We study the efficiency of non-compete agreements (NCAs) in an equilibrium model of labor turnover. The model is consistent with empirical studies showing that NCAs reduce turnover, average wages, and wage dispersion for low-wage workers. But the model also predicts that NCAs, by reducing turnover, raise recruitment and employment. We show that optimal NCA policy: (i) is characterized by a Hosios like condition that balances the benefits of higher employment against the costs of inefficient congestion and poaching; (ii) depends critically on the minimum wage, such that enforcing NCAs can be efficient with a sufficiently high minimum wage; and (iii) alone cannot always achieve efficiency, also true of a minimum wage-yet with both instruments efficiency is always attainable. To guide policy makers, we derive a sufficient statistic in the form of an easily computed employment threshold above which NCAs are necessarily inefficiently restrictive, and show that employment levels in current low-wage U.S. labor markets are typically above this threshold. Finally, we calibrate the model to show that Oregon's 2008 ban of NCAs for low-wage workers increased welfare, albeit modestly (by roughly 0.1%), and that if policy makers had also raised the minimum wage to its optimal level (a 30% increase), welfare would have increased more substantially-by over 1%.