Administrative Discretion in Scientific Funding: Evidence from a Prestigious Postdoctoral Training Program

The scientific community is engaged in an active debate on the value of its peer-review system. Does peer review actually serve the role we envision for it-that of helping government agencies predict what ideas have the best chance of contributing to scientific advancement? Many federal agencies use a two-step review process that includes programmatic discretion in selecting awards. This process allows us to determine whether success in a future independent scientific-research career is more accurately predicted by peer-review recommendations or discretion by program staff and institute leaders. Using data from a prestigious training program at the National Institute of Health (NIH), the Ruth L. Kirschstein National Research Service Award (NRSA), we provide evidence on the efficacy of peer review. We find that, despite all current claims to the contrary, the existing peer-review system works as intended. It more closely predicts high-quality science and future research independence than discretion. We discover also that regression discontinuity, the econometric method typically used to examine the effect of scientific funding, does not fit many scientific-funding models and should only be used with caution when studying federal awards for science.