Abstract The European Research Council (ERC) is one of the most important research funding schemes of the European Union, supporting investigator-driven, excellent, frontier research. The ERC’s governing bodies have upheld the scheme’s ‘bottom-up’ approach, resisting any attempts to promote specific research topics. However, the selection process is not necessarily topic-neutral. Research has provided very limited evidence on the influence of a proposal’s research topic on its chances to be funded. This article addresses this gap by investigating what the research topics of ERC proposals are, how the proposals’ research topics vary across framework programs, countries, type of organization and grant, and whether a proposal’s research topic predicts the chances of being funded. We use state-of-the-art natural language processing techniques to examine the textual content of the abstract of 91,273 proposals for Starting, Consolidator, and Advanced Grants in the Seventh Framework Program (FP7) and Horizon 2020 (H2020)—that is, 95.3% of all the proposals, identifying 188 unique research topics. The proportion of proposals in each topic changed substantially from FP7 to H2020 programs, with one in four increasing or decreasing by more than 20% in relative terms. The research topic of a proposal is related to variation in success rates: 37.4% of proposals fall within topics that display significantly higher or lower chances of being funded. At the same time, the composition of the research topics of the funded proposals is similar to a hypothetical scenario in which topics did not affect success (80.8% overlap). The results suggest that the selection process on what is researched does not overturn the nature of ERC as a bottom-up program.
Seeber et al. (2026) studied this question.