The ARCH-in-Mean Dose Work: Out-of-Sample Predictability of US Stock Market
23 Pages Posted: 25 Jan 2015
Date Written: January 24, 2015
Abstract
The predictability of stock market is of great interest to both reseachers and investors. Despite voluminous evidence of in-sample predictability, the out-of-sample predictability of stock returns remains an ongoing debate. In this paper, motivated by both the financial theories and the well documented facts in financial empirical literature, we employ the ARCH-in-Mean model as a benchmark to scrutinize the out-of-sample predictability of the US stock market returns. Empirical studies performed on the S&P500 stock index demonstrate that the ARCH-in-Mean model does report significant out-of-sample forecasts in both statistical and economic sense. The main conclusions of this paper are that 1) the US stock returns is predictable both in-sample and out-of-sample; 2) the predictability of US stock returns can be traced back to both time-varying risk premia and investors’ underraction to bad news and overreaction to extremely bad news.
Keywords: ARCH-in-Mean, Return Predictability, Time-Varying Risk Premia, Market Inefficiency
JEL Classification: C53, G11, G12, G17
Suggested Citation: Suggested Citation