ABSTRACT
Many studies aim to estimate causal effects of risk factors, interventions, or programs, on outcomes of interest. While randomization is generally seen as the preferred design for estimating causal effects it is not always possible to randomize the “treatments” of interest, especially in the social sciences. Propensity scores are a useful tool that can help yield better estimates of causal effects in non-experimental studies by ensuring that the treatment and comparison groups are similar with respect to the observed covariates. The propensity score itself is defined as the probability of receiving the treatment given a set of observed covariates (Rosenbaum & Rubin, 1983a). It is used to equate (or “balance”) the covariates between the treatment and comparison groups through propensity score matching, weighting, or subclassification (Stuart, 2010). Outcomes can then be compared between the equated groups, with less risk for extrapolation from treatment to comparison group (and vice versa), thus yielding more reliable causal effect estimates (Ho, Imai, King, & Stuart, 2007). For overviews and current methods, see Hernan and Robins (2015), Imbens and Rubin (2015), Stuart (2010), and Rosenbaum (2009).
