The Performance of Characteristic-Sorted Portfolios: Evaluating the Past and Predicting the Future

Abstract:
We study persistent fluctuations in characteristic-sorted portfolio returns through the lens of a statistical model. The model provides a simple formula for adjusting the standard errors of expected return estimates. With plausible parameter values, adjusted standard errors double, casting doubt on the interpretation that the historical performance of characteristic-sorted portfolios represents unconditional return premia. Similarly, maximum likelihood estimates indicate that the historical data are consistent with a wide range of return processes. Finally, our Bayesian analysis shows that investor posteriors about expected returns are highly dependent on their priors about persistence, even after observing close to 60 years of data.