Determinants of Bank Profitability in Selected Balkan Countries: A Combined Econometric and Machine Learning Approach

Abstract

In recent years, the banking system has been affected by several economic and financial shocks, increasing the importance of analyzing bank profitability and its determinants. This study examines bank profitability in selected Balkan countries over the period 2010–2024 by combining econometric panel data methods with machine learning techniques. Bank profitability is proxied by two commonly used indicators, ROA and ROE, while the explanatory variables include bank-specific factors such as efficiency, capital adequacy, non-performing loans, net interest margin, and the credit-to-deposit ratio, as well as macroeconomic variables such as GDP, inflation and unemployment. The econometric results indicate that efficiency and capital adequacy are key determinants of bank profitability, with efficiency negatively associated with ROA and ROE, while capital adequacy is positively associated with both indicators. The machine learning analysis, based on Random Forest and XGBoost, further evaluates the predictive role of the explanatory variables. Overall, the results show that bank-specific variables have a stronger influence on profitability than macroeconomic variables. In particular, feature importance highlights the relevance of the credit-to-deposit ratio, while SHAP values emphasize the contribution of NPLs. Overall, the findings suggest that bank profitability in selected Balkan countries is mainly driven by internal banking factors rather than macroeconomic conditions.