Sir

Another look at the data used by F. Frischknecht in his Correspondence 'Small countries are unexpected winners in ERC grant tables' (Nature 454, 690; 2008) reveals a different story.

Thirty-eight countries are eligible for European Research Council (ERC) grants. Of the 17 that did not receive any grants, 13 had a population size below the median (based on purchasing power parity, using data from the CIA world factbook at https://www.cia.gov/library/publications/the-world-factbook/index.html). At a glance, this might suggest that the smaller countries are the losers.

The reason why this is untrue is obvious, but Frischknecht's figures contain a similar error. He concludes that small countries were successful, because he omitted the small countries that did not receive any grants. Of course, larger countries are less likely to have no successful applicants, because they draw from a larger pool of talent. Consequently, Frischknecht's results are automatically biased.

If we look at the full data and use a log-linear model to regress the number of successful applicants against log-transformed population size and gross domestic product (GDP) per person, we find that the population effect (coefficient of 0.9, standard error 0.11) is not distinguishable from 1 (P = 0.3), indicating no evidence of bias towards smaller countries.

The wealth effect is much more important: doubling GDP per head increases the number of grants tenfold (coefficient of 2.3, standard error 0.49). One might conclude that the winners are the rich countries. Again.

See also European grants: a lifeline in poorly funded countries.