Econometrics Seminar - Juan Rubio Ramirez

Abstract

A long tradition in macro-finance studies the dynamics of aggregate stock returns and
dividends using vector autoregressions, imposing the restrictions implied by the Campbell-Shiller
(CS) identity to sharpen inference. We develop Bayesian methods that encode a priori skepticism
about return predictability while imposing the restrictions. We highlight that persistence in
dividend growth induces “dividend momentum,” a previously overlooked channel for return
predictability. By combining Bayesian shrinkage and the CS restrictions, we obtain more
plausible degrees of return predictability, and superior out-of-sample forecasts and Sharpe ratios,
which cannot be obtained by using either shrinkage or the CS restrictions on their own.