The makers of a simulator that lets users test out different research funding programme designs have said pre-application lotteries are the most cost-effective option for all involved.
GrantInq, built by researchers at the University of Lübeck, allows users to adjust a programme’s budget, grant size, applicants and review biases to see the kind of grantees and costs a competition will produce. A read-out gives the competition’s results, including the quality of ideas funded and the hours and money spent on it by both applicants and reviewers.
Finn Lübber, a doctoral student in Lübeck’s Social Neuroscience Lab, said he hoped the model would give “substance” to debates about the best ways to award research funding and lead to “productive solutions”.
The designers themselves have written that the model shows that pre-application lotteries, in which potential grantees are selected at random before they spend weeks on full applications, were the most cost-effective approach.
In a comment piece in the journal Nature Human Behaviour, they explain how they compared classical competitions, such as the one-stage National Institutes of Health R01 grant, the two-stage European Research Council process and tiebreaker lotteries, as used by the Swiss National Science Foundation (SNSF).
“When considering the costs, sunk costs, effects of self-selection and different biases in each additional review round, we can see that the pre-lottery scenario may have many advantages over others,” they write.
The use of tiebreaker lotteries has grown in recent years, with major funders in Germany and the UK among those trialling them, with broadly positive results.
According to the model, running an SNSF-style tiebreak lottery with a funder budget of 10 million would generate costs of 1.1 million for those involved, whereas using a pre-application lottery would generate just 460,000 in costs and leave 1.35 million of the 10 million funder budget unspent.
In the model’s typical tiebreak scenario, about 14 per cent of winners would be from disadvantaged groups, whereas that share would rise to about 37 per cent for a pre-application lottery. The share was about 18 per cent for typical one- or two-stage processes.
“Systems that implement a lottery at the last stage of the process to alleviate biases may only do so to a limited extent,” said Sören Krach, social neuroscience lead in the Lübeck lab and an author of the Nature paper. “That happens because biases are accumulated during the earlier competitive stages of review.”
Professor Krach and Mr Lübber told Times Higher Education: “The simulator alone cannot change anything, but the people using it are enabled to voice their critique on current systems of funding allocation by simulating their effects and bringing the results into the discussion.”
