Outlier Identification with MFV-robustified Linear Regression in case of Economic Convergence of EU NUTS Regions

Abstract

Despite being questioned, absolute economic β -converge is a widely applied method for investigating cohesion tendencies among EU regions. Corresponding data fur- ther supports the application of the theory. Relying on our previous results the present study utilizes MFV-robustified linear regression in order to identify over- and under- performing regions. For this purpose regional GDP and NDI data of EU NUTS2 and NUTS3 level regions are used from the time period of 2000-2020. Since underlying data distributions have typically long tails, are highly skewed and contaminated by several outliers robust statistical approaches are advised. The outlined procedure suggests that economic convergence tendency among EU regions is less expressed than conventional β - convergence would claim. Furthermore, by substituting mean values by ”Most Frequent Values” introduced by Stener et. al. in corresponding calculations regions can be found to be greatly deviating predicted by conventional convergence theories that would otherwise be masked due to data characteristics.