Unveiling endogeneity between competition and efficiency in European banks: a robust econometric-neural network approach

Research on the European banking industry remains inconclusive concerning how its competitive structure and performance are related, especially given the heterogeneity among countries in the region. We develop a Dynamic Network Data Envelopment Analysis (DNDEA) model formed by three consecutive stages— profit sheet , balance sheet , and financial health efficiency—to assess how market structure and competition impact bank efficiency in European countries. Unlike previous research, a Robust Econometric-Neural Network Approach (RENNA) is used to unveil endogeneity among bank competition, market structure, and overall efficiency scores in European banking. Consistent with the competition-efficiency hypothesis, findings reveal that competition positively affects bank efficiency, particularly its balance sheet dimension. While macroeconomic factors are robust determinants of efficiency for non-GIIPS banks, Bank Z-score is far more relevant in the GIIPS subsample (Greece, Italy, Ireland, Portugal, and Spain). Furthermore, we find only weak evidence of feedback among the variables across subsamples. Our results have critical policy implications since they highlight the heterogeneous relationship between competition and efficiency for the banking sector.