Re-analyzing the results of a community policing study leads to a more positive conclusion about its effects

In the last 30 years, community-oriented policing has been a key innovation in policing. But while this proactive approach boosts citizens' evaluations of police, evidence is sparse that it reduces crime.

In a new study, researchers suggest that this conclusion is likely flawed because of crime reporting sensitivity (CRS) bias. Re-analyzing the results of a prior study, researchers reached a very different, more positive conclusion about community policing, suggesting that future research should consider CRS bias.

The study, by researchers at George Mason University (GMU) and Hebrew University of Jerusalem, is published in the Journal of Law and Empirical Analysis.

"Reviews of community-oriented policing have had a major impact on how we view the crime prevention benefits of such programs, leading to an assumption that community policing does not reduce crime," says David Weisburd, professor of criminology, law, and society at GMU and emeritus professor of law and criminal justice at Hebrew University of Jerusalem, who led the study. Weisburd is an expert whose work is promoted by the NCJA Crime and Justice Research Alliance.

CRS bias occurs because community policing leads to more cooperation with and trust in the police and subsequently increased crime reporting. Such increased crime reporting bias adjusts crime prevention outcomes of community policing downward.

In this study, researchers illustrated this process by reanalyzing data from the Brooklyn Park ACT Experiment, a randomized controlled trial of efforts to boost police-community collaboration in Brooklyn Park, Minnesota. The study, conducted in 2015 and 2016, suggested that the positive effects of community policing on community members' perceptions of the police may lead to what the authors defined as crime reporting inflation, meaning that community members are more likely to report problems in their neighborhoods to the police.

In their re-analysis of the earlier study, the authors identified disorder crimes and domestic crimes as the underlying mechanisms that contributed to its findings. This led them to redefine crime reporting inflation as CRS because the data suggested a greater willingness to report crime to the police rather than changes in crime per se.

Citizens were more likely to report disorder crimes because of the increased community collaboration and trust built in community policing programs. In this case, crimes that were less serious and less likely to be reported before the experiment were reported more often during the experiment.

In turn, the more serious category of domestic crimes, which are generally underreported, were also more likely to be reported during the experiment because of increased collaboration and trust. This increased sensitivity in reporting led to a rise in recorded crime incidents not only in these categories of crime, but also in other categories like violence and property crimes.

In testing this more general mechanism, both in terms of increased sensitivity for crime calls to the police more generally and for disorder crimes specifically, the study found strong evidence of CRS bias in these data. Based on their findings, the authors concluded that adjusted estimates tell a very different, more positive, story about community policing. Simply stated, when previous studies are adjusted for CRS, community policing leads to significant reductions in crime.

The study highlights a contradiction between the goals of community policing and the measurement of its crime prevention impacts. Increasing collaboration and reporting to the police leads to more crime calls to the police—a measure sometimes used to evaluate policing interventions. Increased calls to the police, in turn, lead to higher numbers of crime incidents, which are commonly used as indicators of real crime levels in evaluations of policing programs.

The authors recommend a way to correct crime outcomes reported by previous studies and illustrate how using this method may affect interpretations of the overall impacts of community policing on crime. They also recommend that future research on this topic recognize and adjust for CRS bias or identify other measures not influenced by this mechanism.

Of the study's limitations, the authors note that Brooklyn Park may not be representative of community policing studies more generally.

"Our results suggest that it is time to look critically at the conclusions of previous research on community policing," notes David B. Wilson, professor of criminology, law, and society at GMU, who coauthored the study.

"Adjusting for CRS bias dramatically shifts the size and statistical significance of crime outcomes in other studies, calling into question the prevailing scientific conclusion that community policing does not affect crime."