``Racial'' Profiling Doesn't Work

My forthcoming book (Us and Them, Little, Brown) is about all the categories that people use for people -- categories like race, nation, religion, culture, to name a few, for which the best general term is ``human kinds.’’ One of the themes of the book is that these categories exist because of the way the human mind is organized, not because they reflect some ultimate reality. People tend to assume too much about the merits of these categories as guides in life.

``Profiling’’ in law enforcement is an example. One sense of the word is strictly logical, and involves inferences about cause and effect: If your database shows that all bank robberies in town for the past ten years were committed by tall men wearing blue who bought ski masks on Fridays, you can say there’s a connection between height, clothing, purchasing patterns and the likelihood of being a bank robber. This is the kind of connection-making that various government and private-sector projects are trying to ramp up, in the name of fighting terrorism. It’s the point of what is loosely called ``data mining.’’

But if you use a category like apparent race or culture to decide which people to investigate, you’re not using well-defined, logical concepts like ``people who bought ski masks on Friday.’’ Instead, you’re using a ``human kind’’: A set of beliefs whose roots are in the unconscious workings of your mind. And you’re asserting that this category -- say, Hispanics -- has a connection to the category you’re interested in, which is, say, ``people smuggling drugs.’’

That kind of thinking is sometimes attacked on grounds of ``political correctness.’’ Let’s not assert that ethnic groups differ from one another! And let’s treat everyone as an individual! Cops are right to treat that kind of talk with contempt: Life throws up situations in which people can’t be treated as individuals, and a traffic stop is one of them. And everyone knows that people of different cultures act different.

A much more important argument against ethnic profiling is rooted in the unsound assertion of connection between logical categories, like ``people carrying contraband today’’ and cultural ones, like ``people who look Hispanic.’’ Put simply, ethnic profiling does not work. When you look at the statistics, you find that the connection is not there.

I document a few examples in the book, but here is one that just came out this week: A study of traffic stops in Texas. The bottom line :

The researchers examined reports of several million traffic stops by 1,060 police and sheriff's departments across Texas. It found that about two of every three agencies searched blacks and Hispanics at higher rates than those for non-Hispanic whites.

But the higher rates did not translate into a higher rate of seizing drugs or other contraband, the study found. The agencies that searched blacks at a higher rate were likely to recover drugs from blacks and whites about equally. Those that searched Hispanics at higher rates recovered drugs more often from whites than from Hispanics.