Influential quarters in cross-sectional asset-pricing tests.

摘要

A large body of empirical evidence now shows that certain macroeconomic series explain cross-sectional differences in stocks' average returns. Using a jackknife analysis, I find that, even after making standard small-sample adjustments, the significance of any of these factors is likely to be due to the same small number of quarters in the returns data, which I term “influential quarters.” These quarters occur around recession periods, suggesting a link between the cross-section of average returns and the business cycle and, in turn, providing support for conditional CCAPM models. However, the finding that the same short list of influential quarters, pre-specified from the returns data, determines the statistical significance of any given factor poses two challenges. The first is an inference problem caused by the continual search for new factors using the same sample of returns. The second is that discrimination between factors based on higher t-statistics will merely pick up a factor that is more susceptible to large innovations during these quarters. The results ultimately suggest that the current sample of returns is too small to identify the true economic channel governing stock returns.