Propensity Score Matching and Difference-In-Difference Estimator for Impact Evaluations: A Case Study from Zimbabwe

Development banks, non-governmental organizations, and international institutions regularly implement projects in low-income countries to assist them in their development process. These projects, often called “programs,” are of different natures and target different sectors in the economy, as well as different aspects of society or different demographic groups. This chapter discusses econometric methods based on panel data to evaluate the impact of such programs. Those who participate or benefit from these programs are called the beneficiaries. Often, the evaluation consists of comparing the change in a target outcome (e.g., income) between the beneficiaries and a control group, often called the counterfactual. To do so, both groups should be comparable; therefore, careful attention should be paid when choosing both the beneficiaries and the counterfactual, while avoiding selection bias. This chapter further discusses the sampling strategy for selecting both groups, including the calculation of the sample size, propensity score matching methods to define the final evaluation sample, and ways to handle potential selection bias in panel data using difference-in-difference techniques. The chapter also shows implementation of these techniques in Stata software using a case study and data from the “Training for Rural Economic Empowerment (TREE)” program implemented in Zimbabwe in 2010 by the International Labour Organization.