Quantile Regression for Multi-Factor Stock Selection and Index Enhancement
Summary
The document explains quantile regression as an alternative to ordinary least squares for modeling different points in a return distribution. Because it does not rely on the same normality and constant-variance assumptions, it may be useful with skewed, heavy-tailed, or heteroskedastic data. Unlike a mean regression that produces one fitted value, quantile models estimate conditional outcomes at selected distribution levels.
An analysis of CSI 500 constituents considers market capitalization, prior-month return, and average daily turnover. The reported effect of prior returns varies across quantiles: the median resembles a reversal effect, the lower tail shows a stronger relationship, and the upper tail changes sign. For stock ranking, the article suggests using a statistically significant quantile with a steep fitted slope. In a reported index-enhancement comparison, the 10th-quantile model outperformed mean regression on excess return and Sharpe ratio, with similar volatility, drawdown, and turnover. The summary supplies no dates, numeric performance results, transaction-cost assumptions, or independent validation, so the comparison’s generalizability is unclear.
Key ideas
- Quantile regression estimates relationships at different points of the conditional return distribution.
- It can be more robust than mean regression when data are heavy-tailed or heteroskedastic.
- For CSI 500 stocks, the relationship between prior-month return and future returns reportedly changes across quantiles.
- The article recommends selecting a significant quantile with a steep slope when using predictions to rank stocks.
- Its reported 10th-quantile enhancement results lack enough testing detail to assess generalizability.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.