|
Article Excerpt
The venture capital (VC) market in the United States is a young industry, dating back to a few deals shortly after the close of World War II. Most evidence points to an early history, whereby investors poorly understood the risks attached to this new market (Gompers and Lerner 2001). During the 1960s and 1970s--for which better data exist--the deals still were relatively sparse. By the 1980s, an institutional infrastructure had been put into place that permitted a rapid growth in transactions, in volume and in value. By the start of 2000, over 103,000 transactions had been recorded, and a history of dozens of firms that grew from start-up to the industrial 500 had been written.
This infrastructure is what is called a market. It consists of a chain of investors and brokers who often do not know the start and the end of the chain, but who surely know their adjacent links. From this perspective, capital investment passes through social networks, in which the local character of investor and investee is preserved in the context of a global myriad of intersecting pathways. The key artifact that binds the links in this chain is the transaction, which in formal markets is recorded in contracts. The records of these contractual ties among investors who enter, transact, and exit over time permit a sophisticated analysis of the VC market as an emergent network.
Looking at the emergent network of contractual ties turns the question of financial market integration on its head. The question of integration assumes that local financial markets predate national ones; integration is the joining of existing local markets into larger national ones. We propose a contrary hypothesis, borrowed from the studies of Braudel (1981) on the emergence of markets in medieval Europe, that localities almost from the start are not only connected by bridges through commercial and social ties, but that their own evolution is deeply tied to the dynamics of a wider integrating market. In fact, it is not possible to understand local financial market developments in many important cases without understanding these interregional ties and connections.
The evolution of VC in the United States is an odd context in which to make this claim because past studies have found strong evidence for the local nature of VC lending. In their early article, Florida and Kenney (1988) noted the concentration of VC in a few regional centers, Castilla et al. (2000) trace the dense network linkages among VC firms in Silicon Valley, indicating strong local social ties. Sorenson and Stuart (2001) analyzed the proximity of the VC firm to the target firm and found that investments tended to be within a 500-mile radius. Critical to our analysis below, they showed in their cross-sectional analysis that highly central VC funds bridged these local agglomerations.
By definition, these cross-sectional results do not capture the dynamics by which regions and national networks coevolved. As we show, national brokerage and regional development grew jointly and inter-dependently. The adaptation of the limited partnership agreement to VC investments in the late 1950s was an institutional innovation that suddenly permitted a national market to emerge in two decades. This innovation provided the missing complement for the effective governance of investments in unknown technologies owned by new firms. The explosion in entrepreneurial investments was geographically diverse and indicates a rapid integration of a national financial market in high-risk and new companies.
Our analysis proceeds as follows. After defining terminology, data, and theoretical concepts, we show first that VC deals quickly generated an integrated national market across regions. This central finding begs the question of what microbehaviors could plausibly generate a national market so early in its history. A benchmark hypothesis is that new and incumbent VC firms are drawn to highly connected firms. If this is true, we would expect to find power-law distributions for the number of deals by the VC firm. However, this distribution gives an unsatisfactory fit. The social structure of the network, while poorly explained by the degree distribution, is revealed through an analysis of repeated ties that are impressively distributed by a power law. In these repeated ties, we see a marked preference for VC firms to renew their ties with each other by investing jointly in a sequence of investment opportunities. By partitioning a measure of local density (i.e., clustering) by geography and sector, we show that repeated ties are national, and not local. Hence, the national VC market consists of strong ties in two senses: They are repeated, and they are "one step away." The conservative preference for repeated coinvestments to remain geographically and sectorally local is countered by the opportunities to diversify into new sectors and regions that necessitate new partners. A simple logit model confirms the negative effect of diversification on repeated ties. We thus see a trade-off between trusted expertise and diversity.
The complexity of the sciences challenges us to characterize observations of the global structure of aggregate behavior (e.g., giant components, degree distributions) that are consistent with observations of the motives and strategies of microactors. From simple rules and strategies come surprisingly well-ordered structures. The statistical analysis of these structures is considerably indebted to Newman et al. (2001) for the analytical results for a bipartite graph; to Newman (2005), Mitzenmacher (2004), Gabaix (1999), and Farmer and Geanakoplos (2004) for their explanation of power-law dynamics; to Guimera et al. (2005) for their treatment of the entrant and incumbent link probabilities in relation to percolation; and, finally, to Barrat et al. (2004a, b) for their analysis of weighted graphs. These works permit a rich analysis into the dynamics of VC market formation.
1. Financial Market Development and the Venture Capital Industry
1.1. Historical Background: The Braudel Hypothesis
That social foundations are required to support markets for capital investment is not a new observation and is hardly unique to VC. The history of the development of capital markets after the Great Restoration in England (Carruthers 1999), of credit in post-revolutionary France (Hoffman et al. 2000), of banking in 1800 New England (Lamoreaux 1994), and of stock market trading (Michie 1999, Baker 1984) are recapitulations of patterns that Braudel (1981) saw in the local nature of foreign commerce in late medieval Europe. These histories pose the following question: At what point did these local markets intersect, such that pricing and trade could be determined to be national or international--that is, could no longer be said to be local? The Braudel hypothesis is that local markets developed in response to an evolving global system.
The conventional treatment of financial integration has been to study the narrowing of price variations across local markets. Lance Davis (1965) launched a series of inquiries by studying the process of integration of capital markets in the United States, using implied interest data on both commercial and mortgage bank loans. He found slow convergence over the period of study (1865-1914), with persisting regional differences for long-term rates in regard to the southern states. Part of the reason for this convergence, he noted, was the geographic expansion of life insurance and brokerage companies. Buchinsky and Polak (1993) find that English credit and real estate markets reveal evidence of integration by the Napoleonic Wars, perhaps in consequence of the large government demand for loans; they do not, however, cite by what means such integration occurred. Absent public information on prices and transactions, the integration of prices that is suggested by these studies implies informal mechanisms of communication. However, these mechanisms are not specified.
Evidence on direct contacts and parties to transactions provides greater insight into the social foundations of markets. The benchmark expectation is that social ties are regional, hence, so are capital markets. Lamoreaux's history of bank lending in England reveals the lending of capital even among bank directors; Padgett's and Ansell's (1993) detailed history of the Medici shows that commercial ties lie along local family and patronal lines. Even in financial markets, we find a strong bias for proximity to matter to portfolio holdings, as found by Braggion (2005) for late 19th century England, and by Franks et al. (2005) for their study of 20th century British finance. Goetzmann et al. (2004) found that the proximity bias in Swedish portfolios in the 1990s was related to urban specialization. This role of expertise is also found in our VC data, as discussed below.
However, there are also histories that indicate social ties often bridged spatially noncontiguous transactions, such as those among Maghribi traders (Greif 1993) or among notaries in France (Hoffman et al. 2004). Neal's (1990) superb study of the integration of the London and Amsterdam markets in the 1700s and 1800s notes in detail the movement of people from the continent to England, yet his analysis relies on price data to show convergence. In his study of the London stock market, Michie (1999, p. 125) writes that "by the early twentieth century, a complex web of contacts, clients, agents, and organizers linked the London broker to investors abroad." It is this web, which is captured by the data on VC transactions, that is amenable to analysis as a graph.
1.2. Venture Capital Syndication as a Graph: Definitions
The origin of the VC market is conventionally attributed to the founding of the closed-end fund ARD by George Doriot, a French entrepreneur on the faculty of the Harvard Business School, although there were important precedents (Hsu and Kenney 2005). This fund was imitated by several others in the 1950s--primarily in the northeast United States and in California--but disappointed small investors who were unaccustomed to substantial volatility. The institutional solution was to replace the closed-end fund by private funds that accepted investments only from financially qualified investors. Critical to the legal formation of syndications, the first VC limited partnership founding dates from 1958. The Draper, Gaither, and Anderson fund, founded in Palo Alto, included Rockefeller seed money, and thus was financed by East Coast money from the start (Avnimelech et al. 2004). In this period, the U.S. government promoted VC through small business investment corporations (SBICs), although this organizational form experienced substantial difficulty (Gompers and Lerner 2001). The decision in 1978 to permit pension funds to invest a restricted percentage of their assets in VC provided a substantial inflow of money and also legitimated the market for other investors.
The dominant business model of a VC company (or firm) is the attraction of capital from private or institutional investors for investment in VC funds under the company's management that then invest in a portfolio of target companies. For a fund life of about five years to a maximum of roughly 10 years, the managing VC firm is likely to make deals in the same sector or geography because of the nature of the fund. In many cases, VC firms will join with each other in investing in a single start-up. The combination of investments by multiple VC firms in one start-up is called a syndicated deal, or syndication. Many VC firms involved in syndications invest with numerous VC partners across a range...
|
