Impact Evaluation - Estimation Methods

Estimation Methods

Estimation methods broadly follow evaluation designs. Different designs require different estimation methods to measure changes in well-being from the counterfactual. In experimental and quasi-experimental evaluation, the estimated impact of the intervention is calculated as the difference in mean outcomes between the treatment group (those receiving the intervention) and the control or comparison group (those who don’t). This method is also called randomised control trials (RCT). According to an interview with Jim Rough, former representant of the American Evaluation Assosiciation, in the magazine D+C Development and Cooperation this method doesn't work for complex, multilayer matters. The single difference estimator compares mean outcomes at end-line and is valid where treatment and control groups have the same outcome values at baseline. The difference-in-difference (or double difference) estimator calculates the difference in the change in the outcome over time for treatment and comparison groups, thus utilizing data collected at baseline for both groups and a second round of data collected at end-line, after implementation of the intervention, which may be years later.

Impact Evaluations which is has to be compare average outcomes in the treatment group, irrespective of beneficiary participation (also referred to as ‘compliance’ or ‘adherence’), to outcomes in the comparison group are referred to as intention-to-treat (ITT) analyses. Impact Evaluations which compare outcomes among beneficiaries who comply or adhere to the intervention in the treatment group to outcomes in the control group are referred to as treatment-on-the-treated (TOT) analyses. ITT therefore provides a lower-bound estimate of impact, but is arguably of greater policy relevance than TOT in the analysis of voluntary programs.