Industry Rotation with Post-Lasso and Supply-Chain Networks

Summary

This research tests whether lagged returns in economically related industries can help predict future industry returns, based on the idea that information and shocks may diffuse across sectors over time. To limit overfitting in a large predictor set, its Post-Lasso method first selects lagged industry returns with Lasso, then refits an ordinary least squares regression on the selected variables. The report also describes a network method that maps listed companies’ customer-supplier links into industry relationships, using the resulting network to choose predictors for rolling regressions.

The report presents historical A-share backtests for both methods, comparing selected predictors with simpler models and reporting portfolio returns and other statistics. The network strategy uses a 24-month estimation window and ranks industries into equal-weight long and short groups. Results are tied to the stated samples and data sources; they do not establish future performance. The authors also note limitations in industry-network construction and that the machine-learning method can select relationships that are difficult to explain.

Key ideas

• Lagged returns from related industries may predict target-industry returns if information spreads across sectors gradually.
• Post-Lasso selects predictors with Lasso and then refits a least squares model on those predictors.
• A customer-supplier network can encode prior economic relationships and guide predictor selection.
• The reported strategies rank industries into equal-weight long and short portfolios using historical forecasts.
• The backtests are sample-specific, and data-driven links may be difficult to interpret.

Tags

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