Labor Density and Wages in Spain: Evidence from Geographically Disaggregated Data

Abstract

In this paper, Ciccone's () approach is applied to the Spanish case in 2011 but by estimating it using local labor markets (LLMs) instead of NUTS‐2 or NUTS‐3 regions. It is especially relevant in the case of Spain because the NUTS‐3 (provinces) are large regions in comparison with other cases in Europe. From a sample of income taxpayers published by the Spanish Fiscal Studies Institute, we derive figures on average wages by worker on the scale of LLMs. We argue that working at this level of spatial disaggregation is more in line with the idea of externalities from agglomerations, which are generated on a local scale. We can also observe intra‐regional heterogeneity and how the urban wage premium changes along the entire distribution of cities, including small‐medium size urban areas or rural areas. The empirical analysis is based on several estimation strategies, namely, ordinary least squares, two‐stages least squares, quantile regressions (QR), and instrumental variable quantile regressions (IVQR) estimators; they all find a significantly positive effect of agglomeration in the conditional mean of wages. This result can be estimated along the conditional distribution of wages. According to the QR and IVQR estimates, important variations are found along the distribution.

Geographical scale, industrial diversity, and regional economic stability

“…First, to effectively measure regional industry structure, it is not recommended to use highly disaggregated geographical units for a given spatial extent. Because of the small population problem (Wang & O'Brien, ), the data issue of measurement by place‐of‐residence and place‐of‐job (Dapena et al, ), the UGCoP (Kwan, ), it sometimes can be meaningless to study the economic diversity of a region with a small population and low employment. As illustrated by the ranges and standard deviations in Table , when county‐level data are aggregated, the impact of some extreme values or outliers are excluded.…”

“…Moreover, numerous scholars (e.g., Anselin, ; Harvey, ; Kwan, ; Openshaw & Taylor, ; Ruddell & Wentz, ; Schaeffer, Jackson, & Bukenya, ) have stressed the role of varying geographical scales as a generic issue in impacting empirical results. Meanwhile, scale‐related issues—such as the modifiable areal unit problem (MAUP; Openshaw, ; Openshaw & Taylor, ) and the ecological fallacy (Robinson, )—have been examined in specific research questions in regional science and economic geography (e.g., Briant, Combes, & Lafourcade, ; Dapena, Morollón, Pires, & Gomes, ; Dapena, Vázquez, & Morollón, , ; Resende, ).…”

“…In relation to empirical analyses, Anselin (, p. 26) further suggested that the MAUP can result in two problems: first, the micro–macro aggregation problem always occurs when the spatial units are not homogeneous or spatial heterogeneity exists; and second, the MAUP can also alter the choice of a spatial weight matrix in spatial regression and the corresponding parameter of the spatial lag variables. Furthermore, the MAUP also have been studied for specific topics in regional science related research (e.g., Dapena et al, , , ; Ouyang, Wentz, Ruddell, & Harlan, ; Resende, ). For instance, Resende () examined the determinants of regional economic growth from 1991 to 2000 in Brazil at four spatial scales (including states, municipalities, micro‐regions, and spatial clusters) and indicated that the empirical results vary greatly with the geographical scale.…”

“…For instance, Resende () examined the determinants of regional economic growth from 1991 to 2000 in Brazil at four spatial scales (including states, municipalities, micro‐regions, and spatial clusters) and indicated that the empirical results vary greatly with the geographical scale. Similarly, when studying labor density and wages in Spain, Dapena et al () suggested that agglomeration externalities are generated on a local scale, such as local labor markets rather than NUTS2 or NUTS‐3 regions.…”

“…This inconsistency can greatly impact empirical analysis. To mitigate this problem, Dapena et al (, p. 61) indicated that using functional regions that are defined based on population and commuting flows as the analytical units suffer less from this issue because one's place of residence and place of work “are in the same spatial unit.”…”