Abstract: Certificates of Rehabilitation (CORs) are a promising policy tool to help workers with a criminal record (WCs) gain access to employment and avoid recidivism. While several states have passed relevant legislation in recent years, take-up has been low, in large part because of a lack of awareness.

This paper develops a model of CORs to understand how their increased prevalence could affect outcomes for WCs. Workers differ in unobserved risk type and decide whether to apply for certification, and later, whether to reoffend. Firms cannot observe type and instead rely on certification status as a signal when making hiring decisions, accounting for both productivity and expected litigation risk. In equilibrium, under mild assumptions, low-risk workers are more likely to apply for and receive certification, generating positive selection into the certified pool, and firms statistically discriminate against the uncertified pool.

As awareness of CORs increases, under a single crossing assumption, more low-risk workers obtain certification, improving outcomes for certified individuals through higher employment and lower recidivism. However, this reallocation worsens the composition of the uncertified pool, making it disproportionately high-risk. Firms respond by reducing hiring of uncertified workers, which further increases their recidivism through both direct compositional changes and reduced employment.

Since CORs have had little take-up, one would expect state legislation to have a negligible effect on aggregate outcomes for WCs. Using four different methods on publicly available data, I find no consistent effects. 
In sum, the theoretical results motivate a more thorough empirical analysis with better data. They also highlight an important tradeoff: expanding access to CORs benefits those who obtain certification but imposes negative spillovers on those who do not.