A Critique of the “Outcome Test” in Racial Profiling Research

In racial profiling research, four different research perspectives—legalistic, criminological, economic, and normative—have emerged. The analytical techniques of two of these perspectives, the legalistic and criminological, have been thoroughly detailed in prior research. More recently, the economic perspective has presented an alternative analytical strategy (i.e., the outcome test), to determine racial and ethnic discrimination by police. When applied to police searches, the outcome test is a statistical comparison of search success rates across racial and ethnic groups. The outcome test, however, is based on a number of underlying assumptions regarding police and citizen behaviors that do not coincide with what is known about decision‐making during police–citizen encounters. These underlying assumptions are described and the conclusions based on the use of the outcome test are critiqued. The implications for the use of the outcome test and other analytical techniques for the future of racial profiling research are discussed. Keywords: Outcome Testpolice biasracial profiling researchtraffic stopssearch rates Acknowledgments A previous version of this manuscript was presented at the 2005 Criminology and Economics Summer Workshop, hosted by the University of Maryland. I extend my sincere appreciation to all of the participants of this workshop for their thoughtful comments and interesting discussion. I am also indebted to Rob Worden and John Eck for initially challenging and ultimately clarifying my thoughts on the use of the outcome test. Notes 1. As noted by Hanushek and Jackson (Citation1977, p. 287), "The fundamental problem we face with models containing erroneous or unobserved variables is the absence of information about the true variables and how they covary with the observed variables. This deficiency quickly leads to unidentified models. There are an infinite combination of values for the structural coefficients and error variances that could produce the variances and covariances observed among the measured variables."