Interdisciplinary Researchers Attain Better Long-Term Funding Performance

The paper examines whether researchers who pursue interdisciplinary careers face persistently lower impact and funding success, as suggested by prior studies, or whether these effects differ over longer horizons and broader datasets. Interdisciplinary work is widely promoted for tackling complex problems and fostering innovation, yet evidence indicates mixed outcomes: recognition tends to occur for proximal combinations of fields, while more distal combinations are perceived as risky; citation rates for interdisciplinary outputs are often lower; and funding success is reported to be reduced for interdisciplinary proposals. The authors analyze more than 44,000 UK research grants (2006–2018) from seven discipline-based research councils to compare the career progressions of investigators classified as cross-council (securing funding from at least two different councils) versus within-council (funded by a single council). They investigate collaboration structures, short-term research impact, and long-term funding trajectories to determine whether interdisciplinary engagement confers advantages or penalties over time.

Prior literature documents a rise in interdisciplinary research and increased knowledge transfer across fields. However, outcomes are mixed: recognition and impact accrue more readily to proximal interdisciplinarity, while distal combinations often underperform. Several studies report lower citation rates for interdisciplinary outputs and lower funding success for interdisciplinary proposals relative to specialized ones. Additional findings highlight the growing dominance of team science, multi-institutional collaborations, and the prominence of elite institutions in orchestrating collaborations and producing impactful research. Theories of social networks suggest that brokerage positions (structural holes) may provide access to diverse information and innovation opportunities, potentially benefiting interdisciplinary researchers through central network roles, albeit with coordination and translation costs that may suppress short-term productivity or citations.

Data: The study uses 44,419 research grants funded between 2006 and 2018 by the seven UK research councils (AHRC, BBSRC, ESRC, EPSRC, MRC, NERC, STFC). Projects can be associated with multiple subjects (out of 104). Investigators are categorized as cross-council if they received funding from at least two councils, and within-council otherwise.

Council co-activity network: Nodes are research councils; a weighted link connects two councils if they have both supported at least one investigator. Link weights are the ratio of observed co-funded investigators to the expected number under a randomized null model.

Collaboration network: Nodes are investigators (PIs or CIs); an undirected link exists if two investigators partnered on one or more grants. Grants with only a single investigator (PI only) are excluded. The largest connected component (LCC) comprises 86% of investigators. Analyses include investigators in the LCC with at least two grants during 2006–2018, yielding 6,911 cross-council and 12,563 within-council investigators. Network metrics computed per investigator include degree centrality, closeness centrality, betweenness centrality, and normalized effective size (brokerage), defined for unweighted, undirected graphs as 1 − (2mi)/(ki(ki − 1)) or equivalently 1 − Ci(ki − 1)/ki where ki is degree and Ci is clustering coefficient; higher values indicate more brokerage opportunities.

Linking grants to publications and citations: For each grant, all related publications (titles and DOIs) are recorded and matched to the Microsoft Academic Graph (MAG) to obtain publication metadata and citations. A total of 409,546 publications are matched. Citations are accumulated within 5 years of publication and normalized by the average citations of papers in the same year and discipline (MAG’s 19 top-level disciplines).

Propensity score matching (PSM): To control for confounding, cross-council and within-council PIs are matched using propensity scores from multivariable logistic regression. Nearest-neighbor matching within a 0.01 caliper on the probability scale is used. Balance checks show standardized differences d < 0.1, and no significant differences by t-tests (p > 0.1) and Kruskal–Wallis tests (p > 0.1) across covariates post-matching.

Matched analyses:

• Research outcomes and impact (2006–2013): PIs are characterized by five covariates capturing institutional ranking (by total funding accrued by the PI’s institution over 2006–2018), number of grants, average funding value per grant, average team size, and average project duration (discipline- and time-adjusted). Cross-council and within-council PIs are matched based on 2006–2013 profiles, yielding 958 matched pairs. Outcomes compared (for grants 2006–2013 with complete outcomes before 2018) include: average number of papers per project, average total citations per grant, and average citations per paper per grant (discipline- and year-normalized, 5-year window).
• Long-term funding trajectory: Defines 2006–2010 as the in-sample period for matching, and 2011–2018 as the out-of-sample period for performance comparison. Matching includes both funding profile variables and research performance metrics (institution ranking; number of grants; average grant value; average team size; average duration; average publications; average total citations per grant; average citations per paper per grant) measured in 2006–2010. This yields 709 matched pairs. Out-of-sample performance metrics compared (2011–2018) include number of grants, total funded value, average team size, and project duration.

Additional analyses: Sliding window analyses include only investigators with at least two grants to avoid inflating within-council counts. Robustness checks consider different time windows, institutional stratification criteria, and an alternative interdisciplinarity definition based on the number of distinct publication fields in MAG.

• Collaboration and cross-disciplinarity trends: Team size and the number of affiliations per grant increased over 2006–2018. The average number of subjects per grant rose, with 44% of funded projects linked to at least two subjects. The fraction of cross-council investigators rose from about 0.17 (2006) to 0.26 (2018).
• Council co-activity network: The network became fully connected by 2016–2018, with new links connecting AHRC–MRC and AHRC–STFC. Link weights increased substantially between BBSRC–NERC (+29%) and AHRC–ESRC (+90%), aligning with UKRI policies encouraging cross-council research.
• Institutional stratification: Tier I institutions (top ~40 by total funding) have a higher proportion of cross-council investigators than Tier II in both 2006–2008 and 2016–2018 (chi-square p < 0.0001; odds ratio 1.67 in 2006–2008 and 1.28 in 2016–2018). Tier II showed a larger absolute increase (18% to 26%) than Tier I (27% to 31%).
• Structural network advantage: Cross-council investigators exhibit significantly higher degree, closeness, betweenness centrality, and normalized effective size than within-council peers (Welch’s t-tests p < 0.001), indicating stronger centrality and brokerage roles. Advantages increase with the number of distinct funding councils supported (Nfunder).
• Short-term research outcomes: In matched pairs (n = 958), cross-council and within-council PIs produce similar numbers of publications per grant, but cross-council PIs receive fewer citations: lower total citations per grant (t-test p = 0.0021) and lower mean citations per paper (t-test p = 0.0004), consistent with prior reports of lower short-term citation impact for interdisciplinary work.
• Long-term funding performance: In matched pairs for long-term comparison (n = 709), cross-council PIs outperform within-council PIs in the out-of-sample period (2011–2018) in number of grants, total funded value, and average team size. Results are robust across alternative time windows and to an alternative interdisciplinarity measure based on MAG publication fields.

The study reconciles mixed evidence on interdisciplinarity by distinguishing short-term research impact from longer-term funding outcomes. Cross-council (interdisciplinary) investigators occupy central, brokerage-rich positions in collaboration networks, which confer structural advantages for information flow and partnership formation. However, these advantages do not immediately translate into higher citation impact within a 5-year window; matched analyses show fewer citations for cross-council investigators despite similar publication volumes. Potential explanations include higher coordination and translation costs when bridging disciplines and the possibility that interdisciplinary work garners recognition over longer horizons than captured by a 5-year citation window. Crucially, when tracking subsequent career progression, cross-council investigators achieve superior funding outcomes in terms of grant volume, total funding value, and team size. These findings suggest that although interdisciplinary careers may entail short-term penalties in citation-based impact, they provide longer-term advantages in securing resources, aligning with policies encouraging cross-boundary research and offering guidance for researchers and funders on the strategic value of interdisciplinarity.

The paper demonstrates that interdisciplinary (cross-council) researchers play key brokerage roles in academic collaboration networks and, despite lower short-term citation impact relative to matched within-council peers, outperform in long-term funding performance (grant counts, total value, and team size). This helps explain the sustained push toward interdisciplinarity and its growing prevalence in the UK funding landscape. Future work should examine longer citation windows to capture delayed recognition, refine measures to distinguish interdisciplinary from multidisciplinary activity more precisely, assess generalizability beyond the UK context and public funders, and explore mechanisms by which brokerage positions translate into funding advantages (e.g., network-mediated access to diverse collaborators, proposal competitiveness).

• Citation window: Impact was measured using 5-year citation counts, which may understate delayed recognition common in interdisciplinary work.
• Interdisciplinarity definition: Cross-council status is a coarse proxy that blends interdisciplinary and multidisciplinary activity and does not capture proximity versus distance between fields.
• Data scope: Analyses are limited to UKRI-funded grants and may not generalize to other countries, private funders, or disciplines with different funding structures.
• Matching constraints: Although propensity score matching balances observed covariates, unobserved confounders (e.g., topic novelty, PI career stage nuances) may still bias estimates.
• Network construction: Collaboration networks exclude single-investigator grants and focus on the LCC; measures are unweighted and may not account for intensity of collaboration.
• Temporal choices: Sliding windows and period definitions could influence categorization and dynamics; robustness checks mitigate but cannot eliminate such sensitivity.