Equity premium prediction: A constraint-based predictor decomposition approach

We propose a constraint-based predictor decomposition approach that exploits predictive information in commonly used predictors to improve equity premium forecasts. The approach identifies and quantifies unexpected changes as deviation tendency while bounding values to capture central tendency. Predictions from the two tendencies are then synthesized. Empirical analysis shows this approach outperforms existing methods, producing statistically and economically significant out-of-sample results. These findings validate the ability of our approach to capture both tendencies. We also extend the analysis to multivariate prediction, where results consistently confirm its superiority. Finally, robustness tests and additional analyses demonstrate that the approach delivers stable and reliable forecasting performance.