Forecasting Industry Returns with LASSO and Long-Short Rotation

Summary

The document summarizes research that forecasts stock industry returns using lagged returns from a broad set of industries. Because including all industry predictors creates a high-dimensional regression with overfitting risk, the method uses LASSO to select a sparse set of predictors, then refits the selected variables with OLS to reduce coefficient shrinkage. The reported analysis also addresses post-selection inference and multiple testing, and discusses whether the predictive links make economic sense or could reflect changing risk premia.

For the out-of-sample portfolio, industries are ranked by predicted excess returns and grouped into equal-weighted portfolios; the strategy goes long the highest-ranked group and short the lowest. The summary reports positive historical performance, including returns during recessions, and risk-adjusted results relative to a multifactor model. These are findings from a historical study, not a guarantee of future profitability. The source is a translated secondary account, and some equations and figures are omitted, limiting independent assessment of its setup and reported results.

Key ideas

• Lagged returns across industries serve as predictors of each industry’s future excess return.
• LASSO addresses the regression’s large predictor set, while post-LASSO OLS refits selected coefficients.
• The study applies multiple-testing controls and considers whether selected relationships have plausible economic explanations.
• An out-of-sample rotation portfolio buys the industries with the strongest forecasts and shorts those with the weakest.
• The reported historical results include positive performance in recessions, but they do not establish future returns.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.