New Delhi: US researchers have added a new feature to a nationwide database that identifies counties where police use of lethal force occurs more or less often than statistical models predict. Called the Anomaly Detector, the feature is part of the University of Illinois Urbana-Champaign’s SPOTLITE project, which tracks lethal police action.
According to the researchers, the Anomaly Detector is the first national-scale resource for identifying counties with unusual numbers of cases where police use more force relative to national benchmarks. By identifying counties that record substantially more or fewer incidents than predicted, the tool aims to direct attention to areas that may warrant closer examination. Researchers say such counties could either require reform or may already be doing innovative things to reduce unnecessary uses of lethal force by police.
The Anomaly Detector was developed by the Cline Center for Advanced Social Research at the University of Illinois Urbana-Champaign. The research was led by Scott Althaus, director of the Cline Center and the Merriam Professor of Political Science, along with an interdisciplinary team comprising Joseph Bajjalieh, Jay Jennings, Michael Martin, Jennifer K. Robbennolt and Ajay Singh. The project has also involved dozens of undergraduate and graduate student researchers who have worked since 2017 to build and verify the SPOTLITE database.
Not an evidence of police brutality
The dashboard uses SPOTLITE data from all 3,142 US counties between 2014 and 2023. According to the researchers, SPOTLITE is the most comprehensive nationwide record available of police uses of lethal force. It documents any law enforcement discharge of a firearm that risks human life, regardless of whether anyone is killed or injured, as well as any police pursuit or other use of force that results in a civilian death.
To identify outlier counties, the researchers developed a predictive statistical model using three variables: population size, population density and the level of civilian gun violence. The model estimates how many police uses of lethal force would be expected in a county and compares that prediction with the number of incidents actually recorded by SPOTLITE.
Based on that comparison, counties are classified into five categories: many more incidents than predicted, somewhat more than predicted, about as many as predicted, somewhat fewer than predicted, and many fewer than predicted. Counties at either end of the spectrum are identified as anomalies.
The researchers caution that the anomaly classifications should not be interpreted as evidence that policing in a county is either effective or ineffective. They say the model only accounts for population size, population density and civilian gun violence. Any difference between the predicted and observed number of incidents suggests that other, unobserved factors are contributing to the county’s rate of police action. The dashboard is therefore intended to identify counties that merit further study rather than explain why those differences exist or draw conclusions about what is happening in particular places.
The dashboard also allows users to view anomaly classifications by individual year or averaged across the 2014–2023 period, making it possible to identify whether a county has consistently recorded unusually high or low rates or whether a single year stands out as an outlier.
The SPOTLITE project has been under development since 2017 and was publicly launched in 2022 following calls for greater police accountability and reform after fatal police shootings involving unarmed Black Americans. The researchers describe the project as nonpartisan and nonadvocacy, saying its role is to document police use of lethal force and to provide a public resource for communities, researchers, and policymakers. With the addition of the Anomaly Detector, they say the project now offers a national tool for highlighting counties whose patterns of police use of lethal force warrant closer examination, without concluding the causes of those patterns.