This paper studies the cyclical dynamics of Mortensen and Pissarides' (1994) model of job creation and destruction when workers' effort is not perfectly observable, as in Shapiro and Stiglitz (1984). An occasionally-binding no-shirking constraint truncates the real wage distribution from below, making firms' share of surplus weakly procyclical, and may thus amplify fluctuations in hiring. It may also cause a burst of inefficient firing at the onset of a recession, separating matches that no longer have sufficient surplus for incentive compatibility. On the other hand, since marginal workers in booms know firms cannot commit to keep them in recessions, they place little value on their jobs and are expensive to motivate. For a realistic calibration, this last effect is by far the strongest; even a moderate degree of moral hazard can eliminate all fluctuation in the separation rate. This casts doubt on Ramey and Watson's (1997) "contractual fragility" mechanism, and means worker moral hazard only makes the "unemployment volatility puzzle" worse. However, moral hazard has potential to explain other labor market facts, because it is consistent with small but clearly countercyclical fluctuations in separation rates, and a robust Beveridge curve.