Search Profiling with Partial Knowledge of Deterrence (WP-05-05)
Charles F. Manski
Though economists engaged in normative study of public policy generally assume that the relevant social planner knows how policy affects population behavior, this rarely is the case in practice. Fundamental identification problems and practical problems of statistical inference make it difficult to learn how policy affects behavior. Hence, there is much reason to consider policy formation when a planner has only partial knowledge of policy impacts.
In this paper, I examine a specific and recently debated aspect of law enforcement—the choice of a profiling policy. My concern is not so much to understand the use of personal attributes such as race in profiling policies (though some of my analysis does have implications for detecting discrimination), but rather to understand how a social planner might reasonably choose a profiling policy when he or she only has partial knowledge of how policy affects criminal behavior. I consider both ex ante search, which apprehends offenders before their offenses cause social harm, and ex post search, which apprehends offenders after completion of their offenses. This paper shows how a social planner having partial knowledge of population offense behavior can “reasonably” choose a search profiling policy.