Sequential Regression and Composite Factors for Equity Factor Analysis

Summary

This research summary describes a multi-factor tracking framework built around the pricing logic of factors. It examines historical returns for technical, expectation-based, and financial factors using a modified cross-sectional regression approach. To address correlation among factors, it estimates one factor at a time and uses each regression’s residual as the dependent variable for the next step. The order is set by the researchers’ economic judgment, and the note says changing that order changes the interpretation of later factor returns.

For groups of similar, correlated exposures, the authors recommend combining factors, using equal weights when weight optimization is not being attempted. They compare sequential and single-factor regression results, including return series and explained variation, to assess overlap with earlier factors and propose removing factors whose residual contribution is small. The summary reports these as research recommendations but gives no empirical tables, sample details, or out-of-sample validation. The claim that sequential regression fully avoids multicollinearity should therefore be treated cautiously, and results may depend on factor ordering.

Key ideas

• The framework estimates factor returns sequentially, passing each regression’s residual to the next step.
• Factor ordering reflects economic judgment and affects the interpretation of later factor returns.
• Similar and correlated exposures can be combined with equal weights when weight optimization is not used.
• Comparing sequential and single-factor estimates can reveal overlap with earlier factors.
• Factors with little residual explanatory contribution are candidates for removal.
• The summary offers no sample details or out-of-sample validation.

Tags

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