In the wake of recent events, there has been a growing and impassioned debate on the role of race in police use of force. Critics allege that police act with racial bias, an accusation that has in turn spurned often lethal violence against the police themselves. With a database of highly granular data that includes encounter characteristics from a diverse set of cities across the country, we ask what is the extent, if any, of racial bias in both lethal and non-lethal police use of force?
With the assistance of municipal police departments across the country, we created a dataset consisting of highly detailed police narratives on officer involved shootings and officer involved shooting-like scenarios (police interactions in which lethal force was justified but not used). A team of researchers read each summary and extracted hundreds of pre-determined variables of interest. Additional teams replicated their results.
In a widely published paper released in July 2016, we found no racial bias in officer involved shootings. However, non-lethal police use of force was 50% more likely among blacks and Hispanics, holding all other conditions constant. In order to expand the scope of our findings, we are in contact with agencies in Florida and California to collect more data. Additionally, rather than relying on the time and cost intensive practice of manually classifying police narratives, we are working towards incorporating developments in natural language machine learning to automate the conversion of officer narratives into metadata.
The project is ongoing and the analyses will be forthcoming.