Essays in spatial econometrics

Abstract

This thesis is an attempt to obtain further insight into the role of spatial and dynamic linkages in the field of Economics given the crucial need for a better understanding of the fundamental processes behind the spatial and temporal correlation patterns observable in the economic data.

To date, most theoretical economic models and econometric studies have treated units of analysis as isolated entities, ignoring the spatial characteristics of the data and the potential role of space in modulating the economic evolution of countries, regions, municipalities, etc. In this regard, the essence of spatial economic analysis is that space matters. This implies that what happens in one economic unit of analysis is linked to what happens in neighboring economic units. In a spatial economic modeling framework, the spatial dimension and geographical arrangement of interacting economic agents are key drivers of economic processes and their final outcomes.

The recognition of the wide range of interconnections between the interacting agents in economics requires to accommodate such interdependence in the modeling process and in order to verify models of social and spatial interaction, these spatial effects need to be explicitly accounted for. Failure to take into account spatial dependence and spatial heterogeneity in econometric models leads to major estimation problems because the coefficient estimates will be biased, inconsistent and/or inefficient.

A distinct and innovative feature of this research is the use of static and dynamic spatial panel data estimation techniques for the empirical testing and validation of the theoretical models developed in the different chapters. This methodological approach is particularly appropriate for the analysis of economic phenomena from an integrated space-time perspective because it allows to model spillover, feedback and diffusion effects among the study units. Frequentist Spatial Econometrics modeling tools are complemented with Bayesian Spatial Econometrics and Relative Importance metrics in order to gain knowledge about the type of connectivity structures, the underlying spatial processes behind the observable data and to carry out inference in the relevance of the different factors explaining disparities among spatial units in time.

The structure of this thesis consists of four self-contained chapters.

Chapter 1 analyzes the volatility-regional growth nexus in a sample of European regions. To that end, a model of stochastic neoclassical growth with spatial interdependence is developed. In this framework, the economic growth rate of a particular region is affected not only by its own degree of volatility but also by the output fluctuations experienced by the remaining regions. In order to investigate the empirical validity of this result, the link between volatility and economic growth is examined in a sample of 272 European regions over the period 1991-2011 using a variety of static spatial pane specifications including spatial fixed effects. The results suggest the existence of a robust negative link between volatility and growth.

Chapter 2 investigates regional development dynamics in a sample of 254 NUTS 2 European Union regions over the period 2000–2010. To that end, a new version of the Regional Lisbon Index (RLI) containing changes with respect the index developed by Dijkstra is proposed. The RLI employment, education and R&D indicators. Targets for these indicators are related to an action and economic development plan for the EU regions and have been incorporated into European Regional Policy programming. The analysis of regional development is based on the estimation of the spatial Durbin model. Different specifications of the spatial weights matrix describing the spatial arrangement are compared by means of Spatial Bayesian Econometrics techniques. The salient finding of this chapter is that the main drivers of the RLI growth rate are technological capital, infrastructures and employment growth.

Chapter 3 analyzes unemployment differentials in 241 European regions during the period 2000-2011. To that end, a theoretical model with substantive spatial interactions among regions is developed. The solution implies a Dynamic Spatial Durbin Model specification including regional and institutional level factors as explanatory variables. In conjunction with dynamic-spatial panel estimates, relative importance metrics are used to quantify the effect of regional disequilibrium, equilibrium and national level factors. Relative importance analysis suggests that during the pre-crisis period unemployment disparities were mainly driven by regional level equilibrium factors. Nevertheless, labor market institutions are of major importance to explain increasing disparities during 2009-2011.

Chapter 4 looks into the nature of fiscal policy interactions in local fiscal policy in Spain. This study extends traditional spatial spillover models of government spending by including dynamic effects in order to test the relevance of the incremental budget hypothesis stemming from political science research. The theoretical model developed in this study points out to an empirical specification including simultaneous and lagged endogenous interactions among the sample of municipalities, as well as exogenous interaction effects. To that end, a Dynamic Spatial Durbin panel data model is used to quantify the relevance of spatial spillovers and diffusion effects over time. Using annual data for a sample of 1230 Spanish municipalities during 2000 to 2012, it is observed that: there are significant positive simultaneous spatial spillovers in different government expenditure categories and that the incremental hypothesis stemming from political science has a greater explanatory power than that of spatial spillovers.