External Validity and Generalizability in Program Evaluation: Embracing Complexity

This introduction to the second special issue of <i>Evaluation review</i> on external validity and generalizability opens a dialogue between different ways to think about generalizability in program evaluation. It argues that generalizability in impact evaluation fundamentally is about inferring some form of <i>causality</i> at a level broader than the specific circumstances of the initial study or studies from which these inferences are drawn. The question, then, is about how one apprehends causality: in other words, <i>what</i> is being generalized and <i>how</i>? The first special issue mainly relied on a <i>counterfactual</i> conception of causality, embodied by experimental and quasi-experimental methods, that aimed at impact <i>measurement</i>. The articles in this volume, drawing on mixed methods, also mobilize <i>generative</i> and <i>configurational</i> causal inferences to provide further levers of generalizability, focusing on <i>how</i> the impact is produced. The introduction insists on the specific input of qualitative methods in this respect, as theorized by grounded theory.