This paper studies the earnings and employment penalties associated with criminal records using linked data on criminal justice interactions, state laws on formal labor market restrictions, and wages and employment. We estimate fixed effects models that decompose earnings into a worker's portable earnings potential and firm pay premia, both of which are allowed to shift after a worker acquires a record. We find that sorting into lower-paying firms explains a small share of the record penalty. Legal record restrictions are not concentrated in higher-paying jobs, meaning that sorting away from them contributes little to the earnings gap. Conditional on earnings potential, however, workers with a record are substantially less likely to be employed. The results suggest that criminal record penalties operate primarily by reducing employment and lowering earnings potential at every firm rather than increasing sorting into lower-paying jobs, although the bulk of the gaps can be attributed to differences that existed prior to acquiring a record.