Impact evaluation encompasses the empirical proof of causal effects and their normative evaluation. While science cannot offer the normative ground for an evaluation, it can contribute by providing rigorous methods of causal inference. In this paper, we discuss two causal-analytical approaches, Rubin’s potential outcomes framework and Pearl’s graph methodology, with regard to questions of impact evaluation. From a potential outcomes perspective the advantages of experimental designs become obvious. Thus, in case of non-experimental designs we recommend to approximate this ideal by using statistical procedures. However, the success of all these analytical strategies requires knowledge about the causal structure at hand. Graph methodology facilitates the explication of these theoretical assumptions and the choice of adequate statistical procedure and covariates. We conclude with an illustration for the case of the Bologna reform.

Keywords
Causality, Graph Methodology, Impact Evaluation, Bologna Reform