The Interaction of Race and Place: Predictors of Fatal Police Shootings of Black Victims at the Incident, Census Tract, City, and State Levels, 2013–2018

Abstract

We develop and analyze a multilevel model of fatal police shootings to better understand the interplay between individual-level and contextual-level factors in incidents of police use of deadly force. We use a three-level hierarchical logistic regression model to identify Census tract-, city-, and state-level predictors of fatal police shootings and a four-level model to identify predictors of fatal police shootings of black victims at the incident, Census tract, city, and state levels. We find that higher levels of concentrated disadvantage and interpersonal firearm violence in a Census tract are significant predictors of the likelihood of any fatal police shooting occurring in that tract. We also find that conditional upon a fatal police shooting having occurred in a Census tract, knowing that the victim was unarmed is a significant predictor of the likelihood that the victim was black. Three other predictors of a police shooting victim being black were a higher proportion of black residents in the Census tract, a greater presence of police officers in the city, and a higher level of black-white residential segregation in the state. We conclude that reducing excessive police use of deadly force requires consideration of not only the individual interactions between officers and citizens, but also the quantity of interactions and the characteristics of the place in which these interactions occur. Special consideration must be given to ameliorating the consequences of structural racism, including racial segregation, over-policing of predominantly black neighborhoods, concentrated disadvantage in segregated neighborhoods, and racial inequalities in income and wealth.