Do NSF and UKRI lead each other? A replication of cross-funder topic growth in research awards
Year: 2026
Type: preprint
Abstract: A post-hoc analysis of ten years of computing awards found that when a research topic grew in NSF's portfolio, it grew in UKRI's two years later (b = +0.373, Holm p = 0.005, 38 topics), with little evidence of an association in the reverse direction. Because that specification was chosen after the planned test had failed (Holm p = 0.14), we fixed it in a written plan, committed to a Git repository before any new data was downloaded, and tested it on two new data sets. R1 used the same computing topics with outcomes from 2024-2025 that no earlier estimate had used. The primary replication (R2) used non-computing awards starting in 2014–2025. The NSF and UKRI frames contained 203,233 awards after abstract filtering and deduplication. The analysis covered 1,173 eligible OpenAlex Topics outside the original test. The lead did not replicate. In R1, NSF growth did not predict later UKRI growth (b = -0.064, unadjusted p = 0.75). In R2, growth predicted later growth in both directions (NSF to UKRI b = +0.062, Holm p < 10⁻⁶; UKRI to NSF b = +0.042, Holm p < 0.001), so the decision rule fixed in advance returns "no lead shown". Across the portfolio, topic growth at the two funders moves together, with neither funder clearly ahead. Exploratory analyses after the outcome support this reading: the difference between the two directions is small and not significant (+0.020, p = 0.30); the association is stronger than in 1,000 random re-pairings of topics within the same field or subfield; it appears at horizons of one and two years and is gone at three; and it is strongest within the same year. The case shows that a convincing post-hoc result from a small set of topics can fail a replication with a committed plan, and that such a replication can be built from public award data.
Author: Mamun, S M Abdulla Al
Institution: University of Manchester
Language: English
Subfield: Management of Technology and Innovation
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7236594853
