The Mathematics of Actuarial
Prediction: The Illusion of Efficiency
One of the strongest arguments for the use of actuarial methods is the economic argument based on deterrence and efficiency: assuming that people respond rationally to the costs and incentives of policing, using predictions based on group offending rates will result in greater detection of crime. By maximizing the detection of crime, law enforcement will deter the higher-offending targeted population. This is the most efficient allocation of law enforcement resources.
A number of able economists have turned their attention to demonstrating this more rigorously. They are developing models of criminal profiling and demonstrating that using actuarial methods may be an efficient way to engage in law enforcement—in fact, that profiling based on group offending rates may be the most efficient way to allocate police resources. They are laying out the rational-action argument for profiling and actuarial methods in its most pristine form. The basic idea is that law enforcement should use group offending rates in order to make inferences about individual offending: for example, to pick motorists for searches based on their offending rates by race. They argue that this will reduce the offending of higher offenders and maximize the success rate of police searches. This in turn is viewed as the most efficient allocation of resources. Let me begin by setting forth the economists' models.
Drawing on Gary Becker's groundbreaking work on tastes for discrimination,1a group of economists—notably John Knowles, Nicola Persico,