Information Aggregation for Stock Return Predictability
35 Pages Posted: 24 Sep 2016
Date Written: September 19, 2016
Abstract
The literature on stock return predictability has identified macroeconomic and technical predictors that when combined, leads to out-of-sample outperformance relative to the historical mean null. This paper investigates a new method for aggregating information beyond using forecast combination or principal components. By sequentially layering groups of information, the predictive performance of this new approach outperforms that of prior methods. Applying layering to volatility forecasting yields more mixed results. In all, a mean-variance investor investing in monthly stock returns gains from this new method as much as 4.5% per year.
Keywords: Return Predictability, Forecast combination, Principal Components
JEL Classification: G11, G14, G17
Suggested Citation: Suggested Citation