U.S. flag

An official website of the United States government, Department of Justice.

Against Prediction: Profiling, Policing, and Punishing in an Actuarial Age

NCJ Number
Bernard E. Harcourt
Date Published
336 pages
This book critiques the actuarial method as the basis for making decisions in criminal justice activities and rejects this method in favor of what the author calls "randomization."
Actuarial methods of decisionmaking in administering criminal law involve the identification of genetic, personality traits, background experiences, and past behaviors of individuals so as to predict their risk for future or even past criminal behaviors. Testing instruments place individuals in categories of risk. This book challenges this actuarial practice for three reasons. First, high-risk groups targeted for more intense law enforcement efforts will increase the number of detected offenses in such groups, rather than reduce crime. Second, the reliance on probabilistic methods produces a distortion in the characteristics of the incarcerated population, as populations profiled as high-risk dominate the prison and jail populations. Third, the proliferation of actuarial methods has begun to bias the conception of just punishment. Individuals are punished, not only for specific criminal behavior and the severity of the harms it has caused, but also for what they might do or may have done. The author argues for "randomization" in law enforcement and criminal justice processing, i.e., eliminating judgments based on predictions of future offending or what the individual may have done in the past. Instead, individuals should be judged only on the criminal behavior for which they are convicted and the harms it has caused. Predictions of future or past behavior based on testing should not be an issue in criminal justice decisionmaking. Chapter notes, extensive references, a subject index, and appendixes that retrace the parole-prediction debate and literature, as well as mathematical proofs regarding the economic model of racial profiling.