sufficient incentives for entry, science policy is critical to support such career choices.
On this dimension, lengthening training phases require special consideration. As
research grants and patents come later in the life-cycle, compensation for the delayed
start to the career appears needed.
2. Evaluation. Effective science policy - both in granting patent rights and granting
research support - necessarily relies on the effective evaluation of ideas. The
increasing challenge for an individual scientist to span broad research areas, as
detailed in Section II, suggests the same increasing challenges for government
agencies and the individual scientists within them who are tasked with evaluation.
3. Effort. Even conditional on choosing a scientific/technical career, the rate and
direction of innovator effort responds to the incentives science policy imposes.
Evaluation methods that privilege narrow ideas, whether in research grants, tenure
systems, or elsewhere, will chill efforts to produce ideas with broad impact. Incentive
mechanisms that privilege individual researchers, including high-status individual
prizes like the Nobel Prize, can undermine teamwork, even as teamwork is
increasingly needed for broad impact.
These challenges and possible responses are detailed below.
IV. Rethinking Science Policy: Life-Cycle Issues
The extension of training and decline in early life-cycle innovative output, as detailed in
Section II, raise the cost of becoming a scientist. Labor economics provides a useful framework
for understanding this cost. In particular, let there be some value V to being a scientist once the
necessary training is finished. This value can incorporate the future wage stream, discounted to
the moment one starts actively innovating. More generally, the value V could include the
expected value of research grants, status, or the simple joy of creativity, all viewed from the
moment one begins the active innovative career.
The problem with lengthening training is that it delays receipt of this expected value V.
A standard economic model suggests that the cost of one year’s delay is about 10% of V. That
17