The Sidney Ball Memorial Lecture 2014 given by Professor Tom Cook.
This paper identifies three major assumptions of the current evidence-based policy movement that seeks to use scientific methods to identify effective social policy initiatives: (1) Randomized controlled trials are the best way of identifying what works and should be preferred for building up an evidence base or should even constitute the only legitimate inputs into that base; (2) In policy research it is not difficult to label the causal agent in general language; and (3) Extrapolating from past causal findings (however secure) to future policy contexts depends on formal sampling theory. Empirical evidence is used to address these three assumptions. Each is found to be overstated. A somewhat different model of evidence-based research is briefly proposed. It makes more realistic assumptions about how research can contribute to evidence-based policy and respects the conceptual interdependencies between creating high quality tests of causal propositions, correctly labeling the cause (or effect), and minimizing the extrapolation from past knowledge to future applications.