Table 1 Some characteristics of five alternative models for funding research that do not require detailed applications

Equal allocation

•Avoids peer review biases [8]

•Cannot fund higher cost research

Lottery

•Can increase efficiency by funding riskier research that would rarely be funded by traditional peer review [37]

•Politically problematic [38]

Automated scores

•Harnesses large amounts of existing data on researchers

•Can be gamed [39]

•Takes no account of career disruption

•Scores may have poor sensitivity and specificity

Prediction markets [33]

•Extracts more accurate information by paying reviewers proportional to their ability

•Reviewers may be lobbied to give good predictions

•Potentially more suited to rating departments rather than individual researchers

Peer-to-peer distribution [23]

•Harnesses existing knowledge

•Scientists with the greater respect of their peers have increased decision-making power

•Vulnerable to collusion

•Results may be quite different in the first few years as the distribution mechanism stabilises