Out-of-Sample Equity Premium Prediction: Consistently Beating the Historical Average

While a host of economic variables have been identified in the literature with the apparent in-sample ability to predict the equity premium, Goyal and Welch (2007) find that these variables fail to deliver consistent out-of-sample forecast gains relative to the historical average. Imposing theoretically motivated restrictions on individual predictive regression models, Campbell and Thompson (2007) provide improved forecasts, but the out-of-sample performance is still highly uneven over time. In this paper, we propose two approaches— combination forecasts and covariate estimation—to improve out-of-sample equity premium forecasts based on economic variables. These approaches accommodate structural instability and utilize additional information. We find that despite the failure of individual predictive regression model forecasts to outperform the historical average, combinations of individual model forecasts deliver statistically and economically significant out-of-sample gains relative to the historical average on a consistent basis over time. Forming combination forecasts from individual models estimated using covariates and with Campbell and Thompson (2007) restrictions imposed typically leads to further out-of-sample gains. JEL classifications: C22, C53, G11, G12