Patrones De Convergencia En Las Regiones Españolas: Una Aplicación De La Metodología De Phillips-Sul

Resumen:El objetivo de este trabajo es examinar si las regiones españolas han seguido un proceso de total convergencia entre ellas o si por el contrario presentan un patrón de convergencia en clubs. Para ello se aplica una metodología novedosa basada en el test desarrollado por Phillips y Sul (2007). La convergencia regional se analiza mediante tres variables, la renta per cápita, la productividad del trabajo y el empleo per cápita y para el período 1980-2008. Los resultados obtenidos muestran la existencia de convergencia absoluta en la productividad, mientras que se identifican tres clubs de convergencia para las otras dos variables.Abstract: The literature on economic growth has placed special focus on analysing the convergence processes between countries and regions. Within the growth theories, two alternative approaches have been developed to explain the differences observed in per capita income across countries over time. Neoclassical growth models predict a process of convergence between economies where the relatively poor economies will grow at a faster rate than the relatively rich ones, while endogenous growth models describe a situation of non-convergence. Theoretical developments and empirical studies on convergence have led to the development of different definitions of the term and to the use of different methodologies for its investigation (Islam, 2003). The concepts of sigma and beta convergence have been widely used in empirical papers. Sigma convergence refers to the reduction in the per capita income dispersion across economies over time, while beta convergence refers to the existence of a negative correlation between income growth over time and its initial level. The concept of absolute or unconditional convergence assumes that per capita incomes in the regions will tend to converge in the long term to a single steady state, regardless of their initial conditions. In contrast, the conditional convergence hypothesis holds that each economy converges to its own stationary state, so that economies will converge with one another in the long run if they have similar structural characteristics (Galor, 1996). Neoclassical growth models lead to the hypothesis of conditional convergence between economies, but also to the hypothesis of convergence clubs, which proposes that regions with similar economic structures can converge to different steady states if they start from different initial conditions. Therefore, although certain regions have globally heterogeneous growth paths, they may be gathered into subgroups that exhibit homogeneous growth dynamics. At the international level, the empirical evidence confirms the existence of convergence clubs between countries (Durlauf and Johnson, 1995; Canova, 2004, Phillips and Sul, 2007, Monfort et al., 2013, Borsi and Metiu, 2015), as well as between regions (Postiglioni et al., 2010; Bartkowska and Riedl, 2012; Rodriguez et al., 2016; Tian et al., 2016; von Lyncker and Thoennessen, 2016). However, there is still little empirical evidence for the existence of convergence clubs in the Spanish economy, even though a few papers have been written in this regard. Indeed, some research has provided evidence of convergence clubs between Spanish regions since the late 1970s, clubs that remain to this day (Perez, 2000, Goerlich et al., 2002, Montañes and Olmos, 2014, Brida et al., 2015), although none has used a methodology like the one used in this article. Perez (2000) notes that the convergence process for per capita income in Spain’s Autonomous Communities during the period 1955-1995 can be characterised by subgroups of regions that converge to different stationary states. Goerlich et al. (2002), examining the convergence of Spanish regions during the period 1955-2000, find, by the end of the period, the existence of two convergence clubs both when they use per capita income and labour productivity as a variable. Brida et al. (2015) apply a nonparametric clustering approach to the per capita income data of the Spanish Autonomous Communities to analyse regional convergence during the period 1955-2009. Their results indicate the presence, since the late seventies, of two convergence clubs, one more homogeneous composed by the richer regions, and another more heterogeneous formed by the remaining regions. They also note that there has been more convergence among the regions in the first club and a gap between clubs in the last two decades. However, as the authors point out, these clubs have not remained stable over time, with their numbers ranging from three to five. Finally, Montañes and Olmos (2014), using two different indicators, per capita income and an indicator of human development, study the possible stochastic convergence between Spanish regions for the period 1980-2010. The results show, for the end of the period, the existence of two distinct geographical areas (for the two indicators used), which is interpreted by the authors as evidence of different convergence clubs. Bearing this in mind, this paper contributes to the existing literature by providing some new evidence on the regional converge process in Spain. More specifically, the aim of this work is to analyse whether Spanish regions display a full convergence process among them or if, on the contrary, they form convergence clubs. Phillips and Sul's methodology introduces a cross-sectional study, by means of an analysis of heterogeneous time series in the parameters of a neoclassical growth model, in order to take into account the heterogeneity of the transitional temporary variable analysed. This approach has clear advantages over other alternative methods. Firstly, it can be used to endogenously identify groups of regions converging towards the same growth path, and not by applying a predetermined criterion. Secondly, although a full convergence hypothesis can be rejected, this approach makes it possible to identify convergence clubs among regions, as well as the divergent regions. In addition, the speed of the convergence parameter can also be estimated with this methodology, which allows distinguishing the relative convergence empirically. The regional convergence process is analysed considering three variables: income per capita and its main components, GDP per worker and employment per capita for 17 Spanish regions in the period 1980-2008. Data comes from the regional dataset BD.MORES. The empirical results obtained in this research confirm the existence of full convergence for GDP per worker. However, there is also evidence for the existence of convergence clubs in terms of both income per capita and employment in Spanish regions. Regarding income per capita, our findings suggest the existence of three convergence clubs, which converge to different income levels: high, medium and low; whereas no divergent region was identified. With respect to employment per capita, the results are quite similar to those above for income per capita. We identify three clubs, but no divergent region was detected. The composition of clubs respect both variables, income and per capita employment, which remained relatively stable in the period analysed. Only four regions (Asturias, Cantabria, Castile and Leon and the Basque Country) exhibited differences in the composition of the clubs. The differences in the clubs' configuration may be explained by the different behaviour of labour productivity in these regions. Finally, it is worth noting that this paper is the first step in our research. A deeper analysis of thefactors responsible for the formation of convergence clubs in Spain must be undertaken in order to provide useful insight to policy makers regarding the mechanisms needed to achieve economic and social cohesion amongst regions.