Foreign Versus Local Ownership and Performance in Eastern Versus Western EU: A Random Forest Application

Abstract

Our paper proposes the machine learning Random Forest algorithm for classifying economic activity within the European Union, building on the relevance of a reduced set of variables alongside location and industry of origin for the differences in performance between foreign versus locally-owned companies. We find a diverse landscape of business performance within the European Union that does not indicate a clear-cut dominance of foreign-owned companies against their locally-owned peers. Locally-owned companies from the Eastern European Union have been more dynamic than their foreign-owned peers in the region, which suggests a process of learning from foreign competitors and business partners. The Random Forests model performs surprisingly well given the low number of predictors and indicates that personnel costs per employee is the most important variable that discriminates between foreign and locally-owned companies. The importance of the rest of the variables, including the regional location and the industry, has a relatively uniform distribution.