Out-of-sample equity premium prediction: economic fundamentals vs. moving-average rules

This paper analyzes the ability of both economic variables and moving-average rules to forecast the monthly U.S. equity premium using out-of-sample tests for 1960?2008. Both approaches provide statistically and economically significant out-of-sample forecasting gains, which are concentrated in U.S. business-cycle recessions. Nevertheless, economic variables and moving-average rules capture different sources of equity premium fluctuations: moving average rules detect the decline in the average equity premium early in recessions, while economic variables more readily pick up the rise in the average equity premium later in recessions. When we simulate data with a habit-formation model characterized by time-varying return volatility and risk aversion relating to business-cycle fluctuations, we find that this model cannot fully account for the out-of-sample forecasting gains in the actual data evidenced by economic variables and moving-average rules.