Regional club convergence in the EU: evidence from a panel data analysis

Abstract

We investigate club convergence in income per capita in 194 European NUTS-2 regions using a nonlinear, time-varying factor model that allows for individual and transitional heterogeneity. Moreover, we extend an existing club clustering algorithm with two post-clustering merging algorithms that finalize club formation. We also apply an ordered response model to assess the role of initial and structural conditions, as well as geographic factors. Our results indicate the presence of four convergence clubs in the EU-15 countries. In support of the club convergence hypothesis, we find that initial conditions matter for the resulting income distribution. Geographic clustering is quite pronounced; besides a north-to-south division, we detect highincome clusters for capital cities. We conclude that the main supranational policy challenge is the politically sensitive handling of a multi-speed Europe.

Great Recession and club convergence in Europe: A cross‐country, cross‐region panel analysis (2000–2015)

“…Conversely, the lower level clubs mainly include the southern Italian regions, the Greek regions, and some of the regions of Bulgaria, Romania, and Hungary. Similarly to other studies (i.e., Bartkowska & Riedl, 2012; von Lyncker & Thoennessen, 2017), we find that the first club is typically composed of regions including cities and metropolitan areas, while the intermediate ones pertains to northern and central European advanced regions, and the lower clubs pertain to the peripheral regions of western European members. The larger sample used in our study allows to add further interesting insights to previous research.…”

“…Due to the fact that in presence of transition across clubs (i.e., part of one club tend to move toward another club) or highly conservative values of some parameters (i.e., c* ), there could be an over determination of the groups, the authors proposed a “club merging algorithm” (Phillips & Sul, 2009) that is adopted in this paper. As an improvement, we also apply a different club merging algorithm developed by von Lyncker and Thoennessen (2017). They introduce two innovations to the merging procedure used by PS.…”

“…Convergence clubs have also emerged by adopting the Phillips and Sul (2007, 2009) methodology (Table 2), both at the country (Apergis, Panopoulou, & Tsoumas, 2010; Borsi & Metiu, 2015; Fritsche & Kuzin, 2011; Monfort, Cuestas, & Ordonez, 2013) and at the regional and provincial levels (Bartkowska & Riedl, 2012; von Lyncker & Thoennessen, 2017; Montañés et al, 2018). However, none of these studies analyses simultaneously the national and regional club convergence on the same panel of countries.…”

“…Second, since the geographical level of analysis may lead to different results, we apply the abovementioned procedure simultaneously to the regional and the country levels for the same set of countries, making direct comparison between the two levels. Third, we supplement the merging algorithm for club formation developed by PS with the more recent one by von Lyncker and Thoennessen (2017) and test the two procedures with respect to a broader sample of regions than in previous studies. Indeed, the underlying hypotheses of the two merging algorithms may lead to remarkably different (final) club composition that may affect the club characterization, therefore yielding different conclusions (especially when comparing units with different structural characteristics such as EU‐15 vs. EU‐13 regions).…”

“… 5 5 For additional details see also Phillips and Sul (2009), von Lyncker and Thoennessen (2017) and Sichera and Pizzuto (2019). …”