Evaluation Research

The goal of evaluation research is to establish whether an intervention caused the outcomes it was designed to achieve. This requires isolating the effect of the intervention from other things that changed at the same time, which is methodologically demanding. The findings are used to decide whether to continue, scale, or redesign a programme.

The work starts with defining the research question and identifying what data is available. Where an experiment is possible, that is the preferred approach, since it provides the strongest causal evidence. Where it is not feasible, quasi-experimental methods such as difference-in-differences, regression discontinuity, or matching are used instead. A large part of the work involves assessing the credibility of the causal claims the data supports.

The main tools are R, Python, and Stata. Causal inference packages are widely used, including those for propensity score matching, instrumental variables, and regression discontinuity designs. Power calculations are important when planning an evaluation, since they determine how large a sample is needed to detect a given effect.

Evaluation research projects appear in government, research institutes, international organisations, and healthcare. In the Netherlands, organisations such as the CPB, ZonMw, and government ministries commission or conduct evaluations. The project connects to Economic Researcher, Policy Analyst, and Policy Evaluations work, and the methods overlap closely with those used in academic research.