Police anti-Black discrimination in use of force decisions is a growing, prevalent problem in American cities. This project explores the dual problem of mediation and selection when estimating bias using police administrative datasets. Using data from the Chicago Police Department, I calculate the racial risk ratio of experiencing police force for Black people compared to non-Black people. This study employs a new technique for correcting bias arising from administrative records, in particular, estimating the racial encounter odds through a weighted composite of police shift deployments and the Black residential rate of police beats, finding that police are significantly more likely to use force against Black people. Furthermore, police district-specific analyses show police are more violent towards Black people in districts with a greater white population.