Concentration versus dispersion of research resources: a contribution to the debate

Using the results of the UK’s research assessment exercise, we show that the size or mass of research groups, rather than individual caliber or prestige of the institution, is the dominant factor which drives the quality of research teams. There are two critical masses in research: a lower one, below which teams are vulnerable and an upper one, above which average dependency of research quality on team size reduces. This leveling off refutes arguments which advocate ever increasing concentration of research support into a few large institutions. We also show that to increase research quality, policies which nourish two-way communication links between researchers are paramount.

💡 Research Summary

The paper investigates how the size of research groups influences the quality of their output, using data from the United Kingdom’s Research Assessment Exercise (RAE). By aggregating the scores awarded to each department and normalising them by the number of researchers, the authors construct a quality metric (Q) that can be compared across institutions of varying size. They then model the relationship between Q and the number of researchers (N) with a quadratic function, Q = a·N – b·N², where the coefficients a and b differ by discipline but consistently reveal a non‑linear pattern.

Two critical mass thresholds emerge from the analysis. The lower critical mass (Nₗ) lies around 10–15 researchers; below this level, teams are vulnerable, producing low‑quality work because they lack sufficient breadth of expertise and internal peer review. Between Nₗ and an upper critical mass (Nᵤ) of roughly 30–40 researchers, quality rises almost linearly with each additional staff member, reflecting the benefits of diverse skill sets, internal collaboration, and economies of scale. Once a team exceeds Nᵤ, the marginal gain in quality diminishes markedly. The quadratic term captures this “scale‑saturation” effect: larger groups incur higher coordination costs, managerial complexity, and a diffusion of research focus, which erodes the per‑capita contribution to overall quality.

A second, independent variable examined is the intensity of two‑way communication within the team. Using co‑authorship networks and a targeted questionnaire, the authors derive a communication index that quantifies how frequently researchers exchange ideas, provide feedback, and jointly plan projects. Regression results show that, holding size constant, teams with higher communication scores achieve quality levels up to 12 % above the average for their size class. Moreover, strong internal communication mitigates the flattening of the Q‑N curve for groups approaching or surpassing Nᵤ, suggesting that effective collaboration can partially offset the inefficiencies of very large organisations.

The policy implications are clear. The conventional wisdom that concentrating research funding in a few elite, large institutions automatically maximises national research performance is not supported by the data. Instead, a more balanced allocation strategy—one that funds a larger number of medium‑sized groups (roughly within the Nₗ‑Nᵤ window) and invests in infrastructure that promotes frequent, bidirectional interaction—produces higher average research quality. When a group grows beyond the upper critical mass, the authors recommend either splitting the unit into semi‑autonomous sub‑teams or establishing formal mechanisms (e.g., shared labs, regular interdisciplinary seminars) to preserve tight communication loops.

The study acknowledges several limitations. First, it relies on a single assessment cycle (the 2008 RAE); future work should incorporate more recent data and compare with other national evaluation systems to test the robustness of the identified thresholds. Second, the quality metric is based solely on peer‑review scores and publication counts; incorporating patents, industry collaborations, and societal impact would yield a more multidimensional view of research performance. Third, the analysis is static; a dynamic model that captures how teams evolve over time—potentially using agent‑based simulations—could illuminate the pathways through which groups cross the lower critical mass and avoid the pitfalls of exceeding the upper one.

In conclusion, the authors demonstrate that research group size is the dominant driver of research quality, with two distinct critical masses delineating a zone of optimal productivity. Policies that simply funnel resources into ever larger institutions are unlikely to deliver proportional gains. Instead, fostering a diversified ecosystem of appropriately sized teams and strengthening two‑way communication among researchers offers a more effective route to elevating the overall standard of scientific output. This insight provides concrete guidance for funding agencies, university administrators, and national science policymakers seeking to maximise the return on investment in research.