Propensity Score Matching: An Application in Empirical Finance

Because of the difficulty in multi-dimensional matching and missing variables, investigation of long-run stock performance after major corporate events has been plagued by the “bad model problem”. This study contributes to the literature by implementing the propensity score estimator which is able to match firms in multiple dimensions simultaneously, and illustrate this methodology in the context of seasoned equity offerings (SEOs). We find that firms’ SEO decisions are affected by the following characteristics: size, book-to-market ratio, past returns, and capital surplus. When matching SEO and non-SEO firms in all these dimensions via propensity scores, we are able to achieve much better matches than traditional matching methods. Furthermore, the commonly documented SEO underperformance disappears. Although it is premature to conclude the non-existence of long-run underperformance of SEO firms because of limited data here, this paper undoubtedly establishes a new methodology that can be generally useful for long-term performance study in empirical finance research.