«GONE BUT NOT FORGOTTEN: LABOR FLOWS, KNOWLEDGE SPILLOVERS, AND ENDURING SOCIAL CAPITAL Ajay K. Agrawal Iain M. Cockburn John McHale Working Paper ...»
NBER WORKING PAPER SERIES
GONE BUT NOT FORGOTTEN:
LABOR FLOWS, KNOWLEDGE SPILLOVERS,
AND ENDURING SOCIAL CAPITAL
Ajay K. Agrawal
Iain M. Cockburn
Working Paper 9950
NATIONAL BUREAU OF ECONOMIC RESEARCH1050 Massachusetts Avenue Cambridge, MA 02138 September 2003 The views expressed herein are those of the authors and not necessarily those of the National Bureau of Economic Research.
©2003 by Ajay K. Agrawal, Iain M. Cockburn, and John McHale. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including © notice, is given to the source.
Gone But Not Forgotten: Labor Flows, Knowledge Spillovers, and Enduring Social Capital Ajay K. Agrawal, Iain M. Cockburn, and John McHale NBER Working Paper No. 9950 September 2003 JEL No. F22, O31, R12, R23
Although recognition of the importance of localized knowledge spillovers goes back at least to Alfred Marshall (Marshall, 1920; Krugman 1991), the difficulties of measuring such tacit knowledge flows impeded their study. As pointed out by Krugman3, “they leave no paper trail by which they can be measured and tracked.” The work of Jaffe et al. (1993, hereafter referred to as JTH) pointed, however, to one important exception. They argued that “[K]nowledge flows do sometimes leave a paper trail in the form of patent citations,” which can be followed to “test the extent of spillover localization”.4 Taking patent citations as a proxy for knowledge spillovers, they found strong evidence of geographic localization even after controlling for the tendency of inventive activities to be geographically clustered by technological field.5 If co-location facilitates greater access to knowledge spillovers due to stronger social ties, what happens when an inventor moves? In this paper, we explore the possibility that citations to a patent, and thus knowledge spillovers, also occur disproportionately at locations where inventors were living prior to their current inventive activity. Our hypothesis is that individuals invest in the development of social ties with others with whom they are co-located, and that at least a portion of those ties endure even after the Modern endogenous growth theory casts knowledge spillovers from investments in human capital and research and development as a central character in generating the increasing returns that sustain long-term growth (e.g., Romer, 1986 and 1990).
An important piece of circumstantial evidence for the importance of localized knowledge spillovers is that industries for which new knowledge plays an important role tend to be more spatially concentrated (Audretsch and Feldman, 1996). Another intriguing piece of evidence is that clusters of biotechnology firms developed around academic scientists who published genetic sequencing discoveries in academic journals (Zucker et al., 1998).
P. 53 P. 578 One concern with this result is whether it reflects communication between the inventors and thus true knowledge spillovers. The survey evidence reported in Jaffe et al. (2002, Chapter 12) partially allays this concern, since they find that citations are a signal of communication, albeit a noisy one.
individual has moved. Thus, their past neighbors also have some degree of favored access to the new knowledge generated at their new locations.
To better understand the mechanisms at work, we first develop a simple model of purposeful investments in social relationships that facilitate the exchange of non-rival, non-contractible knowledge. In the model, the opportunities for such investments are limited to co-located inventors. We are particularly interested in how the prospect of mobility, and thus the possible future geographic separation of individuals, alters the incentive to invest in relationships with currently co-located colleagues, and thus has implications for the presence, in equilibrium, of enduring social capital.
Two effects are at work in this model. First, high prospective mobility diminishes the incentive to invest in local relationships when the value of communication is adversely affected by separation. Second, however, high prospective mobility also biases investments towards relationships whose value to the individuals involved is relatively insensitive to their degree of geographic separation. Which of these effects dominates is an empirical question, and in this paper we present evidence consistent with the presence of investments in enduring social capital that facilitates communication between previously colocated individuals even after they become geographically separated.
The essence of our empirical methodology for testing the enduring social capital hypothesis is to seek evidence of disproportionate cites to inventions at locations where the individual lived prior to their invention. Following the pioneering methodology developed in JTH, we compare the extent to which actual citations are disproportionately located in a particular location relative to a distribution of control citations that have the same temporal and technological characteristics. Importantly, this comparison allows us to control for any technology-based clustering of inventive activity, which may otherwise confound any inference drawn from co-location of citations.
Given our interest in inventor mobility, we use citations to inventor-patent pairs as our unit of analysis.
For example, a 1990 patent that has three inventors will generate three inventor-patent pairs. Each inventor-patent pair will be associated with a unique 1990 location (Metropolitan Statistical Area, or MSA) that is based on the inventor’s address as recorded in the 1990 patent. If that inventor had previously patented somewhere else (i.e., is a “mover”), we also record their most recent prior location.
We then examine the proportion of the subsequent citations (1990-2002) that occurred at the inventor’s 1990 location, and, for movers, the proportion of subsequent citations that occurred at the inventor’s prior location.
Our results support the geographically localized knowledge spillover finding of JTH. Like JTH, we find strong evidence of a disproportionate number of cites that are co-located with the inventor. More interestingly, in the context of the present paper, we also find evidence of a disproportionate number of cites in locations where the inventor had previously lived (and patented)—that is, we find evidence of a prior location premium. As discussed above, one plausible interpretation of this finding is that individuals invested in social ties with others at their prior location during their residency there, and at least part of that social capital endured to support above average knowledge flows back to their prior location. In effect, prior co-location allows for investments in social relationships that condition the subsequent distribution of knowledge spillovers from mobile inventors.
Finally, we find evidence that co-location is particularly important for cross-field knowledge spillovers.
We hypothesize that knowledge spillovers within communities of practice (Lave and Wenger, 1993;
Brown and Duguid, 1991) or invisible colleges (Crane, 1965 and 1969) are less likely to be geographically mediated. Groups of researchers interested in similar problem areas are likely to communicate with each other via mechanisms such as conferences, publications, and trade shows such that being co-located is less important for facilitating knowledge spillovers. We find evidence that both current and prior co-location increase the likelihood of cross-field spillovers even more, proportionately, than within-field spillovers.
We think these results are interesting in the context of three literatures. First, they provide additional insight into the processes through which economic knowledge diffuses. The results are consistent with the conjecture that social ties facilitate knowledge spillovers, and they show how the geographic distribution of relevant social capital is determined in sometimes subtle ways.
Second, the results are relevant in the context of measuring what an economic location loses when a portion of its skilled workforce leaves. The migration literature has suggested the importance of knowledge spillovers to the gains and losses of locations from mobile labor (e.g., Borjas, 1995). But these effects generally have been viewed as un-measurable. The JTH findings show that knowledge spillovers are geographically localized, which suggests an important source of location-specific loss when inventors leave. Our results suggest, however, that the losing location can nonetheless retain some degree of favored access to the knowledge generated by the departed inventor from their new location.6 Finally, our results may be of interest to those studying the links between labor mobility and social capital accumulation (e.g., Glaeser et al., 2002). Using patent citation data as a proxy for knowledge flows and modeling knowledge flows as being facilitated by social relationships, our work shows how the rich data that is available on the geographic locations of patenting and citing inventors can be used to empirically examine how prospective mobility affects social capital accumulation.
The rest of the paper is structured as follows. In the next section, we develop a simple model of purposeful investments in social relationships between (prospectively) mobile inventors. The model yields two hypotheses that we test in the remainder of the paper. Section III outlines our general methodology for using patent and citation location data to test these hypotheses. Section IV describes our data and Section V our results. Section VI concludes with a summary of our main findings and some suggestions for future work.
II. Social Relationships, Knowledge Spillovers, and Inventor Mobility:
A Simple Model In this section, we develop a simple model of tacit knowledge communications that are facilitated by social relationships between presently and formerly co-located inventors. Inventor co-location plays two key roles in the model. First, it creates opportunities to develop social relationships. Thus co-location may be thought of as a sort of social treatment, whereby the inventor in exposed to potential social acquaintances.7 And second, conditional on a social relationship existing, the value of the knowledge flowing between the related individuals depends on whether they are currently co-located or separated.
For the purposes of the model, we assume that the knowledge in question is both non-rivalrous and noncontractible. The non-rivalry assumption implies that the knowledge has the public-good characteristic of not losing value when it is communicated to other inventors. Of course, such knowledge is easily excludable; the knowledgeable inventor can simply refuse to communicate the knowledge to others. We Interest in estimating the losses from the out-migration of skilled workers (a.k.a. the “brain drain”) has been growing as the competition for talent between and within national (or regional) economies has increased (Desai et al., 2002).
Although we focus on relationships developed as a result of co-location, we allow for the possibility that social relationships can be developed without co-location.