Using Negative Stock Factors Through Exclusion and Orthogonalization

Summary

This research summary examines how stock-selection factors with strong short-side performance might still be useful in A-shares, where short selling is constrained. For a single factor, it proposes excluding stocks with the undesirable factor characteristics instead of relying on a long portfolio to outperform the market. The summary reports that simulated random portfolios drawn from the remaining universe had a more right-skewed return-quantile distribution than portfolios drawn from the full market, suggesting improved odds of avoiding weak stocks. It cites reversal, average trading value, and idiosyncratic volatility among factors whose short-side contribution was often substantial.

For multi-factor models, the document frames a composite score as a composite factor whose information coefficient depends on individual factor strength, factor IC covariances, and weights. It recommends orthogonalizing factors, using stepwise regression, to address changing cross-sectional relationships and reduce composite or redundant signals. The supplied page is an abstract and links to a PDF, so it omits the full formulas, sample design, transaction costs, and detailed robustness evidence. Its findings should be treated as market- and methodology-dependent rather than universal.

Key ideas

• In a market with short-selling constraints, excluding stocks with unfavorable factor values can use negative factor information.
• The summary reports a more right-skewed return-quantile distribution for simulated portfolios from the filtered universe.
• A multi-factor composite information coefficient depends on factor strength, cross-factor covariance, and model weights.
• The article recommends orthogonalizing factors with stepwise regression to reduce redundancy and support its analysis assumptions.
• The available text is an abstract, so it does not show the full empirical design or costs.

Tags

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