Skip to main content
Open Access Publications from the University of California

UC Berkeley

UC Berkeley Electronic Theses and Dissertations bannerUC Berkeley

Knowledge Spillovers through Networks of Scientists

  • Author(s): Zacchia, Paolo
  • Advisor(s): Moretti, Enrico
  • et al.

In this thematic dissertation, which is divided in three chapters, I discuss new evidence

and insights about knowledge spillovers, that is the process by which new productive ideas benefit individuals and organizations that did not originally devise them. The general theme of the dissertation is the empirical search for one specific micro-level mechanism of information diffusion: that is, social interactions between inventors who are connected in a network of professional collaborations. In the first and second chapter I examine how inventor-level networks function as a channel of R&D spillovers across firms. The two chapters separately introduce two different and novel methodologies to estimate spillover effects in this setting. Both methods stem, however, from a shared theoretical framework, and the resulting econometric estimates are similar across the two approaches. In the third chapter I analyze a different but related issue: that is how geographical proximity facilitates knowledge exchange between inventors, in particular between “superstar” scientists and their coauthors. In the remainder of this abstract I provide a brief summary of each of the three chapters.

The first chapter illustrates a new methodology of quantitatively assessing R&D spillovers across firms in a reduced form fashion. As mentioned, I directly test the hypothesis that interactions between inventors and scientists of different organizations drive knowledge spillovers. To this end, I construct a network of publicly traded companies where each link is a function of the relative proportion of any two firms’ inventors who have patent coauthors in both said organizations. I use this measure of connection to weigh the impact of R&D performed by each firm on the innovation and productivity outcomes of its neighbors in the network. An empirical concern is that the resulting estimates may reflect, rather than genuine spillover effects, some unobserved, simultaneous drivers of both R&D and firm performance, which are common to closely connected firms. By formally modeling the strategic dependence of R&D choices in the network I characterize the conditions under which specific instrumental variables, based on the R&D choices of more distantly related firms (“indirect friends”), solve the problem. Empirical results show that substantial effects of external R&D on firm performance and innovation are identified by this approach. I calculate the resulting marginal social return of R&D to be approximately 24% of the corresponding marginal private return.

In the second chapter I adopt a more structural approach to estimate the parameters of

the model from the first chapter. Instead of quantifying spillovers in terms of their ultimate

effects on firm-level productivity and innovation outcomes, I estimate the effect of connected firms’ R&D in stimulating private R&D investment strategies. On the basis the equilibrium conditions of the R&D investment game, formulated under general assumptions, I establish a set of moment conditions of both the first and second order, relative to equilibrium R&D, that are exploited for estimation. The model separately identifies spillover effects and the variance of common shocks thanks to a zero conditional covariance restriction. Its intuition is summarized as follows: for any two firms that are only indirectly related in the network, their reciprocal strategic dependence is removed by holding constant the choices of their intermediate links. Thus, any residual cross-correlation in R&D must be driven by shared external factors, whose size can be estimated. The empirical results evidence that the R&D of connected firms positively stimulates private investment with an elasticity of about 0.2.

In the third and final chapter I analyze spillover effects on the production of patents following episodes in which superstar inventors relocate to a new city. In particular, in order to determine whether local externalities have a restricted network dimension or a wider spatial breadth, I estimate changes in patterns of patenting activity for two different groups of inventors: the restricted group of coauthors of the superstar, and all other inventors from one specific urban area. The analysis is performed for both the locality where the superstar moves and for the one that is left. I restrict the attention to outcome measures of patent output that exclude any joint work with the superstar, so to isolate spillovers from complementarity effects. The results from the event study evidence a large and persistent positive effect on the coauthors of the superstar who reside in the city of destination (averaging about 0.1 more patents per inventor each year), and a negative trend affecting those who live in the locality of departure. These effects increase with the relative ranking of the superstar in the patent distribution. Conversely, no city-wide spillover effect can be attested, offering little support to place-based policies aimed at generating a positive influx of highly skilled individuals.

Main Content
Current View