Matching Portfolios

Abstract:     
We propose a portfolio performance measure based on comparison with a matched portfolio sharing the same exposures but holding different stocks. Under the framework of the Rubin Causal Model, the matched portfolio provides a view of the counterfactual, an alternative portfolio that the manager could have chosen but did not. We provide a methodology for constructing a matched portfolio using the holdings of the original portfolio and a set of stock characteristics. Treated as a benchmark, a matched portfolio provides a more precise estimate of alpha, or stock-picking ability, because it shares the same characteristics as the original portfolio. It is also more flexible than other benchmarks because it can be matched on any number of characteristics, and can mimic the weighting structure (e.g., long-short, 130/30) of the original portfolio.

Keywords: portfolio, benchmark, performance evaluation, matching, propensity score, rubin causal model

JEL Classifications: C10, G11

Working Paper Series