Reimagining Broken Windows

Introduction

Since its first utterance by Wilson and Kelling (1982), broken windows has become a powerful and enduring part of the criminological nomenclature, with deep influences in both the disciplines’ scholarly and applied traditions. Upon the death of Wilson in 2012, obituaries and personal remembrances often led with his contribution to the broken windows perspective. One clever dual obituary linked Wilson with a famous New York City (NYC)-graffiti artist (Staley 2012). Of course, both Wilson and Kelling are well known for many more contributions to scholarship and public policy.

With more than 30 years having elapsed since the publication of Wilson and Kelling’s classic article, it is an opportune time to take stock of the theoretical, empirical, and policy developments on the broken windows perspective as well as to investigate the progress that has been made from new research on these fronts. It is this duality of interests that frames this special issue. Furthermore, it is stimulated by the sustained scholarly and policy significance of the broken windows perspective. Thirty years have by no means dampened research interest, scholarly and policy debates, and criminal justice practice in this area. We view this special issue as a way to build upon the important body of scholarship and contribute to progress on the specification of the broken windows perspective, advances in the measurement of disorder, new findings on policing disorder and understanding of mechanisms, and priorities for future research and policy development.

Development and Concept Specification

Wilson and Kelling (1982) outlined a developmental sequence where neighborhoods declined into high-crime areas through disorderly conditions. They argued that social incivilities (e.g., loitering, public drinking, and prostitution) and physical incivilities (e.g., vacant lots, trash, and abandoned buildings) cause residents and workers in a neighborhood to be fearful. Fear causes many stable families to move out of the neighborhood and the remaining residents to isolate themselves and avoid others. Anonymity increases and the level of informal social control decreases. The lack of control and escalating disorder attract more potential offenders to the area and this increases serious criminal behavior. Wilson and Kelling (1982) argued that serious crime developed because the police and citizens did not work together to prevent urban decay and social disorder.

Wilson and Kelling invoked the metaphor that a broken window left unrepaired will soon lead to the breaking of all other windows in a building. They drew upon a field experiment by Philip Zimbardo (1973) that demonstrated the process through which seemingly abandoned automobiles, left without license plates and with hoods up, were vandalized in the Bronx, New York, and Palo Alto, California. The abandoned car in the high-crime area of the Bronx was vandalized within 10 minutes and was completely stripped within 24 hours. The abandoned car in upscale Palo Alto remained untouched for a full week until Zimbardo himself damaged the car with a sledgehammer. Within a few hours, the abandoned car was vandalized, stripped, and turned over. To Wilson and Kelling (1982:32), the Zimbardo experiment suggested that vandalism and more serious crimes can occur anywhere once the sense of mutual regard and the obligations of civility are lowered by actions that seem to signal that “no one cares.”

It is important at this juncture to note that the term broken windows has been associated with a wide range of nouns, including thesis, idea, perspective, theory, and approach. In some respects this may be fitting, reflecting the evolution of the term and its many interpretations in the scientific as well as policy and practitioner communities. We refer to it as a perspective, Kelling (2015) refers to it as an idea, and the authors of the other articles in the special issue adopt various usages. We did not seek to apply a standard usage; the freedom of choice seemed important and perhaps necessary in a project on reimagining broken windows.

Broken windows has also been described as an “epidemic” perspective on crime (Gladwell 2000). In essence, crime is contagious and can spread to an entire community via environmental features. Relatively minor problems such as aggressive panhandling and graffiti serve as an invitation to an array of would-be offenders that particular neighborhoods are good locations to commit more serious crimes. As Wilson and Kelling (1982:34) suggested,

… serious street crime flourishes in areas in which disorderly behavior goes unchecked. The unchecked panhandler is, in effect, the first broken window. Muggers and robbers, whether opportunistic or professional, believe they reduce their chances of being caught or even identified if they operate on streets where potential victims are already intimidated by prevailing conditions. If the neighborhood cannot keep a bothersome panhandler from annoying passersby, the thief may reason, it is even less likely to call the police to identify a potential mugger or to interfere if the mugging actually takes place.

The available empirical evidence on the connections between disorder and subsequent crime is voluminous and mixed. Skogan (2015) specifies two bodies of research that have examined the disorder–crime link: One that focuses on community factors related to crime and another that focuses on the direct relationship between disorder and crime. Here, we highlight a few of the key studies. In his classic book Disorder and Decline, Skogan (1990) found disorder to be significantly correlated with perceived crime problems in a neighborhood, controlling for resident’s poverty, stability, and racial composition. Further, Skogan’s analysis of robbery victimization data from 40 neighborhoods in a number of cities found the relationship between crime and economic and social factors were indirect and mediated through disorder. In his reanalysis of these data, Harcourt (1998) removed several neighborhoods with very strong disorder–crime connections (all in Newark, New Jersey) and found no significant relationship between disorder and more serious crime in the remaining neighborhoods. Eck and Maguire (2006) suggested that Harcourt’s analyses did not disprove Skogan’s results; rather, his analyses simply documented that the data were sensitive to outliers. The removal of different neighborhoods from Harcourt’s analysis may have strengthened the disorder–crime connection.

Skogan’s study did not, however, examine if disorder leads to more serious crimes in the developmental sequence proposed by Wilson and Kelling (1982). In his longitudinal analysis of crime and disorder in Baltimore neighborhoods, Taylor (2001) found quite modest support for the proposition that disorderly conditions lead to more serious crime. However, these results varied according to types of disorder and crime. Taylor suggested that other indicators, such as initial neighborhood status, were more consistent predictors of later serious crimes. In the Netherlands, Keizer, Lindenberg, and Steg (2008) conducted six field experiments on the links between disorder and more serious crime and concluded that dealing with disorderly conditions was important to halt the spread of further crime and disorder.

Sampson and Raudenbush (1999) challenged the idea that disorder causes more serious crimes by arguing that crime and disorder have similar roots. They suggested that human intervention to deal with crime and disorder is strongly influenced by neighborhood “collective efficacy,” defined as the trust, solidarity, and shared expectations of residents and their willingness to intervene in support of neighborhood social control (see also Sampson, Raudenbush, and Earls 1997). Using systematic social observation (SSO) data to capture social and physical incivilities on the streets of Chicago, Sampson and Raudenbush (1999) found that, with the exception of robbery, public disorder was not significantly related to most forms of serious crime, controlling for neighborhood characteristics such as poverty, stability, race, and collective efficacy. The authors concluded that the relationship between disorder and crime was spurious.

Sampson and Raudenbush’s findings have been criticized because their social observation data on disorder were collected during the day rather than at night (Sousa and Kelling 2006) and based on their decision to test a model in which disorder mediates the effects of neighborhood characteristics on crime rather than neighborhood characteristics mediating the effects of disorder on crime (Jang and Johnson 2001). Moreover, Xu, Fiedler, and Flaming (2005) observed strong correlations between disorder and serious and less serious crime independent from collective efficacy, thus providing support for the broken windows perspective. Finally, St. Jean (2007) tested both theories in his mixed-methods analysis of robbery, assault, and drug hot spots in Chicago. He found some support for broken windows and collective efficacy perspectives, but neither consistently explained variation in crime within and between neighborhoods. Instead, criminal opportunities, what he explained as “ecological advantages,” seemed to be most important in explaining the existence of crime clusters.

Measurement

The measurement of disorder has evolved since the introduction of the broken windows perspective. It moved from perceptual toward more objective measures, from questions in general victim surveys to resident surveys, from residents as most important observers of disorder of one’s own neighborhood to independent observers or raters, from surveys to mixed methods including administrative and police data on disorder, from simple item measurement to more complex multidimensional scales, and from large neighborhoods to small segments and blocks (Anthony and Perkins 2014). Skogan (2015) provides a useful analysis of the principal strengths and weaknesses of the key measurement approaches.

The first measures of disorder came from general victim surveys like the British Crime Survey. Soon criminologists used typical resident surveys as a method to collect data on the level of social and physical disorder of neighborhoods (Sampson and Groves 1989; Skogan 1990). These kinds of resident surveys have three main methodological limitations: the unknown yardstick with which residents report on disorder; the natural area about which residents report might not coincide with the administrative borders of the neighborhood, which serves as the unit of analysis; and in most cases, the number of respondents in individual neighborhoods is relatively small.

In the context of an innovative laboratory experiment involving students and police officers in Taiwan, Yang and Pao (2015) explore mechanisms behind perceptions of disorder. Intriguingly, the authors find that race and style of dress have a significant impact on how respondents perceive social and physical disorder. This finding is similar to Sampson and Raudenbush’s (2004) observational study that found that as the percentage of Black residents in poor neighborhoods increased, both White and Black residents’ perceptions of disorder in the neighborhood increased after controlling for other factors.

A major advance in the development of more objective measures of disorder came with Sampson and Raudenbush’s (1999) SSO. The authors used videotaping from a vehicle to identify the presence of disorder in 23,816 blocks across 80 neighborhoods in Chicago, which were later coded by independent raters. Steenbeek and Kreis (2015) provide a good example of making use of SSO with independent observers. The observers rated the level of disorder for each 100-meter segment in the city of Amsterdam based on seven physical indicators.

Another effort in the objective measurement of disorder is taking photographs of streets and squares at various moments during the day and night, again rated by independent observers. These photographs can be manipulated experimentally to investigate how and why residents or visitors judge a level of disorder and how disorder levels affect their actual behavior (criminal or crime avoidance). Advances in technology will open new horizons for the measurement of disorder in the future. Google Street View offers to provide a more reliable and cost-effective instrument compared to perceptual measurements (Kronkvist 2013; Odgers et al. 2012).

The second major shift in measuring disorder was the change in unit of analysis (Weisburd, Bernasco, and Bruinsma 2009). In recent decades, the measurement of disorder has focused more on smaller areas like street segments, city squares, and parts of blocks, acknowledging micro variations in disorder within neighborhoods. One of the reasons behind the movement to smaller public spaces has to do with research showing that crime is more clustered at small places (Weisburd, Groff, and Yang 2012). Like crime, disorder is clustered at particular places within neighborhoods (Taylor 1997). Weisburd et al. (2015) and Steenbeek and Kreis (2015) are good examples of using small units of analysis in disorder research. This coincided with the availability of geographical-based data at the address level from police and governmental administrations (Pauwels and Svensson 2014). O’Brien and Sampson (2015) report an innovative use of ecometric methods based on two official databases from Boston: one million 911 calls and 200,000 requests for nonemergency services related to physical disorder. The data allowed the authors to generate various measures of physical disorder, social disorder, and crime and analyze the relationships among these variables with cross-time models.

Effects and Mechanisms

Wilson and Kelling (1982) do not specify how police departments should implement broken windows policing strategies. Instead, they offer general guidance that police should engage order maintenance activities and ameliorate physical blight to reinforce informal social control in neighborhoods. Equally important, they say that police activity on the street should reflect community standards. Residents need to be engaged in the definition of disorderly behavior and integrated into police efforts to alleviate disorderly conditions. In a follow-up article, Wilson and Kelling (1989) made the very broad suggestion that policing disorder activities should be rooted within community policing and problem-oriented policing initiatives.

Perhaps not surprisingly, the research evidence on the crime control effectiveness of broad-based disorder policing strategies, such as quality-of-life programs and order maintenance enforcement practices, is somewhat mixed. For example, using misdemeanor arrests as a proxy for order maintenance activities, Kelling and Sousa (2001) found that the NYC Police Department’s disorder policing strategy was associated with a significant reduction in violent crime in the 1990s, controlling for economic, demographic, and drug use variables. Rosenfeld, Fornango, and Rengifo (2007) analyzed the effects of order-maintenance arrests on precinct-level robbery and homicide trends in NYC between 1988 and 2001, and concluded that the approach generated small but significant crime reduction gains. Using a different analytic approach, Messner et al. (2007) examined homicide trends in 74 NYC police precincts between 1990 and 1999, and found that misdemeanor arrests generated significant reductions in total homicide rates with the largest impacts on gun homicide rates. Policy evaluations, including several randomized controlled trials, implemented in other cities have also found some support for the perspective that dealing with disorderly conditions produces crime control gains (e.g., Berk and MacDonald 2010; Braga and Bond 2008; Braga et al. 1999).

In contrast, other studies have not found significant crime prevention gains associated with broken windows policing. Harcourt and Ludwig’s (2006) reanalysis of Kelling and Sousa (2001) did not find that a generalized broken windows strategy, as measured by increased misdemeanor arrests, yielded significant reductions in serious crimes in NYC. An evaluation of a quality-of-life policing initiative in four target zones in Chandler, Arizona, did not find any significant reductions in serious crime associated with the strategies that focused on physical and social disorder (Katz, Webb, and Schaefer 2001). Finally, a randomized controlled experiment of broken windows policing in three California towns found no significant effects on fear of crime, police legitimacy, collective efficacy, or perceptions of crime and social disorder (Weisburd et al. 2011).

Braga, Welsh, and Schnell (2015) report on the results of a Campbell Collaboration systematic review and meta-analysis of the impacts of disorder policing on crime. Based on 30 randomized experimental and quasi-experimental tests, it was found that policing disorder strategies were associated with modest crime reduction effects. Importantly, community and problem-solving interventions designed to change social and physical disorder conditions at particular places generated the largest crime reduction impacts. Aggressive order maintenance strategies that target individual disorderly behaviors were not associated with significant crime reductions. Weisburd et al. (2015), however, find that very few of these evaluations test the developmental sequence proposed by Wilson and Kelling (1982) and, for the handful of studies that do examine other outcomes, there is little evidence that dealing with disorder reduces fear of crime or enhances community controls.

It remains unclear whether the crime control gains observed by Braga et al. (2015) are generated by the broken windows developmental process or, as suggested by Weisburd et al. (2015), criminal opportunity theories underlying place-based policing strategies. Regardless of the theoretical mechanisms at work, the general idea of dealing with disorderly conditions to control crime is supported by the available evaluation evidence. It is also important to note that aggressive order maintenance strategies, such as “zero-tolerance policing” initiatives do not seem to be effective in reducing crime. Moreover, these kinds of policing practices seem to distance police from the communities they serve (Meares 2015), and have been suggested to cause a host of other negative externalities such as increased complaints against police officers (Greene 1999), increased racial disparities in the criminal justice system (Golub, Johnson, and Dunlap 2007; Harcourt and Ludwig 2006), and the criminalization of homeless and mentally ill populations (Barr 2002).

Priorities for Policy and Research

The articles in this special issue were developed along the lines of the major themes of theory, measurement, and public policy. If there is any imbalance in these themes, it is surely—and perhaps rightly—tilted toward policy. This may have everything to do with the pragmatism of the broken windows perspective, from its introduction and evolution over three decades, to Wilson and Kelling’s writing style and even the publication outlet of their 1982 article, to its coverage in the media and treatment in the scholarly literature. Many of the articles in this special issue have recounted some of these points, noting the visceral connection of the broken windows perspective to day-to-day crime problems and urban policing.

In the course of preparing this special issue, the tragic events of 2014 in Ferguson, Missouri, and NYC as well as the subsequent protests across the country have further elevated the policy relevance of our work. While the project was never meant to be just about if disorder policing can be effective, how best to measure disorder in neighborhoods, or how to make sense of the hypothesized relationship between disorder and more serious crime as well as neighborhood decline, these recent events are intimately linked with the broken windows perspective.

Writing in the Boston Globe, Jackson (2014) called for the complete dismantling of the broken windows style of policing due to it “resulting in racial profiling and unleashing other indiscriminate policies on black people,” whereby “black people themselves became the broken windows, joined in recent years by Latinos” (emphasis in original). The Editorial Board of The New York Times (2014) called for a reappraisal of the policing strategy in the city, noting that it and its variations (e.g., zero tolerance, quality of life, stop-and-frisk) have “pointlessly burdened thousands of young people, most of them black and Hispanic, with criminal records.”

In light of another death of an unarmed Black man by an NYC police officer—this time in the course of a “vertical” check in a pitch dark stairwell of a Brooklyn public housing building—Bellafante (2014) drew attention to what she sees as a complete misreading of the broken windows policing strategy: “We have come to think of ‘broken windows’ in terms of the need to make arrests for minor offenses, the imperative to get rid of squeegee men and other avatars of nuisance. But at the heart of the philosophy is the notion that ruined things—broken windows, cracked sidewalks, littered pathways and so on—must get fixed” (p. 36). The basic implication in this instance is that replacing the burned out lights in the stairwell would have prevented Akai Gurley’s death, let alone likely averted any need for an intervention with “a total innocent.”

Broken windows policing can be quite effective when it instead focuses on problem solving, working in partnership with community residents and businesses, and targeting social and physical disorder conditions at particular places, which includes the use of situational crime prevention measures (e.g., graffiti removal, improved street lighting, and place managers; see Smith and Clarke 2012). As noted previously, this is the major finding of Braga et al.'s (2015) systematic review and meta-analysis of the effects of disorder policing on crime, with the chief implication that police departments should adopt a “community co-production” model rather than drift toward a zero-tolerance model. Policymakers, police leaders, and community leaders should pay close attention to this research evidence. If the broken windows style of policing is desired in a city or neighborhood, it need not be carried out in the manner that has been the cause of long-standing criticism and concern (Bellafante 2015; Harcourt 2001; Taylor 2001).

This is an important start, but it is hardly the full story. On the one hand, the different styles of disorder policing (community problem solving and aggressive order maintenance) sometimes overlap and the targeted crime and disorder problems (e.g., robberies and shootings) may call for tactics that are quite different from those used to address less serious problems (Braga et al. 2015). This draws attention to some issues for future research, which we discuss subsequently. On the other hand, along with reducing crime, the community problem-solving style of disorder policing can go a long way toward improving police–community relations and upholding justice. It may very well be the prospect of achieving these additional, important outcomes that tips the scale toward greater use of this style of broken windows policing.

Meares (2015) takes on directly the importance of procedural justice in the context of broken windows policing. She questions the trade-off that is so often justified by decision makers: short-term, modest crime reductions at the expense of undermining the fairness of law enforcement in the eyes of community residents. Meares proposes a framework for discussing this trade-off in the context of how the criminal justice system educates its citizens. Viewing broken windows policing as a curriculum, she asks: “What is this kind of policing teaching people about how government views them as citizens?” Here, an overt curriculum—one rooted in democracy and fairness—needs to offset a hidden or covert curriculum that discourages individuals from “attributing legitimacy to laws, legal systems, and regimes.” This is no easy task but, as Meares argues, it begins with evaluating broken windows policing as much for its crime reduction potential as for its potential to uphold fairness and legitimacy.

Turning our attention to future research, one immediate priority is the need to test the different styles of disorder or broken windows policing. It should be quite feasible to randomly assign geographically distinct areas within a city (e.g., neighborhoods and police districts) to receive either order maintenance policing or community problem-solving policing. It is useful to note that George Kelling and colleagues (1974) did something very much like this over 40 years ago. The Kansas City Preventive Patrol Experiment remains one of the preeminent place-based policing experiments. One could even entertain a third condition, whereby the focus is not on disorder (i.e., policing as usual). This would be useful as a baseline comparison. If random assignment is not feasible, high-quality quasi-experimental designs should be employed. Importantly, the type of evaluation design needs to be tailored to the research question (Welsh, Braga, and Bruinsma 2013). Whatever the evaluation design, efforts need to be taken to investigate for the possibility of crime displacement or the diffusion of crime prevention benefits. There should also be an effort to assess program effects over longer periods of time to determine if benefits persist, cease, or wane.

In addition to measuring crime and disorder and procedural justice, future evaluations of disorder policing need to measure fear of crime and collective efficacy. Weisburd et al. (2015) draw attention to fear of crime and collective efficacy as fundamental mechanisms in the broken windows model. They conclude that “for the crime prevention gains of disorder policing to be attributed to broken windows policing, it needs to be shown that crime has declined as a result of reductions in fear and increases in collective efficacy.” Few studies of disorder policing have investigated these mechanisms, leaving Weisburd et al. to suggest that observed crime control gains may very well be a product of other crime control mechanisms, including deterrence, opportunity reduction, or incapacitation. These are the active ingredients that Braga et al. (2015) argue need to be investigated experimentally.

Another key research priority—something that is raised in several articles in this issue—is to establish a clearer distinction between crime and disorder. This comes up in the context of observational studies as well as evaluation studies. Here, the matter is about trying to avoid confounding measures of disorder with measures of crime. This remains a major critique of the broken windows perspective. Weisburd et al. (2015) propose that the focus should be on serious crime and violent crime in particular; less serious crime should not be measured. The authors also call for a greater focus on physical disorder and more direct measures, including loitering, disorderly conduct, and drinking or intoxication.

Fittingly, the special issue concludes with a commentary essay by George Kelling. Reflecting on the intent as well as the context of his 1982 article, Kelling (2015) calls attention to the need to put at center stage the perspective’s focus on community residents and their desire to live without fear of crime and violence. He is also unwavering in his calls for innovation in public policy responses to disorder. This may just be reimagining broken windows at its finest.