The research assessment exercise is biased owing to the "small world" effect, according to an analysis by Andrew Colman, an expert in game theory and cooperative reasoning at the University of Leicester.
Professor Colman noticed that the results in small units of assessment were far higher than those in larger units. He analysed the relationship between performance and size and found a "quite large and significant" correlation that allowed him "to rule out chance as the explanation". A quarter of the variance in research ratings is explained by size.
Arguments that smaller units of assessment are genuinely superior at research can also be dismissed, he said. Larger units should be able to select the best researchers from a bigger pool and to train more postgraduates.
Professor Colman said: "In small mutually interdependent communities, people seem more likely to engage in mutually beneficial social behaviour."