Estimating agglomeration economies in Spain: evidence from geographically disaggregated data

In this paper we estimate agglomeration economies in Spain in 2009 basing on Ciccone?s (2002) model, which explains average labor productivity in one spatial unit on employment density and other controls. The novelty of our analysis is that the empirical model is estimated at a highly disaggregated spatial scale, oppositely to the convention of taking as unit of analysis NUTS-2 or NUTS-3 regions. Recent contributions to New Economic Geography (NEG) base their theoretical analysis on geographical units defined at a more disaggregated spatial scale than these administratively defined regions. Specifically, from a sample of income-taxpayers published by the Spanish Fiscal Studies Institute -Instituto de Estudios Fiscales- we derive figures on average wages by worker at the scale of Local Labor Markets (LLMs). The empirical analysis bases on several estimation strategies; namely, Ordinary Least Squares (OLS), Two-Stages Least Squares (2SLS), Quantile Regressions (QR) and Instrumental Variable Quantile Regressions (IVQR) estimators, all they finding a significantly positive effect of agglomeration in the conditional mean of labor productivity. Additionally, the QR and IVQR estimators find a progressively decreasing, but still positive, effect of employment density along the conditional distribution of labor productivity.