I use Mahalanobis distance based on investment opportunity variables to define turbulent periods within financial markets. The distance measure identifies periods of event-driven stress, and not necessarily low returns. CAPM betas estimated from normal sample periods explain vary little variation in cross-sectional returns. However, betas estimated from turbulent subperiods explain a large proportion of full-sample returns. Market betas for small and value portfolios increase during turbulent periods, indicating that the risk of these portfolios is greater than indicated by standard betas, and suggesting an explanation for these anomalies.