Applying economic-based analytical regions: a study of the spatial distribution of employment in Spain

Abstract

Even though there is a general consensus in the regional economics literature about the relevance of agglomeration economies in the distribution of economic activity, analyses of regional differences in employability in Spain have found limited empirical evidence about the role played by population concentrations. Our hypothesis in this paper is that the lack of evidence supporting theoretically expected behavior about the role of agglomerations is mainly explained by the fact that administrative rather analytical regions have been used. To check this, we propose a study based on spatially disaggregated data from the Population Census and its aggregation into analytical regions that incorporate ideas from the new economic geography. Our results regarding the effect of space on employment opportunities with this alternative classification suggest, as expected, that living in large cities or close to metropolitan areas increases the possibilities of being employed. The different results support the need for more highly disaggregated data at spatial level in order to overcome the limitations inherent to empirical analysis based on administrative regions.

Appendix

1.1 Analytical versus administrative regions: testing internal homogeneity and external separation in the Spanish case

In this Appendix, we evaluate the robustness of the analytical regions based on NEG versus the administrative ones commonly used (NUTs at different levels) according to the internal homogeneity and external separation criterion for the study of employment distribution in Spain. The first criterion (internal homogeneity) implies the existence of common characteristics for the workers living in the municipalities considered to be part of a region. As for the second criterion (dissimilarity between regions), the municipalities included in one region should show marked differential characteristics in terms of its labor force in comparison with others belonging to a different region. Specifically, these two criteria can be tested either total or by gender, for the distribution of employment by industry and also by level of qualification using the well-known Theil’s inequality index (Theil 1967), commonly applied to the distribution of income. Its within component will be useful for quantifying intraregional homogeneity when dealing with the spatial distribution of employment.

The formula of Theil’s inequality index is computed as follows:

$$\begin{aligned} T=\sum _{m=1}^n {\frac{\hbox {PopEmp}_\mathrm{municip} }{\hbox {PopEmp}_\mathrm{Esp} }\log \left( {\frac{\hbox {PopEmp}_\mathrm{municip} /\hbox {PopEmp}_\mathrm{Esp} }{1/n}} \right) } \end{aligned}$$

where $n$ is the number of municipalities considered (8,106), $\hbox {PopEmp}_\mathrm{municip}$ is the population employed in municipality $m$, and $\hbox {PopEmp}_\mathrm{Esp}$ represents the Spanish working population.

The Theil’s index can be completely and perfectly decomposed into a between-group component (Tg) and a within-group component (Tw). Intraregional homogeneity can be therefore quantified by the within-group component. Thus:

$$\begin{aligned} T =Tg + Tw \end{aligned}$$

with

$$\begin{aligned} Tg&= \sum _{r=1}^R {\frac{\hbox {PopEmp}_r}{\hbox {PopEmp}_\mathrm{Esp}}\log \left( {\frac{\hbox {PopEmp}_r /\hbox {PopEmp}_\mathrm{Esp} }{n_r /n}} \right) }\\ Tw&= \sum _{r=1}^R {\frac{\hbox {PopEmp}_r}{\hbox {PopEmp}_\mathrm{Esp} }\sum _{m=1}^n {\frac{\hbox {PopEmp}_\mathrm{municip} }{\hbox {PopEmp}_r }\log \left( {\frac{\hbox {PopEmp}_\mathrm{municip}/\hbox {PopEmp}_r }{1/n_r }} \right) } } \end{aligned}$$

where $r$ indexes regions, with $n_{r}$ representing the number of municipalities in region $r$ and $\hbox {PopEmp}_{r}$ the population employed in the region $r$ to which the municipality belongs.

As the within component quantifies the heterogeneity between the individuals of a region, small values indicate the existence of a high degree of internal homogeneity.

Given the characteristics of Theil’s index, an increase in the internal homogeneity of the regions (a decrease in the within component) necessarily implies that the heterogeneity between regions increases (a rise of the between component).

The within component of the Theil’s index (total and by gender) is shown in Table 5 for administrative regions, namely NUTS II (17 Comunidades Autónomas) and NUTS III (50 Provinces), and for the analytical regions suggested (8 regions).

Table 5 Analytical versus administrative regions

Despite the scale effect, i.e., everything else equal, intraregional inequality drops with the number of regions, the within component for the eight analytical regions is clearly lower than for any of the NUTS regions. In other words, the classification proposed shows a higher degree of internal homogeneity in the distribution of employment (therefore, of economic activity) so that the regional labor markets defined under the size and distance criteria are more integrated or coherent (even by gender) than any other political–administrative division of the territory. As expected, we can talk about one labor market for men and another one for women. These are two different labor markets—also at regional level—that show a higher level of homogeneity (and heterogeneity between them) when studied separately.

As for the spatial patterns of distribution of employment by industry, the 2001 Spanish Census offers employment figures for sixteen (16) types of industries. The results for the Theil’s index (total and decomposed) according to the industrial classification used in the Census (Table 6) show higher internal homogeneity within and also heterogeneity between the analytical regions for all industries except “Agriculture” and “Extractive Industries.” One simple explanation for these exceptions is that these particular activities are necessarily linked to the physical location of the natural resources at hand—land, forests, rivers or mines—and therefore, the chances to choose the geographical location of companies and also employees are very limited. In other words, the distribution of these activities is not affected by the existence of agglomeration economies but on the location of the natural resources.

Table 6 Analytical versus administrative regions

In order to test the spatial distribution of high-qualified employees or jobs, the 2001 Census offers information on nine groups aggregated by type of work and level of qualification. In this special classification, qualification is understood as the capacity to carry out the tasks which comprise any given job, and includes two different facets: level of qualification and specialization within this level qualification. To work with a more standard classification, we aggregate the nine groups into three more conventional ones: “High-Qualified Occupations,” “Medium-Qualified Occupations” and “Low-Qualified Occupations,” For detail about categories and their aggregation see Table 7. The results of the Theil’s index by level of qualification are shown in Table 8.

Table 7 Detailed qualification levels

Table 8 Analytical versus administrative regions

For all levels of qualification, the within component is clearly lower for the analytical regions than for any of the administrative divisions. The differences are slightly more pronounced for the high-qualified jobs, where agglomeration economies might play a more effective role.