The study examines the complex relationship between Environmental, Social, and Governance (ESG) performance and bank stability within the context of the increasingly digitalized European banking sector. The research question centers on how changes in ESG scores affect bank stability in the digital era, exploring the link between digitization's impact on bank stability and ESG activities. The study's significance lies in addressing the theoretical debate surrounding the impact of ESG, which centers on two contrasting views: the shareholder view (prioritizing profit maximization and viewing ESG as a cost), and the stakeholder view (emphasizing ethical responsibility and risk mitigation through ESG). Existing literature reveals a lack of consensus and insufficient attention to non-linearity in this relationship, as well as a neglect of the role of digitization. This study contributes by testing the non-linear hypothesis using a Panel Smooth Transition AutoRegressive (PSTAR) model, incorporating the impact of Information and Communication Technology (ICT) diffusion and endowment, and analyzing a diverse sample of European banks over a significant time period (2005-2022) that includes major financial crises. The European context is chosen due to the sector's global importance, the region's focus on ESG activities, and the presence of mature banking sectors within the EU that allow for a clearer understanding of the interplay between digitization and ESG performance.

The literature review analyzes two primary relationships: the impact of digitization on bank stability, and the relationship between ESG performance and bank stability. Regarding digitization, the review discusses the "innovation-stability" hypothesis (where technology improves stability) and the "innovation-fragility" hypothesis (where technology increases risk). Empirical evidence supports both hypotheses, with some studies indicating mixed results depending on the level of development. Gaps in the literature include a focus on financial innovation rather than broad digitization impacts and a lack of consideration for non-linear relationships. Concerning ESG and bank stability, the review highlights the opposing views of shareholder and stakeholder theory. The shareholder view suggests a negative impact of ESG activities on profitability, while the stakeholder view posits a positive impact through risk mitigation, enhanced reputation, and increased stakeholder support. Empirical evidence largely supports the stakeholder view, demonstrating positive effects of ESG scores on risk reduction and stability. However, the literature lacks consideration of non-linearity. Studies suggesting non-linear U-shaped or inverted U-shaped relationships between social responsibility and financial performance are noted, and the review emphasizes the need to incorporate the digital environment in analyzing the relationship between ESG performance and bank stability.

The study employs a Panel Smooth Transition AutoRegressive (PSTAR) model to analyze the non-linear relationship between ESG performance and bank stability in the digital era for a panel of European banks from 15 countries between 2005 and 2022. The PSTAR model allows for regime switching based on ESG scores, capturing the potential non-linearity. The dependent variable is the banking Z-score, a measure of bank stability. Independent variables include indicators of ICT diffusion (internet users and mobile subscriptions), technological endowment (ATMs per capita), ESG scores, and control variables (net interest margin, return on assets, cost-to-income ratio, and GDP growth). Data sources include DATASTREAM, the International Telecommunication Union (ITU), and the World Bank. Before applying the PSTAR model, the study conducts several diagnostic tests to check assumptions. These tests include unit root tests (first and second-generation tests, including tests with structural breaks) to ensure stationarity of the variables, cross-sectional dependence tests (Friedman, Breusch and Pagan, Frees, and Pesaran tests) to assess correlation among countries, and heteroscedasticity and autocorrelation tests. Cointegration tests (first and second-generation) are performed to verify long-run relationships among the variables. The PSTAR model is then estimated, and linearity tests are used to determine the optimal model specification (number of regimes and threshold values). Robustness tests using Granger causality tests (Dumitrescu and Hurlin, and Juodis et al.) are employed to further validate the findings.

The PSTAR model reveals three distinct ESG performance regimes: 1) ESG score ≤ 33.314, 2) 33.314 < ESG score < 63.348, and 3) ESG score ≥ 63.348. In Regime 1 (low ESG), a 1% increase in ESG score decreases the banking Z-score by 0.245%, indicating a negative relationship with stability. In Regime 2 (moderate ESG), this negative effect weakens (-0.171%). However, in Regime 3 (high ESG), a 1% increase in ESG score leads to a 0.173% increase in the banking Z-score, showing a positive relationship. This confirms the non-linearity hypothesis. The transition between regimes is rapid. Analysis of the impact of digitization reveals that ICT diffusion (internet and mobile) positively impacts bank stability across all three regimes, supporting the "innovation-stability" hypothesis. However, technological endowment (ATMs), has a negative impact on bank stability only in Regime 1 (low ESG). In Regimes 2 and 3, the impact is positive. Granger causality tests support bidirectional causality between most variables (NIM, ROA, IUI, MCS, ATM, ESG) and the banking Z-score, indicating mutual influence.

The findings support a contingent perspective on the ESG-bank stability relationship, challenging the simplistic view of a consistently positive or negative effect. The results confirm the non-linearity of the relationship, showing a shift from a negative impact of ESG on bank stability at low to moderate levels to a positive impact at high levels. This transition aligns with the stakeholder theory, suggesting that sufficient ESG investment yields financial returns that enhance stability. Conversely, at lower ESG scores, the cost of ESG investments outweighs the benefits. The positive impact of ICT diffusion on bank stability underscores the importance of embracing digitization. However, the negative impact of technological endowment in the low-ESG regime highlights the vulnerability of banks with weak ESG performance to the risks of technology investments. This emphasizes the moderating role of ESG in the relationship between digitization and bank stability. The study provides evidence that banks with low ESG commitment are more susceptible to risks in the digital environment.

The study offers valuable insights into the non-linear relationship between ESG performance, digitization, and bank stability. The findings emphasize the crucial role of ESG in mitigating the risks associated with technological advancements in the banking sector. European banks should prioritize improving their ESG performance and implementing digital strategies that enhance sustainability. Policymakers need to adapt regulatory frameworks to support digitization and sustainability initiatives. Future research could explore the impacts of individual ESG pillars, incorporate banks' digital investment data, and broaden the geographical scope for enhanced analysis.

The study's limitations include a specific sample of European banks and countries that may limit the generalizability of the findings to other regions or banking sectors. The reliance on self-reported ESG data may introduce potential biases. Exploring the impact of individual ESG sub-pillars and considering other relevant variables (e.g., banks’ investment in digital technology) could provide a more nuanced understanding of the relationships. The inclusion of more specific measures of bank stability would help support the results further.