It should be noted that caution should be taken when understanding the consequence of the test as evidence of officer bias against black individuals. The test conceals heterogeneity in arrest rates across the NYPD precincts. Persico (2013) found that precincts differ significantly in the likelihood that a police stop will lead to an arrest. This heterogeneity may lead to misconception when aggregated together. This misconception can be observed when NYPD precincts are set as distinct jurisdictions such that one police officer from one precinct is unable to stop and search individuals in another precinct.\footnote[4]{In accordance with the Criminal Procedure Law ``a police officer may stop a person in a public place located within the geographical
…show more content…
Persico (2013) found a solution to the problem of precinct aggregation which was to introduce precinct-level fixed effects in the regression that predicts arrest rates. Introducing these fixed effects to the baseline specification allows the effects to absorb the baseline arrest rates. In the theoretical example, the coefficient on the black variable would be estimated as zero. This would be understood as evidence that police officers are unbiased across races. However, if police were biased towards race then the ``black'' coefficient would be non-zero since lower arrest rates would be observed on black individuals and thus the racial difference would be shown in the coefficient, even after controlling for precinct-level fixed effects. …show more content…
The introduction of these effects changes the sign on the ``black" coefficient from negative to positive, therefore when the individual is black the probability of being arrested conditional on being stopped in New York City increases rather than decreases compared to other races. Now this probability of an arrest increases by 0.3253\% to 0.35675\% compared to a stop of another individual. Table 2 shows that precinct-level fixed effects are jointly significant in explaining the probability of being arrested conditional on being stopped in New York City because the p-values are less than 1\%,5\% and 10\%. It should also be noted that 2 out of the three precinct fixed effect specifications are significant, showing that the race indicator variable has a statistically significant influence on the probability of being arrested.
Persico (2013) found a solution to the problem of precinct aggregation which was to introduce precinct-level fixed effects in the regression that predicts arrest rates. Introducing these fixed effects to the baseline specification allows the effects to absorb the baseline arrest rates. In the theoretical example, the coefficient on the black variable would be estimated as zero. This would be understood as evidence that police officers are unbiased across races. However, if police were biased towards race then the ``black'' coefficient would be non-zero since lower arrest rates would be observed on black individuals and thus the racial difference would be shown in the coefficient, even after controlling for precinct-level fixed effects. …show more content…
The introduction of these effects changes the sign on the ``black" coefficient from negative to positive, therefore when the individual is black the probability of being arrested conditional on being stopped in New York City increases rather than decreases compared to other races. Now this probability of an arrest increases by 0.3253\% to 0.35675\% compared to a stop of another individual. Table 2 shows that precinct-level fixed effects are jointly significant in explaining the probability of being arrested conditional on being stopped in New York City because the p-values are less than 1\%,5\% and 10\%. It should also be noted that 2 out of the three precinct fixed effect specifications are significant, showing that the race indicator variable has a statistically significant influence on the probability of being arrested.