Abstract

We investigate the structure of investment bank syndicate networks in Canada. We consider two banks to be connected if they have participated in an underwriting syndicate together, and construct networks of such connections using data drawn from the Record of New Issues (Financial Data Group). We show that these interfirm networks form "small worlds", in which banks are both locally clustered and globally connected by short paths of intermediate banks, and are "scale free", in which the connectivity of the network is highly skewed and with most banks tied to a small set of prominent banks. We examine changes over time in the network's small-world and scale-free properties, and demonstrate their theoretical and practical implications for the structure and operation of Canadian capital markets by linking these properties to the network's cliquey-ness, resilience, and speed of information transmission.

Résumé

Cette étude porte sur la structure des réseaux que forment les syndicats d'émission des banques d'investissement au Canada. Nous posons que deux banques sont liées si elles ont participé ensemble à un syndicat d'émission, et nous retraçons les réseaux de liens en utilisant des données extraites du Record of New Issues (Financial Data Group). Nous montrons que ces réseaux interorganisationnels (RIO) forment des « petits mondes » dans lesquels les banques sont à la fois localement regroupées et mondialement reliées par des courts chemins de banques intermédiaires. Les RIO sont également sans échelle (scale free) : la connectivité dans le réseau est fortement inégale et la plupart des banques sont liées à un petit nombre de banques dominantes. Nous examinons l'évolution des propriétés de petit monde et d'absence d'échelle du réseau et mettons en évidence leurs implications théoriques et pratiques pour la structure et le fonctionnement du marché canadien des capitaux en reliant ces propriétés aux caractères de clique, de resilience et de vitesse de transmission de l'information du réseau.

Networks are ubiquitous; they surround us; we engage dozens of them daily. The global economy is a network of national economies, which are networks of markets, which in turn are networks of firms, which in turn are networks of people. Increasingly, the technologies and social institutions on which we depend are explicitly engineered as networks. Our understanding of networks, however, has not kept up with our dependence on them.

What is a network? A network is essentially anything that can be represented by a graph: a set of points (also genetically called nodes or vertices), connected by links (edges, ties) representing some qualitative relationship. In social networks, the nodes are people or groups of people, "actors" in the jargon of the sociology, with some pattern of interactions or "ties" between them. A social network, then, is a set of people or groups of people (e.g., organizations) linked by acquaintance, friendship, political alliance, professional collaboration, or business relationships.

Network analysis currently forms the core of the "new economic sociology," which rests on the argument that networks generate and structure markets, creating pathways to sources of information and resources (Smelser & Swedberg, 1994). Social networks have, however, been the subject of both empirical and theoretical study in the social sciences for over 50 years, partly because of inherent interest in patterns of human interaction, but also because their structure has important implications for the spread of information, ideas, and disease, as well as social influence and inequality (Smelser & Swedberg). It is clear, for example, that variation simply in the average number of acquaintances that individuals have can substantially influence the propagation of a rumour, fad, fashion, joke, or this year's strain of the flu.

Traditionally, sociologists have viewed networks as static objects, "as given contexts for action" (Madhavan, Koka, & Prescott, 1998, p. 439). More recently, following a surge in interest in network structure among mathematicians and physicists, partly as a result of studies of the Internet and the World Wide Web and partly the broader movement toward research on complex systems, a growing stream of research has investigated the statistical properties of networks and methods for modeling networks. In this work, networks are conceived as dynamic systems that self-assemble and evolve in time through the addition and removal of actors and ties. The techniques of statistical mechanics, it turns out, are well suited to the study of networks. Indeed, graph theoretic analyses have permitted comparison of seemingly unrelated networks, leading to the exposure of deep similarities among social, biological, and technological networks.

Two important families of network structures have emerged from these studies (for a review, see Albert & Barabási, 2002). The first is small-world network structures characterized by the combination of a high degree of clustering, meaning that there is a heightened probability of two actors being acquainted if they have one or more other acquaintances in common, and short characteristic path length, meaning that there exist short paths through a network between most pairs of actors (Watts & Strogatz, 1998). The second is scale-free network structures in which the degree distribution of the network-the distribution of ties among actors-is free of a characteristic scale and highly skewed, with a small number of actors having a disproportionately large number of ties (Barabási & Albert, 1999). Research suggests the widespread presence of the "small-world" pattern (e.g., Adamic, 1999; Davis, Yoo, & Baker, 2003; Kogut & Walker, 2001; Newman, 2001; Powell, White, Koput, & Owen-Smith, 2004; Uzzi, Spiro, & Delis, 2002; Watts, 1999; Watts & Strogatz, 1998) and scale-free degree distributions (e.g., Barabási & Albert, 1999; Jeong, Mason, Barabási, & Oltvai, 2001; Jeong, Tombor, Albert, Oltvai, & Barabási, 2000; Wagner & Fell, 2000) in social, economic, technological, and biological networks.

Why this interest in network structure and evolution? Because structure affects function, and the consequences of small-world and scale-free patterns are far from trivial; they are of great importance to the behaviour and performance of networks. Small-world network structures are highly effective for communication of information across actors, while scale-free network structures are particularly robust to random disruption of network ties and exogenous shocks, and highly searchable based on a simple strategy of seeking heavily connected actors. Although a growing body of research demonstrates the ways in which structural properties of interfirm networks affect firms' economic performance and innovation, this work has focused primarily on a firm's position within its local network rather than the overall structure of the network.

In this paper we analyze the statistical properties of an interfirm network-the syndicate network of investment banks in Canada from 1952 to 1989. Our aim is twofold. First, we hope to gain a better understanding of the way in which the investment banking industry operates in Canada, including how the network evolves over time, how information travels through it, and its underlying structure and robustness. second, and more broadly, we hope to provide further evidence of the generality of the statistical mechanics that appear to govern a diverse array of social, biological, and technological systems. For this network, both the small-world property and the scale-free distribution of ties are found to be present. As we show, the presence of these network properties has important theoretical and practical implications for the structure, operation, and resilience of the interfirm syndicate network and Canadian capital markets.

Statistical Mechanics of Networks