MOD: Estimating the Effect of Exposure to Superstar Scientists: Evidence from Academia and the Biopharmaceutical Sector

National Bureau Of Economic Research Inc, Cambridge MA

Investigators

Abstract

By measuring the "training" effects of research grants and synergies across research efforts, this project provides significant and missing elements in the calculus behind the perennial debates about the benefits from public research and appropriate levels of funding. Measures of geographic influence also inform current policy questions about U.S. competitiveness vis-a-vis the magnetic pull that high quality research exerts on the locational choice of firms in an international context. Moreover, since part of this research explores the mechanisms through which knowledge spillovers operate, the project should generate valuable insights about the allocation of talent across organizations and how the technologies and policies that influence the flow of information between agents has important implications for the level and rate of technological innovation within the economy. The purpose of this project, therefore, is to empirically estimate the importance of spillovers from "superstar scientists" for scientific progress in the biomedical area. The study relies on a unique dataset that the investigators have assembled over the past several years. The dataset has matched publication output, NIH funding and patents for the complete roster of medical school faculty and NIH grantees between the years 1977 and 2003. The central model of this study estimates the effect that exposure to "superstar" scientists exerts has on the scientific productivity of other researchers within the academic life sciences. Exposure is assumed to be a multidimensional construct, with three distinct channels of influence: (a) co-authorship ("social distance"); (b) co-location ("geographic distance"); and (c) overlap/complementarity of research foci ("scientific distance"). Because scientists do not locate or collaborate at random, particular attention is given to the quasi-experiments that can help tease apart causal relationships from mere correlations. As part of this study, open-source software tools are developed to construct measures of social and scientific distance between individual scientists. Beyond intellectual contributions to policy debates, this project will have several additional impacts. First, the software developed to measure social and scientific distance between researchers will be useful to a wide range of scholars interested in science and technology policy. The open-source format is designed to encourage future users to modify and improve the software so that research tools in this area continue to advance. This software along with source code and user manuals will be made publicly available at no charge for use by science policy scholars and other interested parties. Second, the software resulting from this study is expected dramatically to reduce the cost of matching individual-level patent, publication and research funding data in the biosciences by making a number of cross-walk files publicly available. Finally, the findings will be disseminated through various media to a wide range of audiences, thereby facilitating an open dialogue with colleagues in the academy, as well as policy makers and firms in the biomedical industry.