![Sample our Geography journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5937/TOC-Subject-Sample-2021-Geography.jpg"})
Abstract
In a seminal article, Chinitz (1961) considered the effects of industry size, structure, and economic diversification on the performance of firms and regional economies. His inquiry suggested a related but conceptually distinct issue: how does the extent to which a regional industry or industrial sector is concentrated in a small number of firms affect the local performance of that industry? The question has not been addressed systematically in empirical research other than case studies, principally because accurately measuring regional concentration requires firm-level information. This exploratory study uses confidential plant-level data to gauge concentration in manufacturing industries at the regional scale across the continental United States, to explore changes over time in geographic patterns of concentration, and to investigate associations between regional industrial structure concentration and changes in employment. The implications for understanding the impacts of regional industrial structure on economic development processes are discussed.
Acknowledgments
This article is based on research supported by the National Science Foundation (grant BCS-0423900), the Ewing Marion Kauffman Foundation, and the North American Regional Science Council. Any opinions and conclusions expressed herein are those of the author and do not necessarily represent the views of the United States Census Bureau or any of the supporting organizations. All results have been reviewed to ensure that no confidential information is disclosed. Support for this research at the Triangle Census Research Data Center from the National Science Foundation (grant ITR-0427889) is also gratefully acknowledged.
Notes
1 This definition of dominance pertains to relationships among firms interacting in the context of a particular regional industry. Some earlier work used the term dominance to refer to interregional industrial structures, considering competition across regions either within industries or for comparative advantage with respect to particular industrial functions (e.g., CitationStanback and Noyelle 1982; CitationHolloway and Wheeler 1991).
2 This description of regional industrial dominance is analogous to defining competition relative to market structure and influence, rather than to the active or passive nature of the actions of participants (CitationScherer 1980).
3 Results obtained using primary MSAs (PMSAs) instead of CMSAs are substantively equivalent.
4 In the industrial organization literature, empirical work on the distribution of sizes of firms centers mainly on Gibrat’s law, the proposition that the growth rates of firms are independent of the sizes of firms. Nearly all this research examines larger-than-regional scales and ignores location as a relevant factor, so the substantive findings do little to inform this study (CitationSchmalensee 1989; CitationDavies and Geroski 1997).
5 For this aggregation step, industries are defined at the most detailed classification level available in the LRD: four-digit SIC or six-digit NAICS industry categories, assigned according to primary production activity. Establishments belonging to multiunit firms that are located within different regions or that are classified into different industries remain separate “firms.”
6 Qualitatively similar results were obtained with alternative measures: varying the number of firms contained in the concentration ratio numerator, changing the exponent in the Herfindahl-Hirschman index, varying the minimum number of firms in the regional industry for inclusion, substituting shipment value for employment, and calculating a plant-based rather than a firm-based ratio.
7 Cross-sectional tabulations and longitudinal tracking of employment figures for individual plants (not cleared for disclosure) demonstrated little evidence of rising entry rates or increasing growth in employment in small establishments over time. Whereas a given cross section of relatively large establishments may be expected to exhibit greater percentage declines in employment because of reversion to the mean, the categorization of plants into size classes over time revealed that smaller establishments did not experience corresponding greater-than-average gains in employment (or smaller than average percentage reductions) during the period under study.
8 If a few large regions exhibit particularly large values of growth or decline inemployment, or the dependent variable is substantially skewed, then it may be difficult to interpret unambiguously the regression results. (I thank an anonymous reviewer for pointing out this possibility.) Descriptive information regarding the dependent variable has not been approved for public release, but robustness checks suggest that there is little distributional skew and that outliers do not substantially affect the results. The mean and median of the dependent variable are both close to zero; diagnostic tests, such as Cook’s distance and df beta coefficients, indicate only a few influential outliers, and regressions that were calculated omitting these outliers returned an estimated coefficient for the concentration ratio independent variable that is somewhat smaller in magnitude but has the same sign and greater significance than reported in .
9 Alternative regressions calculated with different dependent variables, changes specified as percentages rather than absolute values, and the coverage of different periods yielded weaker results but substantively similar interpretations.
10 The global Moran’s I and LaGrange multiplier statistics diagnose positive spatial dependence in the ordinary least-squares (OLS) regression, favoring a spatial lag interpretation. I implemented a spatial lag model using an inverse distance weights matrix. Because the coefficient estimates from the spatial lag model present no major substantive differences from the OLS estimates save for the census region dummy variables, only the results of the spatial lag model are presented here. The OLS regression results and spatial dependence statistics are available on request.
11 The coefficients obtained from the spatial lag models present few substantive differences from those estimated using OLS. The OLS regression results and spatial dependence statistics are available on request.
12 Manufacturing-wide and industry-specific regional dominance are correlated to a much lesser degree in the positive direction.