The value of research funding for knowledge creation and dissemination: A study of SNSF Research Grants

Scientific research at universities and research organizations is central to knowledge-based societies and economic and social progress. Public research grants have become a key policy tool to promote knowledge creation and diffusion, but rigorous evidence at the individual researcher level is challenging due to selection into funding, multiple and consecutive grants, and field-dependent output patterns. This study quantifies the effect of the Swiss National Science Foundation’s (SNSF) project funding (PF) grants on individual researchers’ future publication outputs and dissemination. Using data on all applicants (winners and non-winners) and their peer-review evaluation scores for 20,476 project grants (2005–2019), we compare researchers with similarly rated proposals to estimate the causal effect of grant receipt on outputs. Beyond peer-reviewed publications and citations, we examine preprints and altmetrics to capture accessibility and public attention. We also explore field-normalized impact via relative citation ratio (RCR) and field citation ratio (FCR), and assess persistence of effects over several years after funding.

Prior work has examined funding effects at institutional, laboratory, and individual levels with mixed findings. At the university level, studies (e.g., Adams and Griliches, 1998; Payne, 2002; Payne and Siow, 2003) report positive elasticities of publications to funding, with diminishing returns at high funding levels (Wahls, 2018). At the lab level, evidence is heterogeneous: Arora et al. (1998) find positive effects in engineering/biotechnology, while Carayol and Matt (2004) report weak links across broader fields. At the individual level, funding can increase outputs, particularly for younger researchers (Arora and Gambardella, 2005). NIH grants yielded about one additional publication over five years (Jacob and Lefgren, 2011); similar magnitudes are observed in Germany (Hottenrott and Thorwarth, 2011) and Canada (Beaudry and Allaoui, 2012). Results on impact/quality are mixed: some find increased publications but no citation gains (Benavente et al., 2012), while others find positive citation effects (Hottenrott and Lawson, 2017; Tahmooresnejad and Beaudry, 2019). Novelty effects can depend on seniority and gender (Wang et al., 2018). In Switzerland, participation in grant competitions can itself foster collaboration and outputs, with nuanced citation patterns (Ayoubi et al., 2019). The Swiss system is characterized by a strong central funder (SNSF) and complementary external sources such as ERC grants.

Empirical strategy combines longitudinal mixed-effects regressions with non-parametric matching to estimate the causal effect of SNSF PF grants on researcher outputs. - Data: SNSF administrative records on project funding applications (PF and Sinergia) and evaluation scores for 2005–2019; linked to Dimensions database for disambiguated publication, citation, preprint, altmetric, RCR, and FCR data. The initial eligible pool encompassed 11,228 researchers; after matching to unique Dimensions IDs, the final sample includes 8,793 researchers and 82,249 researcher-year observations (average panel length 9.35 years; max 15 years). Analyses use 72,738 complete observations from 8,282 researchers after applying lags and completeness filters. Data retrieved September 2020. - Study window: Researcher-specific, starting at first observation as (co-)PI in PF or as a career funding grantee (postdoctoral level), lower bound 2005, ending 2019. A five-year pre-sample period is constructed for each researcher to control for prior performance in models and for matching. - Treatment: Indicator for access to SNSF PF funding, lagged by one year; distinguished between roles as PI and as co-PI. - Outcomes: Annual number of peer-reviewed articles; annual number of preprints; yearly total citations to articles published during the observation window; average yearly citations per article; yearly average altmetric score for articles published in year t; field-normalized citation metrics: RCR (PubMed-indexed, LS only, citation- and time-normalized via co-citation field) and FCR (citations normalized by year and fixed Fields of Research categories); for some analyses, indicator for highly cited (FCR > 3). - Controls/confounders: Gender, age, institution type (Cantonal university, ETH Domain, UAS/UTE/other), research area (Life Sciences, STEM, SSH), time effects, and proposal evaluation scores (six-point scale 1D–6A). Evaluation score used as a rolling four-year average of grades received as PI/co-PI; if absent, all-time average used. - Longitudinal mixed-effects models: For count outcomes (articles, preprints), negative binomial mixed models with random intercepts for researchers; for continuous outcomes (average citations per article; log-transformed altmetrics, RCR, FCR), linear mixed models with researcher random intercepts. Funding is lagged by one year; models include confounders and time trends. - Non-parametric treatment effect estimation: Nearest-neighbor propensity score matching with replacement, augmented by exact matching on funding round year and research field, and Mahalanobis distance on propensity and evaluation scores with a caliper. Propensity scores estimated via probit including pre-sample performance metrics (log pre-sample articles, preprints, average citations, RCR, FCR, altmetrics as applicable), evaluation score, age (log), gender, institution type, field, and year fixed effects. After matching, average treatment effect on the treated (ATT) computed as mean outcome differences; standard errors corrected for repeated use of controls. - Persistence analysis: Effects estimated for t+1, t+2, and t+3 after treatment year, recognizing potential confounding by follow-on grants beyond this horizon. - Heterogeneity: Interaction analyses by age groups (<45, 45–54, 55–64, 65+) and by research area for both counts and continuous outcomes.

- Publication quantity: Mixed-effects negative binomial models show that SNSF funding as PI (t−1) increases next-year publication counts by 21% (IRR=1.21, 95% CI 1.19–1.22, p<0.001); as co-PI by 11% (IRR=1.11, 95% CI 1.09–1.13). Preprints increase by 30% for PIs (IRR=1.30, 95% CI 1.22–1.39) and 10% for co-PIs (IRR=1.10, 95% CI 1.02–1.19). - Persistence: Matching estimates indicate about one additional peer-reviewed article per year in each of the three years following funding (t+1 to t+3). Preprint increases are significant in t+1 and persist into later years. - Citations and impact: Average citations per article increase by 0.33 for funded PIs (p<0.001). Total yearly citations are significantly higher for funded researchers over t+1 to t+3, with some decline in effect size over time. The probability of being highly cited (FCR > 3) is 5.5 percentage points higher in t+1 for funded researchers. - Field-normalized metrics: Mixed-models suggest a positive association with FCR; RCR effects are small and not robust across methods (short-run differences modest; significance at 10% only in some horizons). FCR effects are strongest in t+1 and fade thereafter in matching analyses. - Altmetrics and dissemination: Funded researchers achieve higher attention scores; altmetrics are on average 5.1% higher for funded PIs (by Sept 2020). Matching shows significant increases in t+1; medium-run differences are weaker. - Heterogeneity: • Career stage: Young researchers (<45) see the largest gains in article counts and average citations per article; very senior researchers (65+) also benefit substantially in maintaining productivity. Mid-career groups show smaller differences. • Field: STEM shows the largest publication count gains for funded PIs (~23% more articles vs. unfunded); LS ~15%; SSH ~12%. For average citations per article, SSH shows the largest relative gains (~14–15%), with smaller gains in STEM (~8%) and LS (~6%). - Contextual findings: Male researchers publish more and have higher grant likelihood than females; ETH Domain researchers tend to have higher outputs than those at cantonal universities; field and time trends are substantial (e.g., preprints rising in later years).

The study addresses whether competitive project funding causally increases researchers’ outputs and dissemination. By leveraging applicant-level data, proposal evaluation scores, and robust econometric designs (mixed-effects models and matching with exact controls on field and year), the analysis shows that SNSF project funding leads to sustained increases in publication output and earlier dissemination (preprints), alongside higher citation impact and public attention, beyond mere quantity effects. Effects persist for at least three years post-award, consistent with project lifecycles and time-to-publication. The stronger gains for early-career researchers suggest funding alleviates binding resource constraints and accelerates capacity building. Field heterogeneity indicates higher funding dependence of output in STEM/LS, while quality/impact gains (citations per article) are relatively stronger in SSH, implying that targeted funding can enhance both productivity and influence across diverse disciplines. Elevated altmetrics among funded researchers signal broader outreach to the public and policy communities, aligning with funders’ dissemination goals. Overall, competitive project funding complements institutional core funds by enabling additional, more visible research that contributes to scientific progress and public engagement.

This paper contributes methodologically by comparing grant winners to non-winners within the applicant pool and controlling for proposal quality via evaluation scores, and substantively by incorporating preprints, altmetrics, and field-normalized citation metrics (RCR, FCR) into funding impact assessment. The main finding is that SNSF project funding results in approximately one additional peer-reviewed article per year for each of the three years following the award, along with higher preprint activity, increased citations (including average citations per article), and higher altmetric attention. Effects are particularly pronounced for early-career and very senior researchers and vary across fields (largest quantity gains in STEM; largest per-article citation gains in SSH). Compared to prior literature, effect sizes are larger, likely reflecting the centrality of SNSF in Switzerland and precise identification of truly unfunded controls. Policy-wise, results underscore the value of competitive project grants for fostering knowledge creation and dissemination beyond what institutional funding alone achieves. Future research should investigate the role of funding amounts, project novelty, team composition and dynamics, industry funding interactions, and longer-run impacts beyond three years.

Key limitations include: (1) lack of data on industry/private sector project funding, potentially important in engineering; (2) repeated and overlapping grants complicate simple long-term causal identification and difference-in-differences designs; (3) possible unobserved time-varying factors (e.g., family circumstances, health, administrative/service loads) affecting outputs; (4) absence of detailed team composition and role data limits analysis of team effects and PI/co-PI dynamics; (5) altmetrics and preprints may capture popularity or network effects and can be gamed, and altmetrics were only available as cumulative snapshots (Sept 2020); (6) RCR available only for PubMed-indexed LS publications; (7) the study does not analyze funding amounts, research novelty, or heterogeneity beyond age and field in depth.