February 3rd, 2009

US Science in Relative Decline?


Geoff Davis

Thompson Reuters has a new report out showing US market share of scientific papers in continuing decline relative to Asia. That’s not too surprising given that China and India are both rapidly modernizing their economies. One thing from the report that struck me, though, was this:

But one trend bears watching: in the last two years, the absolute number of Thomson Reuters-indexed scientific papers from the U.S. has edged downward: from a high of roughly 291,500 in 2005 to approximately 286,000 in both 2006 and 2007.

What would cause paper output to decline? I’m sure many will point to the stagnation in the NIH budget as the cause, but I’d be surprised if that were the full story. Surely the insane pressure that people are now under has compensated to some extent for whatever difficulties reduced equipment budgets and the like have caused. And publications are a lagging indicator – the work done for papers published in 2006 was probably done in 2004, right at the end of the doubling.

I wonder if part of the story is related to changes in lab sizes?

Two factors may be contributing to an increase in lab size:

  • The NIH has shifted some funds from individual grants to bigger projects.
  • The NIH budget doubling has substantially increased the variance in the grant making process. Individual researchers are less likely to get grants, but the grants are for more money. One way to cope with increasing variance and uncertainty is to band together. If you partner up with someone doing similar work, submit independent proposals, and agree to share the proceeds (i.e. both people work on the winning proposal), you maintain the same expected return per person but have greatly reduced variability (variance reduction is an important reason that people form companies). I’m guessing that lab sizes have increased at a lot of institutions in part for these kinds of reasons.

One of the drawbacks to increased organization size is increased communication overhead. With N people, you have O(N^2) channels of communication to maintain. If you double the number of people working on a project, you don’t move twice as fast. You have to spend some fraction of your time keeping others apprised of your status, dealing with resource contention, and so on. There is a classic book, The Mythical Man Month that all software engineers are required to read fairly early on in their careers – it’s about an attempt by IBM back in the 70′s to ship an operating system faster by adding more programmers to the project. The resulting explosion of communication overhead ended up delaying things even further.

This is all idle speculation on my part at this point – I don’t really know if lab sizes are tracked – but given some recent anecdotes I’ve seen about the increasing role of Big Science, something to consider.