Systematic Risk Changes, Negative Realized Excess Returns and Time-Varying CAPM Beta

We make two methodological modifications to the method of testing CAPM beta and we show that these significantly affect inferences about the association between CAPM beta and stock returns. While the conventional beta proxy is indeed largely unrelated to realized stock returns (in fact the relationship is slightly negative), using forward-looking beta and eliminating unrealistic assumptions about expected market returns makes it (highly) significant. In addition, we show that complementary empirical factors-size and ratio of the book-to-market value of equity-that are sometimes presented as potential remedies to beta's deficiencies do not seem to outperform beta. This suggests that weak empirical support for CAPM beta is likely caused by complications with implementing CAPM rather than by the weakness of the underlying concept.