Locked In: Cell Level Peer Effects and Recidivism in US Jails

• In Event: Jails, Pretrial Detention, and Reentry: Evidence on Peer Effects, Prosecution, and Pathways to Work

Abstract

Jails process roughly ten million bookings each year in the United States, yet they remain the least-studied part of the carceral system relative to prisons. Within jails, people live at close quarters in involuntary groups, and little is known about whether the composition of one's cellmates affects behavior after release. Prior work on peer effects in correctional settings has largely measured exposure at the facility level, obscuring the daily social environment most likely to matter for behavior.

This paper estimates the causal effect of cellmate composition on offense-specific recidivism using daily cell-level housing assignments from a dozen county jails spanning ten states, linked to administrative records on subsequent arrests and bookings. I exploit the staggered timing of admissions and releases within individual cells to identify peer effects.

I find significant offense-specific reinforcement across all four major crime categories: drug, property, public-order, and violent. A ten percentage-point increase in same-type peer share raises category-specific recidivism by roughly two to four percentage points, with the largest effect on property recidivism. The pattern replicates across jails of very different size, region, and demographic composition. Cross-type effects point in the opposite direction. Exposure to cellmates with unlike criminal histories offsets rather than deepens offending, suggesting a more specialized form of peer learning in which cellmates shape what kinds of offenses people commit, not simply whether they reoffend at all. The core findings survive methods from the competing risks literature that relax assumptions about how different types of rearrests interact; point estimates attenuate modestly but remain economically and statistically significant.

I then run a counterfactual exercise that translates the estimates into implied changes in recidivism under alternative housing assignment rules, quantifying how much aggregate offense-specific recidivism could shift if jails redistribute cellmates across existing cells without changing who is detained, for how long, or under what conditions.

These findings identify a rarely-measured channel through which short-term jail detention can have lasting effects on criminal trajectories. Housing assignments are already made every day during booking and classification, making this one of the few margins in the criminal-legal system where a policy change requires neither new legislation nor new facilities.

Author

• Jake Anderson, University of California - Los Angeles
  Presenting Author