Capturing Spatial Clusters of Activity in the Spanish Mediterranean Axis

Abstract

Concentration of economic activity constitutes a stylized fact of social sciences, giving birth to a very fertile branch of research since the writings of Marshall until today. Many researchers have been working on how to measure and explain concentration patterns of activity, trying to face that challenge and transpose it to a tractable argument in terms of modeling. Cluster analysis is one of the most salient efforts in this direction, with different contributions defining measures of concentration that range from the simplest indexes of inequality of Theil and Gini (Krugman 1992), until the more elaborated measures due to the recent work of Henderson (Henderson 1974, 1988; Henderson and Venables 2009), and Ellison and Glaeser (1997). In general, advances in this literature have focused on refining the construction of concentration indexes for identifying clusters of employment or firms in a certain territory. For example, the pioneer work of Ellison and Glaeser (1997) developed an agglomeration index (EG), together with a co-location one, that has been generalized in the literature as a reference. Its main contribution is that it derives from an explicit theory of firm location behaviour (the random-dartboard approach), controls for differences in the size distribution of establishments among industries, and appears to be robust to the level of spatial aggregation at which industry data are available. Other novel studies in this direction include that of whom evaluates the performance of the EG index but now for different sectors of economic activity, finding that the statistic behaves better for industrial activities than for consumer and business services in measuring concentration levels. Feser and Bergman (2000) that test if the EG index is sensitive to the scale of data employed (at the level of counties, commuting sheds, and zip codes), showing that changes in the spatial scale of data can introduce non-trivial ambiguities in the usual application of the EG index. Because of that, they recommend considerable caution when employing the index in comparative space-time studies about the concentration of industries. Braunerhjelmy and Johansson (2003) employ the EG and Gini locational indexes to evaluate the degree of concentration in 143 industries (at a four-digit level) for Sweden between 1975 and 1993, while Midelfart-Knarvik et al. (2004) use Gini locational index to analyze 36 industrial activities and 5 of services, with both works showing a more disperse pattern for services in comparison with industries. In addition, other locational studies also try to disentangle the forces driving important international flows such as FDI, population, or migrants (see, i.e., Blonigen et al. 2008; Baltagi et al. 2005; Kaushal 2005).

Fernando A. López and Ana Angulo wish to thank financial support from project ECO2009-10534/ECON of the Ministerio de Ciencia y Tecnología of the Reino de España; Andrés Artal-Tur wish to thank financial support from the Spanish Ministry of Science and Innovation (Project ECO2011-27169).