Working Paper Series no. 166: Microeconometric evaluation methods (in French)

Abstract

Our survey covers the recent developments of the microeconometric literature on evaluation methods. In this field, the canonical model is Rubin's causal model, which is close to Roy's selectivity model. This model is the relevant framework for defining and for examining the identifiability conditions of the parameters of interest in any evaluation study. We insist on the definition of these parameters, which include the average effect of the treatment on the treated and on the non-treated individuals. For each set of assumptions (selectivity on observable or unobservable characteristics, conditional independence between outcomes and treatment indicators, etc.), we present the most adapted estimation method. We put a special emphasis on matching estimators in the situation where the selectivity depends only on observables, and on differences-in-differences methods and on regression-discontinuity techniques when the selectivity depends both on observable and unobservable characteristics.

Denis Fougère
March 2007

Classification JEL : C13, C14, C41, J64.

Keywords : evaluation, selectivity bias, matching estimators, differences-in-differences, regression-discontinuity.