Prior research has shown that individual predictors of the market
equity risk premium fail to deliver superior out-of-sample forecast
accuracy relative to the historical average. Combining forecasts of
individual predictors also underperforms the historical average in recent
decades. In this article, we construct an optimal combination of
predictors across two different regimes of economic uncertainty (normal
and high) and show that this optimal combination delivers the
highest out-of-sample forecast accuracy. Specifically, we find that the
earnings-price ratio has the lowest forecast error during periods of normal
economic uncertainty, while inflation has the lowest forecast error
during periods of high economic uncertainty. Investors can use our
economic uncertainty-conditioned predictor combination to generate
more accurate forecasts of the equity risk premium; actionable trading
strategies based on this approach yield utility gains of around 4% and
a haircut Sharpe Ratio of 0.41, which are the largest among all other
predictors.