The presidents of five of Canada’s largest universities (Alberta, British Columbia, McGill, Montréal, and Toronto) recently proposed that postsecondary education be restructured in order to propel Canadian universities to the top tier of international rankings. In an article in Maclean’s magazine of Sept. 21, they encouraged the federal government to increase funding to research. More controversially, they also proposed that research funding be concentrated in an elite set of universities, while the others focus on undergraduate education. Smaller universities were predictably outraged.
This proposal rests on two premises (as I recently argued in a letter published in Nature, 461:1198). The first premise is uncontroversial: that more funding would increase research output. The second premise is dicier: that concentrating research funding in elite institutions would improve total research productivity. The “Big Five” universities together already receive approximately 40 percent of the total federal research funding. Taking money from the poor to fund the already rich – the “Sheriff of Nottingham” model – would work only if bang for the research buck is better when money is concentrated in a few institutions, rather than spread broadly.
This premise is easily testable. One index of research productivity is the total number of research papers published. Another index would be the “H-index”, which tallies the number of highly cited papers. Citation analyses definitely have drawbacks, but they provide a first-pass approximation of research activity,
I calculated the number of papers published from Jan. 1, 2005 to Sept. 1, 2009 for all researchers at several Canadian universities, and university-wide H-indices, using data from the ISI Web of Science. I also tallied total funding from the three federal granting councils – CIHR, NSERC and SSHRC – based on competition results reported on their websites.
Total research productivity of Canadian universities, both in terms of numbers of papers published and in terms of impact, is strongly related to total federal granting council funding (Fig. 1). Funding (log-transformed) can statistically account for more than 94 percent of the variability among universities in numbers of papers published (log-transformed) and H-index (also log-transformed).
Moreover, both relationships are significantly decelerating. Fig. 1 shows that log (productivity)= a + b·log(funding), where a and b are regression coefficients. Equivalently, productivity is proportional to funding raised to the exponent b. In exponential relationships, when b<1, the dependent variable (productivity) increases as a decelerating function of the independent variable (funding). In the present case, b=0.80 ± 0.08 (95% c.i.) for the total number of papers published, and b=0.39 ±0 .03 for the H-index. Both productivity and impact are decelerating functions of dollars invested.
Viewed another way, doubling the funding does not double the number of publications nor the cumulative impact. Indeed, productivity per dollar decreases as total funding increases. Bang for the research buck is better in smaller institutions.
I did find one benefit of being a researcher in a large institution: increased probability of being an author on the highest-impact publications. I searched for the 50 most cited publications in 2006, 2007 and 2008 on which there was at least one Canadian author. Among these 150 articles, 59 percent included one or more authors from a Big Five university. In contrast, only 35 percent had at least one author from some other Canadian university (some papers had no university-based author). This undoubtedly reflects, in part, the fact that 68 percent of these articles dealt with medical subjects. The Big Five all have medical schools, while most of the rest do not. Top-cited articles, medical or otherwise, are strongly dominated by studies with long author lists (e.g., clinical trials at many institutions). It appears that Canadians in prominent universities are more likely to be included in such consortia.
To concentrate funding in a few universities would likely improve their research productivity and their bragging rights when global rankings are considered. However, my analysis suggests that the research productivity and impact of the Canadian university system as a whole would decrease as a result. There are good pedagogic reasons not to separate research and teaching. It appears that that would not even be good for Canadian research.
Dr. Currie is a professor of biology at the University of Ottawa.