Sir

With a view to propelling Canada to the top tier of international rankings, the presidents of five large Canadian universities have proposed a radical restructuring of post-secondary education. This includes the concentration of research in a few elite universities, with the others focusing on undergraduate education. Predictably, the country's other universities are outraged (see http://go.nature.com/6MMcsl).

The 'big five' universities (McGill, Montréal, British Columbia, Alberta and Toronto) already receive about 40% of the country's federal research funding. The 'Sheriff of Nottingham' model — which, like Robin Hood's adversary, robs the poor to pay the rich — relies on the premise that productivity increases faster than dollars invested. If this concept is valid, there should be more bang for the buck when funding is concentrated in a few institutions, rather than spread broadly.

Research productivity can be gauged by number of published papers and by the h-index, which measures highly cited publications, although these indicators have their flaws. I calculated both for all the researchers at 27 Canadian universities using data from the ISI Web of Science, based on papers published from 2005 to 2009. I extracted 2009 research funding per university from the three federal granting councils' websites.

According to these two indices, total research productivity and its impact both relate very strongly to the total funding received per university, but both relationships are significantly decelerating. (Details of these calculations are available from the author.) Expressed per dollar invested, research productivity and impact both decrease as funding per university increases. Bang for the research buck is better in smaller institutions.

Concentrating funding in large universities would probably increase their productivity, and perhaps their bragging rights. However, this would come at the price of reduced total research productivity summed over all Canadian universities. The same conclusion would probably apply to any system in which finite resources for research must be divided among a pool of contenders.