Some two million Americans are currently incarcerated, with roughly six hundred thousand to be released this year. Despite this, little is known about the effects of confinement conditions on the post-release lives of inmates. In this paper we estimate the causal effect of prison conditions on recidivism rates by exploiting a discontinuity in the assignment of federal prisoners to security levels. We find that harsher prison conditions are associated with significantly more post-release crime. We check our identifying assumptions by showing that similar discontinuities do not arise in a control population housed separately from other inmates, and that predetermined correlates of recidivism do not change discretely around score cutoffs. Although our conclusions are somewhat limited due to small sample sizes, in general they are difficult to reconcile with models in which "specific deterrence" or in-prison rehabilitation play a central role, and seem more consistent with models of social interactions or learning.

By M. Keith Chen (Yale School of Management and Cowles Foundation) and Jesse M. Shapiro (University of Chicago)