He applied the “Page 99 Test” to his new book, Suspect Race: Causes and Consequences of Racial Profiling, and reported the following:
I’m a social psychologist and a public policy professor, and I’m also a science geek and a stats teacher, so I don’t believe in fate. Still, I can enjoy a happy coincidence. So when the Campaign for the American Reader asked me to talk about page 99 of my new book, Suspect Race, and I opened it to find it is where I described how I got started down the path of studying racial profiling, I felt lucky.Learn more about Suspect Race at the Oxford University Press website.
From page 99:IT’S EFFECTIVE? ESTIMATING THE EFFECTIVENESS OF RACIAL PROFILINGPage 99 tells the story of how, in 1999 (there’s 99 again!), I read an article by a prominent legal scholar who was arguing against racial profiling on Constitutional grounds, but conceding that profiling is rational and effective. As a stereotyping researcher, my first reaction was, “how can we know?” If police are profiling, then they will be skewing the criminal justice statistics that promote the belief that Blacks and Latinos commit more crime. I set about trying to explore this possibility empirically but quickly ran into “the benchmark problem,” which is that it’s nearly impossible to tell where profiling is happening because we do not have a proper benchmark that tells us what the appropriate stop/search rates would be.
In 1999, Harvard law professor Randall Kennedy published a cogent indictment of racial profiling, arguing in The New Republic that profiling violates basic American constitutional tenets and contributes to Black disenfranchisement and resentment. Notably, like others, Kennedy stipulated that racial profiling made sense on efficiency grounds. It was Kennedy’s excellent essay that sparked my interest in the subject. His legal and moral arguments, like those of many of his colleagues (e.g., Alschuler, 2002; Banks, 2001, 2003), were compelling. But his readiness to concede that profiling is efficient caught my attention.
My initial insight was that if police are already profiling, it would be difficult, if at all possible, to determine its effectiveness because they would have systematically altered the criminal justice landscape. Specifically, profiling would, regardless of any actual differences in offending rates, cause minorities to be overrepresented in the criminal justice system (including prison); if minorities are stopped more because they are minorities (and Whites are therefore stopped less because they are White), more minority (and fewer White) criminals will be caught.
This distortion will have two effects that may result in a self-perpetuating cycle: Stereotypes of minorities as criminals will be reinforced by their overrepresentation in the criminal justice statistics; and those stereotypes will be used to further rationalize continued racial profiling. At the time of this insight, without yet having given the subject much thought, it was only vaguely donning on me that another effect was that, as more minority criminals and fewer White criminals were caught and incarcerated, their proportions in the unincarcerated populations would change&there would be fewer minority and more White criminals at large. Consequently, profiling would lead to a double distortion of…
To get around this problem, I ran mathematical simulations of what would happen if police stopped different groups at varying rates, also varying the rate at which the groups were committing crimes. The simulations showed how merely profiling will cause disparities in incarceration rates. What I hadn’t expected was that most of the scenarios showed that profiling was not good at increasing criminal capture rates. Even when the profiled group had a higher offending rate, profiling yielded modest gains. When the profiling was out of proportion with the offending, it was generally counterproductive.
Now I am working the Center for Policing Equity and some of the country’s biggest city police departments to improve the state of data on policing in America so that we can better determine the rate of profiling and the extent of its impact using real statistics, not just simulations.