I’ve just read a very important paper that I strongly recommend to anyone interested in innovation and growth. The paper, by Ben Jones, an economist at Northwestern, is called “As Science Evolves, How Can Science Policy?”. Jones documents two crucial points. First, as the length of education and training for a scientists gets longer, the value of a scientific career drops sharply.
Second, teamwork has been getting more important. For example, on the issue of teamwork, Jones looks at all science, engineering, and social science journal articles published from 1995 to 2005, and shows that team-written papers have far more impact than solo papers.
The first chart shows that team-written papers end up drawing a lot more cites than solo papers, on average, in both science and engineering, and the social sciences. The second chart shows that the “home run” papers are much more likely to come from teams.
(Related papers include “The Increasing Dominance of Teams in Production of Knowledge” from Science (2007) and “The Burden of Knowledge and the Death of the Renaissance Man: Is Innovation Getting Harder?”)
Jones then goes on to point out that the current incentive structure in science is struggling to deal with a world where scientists have to wait so long to get started:
For example, if careers in finance, management, or law require more static levels of training, then scientific careers will be increasingly costly by comparison. The estimated 6-8 year delay in becoming an active innovator over the 20th century suggests, at a standard 10% discount rate, a compound 45-55% decline in the value to becoming a scientist. This kind of selection effect may not only slow scientific progress but also slow economic growth, should the positive spillovers that follow from idea creation (see Section III) not feature in other white collar careers. The recent finance boom, drawing talented undergraduates into quickly attained,high wage streams, may make this comparison particularly acute.
Teamwork poses several tough problems for the current structure of science. For example, the process of scientific evaluation–grants and patents–is not set up to deal with cross-disciplinary research. Jones writes:
The evaluation of ideas is a central role of government that relies on the correct application of expertise within government institutions. Evaluation is necessary ex-post of innovations, particularly in securing intellectual property rights through the United States Patent and Trademark Office (USPTO). Evaluation is also necessary ex-ante of innovations, particularly in allocating limited research grant support through government agencies such as the National Institutes of Health (NIH) and National Science Foundation (NSF). Traditionally, the USPTO has used a single examiner to evaluate and adjust the property rights claims in a patent. The NIH has employed a panel evaluation model within particular study sections, which cover narrowly delimited areas of science. These evaluation models may be increasingly ineffective for assessing broader ideas. While researchers and innovators themselves increasingly use teams (and teams of growing size) that can span broad bodies of knowledge, their research ideas may be constrained by evaluation systems that bring limited breadth of expertise to bear. … Moreover, as knowledge accumulates, the narrowness of individual expert evaluators will only increase.
I’ve been arguing for a while that we’ve been experiencing an innovation shortfall in recent years, but I’ve stayed away from explanations so far. Jones’ analysis–that scientific incentives have become increasingly misaligned with the realities of the scientific endeavor–has some real possibilities for understanding what’s been going on. It helps explains why research productivity seems to be falling, despite the Internet and more tools for communication. And it helps explain why more and more kids are turning away from science.
This analysis also leads Jones to some novel solutions. For example, Jones writes:
….forcing grant dollars (not wage support) earlier in the life-cycle looks sub-optimal, in the sense that early-life cycle researchers are less likely to produce important ideas…[one alternative] is to accelerate training. This approach may be especially attractive and of increasing importance if an individual’s raw innovation potential is greatest when young….increasing the quality, intensity, and/or focus of training throughout the early life cycle may all bring young scientists more quickly to the knowledge frontier, offsetting the expansion of foundational knowledge and allowing individuals to substitute toward active, high quality innovation at younger ages. The training duration problem thus bears on education policy from childhood and suggests that a central goal of educational policy — and one of increasing importance — is to ensure that future innovators are being trained efficiently from very young ages.
We’re now in the land of cross-national cultural differences, where intensive training in science and math at a younger age may confer surprising advantages.