Coarsened Exact Matching (CEM): A new technique for statistical matching

As empirical tests of causal claims derived from theories become more important in the social sciences, researchers who rely on observational data are confronted with with the inadequacy of their data sets for estimating causal effects. Unlike experimental designs, researchers cannot influence the assignment of the treatment, which leads to biased results. For example, the (self) selection of more talented people into training programs influences the estimation of the programs' efficiency, when we simply compare participants to non-participants.

Statistical matching offers a solution to this problem by finding "statistical twins", one with and one without the treatment. The most common matching technique, Propensity Score Matching, however, is slow and difficult to apply. Coarsened Exact Matching (CEM) offers an alternative solution, which is faster and easier to understand. It temporarily coarsens the data according to the researchers ideas (i.e. in coarse age groups rather than exact birthdays) and then finds exact matches. Yet, is this gain in speed and simplicity exchanged against a lack in validity?

Resources:
Blackwell, Matthew, Stefano Iacus, Gary King, and Giuseppe Porro. (2009). "cem: Coarsened exact matching in Stata". The Stata Journal 9(4):524–546.

Iacus, Stefano M., Gary King, und Giuseppe Porro. (2011) „Causal Inference without Balance Checking: Coarsened Exact Matching“. Political Analysis. (forthcoming)