Using Propensity Scores to Test Stock Concept Effects on Returns

Summary

This research note introduces propensity-score methods as a way to estimate causal effects from observational data, where treated and comparison groups may differ in confounding characteristics. Its workflow is to estimate scores and treatment effects, assess whether the method balances covariates between groups, and run refutation checks to judge the estimate’s reliability. It first applies the framework to the Lalonde dataset, then studies whether membership in selected concepts relates causally to future returns among CSI 800 stocks. Fundamentals and price-volume factor exposures serve as confounders.

For the period from January 2016 through March 2020, the analysis reports a positive estimated effect for seasonally adjusted fund-heavy stocks and a negative one for pledged-stock concept membership. The effects for earnings pre-announcement and moat concepts are inconclusive. Propensity-score weighting performed best in the reported balance and refutation tests. These findings are tied to the selected index, concepts, covariates, and sample period; observational adjustment cannot rule out unmeasured confounding, and the document does not establish that the results generalize to other markets or periods.

Key ideas

• Propensity scores summarize observed confounders to help compare treated and control groups.
• The proposed workflow estimates effects, checks covariate balance, and applies refutation tests.
• The study applies the method to concept membership and future returns for CSI 800 stocks.
• Fund-heavy concept membership had a reported positive effect, while pledged-stock membership had a negative one.
• Results for earnings pre-announcement and moat concepts were uncertain and depend on the study design.

Tags

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