STOP-AND-SEARCH POLICING: NEW EVIDENCE THAT STRATEGIES ARE BASED ON ‘HIT RATE’ MAXIMISATION NOT RACIAL BIAS

New research on how police officers decide which types of drivers to stop and search for drugs and weapons suggests that they choose their search strategies to maximise successful searches and not necessarily out of racial bias. The study by Professors Nicola Persico and Petra Todd, published in the November 2006 issue of the Economic Journal, looks at evidence from Wichita, Kansas, and finds that rates of detection of contraband – ‘hit rates’ – are remarkably similar for drivers of all races/ethnicities.

The researchers also find that hit rates are virtually identical for drivers aged over 50 and for 18-24 year olds, as well as for men and women. But, they point out, equalising the hit rates in this way does not guarantee ‘maximum deterrence’. If the police wanted to maximise deterrence, they would have to go after groups that are more easily deterred, if there are any such groups, and thus might have to accept unequal hit rates.

Over the last ten years, numerous lawsuits have been brought against US law enforcement agencies alleging racial profiling. The pattern that prompts these lawsuits is familiar: minorities are disproportionately the target of interdiction by law enforcement. In ordering an investigation into possible racial profiling, President Clinton condemned the practice as ‘the opposite of good police work where actions are based on hard facts, not stereotypes’.

But precisely what police work should be considered lawful because it is based on ‘hard facts’ as opposed to unlawful because it is based on ‘stereotypes’? Recent economic research has make progress in this direction. The test case is that of Wichita, a town in Kansas that exhibits the type of interdiction pattern that gives rise to concerns of racial profiling: black people are searched for contraband three times as often as their representation in the Wichita population, and three times as often as white people.

The disparity, of course, is not only between black and white people. Compared with older people (those over 50 years old), for example, young people (those aged 18-24) are 9 times as likely to be searched. What to make of these disparities? Are they evidence of racial bias on the part of the police or do they reflect goal-oriented policing?

To assess which of these is the correct motive, Professors Persico and Todd develop a behavioural analysis of policing that is designed to simulate the decisions made by police officers and drivers in Wichita.

In their analysis, officers choose which types of drivers to search. Drivers decide whether to carry illegal contraband (such as drugs or weapons), taking into account the probability that they will be detected through a search. As the probability of detection increases, drivers are less likely to carry illegal contraband.

The key insight obtained from this analysis is that unbiased police should sample drivers to the point where detection rates are equalised across different observable categories of drivers. If the rates were not equal, then an officer could do better by reallocating efforts to a category of driver with a higher detection rate.

On the other hand, racially biased police will tend to ‘over-search’ certain types of drivers, for example, minority drivers, to the point where the carrying rate (and thus the detection rate) is lower for that type. One can therefore infer racial bias by testing whether detection rates (‘hit rates’) are equal across different categories of drivers.

When this ‘hit rate test’ is employed, the data show a remarkable equalisation in the detection rates (about 22%) for drivers of all races/ethnicities. This finding is consistent with the notion that police in Wichita choose their search strategies to maximise successful searches and not necessarily out of racial bias. Moreover, the hit rate is virtually identical for drivers aged over 50 (20.6%) and those aged 18-24 (21.7%), as well as for a variety other characteristics (such as gender of driver).

This finding is consistent with earlier work by these researchers with John Knowles on Maryland data and with a review of racial profiling studies for a number of other cities and states. The overall picture that emerges from employing the hit rates analysis provides support for the conjecture that the difference in search rates among categories of drivers arises mainly from a goal-oriented policing motive as opposed to racial bias.

ENDS

Notes for editors: ‘Generalizing the Hit Rates Test for Racial Bias in Law Enforcement, with an Application to Vehicle Searches in Wichita’ by Nicola Persico and Petra Todd is published in the November 2006 issue of the Economic Journal.

The authors are at the University of Pennsylvania.

For further information: contact Nicola Persico on +1-215-898-7711 (email: nicola.persico@nyu.edu); or Romesh Vaitilingam on 07768-661095 (email: romesh@compuserve.com).