Star Power

Where is the most important science produced?

A study (abstract available here) being released today suggests that it may be coming from a broader range of academic departments, but from a smaller number of elite scientists. The study -- from the National Bureau of Economic Research -- points to a new way of thinking about so-called "star" scientists.

The analysis is based on a look at the top-ranked departments and the top scientists (as judged by output of citation-weighted papers) in evolutionary biology from 1980 through 2000. The research found two apparently contradictory trends:

• The share of citation-weighted publications produced by the top 20 percent of departments fell from approximately 75 percent to 60 percent.
• The share of papers produced by the top 20 percent of individual scientists increased from 70 percent to 80 percent.

In other words, the role of the individual star became more important at a time that the role of the star department (while still significant) fell.

While one might expect stars to be concentrated at top departments, the paper finds that the current generation of scientists has seen increased collaboration -- and increased collaboration across levels of departmental rank. First, there are more people involved. The average number of authors on an evolutionary biology paper increased from 2.3 in 1980 to 3.8 in 2005.

But the study also finds collaborators spread among greater geographic distance -- perhaps not surprising in a period in which digital communication greatly improved the ability to collaborate remotely. From 1980 to 2005, the average distance (geographically) between collaborators increased from 325 to 500 miles.

And the average distance in rank of institutional departments increased as well. In 1980, it was about 30 (meaning someone at an institution ranked 20th, say, was collaborating with someone at an institution ranked 50th). By 2005, the average rank gap was 55. (Appropriately, the NBER study was conducted by business school researchers at three institutions in two countries: Ajay Agrawal of the University of Toronto, John McHale of Queen's University in Ontario, and Alexander Oettl of Georgia Institute of Technology.)

How are universities likely to respond to these trends? One possibility the authors note is that, in addition to focusing on recruiting stars, departments may place additional value on professors who collaborate with stars elsewhere -- given that such collaboration appears linked to the influence of the individual researcher.

But even as the research shows the ability of top scholars to thrive at places other than in top-ranked departments, the authors suggest that this type of analysis will encourage more efforts by top departments and those that aspire to that status to recruit stars. The authors note "a concern about the possibility of reputation-driven positive sorting at top institutions, with stars attracting stars irrespective of productivity-increasing knowledge spillovers. This in turn raises a concern about lost opportunities for stars to seed focused and dynamic research clusters at lower-ranked institutions."

But, the authors add: "Given the apparent role of star recruitment in department building -- which our evidence suggests would be particularly effective where the institution already has a cadre of incumbents working in related areas to the star and has a sufficient flow of new openings to take advantage of star-related recruitment externalities -- an offsetting force to excessive concentration could come from the incentive of lower-ranked institutions to use star-focused strategies to ascend departmental rankings."