Tree-based approaches for understanding growth patterns in the European regions

• Published in: Region (Louvain-la-Neuve) Vol. 3; no. 2; pp. 23 - 45
• Main Authors: Annoni, Paola, Rubianes, Angél Catalina
• Format: Journal Article
• Language: English
• Published: Louvain-la-Neuve European Regional Science Association (ERSA) 01.01.2016; ERSA
• Subjects: data mining; decision trees; European Union regions; multivariate adaptive regression splines; Regional economic growth
• ISSN: 2409-5370; 2409-5370
• DOI: 10.18335/region.v3i2.115

The paper describes an empirical analysis to understand the main drivers of economic growth in the European Union (EU) regions in the past decade. The analysis maintains the traditional factors of growth used in the literature on regional growth - stage of development, population agglomeration, transport infrastructure, human capital, labour market and research and innovation - and incorporates the institutional quality and two variables which aim to reflect the macroeconomic conditions in which the regions operate. Given the scarcity of reliable and comparable regional data at the EU level, large part of the analysis has been devoted to build reliable and consistent panel data on potential factors of growth. Two non-parametric, decision-tree techniques, randomized Classication and Regression Tree and Multivariate Adaptive Regression Splines, are employed for their ability to address data complexities such as non-linearities and interaction eects, which are generally a challenge for more traditional statistical procedures such as linear regression. Results show that the dependence of growth rates on the factors included in the analysis is clearly non-linear with important factor interactions. This means that growth is determined by the simultaneous presence of multiple stimulus factors rather than the presence of a single area of excellence. Results also conrm the critical importance of the macroeconomic framework together with human capital as major drivers of economic growth of countries and regions. This is overall in line with most of the economic literature, which has persistently underlined the major role of these factors on economic growth but with the novelty that the macroeconomic conditions are here incorporated. Human capital also has an important role, with low-skilled workforce having a higher detrimental eect on growth than high-skilled. Not surprisingly, other important factors are the quality of governance and, in line with the neoclassical growth theory, the stage of development, with less developed economies growing at a faster pace than the others. The evidence given by the model about the impact of other factors on economic growth such as those on the quality of infrastructure or the level of innovation seems to be more limited and inconclusive. The analysis conclusions support the reinforcement of the EU economic governance and the conditionality mechanisms set in the new architecture of the EU regional funds 2014-2020 whose rationale is that the eectiveness of the expenditure is conditional to good institutional quality and sound economic policies.