Geography and Agglomeration

Contributed by Juan Alcacer and Joanne Oxley

This reading list explores the relationship between economic geography and firm strategy, with a particular focus on agglomeration externalities. Agglomeration refers to the dynamic process underpinning the observed phenomenon that economic activity tends to concentrate geographically, generating localized “clusters,” particularly within certain industries (e.g., software in Silicon Valley or movie production in Hollywood). Agglomeration is of particular interest to scholars of firm strategy because choosing appropriate locations for operations is one of the most basic decisions that managers must make, and the literature on agglomeration provides insights into the interdependencies and dynamics of location decisions for firms within an industry, and the likely impact on competitive advantage.

The six core readings provide a good introduction to the theoretical underpinnings and recent developments in the literature on geography, agglomeration and firm strategy. This commentary traces the development of these ideas as they relate to firm strategy, highlights some of the main challenges that have emerged for research in this area, and discusses recent advances that address these challenges; the supplementary readings provide additional background and further development of core ideas for the interested reader. Articles referred to in the commentary that also appear in the supplementary readings list are marked with an asterisk (*).

Most people trace the beginning of the modern era of economic geography to the work of Alfred Marshall. In his path-breaking work, Marshall (1920)* focused on three sources of positive externalities or agglomeration benefits—inter-firm technological spillovers, access to specialized labor, and access to specialized intermediate inputs—which have since been at the heart of theoretical and empirical developments in the field. Demand-side economies associated with customer search costs have also been explored in subsequent work (see Fischer and Harrington, 1996*).

Core reading #1, (Hanson, 2001), provides a concise and accessible review of the main strands of the theoretical literature that builds on Marshall’s insights. The lines of argument in this theoretical literature are quite rich and increasingly sophisticated, particularly since the emergence of the “new economic geography” in the early 1990s (for a review, see, Krugman, 1998*). However, as Hanson explains, the challenge in this research domain lies less in developing new theoretical mechanisms that might explain agglomeration and more in empirical assessment of (i) whether agglomeration actually exists and (ii) the importance of different agglomeration mechanisms.

The significant empirical challenges faced by scholars of agglomeration are rooted in the fact that the basic empirical phenomenon - firms clustered in one location - may be generated by agglomeration economies or by location- specific traits that are quite unrelated to the presence of other firms in the cluster. Establishing causal links between agglomeration and organizational performance thus requires that we are able to differentiate empirically between two possible causes of location choice —physical location traits and proximity to other firms. If one were able to control for all location traits, one could of course separate these two elements. However, this is practically impossible. In the last few years the literature has followed several approaches to address this problem. The first, and arguably the most important approach is that illustrated in Core reading #2 (Glaeser and Ellison, 1997). This approach, still under development (see, e.g., Ellison, Glaeser & Kerr, 2007*) is known as the dartboard approach. The basic idea is that not all locations are equal and there is a natural underlying propensity for firms to flock to a given location given its geographical traits. This propensity is the null hypothesis against which one should compare actual data to determine whether there is agglomeration. In the analogy adopted by Glaeser and Ellison, each location is a ring within a dartboard, with more inherently attractive locations located towards the exterior of the board. Since the external rings of a dartboard are larger it is therefore more likely that a random dart thrown at the board will hit these locations.

Other approaches to overcoming the empirical challenges of assessing agglomeration effects include the use of location fixed effects (see Head, Ries, and Swenson, 1995*); analysis of cluster dynamics, which allows researchers to control for time-invariant location traits, (e.g., Davis and Weinstein, 2002*; Dumais et al., 2002*); and more qualitative research aimed at uncovering specific agglomeration mechanisms (e.g., Saxenian, 1991*).

Reflecting its roots in economics, most of the research on agglomeration through the late 1990s assumed that all firms benefit equally from clustering. Core reading #3, (Flyer & Shaver, 2000) introduces firm heterogeneity and thus brings agglomeration more firmly onto the strategy terrain. The argument offered in this canonical reference in agglomeration and firm strategy is quite simple: given heterogeneity in firm-specific capabilities and market positions, the benefits and costs associated with locating in a cluster are also likely firm-specific; as a result, while many firms will be attracted to locating in a cluster, some firms will prefer to choose an isolated position and locate away from a cluster. For example, more capable firms may contribute disproportionately to other firms in the cluster through knowledge spillovers while having little to gain from the spillovers generated by inferior firms. Consistent with this argument, Flyer and Shaver (2000) find evidence of adverse selection in clusters (as smaller and weaker firms disproportionately agglomerate) in a sample of foreign greenfield investments in U.S. manufacturing industries. Critics have nonetheless noted that the “flee and follow” logic proposed here suggests complex cluster dynamics that are underspecified in the natural language arguments advanced in this paper, so that the implications for agglomeration patterns remain unclear. In a follow-on paper Flyer and Shaver (2003)* address this concern by developing a formal model which provides important additional results and qualifications. Subsequent empirical work has further refined our understanding of this basic phenomenon.

Core reading #4, (Alcacer, 2006) takes the idea of heterogeneity in agglomeration costs and benefits to a deeper level, exploring the strategic implications of agglomeration across different value chain activities in the face of heterogeneous capabilities. At the functional level, Alcacer argues that agglomeration is most likely to be beneficial for R&D activities, while locating sales operations within clusters may generate few benefits and increase inter-firm competition, so reducing profitability. Evidence from an empirical study that applies and builds on the Ellison and Glaeser methodology (Core reading #2) finds that, consistent with Flyer & Shaver (2000), more-capable firms collocate less than less-capable firms, and that this is most pronounced among sales subsidiaries, followed by production and then R&D.

Core reading #5, (Sorenson & Stuart, 2001) represents a departure from the economic foundations of the prior readings and focuses instead on the social and institutional structures that shape spatial patterns of exchange. The paper provides a brief review of the sociological view of the importance of proximity in determining the likelihood of social or economic exchange, and demonstrates how social networks in the venture capital community diffuse information and expand the spatial radius of exchange. This paper also provides an entry point to the emerging research exploring effects of agglomeration on entrepreneurship (see, e.g., Michelacci and Silva, 2007*).

The idea that knowledge flows (and knowledge spillovers) may be mediated by social networks points to a potential limit or boundary condition for agglomeration benefits: To the extent that social networks span firm boundaries, (or become decoupled from firm identities altogether), then this may undermine or substitute for agglomeration benefits based on firm location, with intriguing implications for firm strategy. Core reading #6 (Agrawal, Cockburn & McHale, 2006) explores this possibility. By estimating the knowledge “flow premium” associated with an inventor’s prior location Agrawal et al provide evidence that indeed social relationships, not just physical proximity, are important for the types of knowledge flows that constitute one of the fundamental externalities posited to lead to agglomeration. Sitting at the intersection of economics, social networks and innovation research, this paper also provides a useful perspective on the relationship among these related but distinct lines of research and an entry point into the literature on location and innovation (e.g., Audretsch and Feldman, 1996*; Zucker, Darby, and Brewer, 1998*).

The study of geography and agglomeration as contributors to industry and organizational performance represents a well-developed and vibrant stream of research within strategy. The theoretical underpinnings of models of agglomeration and location choice are strong and continue to develop as additional mechanisms and sources of heterogeneity are explored. Empirical work continues to grapple with the thorny challenge of designing and implementing empirical studies that can isolate causal mechanisms; the articles in this reading list represent some of the most successful efforts to date, and should provide inspiration and guidance for future work.