‘Evidence-based policy’ has become a mantra of many governments that may determine, justify, illuminate or act as smoke screen for projected changes to a specific policy. In economic terms this means that net benefits should be calculated by estimating the expected future returns from a potential change discounted into the present value and compared with estimated costs to those affected: this is sometimes called ‘impact analysis’. Its ‘shadow’ side is ‘policy-based evidence’, whereby selective facts are offered in support of a predetermined government position. Actual practice probably veers between the two. The increased focus on evidence has been a boon to economic consultancies and to lobbyists. But to many the meaning of ‘evidence’ remains unclear. Put bluntly: Is it just a rhetorical device? (it’s ‘evidence’ if you like it, ‘lobbying’ if you don’t); Or is ‘evidence’ a useful social science concept for understanding human behaviour, and evaluating different normative paths?