Did you match? An introduction to matching techniques

Abstract

How can we infer causality from observational data? This article provides an intuitive introduction to matching, a useful statistical technique for identifying causal relationships in non-experimental research designs. Methodologically, we present the main features of matching through three examples: a) the effect of military schools on learning; b) the impact of Bolsa Família on the propensity to vote for the Workers' Party; and c) the effect of gender in electoral results. We show the computational implementation in R and explain the substantive interpretation of the results. In order to increase the pedagogical potential of the research, we provide all replication materials, which makes it easier for students and professionals to use the data and scripts in their study and work activities. We hope to foster the use of quasi-experimental techniques in the Social Sciences with this article and encourage replicability as a teaching strategy in data analysis.

Keywords:
matching; causal inference; observational data; quasi-experimental techniques