Foreign bank presence, threshold effects, and institutional moderation: a causal machine learning approach to financial stability in Vietnamese banks

Purpose The study aims to investigate the impact of foreign bank presence (FBP) on financial stability in emerging and developing economies, focusing on the dual dimensions of FBP, foreign bank numbers (FS1) and asset shares (FS2), and the moderating role of institutional quality within Vietnam's evolving banking system. Design/methodology/approach A balanced bank-year panel of 28 Vietnamese commercial banks from 2010 to 2024 is analyzed using Hansen's threshold regression and Causal Forests with Double Machine Learning. The Enhanced Worldwide Governance Index (EWGI), combining traditional and digital governance indicators, is employed to capture institutional quality effects. Findings Results reveal nonlinear and heterogeneous effects. FS1 enhances financial stability only beyond a critical threshold, whereas excessive FS2 values generate destabilizing outcomes, confirming the liability of foreignness. Heterogeneity analysis indicates stronger positive responses among small and privately owned banks, with diminished FS1-related responses following the COVID-19 shock. Institutional quality significantly moderates these relationships: under weak governance, foreign banks enhance stability by filling institutional voids, while their marginal contribution diminishes in strong institutional settings. Research limitations/implications The analysis is limited to Vietnam's banking sector and may not generalize to economies with differing institutional architectures. Future cross-country extensions could validate threshold heterogeneity. Practical implications Results suggest that financial liberalization policies should be accompanied by improvements in institutional and regulatory quality to ensure stable integration of foreign banks. Originality/value This study provides novel empirical evidence linking foreign bank presence, digital-era institutional quality, and financial stability through an integrated nonlinear and machine-learning-based heterogeneity framework. The development of the EWGI and the use of Causal Forests offer methodological and conceptual innovations relevant to emerging-market financial governance.