Unveiling the impact of depolicing on crime: Insights from Denver’s neighborhoods

A new research study published in Criminology examined the impact of sudden and sustained reductions in proactive policing (i.e., “de-policing”) on violent crime across neighborhoods in Denver, Colorado, leveraging the onset of the COVID-19 pandemic and social unrest after the murder of George Floyd as exogenous shocks. The study utilized multilevel models to analyze the impact of large-scale reductions in police stops and drug-related arrests on violent and property crime in the city.

Methods

The researchers examined weekly patterns in crime and policing from 2020 across Denver’s 78 neighborhoods. Data on reported crimes, policing activities, and neighborhood characteristics were obtained from various sources, including the American Community Survey, Denver’s Open Data Catalog, and OpenTable. The unit of analysis was the neighborhood-week, where authors examined Denver’s 78 neighborhoods over a period of 52 weeks, generating a total sample size of 4,056 neighborhood-weeks.

Outcome variables were weekly counts of violent crimes and weekly counts of property crimes. The predictor variables included two exogenous shocks—COVID-19 and George Floyd’s murder—in addition to four measures depicting weekly patterns in police discretionary behavior: pedestrian stops, vehicular stops, drug-related arrests, and public disorder-related arrests. For each outcome, weekly counts were compared to weighted averages based on data from prior years (2016–2019), such that “deviations from baseline” for each week could be calculated. These deviations from baseline were considered the primary outcome and were used to generate the main results.

The study also included control variables that varied within and between neighborhoods, such as demographics and population mobility. The authors measured the latter based on the number of motor vehicle accidents and the number of restaurant reservations in a neighborhood. Additionally, the violent and property crime models included spatial weights.

Findings

Regression models revealed significant reductions in discretionary policing activities, with a notable decrease in pedestrian and vehicle stops, drug arrests, and disorder arrests. These reductions were particularly pronounced during the COVID-19 and Floyd periods. Pedestrian stops reduced by 83 percent (85% during COVID-19, 87% during Floyd), vehicle stops by 85% (88% during COVID-19, 88% during Floyd), drug arrests by 76% (81% during COVID-19, 81% during Floyd), and disorder arrests by 62% (61% during COVID-19, 65% during Floyd).

Meanwhile, property and violent crime counts increased over the same time frame. Specifically, violent crime increased by 14.3%, while property crime increased by 27.1%. These increases amounted to a median increase of 4 violent crimes and 70 property crimes per neighborhood. Again, this was not uniform across neighborhoods. Violent crime increased in 55 neighborhoods (max increase of 53 incidents) and decreased in 23 (max decrease of 12 incidents). Property crimes increased in 70 neighborhoods (max increase of 411) and decreased in eight (max decrease of 101).

The findings revealed that drug arrests were associated with property crime, with a one-unit increase in drug arrests corresponding to a 2.7 percent decrease in property crime. However, no evidence was found to support a relationship between pedestrian stops, vehicle stops, or disorder arrests and variations in property crime across neighborhoods. The study also highlighted that the effects of policing were not consistent across neighborhoods, with significant variation observed. For example, the effects of depolicing on violent crime were stronger in more affluent neighborhoods.

Conclusion

The study utilized multilevel models to analyze the impact of large-scale reductions in police stops and drug-related arrests on violent and property crime in the city. Overall, the findings showed that depolicing impacts crime. Reduced stops were associated with more violence in most Denver neighborhoods. Reduced drug arrests were likewise associated with increases in property crimes. These results are consistent with other research suggesting that proactive policing can decrease crime.