This paper examines the effectiveness of traditional regulatory schemes and newly emerging social information schemes for achieving compliance. Our experiment focuses on two stochastic audit schemes for enforcing regulatory compliance. In the Random Audit mechanism firms are randomly chosen for inspection. In the Tournament Audit mechanism the probability of inspection increases with the degree of estimated underreporting. To study the effects of social information, the experiment varies the observability of identity, output, and compliance decisions. Optimal output is theoretically independent of the auditing scheme, but equilibrium reporting is higher under the Tournament mechanism than Random auditing. Experimental findings are broadly consistent with the theoretical predictions for reporting, but deviate modestly for output. In particular, we find that average output is lower and reporting is higher in the Tournament treatment compared to the Random Audit treatment. At the individual level, a majority of participants misreported in most periods. Social observability does not affect output or reporting significantly in either of the audit treatments.