Dissecting Anomalies in Conditional Asset Pricing

We develop a methodology for estimating and testing the effect of anomalies in conditional asset pricing models when premia are time-varying. Our method, which builds on the two-pass methodology, is developed for ordinary and weighted least-squares estimation, considering both cases of correct specification and global misspecification of the candidate asset pricing model. A cross-sectional R-squared test to dissect anomalies is proposed, establishing its limiting properties under the null hypothesis of no effect of anomalies and its alternative. Using a dataset of 20,000 individual US stock returns, we find that although anomalies are statistically significant in about half the cases (out of 170 anomalies), they explain a small fraction (less than 10%) of the cross-sectional variation of expected returns. Anomalies tend to be more important during economic and financial crises