Mapping Economic Activity in the European Union: Do Ownership, Industry and Location Matter?

The paper proposes a new method for analysing the structure and dynamics of economic activity undertaken by locally owned and foreign-owned companies within the European Union. We employ an unsupervised learning algorithm that generates a neural network depicted on Kohonen maps and offering a clustering of companies with a different ownership (local and foreign) from various industries and countries of the European Union during 2009–2016. The research methodology, based on a self-organizing map (SOM) algorithm, belongs to a class of neural networks trained to organize data so that unknown patterns may be discovered, thus leading to results that cannot be attained by more traditional clustering methods. Each type of company (locally owned and foreign-owned) from a specific industry and country is characterized by a series of performance indicators that are included in the SOM algorithm, i.e. indicators at the average enterprise and employee level (turnover, value added at factor cost, gross operating surplus, personnel costs, gross investments) and comprehensive indicators, such as labour productivity and profitability (the latter through the gross operating rate). We detect various clusters of companies based on Euclidian distances that provide similarities and differentiation between companies’ common production activities by taking into account their ownership (foreign versus local), industry and country of location, and related performance results, as well as their interrelationships. The resulting classification can be used to understand the linkages between European Union companies and the different branches of economic activities across EU countries, as well as to investigate the performance gap between locally owned and foreign-owned companies.