Propensity-Score Matching as Panacea for Correcting Self-Selection Bias in Observational Studies

Observational studies often suffer from the self-selection problem which is due to the non-randomness of selecting objects into treatment and control group. As result, the precise estimation of treatment effect becomes impossible by relying on a conventional approach, such as ordinary least squares (OLS) regression. Many researchers in public health use propensity-score matching to correct the bias related to selfselection problem in order to obtain the exact estimate of treatment effect. However, propensity-score matching is not a panacea for self-selection problem. This article explains why sometimes propensity-score matching may not work, and gives empirical researchers in public health a gentle introduction to other methods, which is not so well-known in the field, for correcting self-selection problem.