Publisher's Synopsis
Previous attempts to test the CAPM using average realized returns as estimates of expected returns were not very successful nor are they likely to be. Even if the CAPM is true, any evidence of beta/return linearity that might exist in the historical record is obscured or obliterated when the average market return is near or even modestly above zero. On the other hand, estimating expected returns using average realized returns further out in the tails of the return distribution (in down markets, for example) provides evidence that strongly supports the CAPM. Using 24 years of daily return data for 90 high capitalization stocks the estimation technique developed in the paper results in expected returns that are highly correlated with beta, not just over the full 24 years, but over twelve 2-year sub-periods as well. Moreover, in a Monte Carlo experiment it is demonstrated that this alternative estimation technique delivers estimates that are closer to "true" expected returns than those resulting from the usual practice of estimating expected returns as average realizations. The robust relationship between beta and expected returns leaves little room for the so-called return anomalies to play much of a role in explaining expected returns. In particular, the statistical significance of the four non-market (i.e., non-beta) return factors of the Fama/French five factor model nearly slips away entirely. In addition, with this new approach to estimating expected returns there are many fewer tangency portfolio short positions than other researchers have documented.
