Using the conditional volatility channel to improve the accuracy of aggregate equity return predictions

In a recent study, Maheu et al. (Int J Forecast 36: 570–587, 2020) suggest a predictive regression model, where besides the conditional mean, the lagged value of the predictor of interest can also impact the dependent variable through the conditional volatility process. Their out-of-sample study focusing on predicting the conditional distribution of the US real GDP growth rate by conditioning on the price of crude oil finds strong evidence in favor of the suggested specification with respect to density forecast accuracy. In this study, we demonstrate that their framework is also very useful with regard to predicting aggregate equity returns by conditioning on macroeconomic variables. Using the well-known Goyal and Welsh dataset, we show that the suggested framework results in statistically significant more accurate density predictions relative to the stochastic volatility benchmark as well as competitors, where the lagged value of the predictor of interest impacts aggregate equity returns exclusively through the conditional mean process. Evidence of statistical predictability also results in VaR accuracy gains.