Racial Bias in Motor Vehicle Searches: Additional Theory and Evidence

Knowles, Persico, and Todd (2001) develop a model of police search and offender behavior. Their model implies that if police are unprejudiced the rate of guilt should not vary across groups. Using data from Interstate 95 in Maryland, they find equal guilt rates for African-Americans and whites and conclude that the data is not consistent with racial prejudice against African-Americans. This paper generalizes the model of Knowles, Persico, and Todd by accounting for the fact that potential offenders are frequently not observed by the police, and by including two different levels of offense severity. We show that the data is consistent with prejudice against African-American males, no prejudice, and reverse discrimination, depending on the type of equilibrium that exists. Additional analyses, based on stratification by type of vehicle and time of day, do not shed any light on the nature of the equilibrium.