A Review of Propensity Score Matching Techniques: Estimation of Income Effect with an Illustration Using Lalonde Dataset

Abstract

For evaluating the success of social and economic programmes, impact evaluation of such is necessary. In randomization (Experimental studies) results of the estimates are free from bias, but randomization is not possible in every cases then we access the result of such programmes through observational study. Selection bias frequently makes it more difficult to estimate causal effects in observational studies because treatment assignment is not random. One of the most popular quasi-experimental methods to deal with this problem is Propensity Score Matching (PSM), which creates similar treatment and control groups based on observed characteristics through similar p-score. The conceptual framework of PSM, its underlying assumptions, matching algorithms, and estimation techniques are all covered in this paper. The estimation of treatment effects on income is explained with an example that makes use of fictitious calculations and dataset structure. This paper presents a methodological review of PSM and uses the Lalonde Training Program Dataset to illustrate its use. The advantages and limitation of PSM in empirical research are also discussed in the study. The results explains that PSM is a powerful and useful tool for assessing program impacts in situations where randomized experiments are impractical, though its efficacy is dependent on the appropriate model specification and availability of strong covariates.