Do Manufacturing Plants Cluster across Rural Areas? Evidence from a Probabilistic Modeling Approach

A statistical procedure for detecting “contagious” location patterns for manufacturing establishments is presented. Manufacturing industries’ establishment clustering tendencies are ranked based on the “dispersion parameter” of the negative binomial distribution. Establishment data are for three-digit SIC manufacturing industries, nonmetro counties of BEA Component Economic Areas, 1981 and 1992. Findings indicate that virtually all manufacturing industries cluster establishments in nonmetro areas. Approximately two-thirds of the industries had dispersion parameters indicating a high or moderate level of spatial concentration. The propensity to cluster plants in nonmetro CEAs was evident for both 1981 and 1992, though weaker in 1992. Much of the industry clustering in nonmetro areas appears to be attributable to local “natural advantages” and not inter-firm spillovers. Contact Person: David L. Barkley, Professor Department of Agricultural and Applied Economics 255 Barre Hall Clemson University 29634-0355 Phone: (864-656-5797) FAX: (864-656-5776) e-mail: [email protected] Do Manufacturing Plants Cluster Across Rural Areas? Evidence from a Probabilistic Modeling Approach Introduction The new economic growth theory of Romer, Krugman, and Venables has stimulated a renewed interest in the spatial concentration of industrial activity and the advantages that concentrations provide member establishments. The focus of this recent research is the extent of industry clusters or agglomerations at state or metro levels and evidence of static and/or dynamic locationalization economies attributable to these concentrations. Research on industry clusters in nonmetropolitan areas is also available but currently limited to case studies of specific industries (Rosenfeld) and analysis of the association between nonmetro concentrations and industry employment growth (Henry and Drabenstott) and wage rates (Gibbs and Bernat). Yet industry clusters may be especially important to nonmetro areas if these agglomerations provide (through external economies) the means to overcome disadvantages inherent with nonmetro locations (sparse local markets, geographic isolation, and lack of economic diversity). The purpose of this study is to provide an overview of establishment clustering tendencies and trends for nonmetropolitan manufacturing industries (three-digit SIC level). Three research questions are of particular interest. Do manufacturers cluster establishments in nonmetro areas, and if so, are these agglomeration propensities relatively strong? Which manufacturing industries exhibit the greatest or least spatial concentration in nonmetro areas? Have nonmetropolitan clustering propensities increased or decreased over time for manufacturers? Answers to the above questions 2 will provide insights into the determinants of industry clustering in nonmetropolitan areas (natural advantages versus inter-industry spillovers). Documentation of industries’ agglomeration tendencies also will be useful to nonmetro communities that are attempting to develop or expand industry clusters through targeted industrial recruitment efforts. The development of an industry cluster provides greater local economic development benefits than a less focused industrialization strategy because establishment clustering promotes external economies, facilitates industrial restructuring, stimulates inter-firm networking, and permits greater focusing of public resources (Barkley and Henry). However, all industries are not equally attractive candidates for industry clusters, and industry cluster development programs will have greater success if the targeted industries tend to spatially concentrate their establishments. In this paper, industry clustering is addressed through analysis of a statistical measure of geographic dispersion based on the spatial distribution of nonmetro establishments in three-digit SIC manufacturing industries. Thus, an industry cluster is defined to be a group of establishments in the same or closely related industry, located in close proximity to one another. As noted by Bernat, this definition represents an intermediate view of industry clusters. Broader interpretations include other industries linked to the “core” industry through actual or potential buy or sell relationships. Narrower interpretations, on the other hand, restrict industry clusters to establishments in close proximity that are closely connected through networks. Bernat notes, however, that neither intra-industry buyor sell-linkages nor networking are necessary for the existence of establishment clusters since establishments may be responding to cluster3 related externalities provided through the market. Our analysis of nonmetro establishment concentrations is organized as follows. First, we provide a summary of the reasons why establishments in an industry may locate near one another. Second, we present a statistical methodology for detecting “contagious” establishment location patterns, and rank industries’ nonmetro clustering tendencies based on the “dispersion parameter” of the negative binomial distribution. Third, industries with high or low establishment concentrations are compared to provide insights into the determinants of nonmetro clusters and implications for nonmetro industrial development policy. Why Do Industry Establishments Cluster? Ellison and Glaeser find widespread evidence of industry clustering and attribute this to two principal forces: industry-specific spillovers and natural advantages. Industry-specific spillovers are economies external to the firms but internal to the regional industry cluster. These external economies are referred to as static localization economies if they are attributable to the current scale (e.g. employment or number of establishments) of the industry cluster or Marshall-Arrow-Romer (MAR) dynamic externalities if they result from a historical presence and regional specialization in a particular industry. More specifically, Henderson (1986) attributes static localization economies to: (1) economies of intra-industry specialization where increased industry size permits 4 greater specialization among industry firms in addition to a greater availability of specialized intermediate input suppliers, business services, and financial markets. (2) labor market economies resulting from a larger pool of trained, specialized workers and reduced search costs for firms looking for workers with specific skills. (3) scale for networking or communication among firms to take advantage of complementarities, exploit new markets, integrate activities, and adopt new innovations. (4) scale in providing public goods and services tailored to the needs of a specific industry. Alternatively, Marshall-Arrow-Romer externalities are derived from the accumulation of knowledge and knowledge spillovers among local firms in the same industry (Glaeser et al.; Henderson, Kuncoro, and Turner). The build-up and sharing of knowledge among area firms in the industry are enhanced by a local legacy of and specialization in a particular industry. Both static and dynamic externalities encourage the clustering of industry establishments in a limited number of locations. Yet, Ellison and Glaeser (p. 921) suggest that “some of the most extreme cases of concentration are likely due to natural advantages.” Natural advantages include climate, topography, proximity to locationspecific inputs, locations that minimize transportation costs associated with shipping inputs and outputs, and locations with access to pools of labor with desired characteristics (e.g., lower labor costs or amenities attractive to skilled labor). McCann suggests that spatial industry agglomerations resulting from “natural advantages” may be purely the incidental result of individual firm optimizing behavior. The presence of other establishments in the industry at the location may not provide any benefits in terms 5 of external economies. Insights into the role of industry-specific spillovers versus natural advantages in nonmetro clusters may be provided by an investigation of clustering propensities across industries. A comparison of alternative methodologies for estimating spatial concentration is presented in the following section. Methodology for Measuring Industry Agglomerations Estimating Spatial Concentrations. Four principal indices are used in previous research to measure the spatial concentration of industrial activity: spatial concentration ratio, spatial Hirschman-Herfindahl index, locational Gini coefficient, and the Ellison and Glaeser concentration index. The spatial concentration ratio is generally the percentage of an industry's employment in the most concentrated four or eight geographic areas (e.g., states, metro areas, or counties). Spatial concentration ratios provide only limited information on differences in the spatial distributions of industries because the industry's ratios may be sensitive to the number of regions selected and information on the distribution of employment outside the selected four or eight regions is not considered. The spatial Hirschman-Herfindahl index is preferred to the concentration ratio because the index includes information from all relevant regions. The HirschmanHerfindahl index generally is estimated as: