By exploiting discontinuities in the assignment of inmates to different security levels, this study attempted to isolate the causal impact of prison conditions on recidivism. The findings suggest that harsher prison conditions do not reduce post-release criminal behavior, and may even increase it. While the estimates are imprecise, they are large in magnitude and appear larger than benchmark estimates of deterrence effects. The results highlight the potential importance of research aimed at determining which aspects of incarceration increase or reduce recidivism. A richer understanding of the ways inmates respond to both harsher prison conditions and exposure to more violent peers would likely allow policymakers to suppress socially costly recidivism by adjusting conditions and redesigning assignment (classification) systems. With over 2 million inmates currently incarcerated and 600,000 inmates released per year, the demographic impact of American prisons can hardly be overstated. Theory alone cannot tell us whether an increase in the severity of prison conditions will increase or decrease the propensity of inmates to commit crimes after release. This article attempts to understand the impact of incarceration on inmates’ subsequent criminal behavior. It estimates the causal effect of prison conditions on recidivism rates by exploiting a discontinuity in the assignment of Federal prisoners to security levels. Figures, tables, appendix, and references