Economic Research on Profiling
IPR economist leads effort to show how economic thinking can illuminate the nature and consequences of profiling

Spring 2007, Volume 29, Number 1

Charles F. Manski (far right) meets with graduate student Ben Handel and colleague Joel Horowitz.

Allegations of racial and ethnic profiling abound in the media—from claims of Rhode Island police using racial profiling in the unconstitutional detention of 14 Guatemalans on I-95 to loan officers redlining poor minority families and airlines removing Muslim passengers from flights. Yet until recently, racial and ethnic profiling has received little evidence-based scrutiny.

Starting in 2002, a group of 11 economists met periodically to explore substantive and methodological issues in analyzing social interactions. They quickly turned to racial and ethnic profiling, seeing “a pressing need to bring serious theoretical and empirical analysis to bear on a subject of enormous controversy,” said Charles F. Manski, Board of Trustees Professor in Economics and an IPR faculty fellow. He successfully applied for National Science Foundation funding and subsequently led and organized the resulting research network until it ended in 2006.

Manski points out that the term profiling, which is a relatively recent lexicon entry, comes from public discourse and is not formally defined in economics. The Oxford English Dictionary (OED) first published its modern definition of profiling (see (b) below) in its online edition in 2004.

According to Manski, the definition of profiling indicates why it is so controversial. One definition (a) is broad and neutral; it describes cataloguing and analyzing a person’s “known characteristics” to evaluate his or her competence. The second definition (b), however, is a narrow statement that focuses on identifying a person by his or her “superficial characteristics” such as race or ethnicity for the sole purpose of further scrutiny with no evidence of wrongdoing.

Though economists have long studied problems related to the OED’s broad definition (a) of profiling, Manski notes, it was a 2001 study by John Knowles, Nicola Persico, and Petra Todd of the University of Pennsylvania on racial profiling in traffic stops that prompted more serious scrutiny of search profiling as defined under (b). Their model of police and driver behavior shows that if a police officer’s goal is to catch as many guilty motorists as possible, then lower rates of crime detection, or hit rates, for minority groups such as blacks would signal that the police officer was racially motivated. This is because an unbiased police officer would see that he or she could catch more wrongdoers by searching white drivers instead. “Their work did much to shed light on profiling, but we knew it was only a beginning,” Manski said.

The resulting papers, which cover a diverse range of areas including university admissions, welfare programs, and loans, were published as a special feature of the U.K.-based Economic Journal in November 2006. The National Science Foundation provided financial support for the workshops and conferences.

The first two articles, by Nicola Persico and Petra Todd and by Jeff Dominitz and John Knowles, shed new light on the validity of the hit-rate test for discrimination. Persico and Todd find that hit rates are similar for all races and ethnicities. Thus, police officers searching for drugs and weapons appear to pick strategies that will maximize successful searches, rather than ones that demonstrate racial bias. Dominitz and Knowles consider an alternative to the Knowles, Persico, and Todd basic assumption on police behavior. They assume that an unbiased police officer aims to minimize crime rather than maximize successful searches. They evaluate the hit-rate test under this assumption and find that it is valid in some cases but not others.

The next two articles are concerned with normative aspects of search profiling policy. Previous research assumes that a policy planner knows the deterrent effect of search on offense rates and thus can select an optimal search policy. Manski observes that this is an unrealistic assumption, and he shows how a policy planner with only partial knowledge of deterrence can reasonably choose a strategy. Steven Durlauf argues that incomplete knowledge of the deterrent effects of search profiling should lead one to reject a public policy of using profiling in traffic stops because of the uncertainty as to whether it reduces crime and the harm that it causes to innocent African Americans.

The final four articles consider profiling problems that arise in areas outside of police searches—mortgage lending, means-tested transfer programs, college admissions, and the operation of labor markets. William Brock and Robert Moffitt examine social planning problems. Dennis Epple, et al., and Lawrence Blume show how economic analysis can shed light on the version of profiling emphasized in OED definition (a); that is, the use of data on a person’s attributes to assess his or her capabilities.

“We hope that the reports in this feature demonstrate how basic economic thinking about the interaction of profilers and profilees who respond rationally to incentives can illuminate controversial issues of public policy and contribute to their resolution,” Manski said.

The articles can be found in the online edition of The Economic Journal at www.res.org.uk/economic/ejbrowse.asp.

Related Articles

Knowles, J., N. Persico, and P. Todd. 2001. Racial bias in motor vehicle searches: Theory and evidence. Journal of Political Economy 109(1): 203-29.

Manski, Charles, F., et al. 2006. The economics of “profiling” (special feature). The Economic Journal 116(515): F347-F498.