Model Uncertainty for Long-Term Investors

We develop a method to identify the most important predictors of long-term asset returns and use it to analyze the impact of model uncertainty on long-term investors. We find that the impact of model uncertainty changes a lot over time which leads to considerable time-variation in all moments of the predictive stock return distribution. Even though stocks are safer in the long-run than short-run when model uncertainty is low, stocks are actually riskier in the long-run when model uncertainty is high. For buy-and-hold and dynamic portfolios with long investment horizons, incorporating model uncertainty substantially lowers the equity allocation.