78 Pages Posted: 3 Nov 2015 Last revised: 10 Jul 2016
Date Written: July 9, 2016
We show that low-order autoregression models for short-term expected returns imply long-term dynamics that have a (too) fast vanishing persistence when compared with the evidence from long-horizon predictive regressions. We then propose a novel modeling framework that exploits the low-frequency information in the predictors as a prior to update the high-frequency distribution of expected returns. Our framework shows that, in order to restore consistency with the empirical evidence from predictive regressions, the short-term dynamics of expected returns need to have long-range dependence. In turn, these long-memory type of dynamics generate first-order effects on forecasting and investment decisions, especially in the long-run. We quantify these effects along several dimensions.
Keywords: Expected Returns, Long-Horizon Predictability, Multi-Scale, Bayesian Methods
JEL Classification: G17, G11, C53, C58
Suggested Citation: Suggested Citation
Bianchi, Daniele and Tamoni, Andrea, The Dynamics of Expected Returns: Evidence from Multi-Scale Time Series Modeling (July 9, 2016). Available at SSRN: https://ssrn.com/abstract=2684728 or http://dx.doi.org/10.2139/ssrn.2684728