a more likely outcome. We briefly review the existing literature on crime displacement before discussing our methodology for the review. We then turn to a more detailed description of our results before concluding with a discussion of the implications of our findings for future place-based crime control efforts and research on displacement and diffusion.

Background literature

Although there is growing evidence that formal social control, primarily in the form of police activity, can have an impact on crime at the specific areas where efforts are focused (Telep and Weisburd 2012; Weisburd and Eck 2004), such approaches risk shifting crime or disorder to other places where programs are not in place or to other times, targets, offenses, tactics, or offenders. This phenomenon is usually termed displacement, and it has been a major reason for traditional skepticism about the overall crime prevention benefits of place-based prevention efforts (see Reppetto 1976). The

516 C.W. Telep et al.

majority of research has focused on spatial or place-based displacement. The idea of spatial displacement can be traced to early work by sociologists who noted the role of opportunities for crime at places, but at the same time assumed that the concentration of crime prevention efforts at places would simply shift crime events from place to place without any clear long-term crime prevention benefit. Crime opportunities provided by places were assumed to be so numerous as to make crime prevention strategies targeting specific places of little utility for theory or policy. In turn, criminologists traditionally assumed that situational factors played a relatively minor role in explaining crime as compared with the “driving force of criminal dispositions” (Clarke and Felson 1993: 4; Trasler 1993).

The assumption that displacement is an inevitable outcome of focused crime prevention efforts has been replaced by a new assumption that displacement is seldom total and often inconsequential (e.g., see Clarke 1992; Gabor 1990; Weisburd et al. 2006). Since 1990, there have been five main reviews of empirical studies that report on displacement: Barr and Pease (1990), Eck (1993), Hesseling (1994), Guerette and Bowers (2009) (updated in Johnson et al. 2012), and Bowers et al. (2011a). All five reviews arrive at the same basic conclusions: there is little evidence that crime prevention strategies lead to displacement, and if displacement does occur it is usually offset by the amount of crime prevented.

Clarke andWeisburd (1994), moreover, suggest that scholars need to be cognizant of the reverse of displacement. They point to evidence indicating that situational and place-oriented crime prevention strategies often lead to a diffusion of crime control benefits to areas or contexts that were not the primary focus of crime prevention initiatives. Such spatial diffusion of crime control benefits has now been noted in a number of studies (e.g., Braga et al. 1999; Weisburd and Green 1995; Weisburd et al. 2006). The Weisburd et al. (2006) study, in particular, was designed explicitly to examine displacement and diffusion effects, and a wealth of data was collected in the intervention target areas and surrounding catchment areas, approximately two blocks surrounding each target area.

Only two reviews have focused explicitly on displacement and diffusion effects. Guerette and Bowers (2009) reviewed situational crime prevention studies, finding some displacement in 26 % of the 574 observations from 102 studies they examined, and a diffusion of crime control benefits in 27 % of the examined studies. Focusing only on studies reporting on spatial displacement and diffusion, they found that 37 % of the observations showed evidence of spatial diffusion while only 23 % showed evidence of spatial displacement (see also Johnson et al. 2012). As situational crime prevention tends to focus on specific situations in specific places, this review concen- trates observations on what might be termed “micro” areas of geography, usually a single facility or location, or sometimes a small cluster of buildings (for example, a housing project). Bowers et al. (2011a), in a Campbell Collaboration systematic review of crime displacement in police interventions, examined a number of studies focused on smaller geographic areas such as crime hot spots (e.g., Sherman et al. 1989; Sherman and Weisburd 1995).1 They included 44 studies in a narrative review, 16 of which also

1 Their specific definition of a place was “a specifically defined area that is smaller than a city or region,” including census blocks, police areas, housing estates, districts, suburbs, block areas, series of roads, neighborhoods, or hot spots (Bowers et al. 2011a: 16).

Displacement and diffusion in large-scale geographic areas 517

contained sufficient quantitative data on treatment, control, and catchment areas to perform a meta-analysis. They also found little evidence of displacement of crime, reporting that on average police interventions at micro-places are associated with significant reductions in crime, and while changes in crime in catchment areas were non-significant, the trend favored diffusion of benefits rather than displacement. In their analysis of 36 studies that contained treatment and catchment area outcomes, they also found evidence in favor of diffusion of benefits over displacement, although the finding could not be statistically tested.

Like these two reviews, much of the primary research on displacement has focused primarily on local area (“micro-place”) displacement. That is, many studies have been concerned with geographically focused police initiatives at crime hot spots of a single street block, or clusters of street blocks with high intensities of specific types of crime. Indeed, some of the strongest and most persuasive evidence against the assumption of immediate spatial displacement has come from recent studies of focused interventions at crime hot spots (see Braga et al. 2012). However, displacement may also occur across larger areas (“macro-places”), such as police beats, neighborhoods, cities, regions, states, and even nations. Displacement in these contexts involves the movement of crime across administrative, governmental, and/or social boundaries as a result of larger scale interventions of formal social control (such as policing strategies and changes in laws or policies) implemented by governmental or private agencies (McIver 1981).

Teichman (2005), for example, argues such large-scale displacement can occur as a result of efforts by jurisdictions to push criminal offenders to neighboring locations (see also Broude and Teichman 2009; Marceau 1997). By increasing sanctions or the probability of detection, for example, a jurisdic- tion could change a criminal’s opportunity costs and, for certain financially motivated crimes, make it worth the offender’s effort to displace to a neigh- boring jurisdiction with less severe sanctions. Teichman (2005) pointed to the Michigan Auto Theft Prevention Authority as an example, noting that increased enforcement efforts against auto theft and chop shops displaced auto thieves to neighboring and nearby states such as Wisconsin and Illinois.

Anecdotal evidence of large-area displacement can even be found at a global level. The United Nations World Drug Report (2007: 16), for instance, de- scribes displacement on a larger scale in regards to international methamphet- amine markets, noting that “[i]mproved controls in Canada and further tighten- ing of controls in the USA have led to a decline in the number of clandestine laboratories operating within the USA and a shift of production across the border to Mexico. However, Mexico has now also improved its precursor control regime, prompting drug trafficking organizations to exploit other areas, such as Central America and possibly Africa.” National drug control policy may thus have been responsible for pushing methamphetamine laboratories across international borders. These examples suggest that displacement in larger areas could be more likely for crimes in which there is a strong potential for financial gain. Interviews with drug smugglers, for example, suggest that potential massive payouts (even if the actual gain is far less) are one major motivator for offender involvement (Decker and Chapman 2008). Thus, this

518 C.W. Telep et al.

activity may be more likely to be displaced in the face of enforcement activity because of a recognized potential for continued significant financial gains due to strong user demand for illegal drugs.2

The belief that displacement may be occurring at larger units of geography is not universal, however. In one of the first works to discuss spatial displace- ment, Reppetto (1976) argues that displacement in large-scale interventions might be even less common than more micro-scale displacement. He notes, “Probably the programs least subject to displacement would be those based on large areas rather than on individual targets, since securing only buses, stores, or particular streets while leaving nearby subways, homes, and other streets unprotected is likely to be unproductive” (Reppetto 1976: 176). It could be the case then that well-implemented and comprehensive larger-scale programs lead to less displacement because offenders would have to travel a great deal to find the same opportunities to offend.

Weisburd et al. (2012) have recently argued that the “tight coupling” of crime to place helps explain the stable concentration of crime in micro-units of geography. The stability of street-level factors explaining crime concentrations and the heterogeneity of surrounding streets helps explain why crime may not easily move around the corner. These ideas of tight coupling are especially relevant to micro-units of geography, but may also be applicable to more macro-units. If the police increase enforcement in one high crime neighborhood, it may not be easy for offenders to move their activity to surrounding neighborhoods if these places do not provide the same opportunities to offend. Additionally, as Weisburd et al. (2006) found, offenders are often reluctant to relocate criminal activity because of a lack of familiarity with surrounding areas or because of a recognized danger of infringing on the territory of other offenders or groups.

While less attention has been given to diffusion at larger units of geography, Weisburd and Telep (2012) review a number of arguments for why diffusion may occur, some of which are applicable at the neighborhood level. Clarke and Weisburd (1994), for example, argue that offenders may be unaware of the boundaries of interventions and thus may overestimate their risk of apprehen- sion in surrounding areas and also avoid offending in these places. This remains possible in larger units of geography. Mears and Bhati (2006: 537) point to another potential mechanism and argue that “the influence of an initiative aimed at reducing community disadvantage may have positive ripple effects that extend to other communities in geographic and social space, especially insofar as social networks and ties are not constrained by neighbor- hood boundaries.” Thus, a diffusion of crime control benefits could occur as residents in one neighborhood communicate with those in nearby areas about their positive experiences, potentially enhancing levels of social organization in both places. Others have focused on market-based explanations for a diffusion of benefits (see Taniguchi et al. 2009). Zielenbach and Voith (2010), for example, examine how the redevelopment of public housing projects affected

2 These same levels of motivation likely do not exist for crimes with less potential rewards. As one anonymous reviewer noted, “it seems unlikely that a bag thief will travel 100 miles to commit a crime denied to them.”

Displacement and diffusion in large-scale geographic areas 519

crime and property values in surrounding neighborhoods. They find evidence suggestive of a diffusion of crime control benefits and frame these findings in the context of overall economic spillover benefits resulting from improved housing markets surrounding the newly developed public housing sites.

The study of large-area displacement is important because, despite the extent of research on micro-places, many police interventions take place at geographic units larger than hot spots. For example, the Evidence-Based Policing Matrix, a compilation of rigorous policing evaluation studies (Lum et al. 2011), suggests that police are frequently targeting crime and disorder at the neighborhood level. While 29 of the 124 studies included in the Matrix (23.4 %) focused on crime micro-places, 42 (33.9 %) used the neighborhood as the unit of analysis. Six studies used the jurisdiction as the unit of analysis. While the Matrix did not systematically assess displacement and diffusion effects in these studies, these results suggest the importance of understanding the relationship between formal social control interventions at larger geographic units and displacement/diffusion of crime.

Does displacement operate differently at these larger geographic levels? There are several reasons to question the applicability of the findings on displacement and diffusion of crime control benefits in micro-places to larger geographic units. First, the types of interventions used in larger areas are often different from those applied to smaller places. For example, changes in the law that could result in crime (or benefits) shifting across boundaries clearly apply to an entire jurisdic- tion, rather than a small group of street blocks. Further, any type of intervention applied over a larger area will necessarily differ from those applied to small areas in terms of intensity, focus, and dosage, which could affect displacement and diffusion outcomes. Second, displacement across administrative areas may differ from displacement across hot spots or other small areas that are not “officially” defined. For example, offenders may find it more difficult to move to a different area when the area is large (it may not be practical to move to a new city), but there could also be benefits in such moves, such as avoiding social control agents like the police by crossing administrative borders. A third and related point is that differences in enforcement are more clear-cut across administrative boundaries than small areas. Finally, socioeconomic composition and behavioral norms may vary more widely across large places than small ones, and could also be related to the mechanisms by which displacement and diffusion occur. The theory of crime displacement is still in an early stage of development and the reasons for these potential differences are not yet fully understood; however, investigating whether there is evidence for these differences could help to inform theoretical advances.

Thus, we felt it important to undertake a systematic assessment of what we know about displacement and diffusion of crime control benefits in broadly targeted place-based interventions. We wanted to more rigorously assess the claim, largely from anecdotal evidence, that large-scale displacement may be a greater problem than displacement resulting from micro-scale interventions. The objective of this review was to synthesize the extant empirical evidence (pub- lished and unpublished) on crime displacement and diffusion of crime preven- tion benefits across medium and large geographic units as a result of formal social control interventions.

520 C.W. Telep et al.

Methods

Criteria for inclusion

To be included in the review studies had to meet five main criteria:

1. The main intervention must have been an instance of formal social control, such as a law enforcement strategy targeted at a particular beat or neighborhood; or a legal or policy change. The instance of formal social control must have been imple- mented with the explicit purpose of controlling or preventing crime or disorder.

2. The intervention must have been targeted at a “meso-” or “macro-” geographic area. We define meso- and macro-areas as larger geographic units at which crime prevention resources are organized and can be distributed. That is, these units have some form of administration or government with at least some control over how crime is addressed in that unit. These include police beats, police districts or precincts, cities, jurisdictions or counties, states, and countries. We included non- administrative units only when they are specifically defined in the intervention as representing “neighborhoods” or “communities.”We recognize our examination of macro-level displacement at geographic areas overlaps somewhat with the Bowers et al. (2011a) review described above. Bowers and colleagues included some police interventions that focused on neighborhoods or police precincts within a city. Our review takes a more expansive view of the macro-level, so, while there is overlap that we discuss more below, we also collect studies using larger geographic units than Bowers et al. considered. Additionally, we do not limit our review to only police interventions, so we also take a more expansive view of what consti- tutes crime prevention efforts at geographic areas.

3. The intervention must have been assessed using at least one crime- or disorder- related outcome. This could include measures related to total crime or disorder or total amount of a particular crime or disorder type.

4. The study must have measured spatial displacement and/or diffusion effects. Displacement and diffusion effects need not be the sole focus of the evaluation, but they must be explicitly measured as part of the evaluation.

5. We included randomized experiments or quasi-experiments with a comparison group that did not receive the intervention or change in conditions, as well as quasi-experiments that adjusted for secular trends (e.g., citywide crime rates), in our main analysis.

We recognized that many studies of displacement and diffusion are likely to simply look at pre-post changes in the target and surrounding areas (i.e. not make use of a comparison group). We think these studies are highly vulnerable to historical validity biases. It is sometimes argued that a decrease in crime in the target area but an increase in the surrounding areas would provide a reasonable case for a displacement effect even without a comparison group or adjustment for secular trends. However, even here, the displacement effect could simply represent a secular trend, while the target area effect represented the success of the intervention in offsetting a general secular trend. Because of our concern with drawing conclusions from such studies, we made an initial decision to avoid including these studies in our main analysis.

Displacement and diffusion in large-scale geographic areas 521

Search strategy for identification of relevant studies

Several strategies were used to perform an exhaustive search for literature fitting the eligibility criteria. First, a keyword search was performed on an array of online abstract databases.3 Second, we reviewed the bibliographies of past reviews of crime displace- ment (e.g. Barr and Pease 1990; Eck 1993; Hesseling 1994; Guerette 2009; Guerette and Bowers 2009; Bowers et al. 2011a). Third, we performed forward searches for works that have cited seminal displacement studies.4 Fourth, we reviewed abstracts of leading journals in the field.5 Fifth, we searched the publications of several research and professional agencies. Sixth, we emailed a preliminary list of eligible studies to leading scholars knowledgeable in the area of crime displacement and diffusion of crime control benefits in an effort to identify any additional relevant studies. Our initial searches were conducted in the spring and summer of 2011 with supplemental searches conducted in the fall of 2013.

Details of study coding categories

All eligible studies were coded on a variety of criteria. A full coding sheet is available in Weisburd et al. (2011). After coding basic reference information, we recorded information on the nature of the target and comparison sites, the unit of analysis for the intervention, the sample size for the intervention, and what exactly the intervention entailed. We also coded the strategy used for measuring displacement and diffusion and the nature of the catchment area(s). We described any implementation difficulties described by the authors and coded the methodology of the evaluation. We noted any statistical tests completed, and we coded any results from tests of statistical signifi- cance. For calculating effect sizes, we included pre and post counts or rates of crime in treatment, control, and catchment areas when these were available. Finally, we detailed the conclusions drawn by the authors about the main effects of the intervention (was crime reduced?) and whether there was evidence of displacement or diffusion.6 If there was evidence of displacement or diffusion, we recorded any explanations provided by the authors.

Statistical procedures and conventions

When possible, meta-analytic procedures were used to combine data from studies. For eligible studies with enough data present, effect sizes were calculated using an ap- proach first described by Farrington and colleagues (2007). Reviews of displacement offer a special challenge in meta-analysis because rather than a simple treatment– control comparison of outcomes, we are interested in whether the change in crime in the catchment areas (the places to which crime or crime control benefits might be

3 See Telep et al. (2014) for a list of databases searched and keywords used. 4 The seminal pieces used were: Clarke (1995), Clarke and Weisburd (1994), Cornish and Clarke (1987), McIver (1981), Reppetto (1976), and Teichman (2005). 5 See Telep et al. (2014) for a list of these journals. 6 While we coded any type of displacement noted by study authors, we focus here only on spatial displace- ment as this was the most common type of displacement examined and often the only type examined quantitatively.

522 C.W. Telep et al.

displaced or diffused as a result of an intervention in another area) differs to a greater or lesser extent from the change in the control areas relative to the treatment areas. This challenge makes it difficult to use many of the standardized measures of effect sizes suggested in the meta-analytic literature (see Lipsey and Wilson 2001). An optimal comparison would be the simplest one that compares treatment catchment areas to control catchment areas. In this scenario, one would assess whether areas around treatment sites experience greater crime increases or decreases than those around control sites. Most of our eligible studies, however, have only catchment areas sur- rounding the treatment site(s). Thus, we focus on the overall difference between treatment and catchment areas to compare the change that has occurred in a catchment area relative to a treatment control area.

Following Bowers et al. (2011a, b), we use before and after data in treatment, comparison, and catchment areas to calculate a modified odds ratio, based on the method proposed by Farrington et al. (2007) in their meta-analysis on CCTV interven- tions. We use the following equations to calculate the effect size (ES) and standard error (SE) for the main results (was crime reduced in the intervention area compared to the comparison area?) and for the displacement and diffusion analysis (did crime change in the catchment area relative to the comparison area?)

ES ¼ ad

bc SE ¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

1

a þ 1

b þ 1

c þ 1

d

r

Where a, b, c and d represent crime counts, average crime counts, or crime rates:

Pre-intervention Post-intervention

Intervention area OR Catchment area

a b

Comparison area c d

For the main analysis, we use the effect size equation above, so that odds ratios greater than 1 are indicative of a crime decline in the intervention area relative to the comparison area. For the displacement and diffusion analyses, we compare the catch- ment area to the comparison area. Here, odds ratios greater than 1 indicate a greater crime decline in the catchment area compared to the control area and suggest a possible diffusion of crime control benefits. Odds ratios less than 1 indicate possible crime displacement. We use the term odds ratio to describe these results, while recognizing that this is not truly an odds ratio as commonly calculated for meta-analyses and so should be thought of as a modified odds ratio or, as Farrington et al. (2007) call it, a relative effect size. Following Bowers et al. (2011a), we note that that this measure is not without potential limitations, but it allows for comparisons across a number of our eligible studies and we felt it was the best way to create a meta-analytic summary of our main findings.

Bowers et al. (2011a) also discuss in some detail the potential issues with the calculation of the standard error for these effect sizes. As they note, the variance estimates may be too small, particularly because the assumption that these data are distributed Poisson may be less realistic with place-based interventions, as compared to interventions focused on individuals. Thus, we follow the Bowers et al. (2011a)

Displacement and diffusion in large-scale geographic areas 523

approach of multiplying the standard error by an inflation factor of 2 to increase the size of confidence intervals and make any estimations of statistical significance more conservative (see Farrington et al. 2007).

Mean effect sizes were computed across studies and weighted (using the inverse variance weighting procedu