Abstract
There is wide agreement that the high social rate of return to research and innovation justifies government support for research. There is, however, only limited evidence on the effectiveness of different public research programmes. Reliable measurement of programme effectiveness is hampered by the ‘selectivity’ problem (public funding goes to proposals judged in advance to be likely to succeed) and the question of ‘additivity’ (whether public funding increases total spending on research or merely displaces funding from other sources). The selectivity problem can be mitigated by building evaluation into programme design, either by partial randomization of the grant process, or by recording the rankings used in grant evaluation and making this information available to researchers. The additivity question reflects the more fundamental problem that the ultimate objective of these programmes is to have long‐term effects that are inherently very difficult to measure and attribute to particular programmes.