Stuart Rosenthal
Department of Economics
426 Eggers Hall, Center for Policy Research
Syracuse University
Syracuse, NY 13244-1020
Phone: (315) 443-3809
Email:

POSITION STATEMENT

In recent years I have worked on several different topics in which spatial dimensions affect the degree and nature of externalities. The most recent portion of this work is joint with Will Strange (University of British Columbia) and focuses on agglomerative externalities arising from spatial concentrations of firms. I will focus my comments on that research.

The central purpose of the research program with Will Strange has been to shed light on three critical questions: why do cities exist? What makes them grow? What role will they play in the future? Since concentrating people and capital is costly, there must be some benefit that encourages profit maximizing firms to incur the costs of city building. These benefits are referred to as agglomeration economies. Broadly defined, agglomeration economies arise when spatial concentration of industry and employment provides firms with improved access to valuable human and non-human capital. The research program with Strange adds to the growing literature on urban development by incorporating detailed geographic information into a series of analyses designed to evaluate the determinants and geographic scope of agglomeration. As a result, our research focuses on several questions that have explicitly spatial dimensions including the following.

1. What is the geographic scope of agglomerative externalities? Although the theoretical literature argues that interaction between establishments declines with distance, most previous empirical studies have implicitly assumed that all activity in a rather large area such as a state, city, or county affects everyone in the area symmetrically.
   
   
2. What are the principal determinants of agglomeration? Although this issue has been examined at the state and county levels by Audretsch and Feldman (1996) and Dumais, Ellison, and Glaeser (1997), it has not been considered at a refined level of geography.
   
   
3. What are the implications of the information and telecommunications (IT/T) revolution for the future growth of cities, and more generally, for the mix of industries doing business in urban and non-urban areas? The IT/T revolution has greatly increased our ability to access valuable resources from remote locations. Access to human and physical capital, however, is the main reason firms choose to cluster in cities despite the high overhead costs of such locations. Thus, it certainly makes sense to consider the future of cities in this light. Gaspar and Glaeser (1998) provide a nice way of thinking about this issue, and also provide some suggestive evidence. However, they do not consider the relationship between IT/T and location directly.

To address these and other questions we have been working with data from the Dun and Bradstreet Marketplace file in conjunction with various MapInfo mapping software products. The D&B data provide establishment-level information on over twelve million establishments in the United States. These data include information on establishment location down to the 5-digit zip code level, SIC codes to the 8-digit level, number of employees at the site and at the parent firm, sales, ownership status (e.g. corporate, subsidiary, public, etc.), age, various measures of IT/T use, and much more. With mapping software, we have geocoded the data and computed the spatial distribution of different types of employment and establishment attributes across the United States. For example, in a couple of our projects, we calculate a set of concentric ring variables that measure the levels own-industry and other-industry employment at a given distance from an establishment's own zip code, such as less than 1 mile, 1 to 2 miles, and on out to k miles.

Our first effort examined the geographic scope of agglomerative externalities. This was accomplished by modeling the number births of new establishments and related employment in a given zip code as a function of concentric rings of own- and other-industry employment associated with that location at the time the location decision was made. Results suggest that entrepreneurs in the six manufacturing industries studied care much more about the level of own-industry employment within a few miles than just 5 or 10 miles away. From this and related results we conclude that agglomeration economies attenuate rapidly at first, but then level off and attenuate much more slowly beyond 5 miles. For many applications, therefore, agglomerative externalities should ideally be studied at a relatively refined geographic level.

In a related project, Strange and I examine the microfoundations of agglomeration economies for 3- and 4-digit manufacturing industries in the United States. We regress the Ellison-Glaeser (1997) measure of spatial concentration for each industry on industry characteristics that proxy for the presence of knowledge spillovers, labor market pooling, shared intermediate inputs, and natural advantages. The analysis is conducted separately at the zip code, county, and state levels, and for both old and newly established enterprises. Results indicate that input sharing and labor market pooling are both associated with increases in agglomeration. In addition, there is suggestive evidence that knowledge spillovers further contribute to agglomeration, with the greatest effect stemming from innovation at small firms. Furthermore, there is evidence once again that agglomerative externalities attenuate geographically, especially among industries where input sharing and labor market pooling are important.

In work in progress, Strange and I are examining the impact of technological changes in information technology and communications on urbanization. Since these changes permit access to valuable resources from remote locations, they may impact the benefits from agglomeration. Ultimately, this means that they may change the role of the city in the economy, and the mix of industries and establishment-types choosing to locate in urban versus non-urban areas. For example, one question we are actively studying concerns the impact of IT/T on the industrial organization and size distribution of establishments in and outside of urban areas. In cities, firms can make use of a broad and deep network of outside suppliers, allowing for both a good match between a firm's needs and supplier competency, and for competition among suppliers. As noted in Vernon (1962, 1972) and Holmes (1999), this means that more vertically integrated establishments tend to locate outside the central city while less vertically integrated establishments tend to locate in downtown areas, ceteris paribus. As noted in Holmes and Stevens (2000), it is not necessarily true that the centrally located firms are smaller, however. With improvements in IT/T, distant firms are more accessible, so it may be possible to avoid the costs of vertical integration in industries able to take advantage of IT/T. That in turn has implications for the mix of industries and size distribution of establishments in and outside of urban areas.

With these thoughts as a backdrop, I would recommend that the CSISS consider offering a future workshop on the potential role of IT/T on spatial patterns of economic activity. In addition, further advertisement of underutilized data sources such as the Dun and Bradstreet MarketPlace file would be valuable. Training seminars on mapping software and related products - skills that are relatively uncommon among economists - would also be of considerable value.