RACE ON POLICING: Introduction 2
Most of the analysis to date has taken the form of ethnographic research and case studies (Groves and Rossi 1970, Mast 1970, Bordua and Tifft 1971, Skolnick and Bayley 1986, Alpert and Dunham 1987). Although these studies are immensely useful, more systematic quantitative evidence is also needed. The quantitative research that has been done typically has focused on police and community perceptions (Decker and Smith 1980, Lasley 1994), but has stopped short of looking for impacts on tangible social outcomes such as reductions in crime or patterns of arrest.3 The lone exception that we are aware of is Lott (1997) which focuses on the impact of affirmative action in policing on aggregate crime ratesfurther.
In this paper we first develop a simple model for analyzing the impact of race on police and the communities they serve, allowing for the possibility of false arrests, different standards of guilt for making arrests within and across racial lines, and police corruption. As the model demonstrates, different patterns of arrests and crime provide evidence about how race influences policing.
We then analyze empirically the relationship between minority representation in policing, arrest rates, and crime rates. The analysis combines publicly available city-level Federal Bureau of Investigation (FBI) data on crime and arrest rates and data on the racial composition of municipal police departments compiled over the last twenty years by the Equal Employment Opportunity Commission (EEOC). The EEOC data set is far superior in both the breadth of cities covered and the number of years for which data are available relative to data sets used in earlier studies of minority policing (Walker 1983, Hochstedler and Conley 1986, Stokes and Scott 1993). While aggregated statistics have been published by the EEOC, the city-level data have not previously been exploited for scholarly research.
We find evidence across a wide range of crime categories that own-race policing is associated with lower numbers of arrests than cross-race policing. The results with respect to crime rates are less clear-cut and far more sensitive to the choice of specification. Own race policing appears more effective in lowering property crime rates, but does not differ systematically from cross-race policing with respect to violent crime.4 Taken together, these results suggest that for property crime own-race policing is more “efficient” than same-race policing, i.e. similar or better crime outcomes are obtained with fewer arrests required.
This implies either that deterrence is greater with own-race police (due possibly to greater community cooperation) or that fewer false arrests are made. The magnitude of our results is substantial. For the mean city in our sample, reallocating police from random assignment by race to an assignment that maximizes own-race policing (holding the number and racial composition of the police constant) is predicted to reduce arrests by more than 10 percent while decreasing property crime by as much as 20 percent. Violent crime is unaffected.