We test the prediction of recent theories that stocks with high idiosyncratic skewness should have low expected returns. Because lagged skewness alone does not adequately forecast skewness, we estimate a cross-sectional model of expected skewness that uses additional predictive variables....