Industry Rotation Using Post-Lasso and Supply-Chain Knowledge Graphs

Summary

This report studies whether one industry’s past returns can help predict another industry’s future returns. It argues that information may spread gradually across related industries because investors cannot immediately assess every impact of a new shock. The strategy models target-industry returns using lagged returns from other industries, addressing the risk of overfitting a broad regression and the subjectivity of manually selecting predictors.

One approach uses Lasso to select predictors, then fits a regression using those selections. A second uses customer-supplier links among Chinese A-share companies, combined with an input-output network, to define related industries; it fits a rolling 24-month regression using those industries’ lagged returns. The report gives historical results for both approaches, including long-short returns and, for the knowledge-graph strategy, maximum drawdown. These are reported backtest results over different sample periods, not evidence of future performance. The summary does not detail transaction costs, implementation effects, or robustness across other periods.

Key ideas

• Lagged returns from related industries may forecast a target industry’s returns as information diffuses.
• Post-Lasso first selects industry return predictors and then estimates a regression using them.
• A supply-chain knowledge graph supplies prior relationships for choosing industry predictors.
• The knowledge-graph approach uses a rolling 24-month estimation window.
• The reported performance comes from historical samples and does not establish future results.

Tags

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