Recidivism as a feedback process: An analytical model and empirical validation

Abstract A feedback model of the criminal justice system (CJS) incorporates the continuing input of people arrested for the first time (virgin arrests) and the recycling of individuals with prior arrests (recidivists). Such a model is needed to enable CJS planners to assess the impact of possible actions on the future arrests and system workloads. Using an empirically determined estimate of the number of virgin arrests in the U.S. as input to a feedback model of the CJS, recidivism parameters, probability of rearrest, and average time between arrests were estimated by matching the output of the model to the total arrests in the U.S. in the period 1960–70. The average deviation between the model output and total U.S. arrests was minimized at less than 4% when the probability of rearrest is equal to 0.875 and the average time between arrests equal to 1.1. years. The relative sensitivity of total arrests to changes in virgin arrests and the probability of rearrest are also presented.