A0703
Title: Forecast encompassing in modal regression for equity premium prediction
Authors: Tae-Hwy Lee - University of California Riverside (United States)
Yaojue Xu - University of California, Riverside (United States) [presenting]
Abstract: While the equity premium may not be predictable in mean using many macroeconomic and financial predictors, it may well be predictable in some quantiles especially in tails or in the mode. The mode has its own merits relative to mean and quantile. It is robust when the distribution is skewed. Is the mode of the financial returns more predictable than the mean? With this in mind, we develop a novel framework for Granger-causality (GC) test in the predictive regression for the conditional mode based on the forecast-encompassing principle. We show that the encompassing statistic is asymptotically standard normal with zero mean under the null hypothesis that there is no GC in the modal regression. The Monte-Carlo simulation shows the encompassing test has a proper size and excellent power in finite samples. We apply the encompassing statistic to test if financial and macroeconomic variables Granger-cause the mode of the equity premium distribution. The mode prediction results for the equity premium are generally more significant than the mean prediction results.
