Measuring Racial Profiling
26 Pages Posted: 7 Jul 2011 Last revised: 4 Oct 2011
Date Written: July 7, 2011
What if racial profiling were useful? Even, perhaps, very useful? Are the costs of racial profiling so significant that racial profiling should be banned? Courts and commentators do not ask these questions; indeed, they rarely, if ever, acknowledge that racial profiling may be useful. This paper explores what costs police would have to justify, and how they might do so in order to use racial profiling legally.
Using race as a part of a profile of criminal has some efficiency value; it is not solely (or even at all) about racial animus. Race is a marker of criminal behavior; it would be miraculous if it where not. Height, weight, gender, hair style, ear piercings, all of these characteristics almost surely delineate some difference in the commission rate of at least one crime. This is, essentially, a mathematical truism: race and crime are not independent. Because race is “useful” when “useful” is defined as “information worth more than zero,” it is important to move beyond the initial question - is it useful - to the more important question - how useful is it? What do we give up in allowing or banning racial profiling? What are courts missing when they discuss racial profiling? This paper is an attempt to get at least some of the empirics correct regarding the direct costs and benefits of racial profiling; it leaves the important research of mapping the collateral consequences of racial profiling to others.
But even when one puts aside the collateral consequences of focusing the might of the criminal justice system on one specific race, there are many complicating factors in measuring how useful racial profiling. First, how do you measure the gap in crime rates (that is, the different crime rates for different racial groups and different crimes)? Without this, no one could calibrate their behavior: police would not know how much to profile, and the courts would not know how much reliance on race in police decision-making is too much. This is also important as a policy matter, to determine what society loses if racial profiling were not allowed. To be more concrete, what if whites were twice as likely to sell cocaine as Blacks were? Does this justify profiling them? Perhaps the answer to this depends on the base rate: how much more cocaine do white people sell? Is this a difference between one and two percent, or fifteen and thirty percent of individuals who could be targeted. One needs to know more than just a single statistic. A second complicating factor, beyond the initial measurement question, is whether race is too easy to over-use. People see race quickly, and believe it is a highly salient part of a persons physical description; police might over-rely on this characteristic because it is easy to determine (or at least perceived to be so). A third complication is that race is rarely the only available information about a potential suspect, and a better question than whether race is correlated with crime rates is whether information about a suspects race adds any useful information to a profile. But, again, the chance that these two probabilities are always equal, for every type of crime and any other possible set of information is zero. Race matters. This bears emphasis not because it isn't obvious, but because many opposed to racial profiling deem it inefficient automatically; that is, they argue that racial profiling does not work because race is not a predictor of criminality.
Answering these questions is a difficult, but not insurmountable task. But these questions also suggest that a simple focus on whether race is a predictor of criminality is misplaced; the answer to that first question is yes. The more important question, which legislatures, courts, legal and economic researchers alike have generally ignored, is when and how useful racial profiling is in different situations. To answer this question, one needs to investigate the costs and benefits of racial profiling, which the above complicating factors suggest can be done, but needs more attention to the detail of the reality of the relationship of race and crime to answer. This paper seeks to quantify the costs and benefits of racial profiling, and in doing so creates new methods to control for the problem of selection bias in the data that are generally available to evaluate this question.
Suggested Citation: Suggested Citation