Predicting Market Risk Premiums with Historical Patterns

Abstract

Several studies have supported predictability of stock market returns, but others have questioned the evidence. Some researchers have indicated that returns predictability reflects risk aversion fluctuating with business cycles. This study investigates whether historical patterns in market risk premiums, which indicate variations in risk aversion, can predict risk premiums. Eight forecasting methods are used to identify optimal monthly forecasts of US market risk premiums for 70 years, with 95 years of data. Double moving averages of historical market risk premiums, reflecting nonseasonal data with trend, consistently provide optimal forecasts. The forecasts match the distribution of risk premiums more closely than historical averages and, unlike historical averages, they have significant predictive power for risk premiums. Years with higher forecasts provide higher risk premiums and the forecasts produce substantial utility gains in recessions and in months with negative forecasts. Four performance measures show that two investment strategies using the forecasts outperform a passive stock market investment, by enhancing risk premiums and reducing both systematic and total risk.