Recidivism prediction instruments provide decision makers with an assessment\nof the likelihood that a criminal defendant will reoffend at a future point in\ntime. While such instruments are gaining increasing popularity across the\ncountry, their use is attracting tremendous controversy. Much of the\ncontroversy concerns potential discriminatory bias in the risk assessments that\nare produced. This paper discusses a fairness criterion originating in the\nfield of educational and psychological testing that has recently been applied\nto assess the fairness of recidivism prediction instruments. We demonstrate how\nadherence to the criterion may lead to considerable disparate impact when\nrecidivism prevalence differs across groups.\n