Proposal to the Bechtel Initiative: A Network Study of Silicon Valley
Submitted by Principal Investigator: Prof. Mark Granovetter, Department of Sociology
April 5, 1999
"The most crucial aspect of Silicon Valley is its networks." There is no proposition so universally agreed upon and so little studied. Why is this? I see two main reasons.
The first is that social network analysis has been mainly the province of sociologists (less likely methodologically reductionist than psychologists, economists or political scientists), but only since the 1980s have sociologists become interested in industrial organization.
The second is that methods for systematic study of social networks are of very recent origin (for a detailed historical account as well as a comprehensive inventory of current knowledge, see S. Wasserman and K. Faust, Social Network Analysis: Methods and Applications, Cambridge University Press, 1994). The first glimmer of understanding came in the 1940s, with the idea of representing a social network as a matrix of 1s and 0s, indicating a tie or its absence between elements i and j. Standard matrix manipulations could then yield detailed accounts of potential network dynamics (e.g., squaring the matrix produces a new matrix of two-step connections between elements i and j). But it was only from the 1970s on that more complex analyses began to be developed.
Sociologists’ earlier theoretical concerns led them to focus network analysis on relatively small, informal groups (like children in schools) and this was reinforced by the unavailability of methods and software suited for larger-scale studies. The central limiting problem in network analysis has been that the number of possible connections among n individuals is n(n-1)/2, i.e., on the order of n2, which presents daunting computational problems.
These limitations are beginning to recede. The first major study treating an entire industry from a social network point of view is currently being carried out by myself and various collaborators. It is a National Science Foundation funded study of the electricity industry in the United States from its origins in the early 1880s to the middle 1920s, when, after many conflicts and transformations, it reached a form that was more or less stable for the next fifty years. We are coding vast amounts of archival data on central station firms, their directors and officers, equipment providers, banks, trade association committees, meetings and proceedings, and technical journals, using the power of a sophisticated relational database program (Microsoft Access) and a program (UCINET) allowing such data to be input into social network analyses, such as the computation of relative centrality of various network nodes. Even though none of the many thousands of individuals whose biographical and career data we have coded is still living, we reconstruct social networks through imputation -- e.g., assuming a tie between individuals who sat on the Executive Committee of a trade association during the same year. By following such networks over time, and tracking the evolution of such quantities as centrality for key individuals, we have already achieved new insight as to how the technology and organizational forms changed as they did.
A network study of Silicon Valley could be constructed along lines similar to those of the electricity project, with the advantage that once publicly available data had been mined, the principal participants are still living, in many cases in the Bay Area, and could fill in key missing information, more easily identified once a network structure has been mapped. We would give high priority to the strategic scheduling of such interviews. Because we cannot expect unlimited and frequent access, we would want to accomplish enough background research first, in most cases, to approach the interviews with a focused agenda, rather than with a journalist’s request to “tell us what happened”; but such focus will have to be traded off against the possibility that, for whatever reasons, any given figure might become unavailable or less available with the passing of time.
The project should begin with some systematic attempt at mapping the network and its evolution over time. A series of snapshots could be constructed, with comprehensive inventories of companies at, say, five year intervals, as we have done for electricity. A key step would be to track the affiliations of company principals, as they shift among companies, as their companies are merged or acquired, and as spinoffs and completely new ventures appear. Although the business press has often discussed the "genealogy" of Silicon Valley companies, and there are some valuable narrative accounts, as in Saxenian (1994), the companies studied have been almost entirely those which became successful and important. This selection bias obscures understanding of the overall process, which can only be built on a more comprehensive account, in which failures and obscure cases also receive attention.
We should be able to isolate and identify structures of relations among firms as given by genealogies of firm origin, and by tracking how firms are linked by having had interchange of personnel from one to the other. Having separate tables for people and companies in a relational database also allows mapping the complex linkages among them, widely assumed to be central to the apparently chaotic flux that gives the Valley its dynamic edge. The previous network location of entrepreneurs should also be studied in relation to their success or failure, moving us away from the misleading portrayal of outcomes as resting entirely on technical or entrepreneurial brilliance, tempting as such an account is to the winners.
As with any industry, it is also necessary to track how networks of individuals literally outside the industry but playing a vital role, articulate with and sometimes become "insiders". The most obvious such groups are venture capitalists, lawyers, headhunters, engineers and their associations and trade groups. The power of relational database programs is that each group can be entered into its own set of tables, which makes the task manageable, but relations among groups can then be studied easily by systematic linkages and queries among the tables within the overall database. More generally, it is important to construct as background some account of how financial, commercial, educational and political institutions are linked not only to information technology firms but also to one another in this region, since industries do not arise and exist in a vacuum, but in a distinct institutional context. Variations in these contexts may well explain why the myriad attempts to replicate Silicon Valley in utterly different contexts, by copying only the features of its firms, are rarely fruitful. This broader account is essential for understanding of how regional economies are articulated. The comparative advantage of sociologists for such a study is that their theoretical arguments point them toward investigating how different institutional arenas mesh with one another, rather than focusing exclusively on the technical, economic, legal, educational or political aspects.
Current observers of the Silicon Valley scene would also concur that international connections have increased dramatically in importance. Connections to India, Israel, Southeast Asia and Western Europe have escalated, and have assumed important roles in both financing and technical innovation. While it is clear from casual observation that these global connections operate through well-defined and ethnically bounded networks, we have no systematic accounts of this process. Another aim of this project would be to begin mapping these networks and processes in a serious way.