An FWCI decomposition of Science Foundation Ireland funding

TL;DR

The study asks whether prioritizing socioeconomic impact in Irish funding preserves academic impact. It decomposes 3,243 publications linked to Science Foundation Ireland Investigator Awards (2012–2016) into SCOPUS FWCI values, fits a lognormal model to the FWCI distribution, and uses simulations to benchmark small-sample awards against an international median. Key findings show the FWCI distribution is consistent with a lognormal form, with a median near $e^{\mu} \approx 0.93$ and a mean near $e^{\mu + \frac{1}{2}\sigma^2} \approx 1.43$, indicating overall academic impact exceeds the global average; about 67% of awards surpass the international median after correction. The results support the view that socioeconomic-targeted funding can still yield substantial academic interest and provide a practical framework for national benchmarking and policy evaluation of research impact.

Abstract

In response to the 2008 global financial crisis, Science Foundation Ireland (SFI), now Research Ireland, pivoted to research with potential socioeconomic impact. Given that the latter can encompass higher technology readiness levels, which typically correlates with lower academic impact, it is interesting to understand how academic impact holds up in SFI funded research. Here we decompose SFI \textit{Investigator Awards} - arguably the most academic funding call - into $3,243$ constituent publications and field weighted citation impact (FWCI) values searchable in the SCOPUS database. Given that citation counts are skewed, we highlight the limitation of FWCI as a paper metric, which naively restricts one to comparisons of average FWCI ($\overline{\mathrm{FWCI}}$) in large samples. Neglecting publications with $\textrm{FWCI} < 0.1$ ($8.8\%$), SFI funded publications are well approximated by a lognormal distribution with $μ= -0.0761^{+0.017}_{-0.0039}$ and $ σ= 0.933^{+0.011}_{-0.012}$ at $95 \%$ confidence level. This equates to an $\overline{\mathrm{FWCI}} = 1.433^{+0.029}_{-0.015}$ well above $\overline{\mathrm{FWCI}}=1$ internationally. Broken down by award, we correct $\overline{\mathrm{FWCI}}$ for small samples using simulations and find $\sim 67\%$ exceed \textit{median} international academic interest, thus exhibiting a positive correlation between the potential for socioeconomic impact and academic interest.

Figures (5)

• Figure 1: A breakdown of the 148 awards by number of papers traceable using funding number metadata in SCOPUS. We removed awards with no original publications and provide a zoom-in for awards with less than 10 papers.
• Figure 2: A histogram of FWCI values from 3,243 publications from SFI awards with $\mathrm{FWCI} < 0.1$ publications highlighted in red. The green curve is the best fit lognormal to the blue bins.
• Figure 3: A histogram of the natural logarithm of FWCI values from Fig. \ref{['fig:lognormal']}. $\mathrm{FWCI} < 0.1$ publications are highlighted in red where we have shifted $\mathrm{FWCI} =0 \rightarrow 0.01$ papers for visual purposes. The green curve is the best fit normal to the blue bins.
• Figure 4: A breakdown of the average $\textrm{FWCI}$ for each of the 148 awards in Table I. 87 from 148 awards ($59 \%$) have $\overline{\mathrm{FWCI}} \geq 1$.
• Figure 5: The median $\overline{\mathrm{FWCI}}$ from the $\overline{\mathrm{FWCI}}$ of $n$ papers drawn randomly from lognormal distributions with $\sigma^2 = - 2 \mu$ across 100,000 simulations. For $n=1$ one recovers the median of a lognormal $e^{-\frac{1}{2} \sigma^2}$.