Broken Windows Policing

Resources

What is Broken Windows Policing?

What is the Evidence on Broken Windows Policing?

Broken Windows Policing

Seattle Police officersmage courtesy of Flickr user jmabel and used under a Creative Commons license.

What is Broken Windows Policing?

The broken windows model of policing was first described in 1982 in a seminal article by Wilson and Kelling. Briefly, the model focuses on the importance of disorder (e.g., broken windows) in generating and sustaining more serious crime. Disorder is not directly linked to serious crime; instead, disorder leads to increased fear and withdrawal from residents, which then allows more serious crime to move in because of decreased levels of informal social control.

The police can play a key role in disrupting this process. If they focus in on disorder and less serious crime in neighborhoods that have not yet been overtaken by serious crime, they can help reduce fear and resident withdrawal. Promoting higher levels of informal social control will help residents themselves take control of their neighborhood and prevent serious crime from infiltrating


What is the Evidence on Broken Windows Policing?

Broken windows policing is listed under “What do we need to know more about?” on our Review of the Research Evidence 

A recent systematic review by Braga, Welsh and Schnell (2015) found that policing strategies focused on disorder overall had a statistically significant, modest impact on reducing all types of crime.  This positive effect was driven by the success of place-based, problem-oriented interventions.  In contrast, there was no significant overall impact of aggressive order maintenance strategies.  Thus, they conclude that police can successfully reduce disorder and non-disorder crime through disorder policing efforts, but the types of strategies matter.

Not all of the interventions included in the Braga et al. (2015) review, however, are true tests of broken windows theory.  Indeed, the broken windows model as applied to policing has been difficult to evaluate for a number of reasons.

First, agencies have applied broken windows policing in a variety of ways, some more closely following the Wilson and Kelling (1982) model than others. Perhaps the most prominent adoption of a broken windows approach to crime and disorder has occurred in New York City. In other agencies though, broken windows policing has been synonymous with zero tolerance policing, in which disorder is aggressively policed and all violators are ticketed or arrested. The broken windows approach is far more nuanced than zero tolerance allows, at least according to Kelling and Coles (1996) and so it would seem unfair to evaluate its effectiveness based solely on the effectiveness of aggressive arrest-based approaches that eliminate officer discretion. Thus, one problem may be that police departments are not really using broken windows policing when they claim to be.

A second concern is how to properly measure broken windows treatment. The most frequent indicator of broken windows policing has been misdemeanor arrests, in part because these data are readily available. Arrests alone, however, do not fully capture an approach that Kelling and Coles (1996) describe as explicitly involving community outreach and officer discretion. Officers must decide whether an arrest is appropriate and many police stops and encounters with citizens in broken windows policing do not end in arrest. As opposed to a zero-tolerance policy focused only on arresting all minor offenders, Kelling and Coles (1996) describe a more community-oriented approach to partnering with residents and community groups to tackle disorder collectively in a way that still respects the civil liberties of offenders. Whether the NYPD was able to adopt this model successfully remains up for debate but it does suggest that the intervention is complex and difficult to evaluate.

Third, the broken windows model suggests a long term indirect link between disorder enforcement and a reduction in serious crime and so existing evaluations may not be appropriately evaluating broken windows interventions. If there is a link between disorder enforcement and reduction in serious crime generated by increased informal social control from residents, we would expect it would take some time for these levels of social control in the community to increase. Policing studies usually use short-follow up periods and so may not capture these changing neighborhood dynamics.

There is also no consensus on the existence of a link between disorder and crime, and how to properly measure such a link if it does indeed exist. For example, Skogan’s (1990) research in six cities did suggest a relationship between disorder and later serious crime, but Harcourt (2001) suggested in a re-analysis of Skogan’s (1990) data that there was no significant relationship between disorder and serious crime. Hence, there is no clear answer as to the link between crime and disorder and whether existing research supports or refutes broken windows theory.

There is much debate over the impact of New York policing tactics on reductions on crime and disorder in the 1990s. Broken windows policing alone did not bring down the crime rates (Eck & Maguire, 2000), but it is also likely that the police played some role. Estimates of the size of this role have ranged from large (Bratton & Knobler, 1998, Kelling & Sousa, 2001) to significant but smaller (Messner et al., 2007; Rosenfeld et al., 2007) to non-existent (Harcourt & Ludwig, 2006).

Tackling disorder has frequently been a tactic chosen by police in crime hot spots. For example, in the Braga et al. (1999) problem-oriented policing hot spots study in Jersey City, NJ, officers used aggressive order maintenance as a strategy to reduce violent crime and results suggested significant positive results. Thus, we suspect that the tactics common in broken windows policing will be most successful when combined with knowledge about the small geographic areas where crime is highly concentrated. These hot spots approaches, however, should not be viewed as direct tests of broken windows theory.  A number of other strategies were also used in these interventions, including situational crime prevention efforts, which were shown to be the most effective strategy for reducing crime in the Braga and Bond (2008) study.

Finally, there is concern that any effectiveness of broken windows policing in reducing crime (where the evidence, as noted above, is mixed) may come at the expense of reduced citizen satisfaction and damage to citizen perceptions of the legitimacy of police.  See the Community Policing and Procedural Justice page for more on the importance of implementing effective strategies in ways that are viewed by citizens as fair.

Evidence-Based Policing Matrix

 

Studies from the Evidence-Based Policing Matrix incorporating some of the principles of broken windows policing:

Author
Intervention
Berk & MacDonald (2010) Broken windows approach to deal with homeless encampments associated with meaningful reduction in violent, property, and nuisance crimes full-circle
M
N
F
P
Braga et al. (1999) Problem-oriented policing in violent crime hot spots leads to reductions in violent and property crime, disorder and drug selling full-circle
VR
MP
F
HP
Braga & Bond (2008) Focus on hot spots of crime leads to reductions in crime and disorder calls for service full-circle
VR
MP
F
P
Pate & Skogan (1985a) Proactive disorder arrests associated with significant reductions in total Part I crimes, personal crimes, and burglary. grey-circle
M
N
G
P
Novak et al. (1999) No decline in burglary or robbery following proactive disorder enforcement empty
M
N
F
P
Pace (2010) Order maintenance unit had no effect on offenses empty
M
N
F
P
Weisburd et al. (2012) Broken windows policing had no evidence of an effect in calls for service empty
VR
MP
G
P

Resultfull-circle =successful intervention; grey-circle = mixed results; empty = nonsignificant finding; backfire = harmful intervention

Rigor: M = moderately rigorous; R = rigorous; VR = very rigorous

X-axis: I = individual; G = group; MP = micro place; N = neighborhood/community; J = jurisdiction

Y-axis: F = focused; G= general

Z-axis: R = reactive, P = proactive, HP = highly proactive