Hit Rate Analysis Examples

Earlier work by Gelman et al. (2007) presented concern that the arrest outcome of “hit rate analysis” may be an issue. They stated that a perfect outcome of the analysis would be a measure of officer productivity which the officer aims to maximise, this objective is impartial to racially bias behaviour and cannot be influenced by police officer bias of black individuals. The arrest outcome may not be impartial to officer bias because arrests are subject to the police officers decision and thus could be subject to racial bias. This matter of interest could invalidate “hit rate analysis”. By studying the behaviour of the officer after the stop has been made can focus on the invalidation concern. It is possible to determine whether the police

An African-American individual is 0.356\% less likely to be arrested given the type of crime committed relative to other races. When precinct-level fixed effects are included in the regression, the sign of the coefficient changes from negative to positive, this is the same as Table 2. When the regression also conditions for precincts, an African-American individual is between 0.265\% and 0.294\% more likely to be arrested conditional on being stopped and on the type of crime committed. All six specifications of the regression adjusted for crime are insignificant, therefore the race of the individual does not statistically influence the likelihood of being arrested when conditioned for the type of crime committed. This result follows Persico (2013) where African-American individuals are 0.495\% less likely to be arrested given the type of crime committed relative to other races and when precincts effects are included, between 0.228\% and 0.234\% more likely to be arrested given the type of crime committed relative to other races. Persico (2013) finds that a significant race effect does not exist. It can, therefore, be concluded that after conditioning on the type of crime there is not a significant interaction between the race of the individual and the probability of being arrested. Hence police officers do not use discretion during the arrest decision, or discretion which is uncorrelated with race is used. Given this, the arrest outcome of hit rate analysis is valid.