Confident Risk Premiums and Investments using Machine Learning Uncertainties

This paper derives ex-ante (co)variances of stock-level and portfolio-level risk premium predictions from neural networks (NNs). Considering precision, I provide improved investment strategies. The confident high-low strategies that take long-short positions exclusively on stocks with precise risk premiums deliver superior out-of-sample returns and Sharpe ratios than traditional high-low strategies because precise measurements have low squared forecast errors. Mean-variance strategies incorporating covariances of return predictions outperform existing strategies too. Risk premium variances reflect time-varying market uncertainty and spike after financial shocks. Cross-sectionally, the level and precision of risk premiums are correlated, thus NN-based investments deliver more gains in the long positions.