Racial profiling to limit terror attacks is fundamentally flawed, expert says

This is the topic of a paper publishing in Significance, the magazine of the Royal Statistical Society and the American Statistical Association.

Racial profiling rests on the idea that people from particular racial or ethnic groups are more likely to be involved in acts of terror than people from other groups. The theory then suggests that law enforcement officers should spend a greater proportion of their time scrutinising people from the 'high risk' group. One problem with this approach is that innocent people who also belong to the targeted group rapidly become offended, and some may even become radicalised as a result.

"Racial profiling is as indiscriminate as deciding that people named Patrick are more likely to drink and drive, and so everyone who is named Patrick should be stopped and breathalysed more frequently than people with other names," says Press. Although it may be the case that some people named Patrick do drink and drive, he points out that there are clear problems in this profiling strategy. First, many Patricks don't drink and drive and will be unfairly detained. Secondly, this sort of profiling can have the appearance of success because if you keep testing more people named Patrick than other people, you are almost bound to find more Patricks who have drunk alcohol than people with other names. This then leads you to think the problem is even worse than you first suspected, and so you further increase the targeting of Patricks.

In the Significance paper, Press, based in the departments of Computer Science and Integrative Biology at the University of Texas at Austin, takes a thorough mathematical and statistical view of the process that underlies racial profiling, and concludes that some forms of racial profiling may even result in a smaller chance of detaining a terrorist than carefully conducted standard sampling.

In a world threatened by terrorists from a small number of countries, it is tempting to think that racial profiling for security purposes, even if morally objectionable, might save lives. "But uniform sampling, without the use of profiling, is surprisingly good. It is robust against false assumptions, it is deterrent, it is easy to implement, it is about as effective as any real-life system can be -- and it is devoid of moral and political hazard," says Press.

He believes that the choice between a strategy of profiling and one of uniform random sampling should not be viewed as difficult; uniform random sampling wins.