Imputation Based Estimators for Treatment Effects in Impact Evaluation Framework

Abstract

The problem of counterfactual has been at the core of impact evaluation frame- work. Almost all existing methods aim to find the best way to estimate efficiently the counterfactual. A solution for estimation of counterfactual was proposed and investigated in this study. The objective of this research was to use classical imputation methods to estimate counterfactual, then derive treatment effect estimators from the data sets completed using the basic definition of treatment effect described in Rubin framework. The estimators obtained, called Imputation Based Treatment Effects estimators, and were theoretically unbiased, convergent and consistent. Using simulation, the results revealed that they were asymptotically unbiased and convergent as well. Results from the data application showed that they performed as well as the classic estimators and sometimes better in cases of shortage in data. To conclude the research, a hypothesis testing procedure was proposed to test the significance of the treatment effect. The results showed that the three approaches proposed were efficient, and could detect any change between two distributions, even slight changes.