The Impact of ESG Performance on Non-Performing Loans, Capital Adequacy, Liquidity Risk, and Net Balance Sheet Position in Banks

Abstract

This study examines the relationship between banks’ ESG performance and core risk and balance sheet indicators in the Turkish banking sector. Using an unbalanced panel of eight banks listed on Borsa Istanbul over the period 2008–2023, we estimate bank fixed-effects models with one-year-lagged ESG measures and controls and report Driscoll–Kraay standard errors. Two complementary specifications are employed: one based on the composite ESG score and another based on its environmental (E), social (S), and governance (G) pillars. The findings suggest that the composite ESG score is positively associated with non-performing loans and capital adequacy, while its relationship with liquidity risk and net balance sheet position/equity is less stable across specifications. When the ESG pillars are examined separately, substantial heterogeneity emerges across the E, S, and G dimensions. In particular, the environmental score is negatively associated with capital adequacy, whereas the social score is negatively associated with net balance sheet position/equity. Governance-related results appear weaker and more sensitive to specification choice. Overall, the findings indicate that ESG does not operate through a uniform risk channel in banking and should be interpreted as associational rather than causal. The study contributes evidence from an emerging-market banking system and highlights the importance of disaggregated ESG analysis.

1. Introduction

Banks, due to their central position within the financial system, are key actors that significantly influence the sustainability orientation of the real economy through their lending and investment decisions. Consequently, environmental, social, and governance (ESG) factors in the banking sector have moved beyond being issues confined to corporate social responsibility or voluntary disclosure practices; instead, they have evolved into strategic variables directly linked to risk management, credit allocation processes, and balance sheet structures. From this perspective, ESG practices are no longer merely instruments of reputation management or regulatory compliance for banks, but have increasingly gained importance as integrated components of risk management that shape banks’ risk-taking behavior and overall financial stability (Jeucken, 2001; Gangi et al., 2019; Chiaramonte et al., 2022).

Indeed, it is reported that banks with higher ESG engagement experience more limited increases in credit and asset risks during crisis periods, while facing smaller declines in profitability. Moreover, these banks are found to encounter less contraction in market-based funding, allowing them to reduce their balance sheets to a more limited extent during crises and thereby allocate resources more efficiently (Danisman & Tarazi, 2024). These findings highlight the importance of ESG practices for stakeholders and, in particular, demonstrate that social and governance performance plays a complementary role both in mitigating risks and in facilitating the institutional adoption of environmental practices (Buallay, 2019; Chollet & Sandwidi, 2018). Consequently, ESG performance can also be viewed as a “resilience component” that contributes to the preservation of market confidence and financial stability during periods of crisis (Yuen et al., 2022). These developments have also been reinforced by international regulations, which have placed ESG on an institutional footing within the banking sector.

The EU Taxonomy provides a comprehensive classification system aimed at directing financial resources toward environmentally sustainable economic activities. In line with the European Green Deal and the 2050 net-zero emissions target, the promotion of investments consistent with this objective increasingly requires ESG factors to be incorporated more systematically into credit and investment decisions (European Commission, 2025). In parallel, the Non-Financial Reporting Directive (NFRD) and its successor, the Corporate Sustainability Reporting Directive (CSRD), have made banks’ ESG risk management processes more visible within a standardized reporting framework. This enhanced transparency goes beyond merely increasing the level of disclosure; by encouraging the integration of ESG risks into measurement, monitoring, and management processes, it facilitates a clearer linkage with core prudential areas such as capital adequacy, credit risk, and liquidity management (Esposito et al., 2021; Neisen et al., 2021; Intonti et al., 2022). From this perspective, ESG should not be regarded by banks solely as a matter of regulatory “compliance,” but rather as part of a broader risk management agenda encompassing regulatory risk, balance sheet risk, and reputational risk.

Beyond these regulations, the post Paris Agreement period demonstrates that ESG-oriented banking practices have been shaped not only by internal institutional preferences but also by international policy dynamics. During this period, financial institutions increasingly began to account more explicitly for environmental impacts in their credit and investment decisions, leading to a gradual reduction in the financing of environmentally harmful projects. The responsible investment approach likewise emphasizes the assessment of ESG factors as a means of maintaining an appropriate risk–return balance. While the integration of ESG principles into banks’ strategic frameworks supports the development of a sustainable financial system, inadequate management of these principles may generate adverse consequences in terms of both risk exposure and corporate reputation (Galletta et al., 2023).

Within this framework, banks occupy a “dual role” in the ESG domain. On the one hand, through their own ESG disclosures, banks provide the market with information regarding their institutional quality and risk management standards; on the other hand, as creditors and investors, they use ESG-related information to assess the comprehensive risk profiles of borrowers and investee firms (Gyura, 2020). Indeed, Tóth et al. (2021), drawing on signaling theory, argue that ESG disclosures reduce information asymmetry and convey stronger signals to stakeholders about banks’ institutional quality and risk management practices. Empirical evidence suggests that ESG performance can enhance financial stability, particularly by improving banks’ risk profiles through the channel of reduced non-performing loan (NPL) ratios. There is also evidence indicating that ESG performance may influence bank valuation (Di Tommaso & Thornton, 2020). Accordingly, ESG should not be viewed merely as a reporting exercise for banks; rather, it constitutes a strategic decision input that shapes capital allocation and credit policy choices.

The impact of ESG on banking risks can be understood more clearly when the distinct transmission channels of each of its three sub-components are taken into account. The environmental dimension (E) may affect banks’ balance sheet structures and capital requirements through climate-related risks, environmental regulations, and green investments. In this context, environmental performance is often associated with short-term capital pressures, particularly due to transition risks; however, it is also emphasized that, in the longer term, it may enhance resilience by transforming banks’ overall risk profiles (D’Orazio & Popoyan, 2020; Alessi et al., 2024). These short-term pressures may reflect the adjustment costs of ESG integration, including compliance investments, staff training, operational restructuring, and the gradual incorporation of sustainability criteria into lending and risk-management processes (Aldousari & Alsabah, 2025). The social dimension (S) influences operational costs and customer behavior through areas such as employment policies, customer relations, corporate reputation, and financial inclusion, with these effects potentially transmitting to liquidity dynamics via deposit stability (Buallay, 2019; Chiaramonte et al., 2022; Jaiwani & Gopalkrishnan, 2025). The governance dimension (G), in turn, directly shapes banks’ risk-taking behavior and capital management through board structures, internal control mechanisms, and transparency (Chollet & Sandwidi, 2018; Yuen et al., 2022). Accordingly, banks with stronger ESG performance are expected to exhibit more manageable risk levels as a result of adopting more prudent and sustainable banking practices.

Although ESG investments are reported to strengthen reputation and customer trust—thereby supporting market performance—they may also generate cost pressures in the short run (Buallay, 2019). With respect to liquidity, strong ESG performance has been found to reduce risks and improve risk management practices (J. Liu & Xie, 2024); however, some studies argue that ESG factors may also exert adverse effects on financial, operational, liquidity, and funding risks (Menicucci, 2025). Accordingly, the evidence in the literature is not unidirectional. It is suggested that ESG implementation may increase operational costs and lead to restructuring within loan portfolios in the short term, while, in the longer term, ESG integration functions as a strategic mechanism that enhances financial resilience (Cornett et al., 2016; Galletta & Mazzù, 2023; Magazzino et al., 2025). For this reason, beyond the question of the “overall” effect of ESG, increasing attention is being paid to identifying which ESG component (E, S, or G) affects which risk indicators more prominently and over which time horizons.

The literature has generally evolved around two main streams: studies examining the relationship between ESG and bank performance (Bătae et al., 2020; Menicucci & Paolucci, 2023; El Khoury et al., 2023; Kouzez et al., 2024; Lamanda & Tamásné Vőneki, 2024; Yudaruddin et al., 2025), and research focusing on the effects of ESG on individual risk indicators—such as liquidity risk, non-performing loans (NPLs), systemic risk, overall bank risk, and stability/risk measures (Di Tommaso & Thornton, 2020; Aevoae et al., 2023; J. Liu & Xie, 2024; Cantero-Saiz et al., 2024; Baek & Kang, 2025).

In contrast, comprehensive studies that jointly assess the simultaneous effects of ESG performance on banks’ core balance sheet variables and risk indicators within a single analytical framework remain relatively limited. Accordingly, it is widely recognized that the ESG–risk literature in the banking context is still evolving (Citterio & King, 2023). Moreover, empirical evidence on the ESG–bank risk nexus varies across countries, time periods, and types of risk, indicating that the relationship is inherently context-dependent. Therefore, there is a clear need in the literature for approaches that jointly evaluate banks’ balance sheet dynamics and key risk indicators within a unified analytical framework.

This study aims to address this gap by simultaneously examining the effects of ESG performance on non-performing loans (NPL), capital adequacy (CAR), liquidity risk, and net balance sheet position to equity (NBSP/E) for banks listed on Borsa Istanbul in Türkiye. Türkiye provides a particularly suitable setting for analyzing the ESG–risk nexus due to its characteristics as an emerging market and its bank-centered financial system. The analysis is based on an unbalanced panel dataset covering the period 2008–2023 and comprising eight banks for which ESG data are available.

The main limitation of the study is the relatively small sample size, stemming from the limited long-term availability of ESG data for a restricted number of banks. To mitigate the potential impact of this limitation, bank fixed effects and Driscoll–Kraay standard errors are employed. The primary objective of the study is to empirically assess whether ESG performance functions as an institutional risk management component that supports financial stability in the banking sector by influencing banks’ credit quality, capital resilience, and balance sheet risk indicators.

The paper is organized into four sections. The first section presents the aim and scope of the study. The second section reviews the theoretical framework and the relevant literature. The third section reports the empirical analysis and results. The final section discusses the findings in light of the existing literature and offers conclusions and policy implications.

2. Literature Review

In the global financial system, environmental, social, and governance (ESG) factors are no longer a concern confined to the domain of corporate social responsibility for banks; rather, they have become a strategic element embedded in decision-making processes, influencing banks’ risk management approaches, funding conditions, and balance sheet composition (Gangwani & Kashiramka, 2024; Wu et al., 2024; Magazzino et al., 2025; Zournatzidou et al., 2025). At the same time, the ESG–risk relationship is not unidirectional nor uniform across contexts. Existing evidence highlights that this relationship may display heterogeneous and, at times, non-linear patterns depending on factors such as the regulatory framework, banks’ business models, profitability pressures, and ownership structures (Azmi et al., 2021; Gaies & Jahmane, 2022; Jaiwani & Gopalkrishnan, 2025; Magazzino et al., 2025).

Accordingly, structuring the literature around specific risk indicators to elucidate how ESG performance translates into core banking risk channels is useful both for enhancing the comparability of empirical findings and for clarifying the analytical framework of the study.

This section is structured into six subsections in line with the dependent variables of the study: (i) ESG and non-performing loans (NPL), (ii) ESG and liquidity risk, (iii) ESG and capital adequacy (CAR), (iv) ESG and net balance sheet position to equity (NBSP/E), (v) an overall assessment of the literature and the identified research gap, and (vi) the institutional structure of the Turkish banking system and the rationale for country selection.

2.1. ESG and Non-Performing Loans (NPL)

Non-performing loans (NPLs) are a core indicator of banks’ asset quality and realized credit risk. The literature on the relationship between ESG performance and NPLs largely converges on the argument that ESG practices improve borrower screening and credit monitoring processes, thereby reducing information asymmetry and lowering default probabilities through this channel (S. Liu et al., 2023; J. Liu & Xie, 2024; P. M. U. Nguyen & Pham, 2025).

A substantial body of cross-country panel studies documents a negative relationship between ESG scores and NPL ratios. For U.S. banks, S. Liu et al. (2023), using OLS and 2SLS estimations, report that ESG ratings are inversely related to NPLs, with the combined contribution of the E, S, and G pillars supporting credit quality. In the European context, Tóth et al. (2021), employing Panel VAR and panel regression techniques on a sample of 243 banks from the EU and EFTA, find that ESG performance significantly reduces NPLs, while Di Tommaso and Thornton (2020), using system GMM for 19 European countries, show that higher ESG scores are associated with lower risk-taking, including lower NPLs. Similarly, using an OECD sample, Maztoul (2025) confirms that banks with higher ESG scores exhibit lower NPL ratios, based on recursive modeling and structural equation modeling (SEM) approaches. In emerging markets, the evidence appears even stronger, suggesting that ESG disclosures enhance transparency and thereby mitigate credit risk (Gangwani & Kashiramka, 2024; Hussain et al., 2024; Gissay & Majid, 2025).

When ESG is decomposed into its sub-dimensions (E, S, and G), the results become more heterogeneous. For Vietnamese banks, P. M. U. Nguyen and Pham (2025) find that overall ESG performance, as well as the environmental (E) and governance (G) pillars, exert a mitigating effect on NPLs. In contrast, evidence from ASEAN countries indicates that environmental and governance performance reduce NPLs, whereas social (S) performance may be associated with higher NPL ratios (Ananta & Anwar, 2025).

At the same time, the findings are not uniformly one-directional, and differences in measurement and institutional context can alter the observed relationship. Bruno et al. (2024), analyzing listed European banks using fractional logit and instrumental variable approaches, show that, in some specifications, higher ESG scores may coincide with higher NPLs, potentially reflecting measurement issues or the short-term cost burden of ESG investments. Moreover, within the framework of the “skimping hypothesis,” which suggests that banks under profitability pressure may curtail credit monitoring resources, threshold-type behavior has been reported whereby NPLs may increase despite high ESG scores (Cantero-Saiz et al., 2024). Overall, while the bulk of the literature indicates that ESG performance tends to reduce NPL ratios, both the direction and magnitude of this relationship appear to be shaped by ESG measurement methodologies, bank profitability and business models, and country-specific institutional environments (Tóth et al., 2021; S. Liu et al., 2023; Maztoul, 2025; Bruno et al., 2024; Cantero-Saiz et al., 2024).

2.2. ESG and Liquidity Risk

Liquidity risk reflects banks’ capacity to meet maturing obligations and to sustain market-based funding that relies on confidence. In the literature, ESG performance is commonly viewed as a factor that enhances resilience to liquidity shocks by improving funding conditions through greater corporate transparency and reputational capital (Zhao et al., 2021; Gangwani & Kashiramka, 2024; J. Liu & Xie, 2024). The impact of ESG on liquidity appears to be more pronounced in emerging markets. Gangwani and Kashiramka (2024), analyzing 178 banks across 20 emerging economies using system GMM, find that higher ESG disclosure scores significantly reduce liquidity risk. Evidence from China similarly indicates that ESG performance lowers liquidity risk both directly and indirectly, through channels such as reduced NPLs and improved financial performance (J. Liu & Xie, 2024; Tang & Liu, 2024). Yang (2024) further shows that ESG investments improve banks’ net stable funding ratio (NSFR) over the medium and long term.

Within the liquidity channel, the effects of ESG sub-dimensions are most often interpreted through environmental reputation and governance quality. U.S. evidence suggests that banks with weak environmental reputations operating in regions exposed to high climate risk experience deposit outflows and a diminished capacity to generate liquidity, highlighting the critical role of the environmental (E) pillar for funding stability (Choi et al., 2023). Moreover, by reducing information asymmetry and strengthening banks’ reputational standing, ESG activities can facilitate access to funding and lower funding costs (Zhao et al., 2021; Gupta & Kashiramka, 2024).

Climate-related shocks further underscore the role of ESG—particularly the environmental dimension—as a “risk-mitigating mechanism” in the liquidity context. Mukharjee and Barai (2025) show that temperature and precipitation anomalies in Indian commercial banks weaken holdings of high-quality liquid assets (HQLA), thereby increasing liquidity risk, while strong ESG integration enhances banks’ capacity to manage such shocks. These findings are consistent with the broader banking literature, emphasizing the joint interaction of liquidity and credit risks, suggesting that ESG’s influence on risk is not confined to a single channel but may operate through multiple, interconnected mechanisms (Bandyopadhyay & Saxena, 2023; Danisman & Tarazi, 2024).

Overall, the literature supports the view that ESG performance has the potential to reduce liquidity risk; however, the magnitude of this effect appears to vary with country-specific risks, market discipline, climate vulnerability, and banks’ business models (Di Tommaso & Thornton, 2020; J. Liu & Xie, 2024; Mukharjee & Barai, 2025).

2.3. ESG and Capital Adequacy (CAR/SYR)

The capital adequacy ratio (CAR) represents the equity buffer that banks hold against their risk-weighted assets and constitutes a core prudential metric of resilience to shocks. The literature on the impact of ESG on CAR suggests that ESG performance can strengthen risk management practices, thereby affecting risk weights and the cost of capital and, through these channels, supporting capital buffers (Wu et al., 2024; Lippai-Makra et al., 2021; Magazzino et al., 2025). In a large cross-country study, Wu et al. (2024) examine 583 banks from 57 countries and report that higher ESG performance significantly increases both CAR and Tier 1 ratios, with the underlying mechanism operating through reduced information asymmetry and a lower cost of equity. Focusing on European banks, Lippai-Makra et al. (2021) find that improvements in ESG scores are positively associated with capital adequacy, particularly among mid-sized banks.

At the sub-component level, governance quality is argued to support CAR through more effective management of risk-weighted assets, stronger internal controls, and well-functioning risk committees, while environmental responsibility may influence capital dynamics by shaping banks’ risk profiles and regulatory compliance costs (Gangi et al., 2025; Neitzert & Petras, 2022). Gangi et al. (2025) show that board gender diversity and environmental responsibility can increase Tier 1 and total capital ratios while reducing risk-weighted assets. Recent studies further indicate that specific ESG indicators may serve as powerful predictors of capital-to-asset ratios, suggesting that the ESG–CAR relationship may entail not only correlation but also predictive content (Magazzino et al., 2025; N. B. Nguyen, 2025).

Overall, the literature generally points to a positive role of ESG in enhancing capital adequacy; however, the magnitude and significance of this relationship appear to be sensitive to country-specific regulatory frameworks, bank segments, and differences in data and measurement approaches (Adelopo et al., 2022; Wu et al., 2024; Magazzino et al., 2025).

2.4. ESG and Net Balance Sheet Position/Equity (NBSP/E)

NBSP/E reflects banks’ asset–liability composition and the efficiency of resource allocation and represents a strategic performance and risk domain that can be monitored through indicators such as net interest margin (NIM), funding costs, and on balance sheet positions (Azmi et al., 2021; Menicucci & Paolucci, 2023; Jaiwani & Gopalkrishnan, 2025). The impact of ESG performance on this domain is commonly discussed through three main channels: (i) lower funding costs, (ii) improved asset quality, and (iii) a more stable income structure (Azmi et al., 2021; Abdullah et al., 2023; N. B. Nguyen, 2025).

Using a sample of 251 banks from emerging economies, Azmi et al. (2021) report, based on System GMM estimates, a positive relationship between ESG activities and net interest margins, arguing that ESG facilitates the attraction of lower-cost deposits through enhanced transparency and reputational capital. Abdullah et al. (2023) find that ESG practices can support operational performance indicators even during crisis periods in emerging Asian economies. Focusing on 153 banks across 12 emerging countries, N. B. Nguyen (2025) shows that ESG performance has predictive power for NIM and bank size, suggesting that ESG can influence balance sheet efficiency. In addition, strong ESG performance, when combined with interest-income-based intermediation activities, has been found to improve asset quality and reduce funding costs (Baek & Kang, 2025).

Country-specific evidence further indicates that ESG may steer balance sheet structures toward more sustainable and lower-risk assets. For Italy, Menicucci and Paolucci (2023) document that emission- and waste-reduction policies (the environmental component) are positively associated with operational efficiency and balance sheet profitability. In the Indian context, Jaiwani and Gopalkrishnan (2025) identify heterogeneity linked to ownership structure, showing that environmental scores are positively related to balance sheet and profitability indicators in public banks, whereas social scores may be negatively associated with certain outcomes in private banks. These differences suggest that the effect of ESG on balance sheet positions is not uniform but varies with bank type and governance structures.

2.5. Overall Assessment, Synthesis of Findings, and the Research Gap in the Literature

The synthesis of the reviewed studies indicates that ESG performance can influence banking risk indicators through multiple, mutually interacting channels. With respect to asset quality, the majority of the literature reports a negative relationship between ESG performance and non-performing loans (NPLs) (Tóth et al., 2021; S. Liu et al., 2023; Maztoul, 2025). However, several studies show that this relationship may weaken or even reverse due to measurement differences or behavioral mechanisms such as “skimping” under profitability pressure, whereby banks reduce monitoring efforts despite high ESG scores (Bruno et al., 2024; Cantero-Saiz et al., 2024). In the liquidity dimension, ESG performance is generally found to support funding stability and mitigate liquidity shocks, while climate-related shocks tend to erode liquidity buffers particularly in banks with limited ESG integration (Zhao et al., 2021; Choi et al., 2023; Mukharjee & Barai, 2025). Regarding capital adequacy, growing evidence suggests that ESG performance can strengthen capital buffers by reducing information asymmetry, lowering the cost of capital, and improving risk management practices, thereby supporting capital adequacy ratios (CAR) (Wu et al., 2024; Lippai-Makra et al., 2021; Magazzino et al., 2025).

Despite these insights, two notable gaps remain in the literature. First, many studies examine risk channels such as NPLs, liquidity risk, and capital adequacy in isolation, providing only limited evidence on their simultaneous interaction and the implications for overall balance sheet composition (Aevoae et al., 2023; Lupu et al., 2022; Danisman & Tarazi, 2024). Second, existing empirical evidence is heavily concentrated on markets with high data transparency, such as the United States, the European Union, and China, while ESG–risk dynamics in emerging, bank-dominated financial systems remain comparatively underexplored (Abdullah et al., 2023; Gangwani & Kashiramka, 2024).

Against this backdrop, the present study aims to contribute to the literature by analyzing the effects of ESG scores on NPLs, capital adequacy (CAR), liquidity risk, and the net balance sheet position relative to equity (NBSP/E) within a single, integrated analytical framework. By jointly assessing these core balance sheet and risk indicators, the study seeks to move beyond the fragmented evidence in prior research and provide a more holistic understanding of the ESG–risk nexus.

Overall, while the literature documents meaningful relationships between ESG performance and individual banking risk indicators, empirical studies that address these risk channels jointly and in a comprehensive manner remain scarce. This limitation is particularly salient for emerging economies with bank-based financial systems, such as Türkiye, highlighting a clear and policy-relevant research gap that this study seeks to address.

2.6. Institutional Structure of the Turkish Banking System and the Rationale for Country Selection

The primary reason for selecting Türkiye as the focus of this study is the country’s distinctly bank-oriented financial structure. As of December 2024, the number the number of banks in operation was 67, comprising 37 deposit banks, 20 development and investment banks, and 9 participation (Islamic) banks (BAT, 2025). Within the deposit banking segment, ownership is diversified: there were 4 state-owned, 11 privately-owned, 22 foreign-owned, and 1 Savings Deposit Insurance Fund deposit bank at end-2024 (BAT, 2025).

The numerical dominance of deposit banks indicates that the backbone of the financial system continues to operate primarily along the deposit-taking and lending axis, implying that the ESG–risk relationship can be meaningfully examined through core transmission channels such as credit risk (NPLs) and liquidity risk. In parallel, group-level market shares confirm the bank-based structure: as of 2024, deposit banks accounted for 86% of total banking sector assets, while the shares of development and investment banks and participation banks were 6% and 8%, respectively (BAT, 2025). Moreover, within deposit banks, asset shares were 38% (state-owned), 28% (private), and 20% (foreign-owned), underscoring the economically material role of ownership structure in the Turkish banking landscape (BAT, 2025).

The capital adequacy ratio stood at 19.7% (standard ratio) and 15.6% (core ratio), while the NPL ratio was 1.8% at year-end (BAT, 2024). These institutional and prudential features help contextualize the mechanisms through which ESG-related exposures may propagate into risk outcomes in a bank-dominated financial architecture.

Beyond this structural background, regulatory and sector-level initiatives supporting the banking sector’s transition toward sustainability have gradually strengthened over the past decade in Türkiye. In this context, the BRSA issued guiding documents on environmental and social risk management as early as 2014, while the Green Deal Action Plan, introduced in 2021 in alignment with EU policy orientations, reinforced the framework at the level of public policy. In subsequent years, the Sustainable Banking Strategic Plan published by the Turkish Banks Association (TBA) in 2022 enhanced sectoral coordination, while banks accelerated their compliance efforts in line with evolving global reporting expectations shaped by standards such as the CSRD and the ISSB’s IFRS S2 (2023–2024).

Moreover, the BRSA’s Guideline on Credit Allocation and Monitoring Processes explicitly emphasizes the incorporation of ESG factors into banks’ credit risk appetite and lending processes, highlighting the use of tools such as heat maps to identify exposures (Banking Regulation and Supervision Agency (BRSA), 2021). This guidance translates ESG integration into a concrete and operational dimension of credit processes.

Overall, the sustainable finance and ESG agenda in Türkiye extends beyond mere policy declarations, pointing instead to a framework that is increasingly institutionalized within the sector and supported by both regulatory and industry-level components. This evolution is consistent with Türkiye’s progression from the “Emerging” stage to the “Advancing” level in the Sustainable Banking and Finance Network (SBFN) matrix (IFC, 2022; BAT, 2023).

From a representativeness standpoint, the eight listed banks analyzed in this study are economically central: over 2008–2023, they represent, on average, 63.8% of total banking sector assets (min–max: 61.2–66.1%), supporting the external relevance of the empirical findings for the core of the Turkish banking system.

The findings derived from the Turkish case may therefore contribute to a broader understanding of the role of ESG integration in bank risk management across other emerging economies characterized by similar institutional structures and macro-financial volatility, thereby offering insights with relevance and potential generalizability for the international literature.

3. Materials and Methods

3.1. Sample and Data

The sample of this study is constructed from banks listed on Borsa İstanbul (BIST) in Türkiye for which ESG_Combined_Score, the E, S, and G sub-component scores, and the ESG_Controversies_Score are reported in the LSEG Refinitiv Workspace database over the 2008–2023 period. Financial ratios and bank-specific balance sheet variables are obtained from the Banks Association of Türkiye (BAT) database, while macroeconomic control variables—inflation rates and economic growth rates—are sourced from the Central Bank of the Republic of Türkiye (CBRT) (2025) and the World Bank (2025), respectively.

As a measure of sustainability performance, a comprehensive screening of the Refinitiv database indicates that ESG data are available for nine publicly listed banks, one of which operates as a participation (Islamic) bank. Participation banks differ structurally from conventional (interest-based) deposit banks due to their distinct funding and income-generation mechanisms based on profit-and-loss sharing principles. These differences may lead to systematic inconsistencies in balance sheet composition and in the definition and interpretation of financial ratios. Moreover, some control variables employed in this study—such as interest income/expense ratios—are inherently incompatible with participation banking, which would compromise cross-bank comparability. To enhance sample consistency and measurement homogeneity, the participation bank is therefore excluded from the analysis. This exclusion reflects a business model comparability and measurement consistency criterion, rather than any consideration related to bank performance or risk levels.

Following the LSEG ESG scoring methodology, which is based on letter-grade percentile ranges, each ESG score is converted into a 0–100 scale by taking the midpoint of the corresponding letter-grade interval and multiplying it by 100 (LSEG, 2024).

Accordingly, the final sample consists of eight banks: Akbank T.A.Ş., Türkiye Garanti Bankası A.Ş., Türkiye İşBankası A.Ş., QNB Finansbank A.Ş., Yapı ve Kredi Bankası A.Ş., Türkiye Halk Bankası A.Ş., Türkiye Vakıflar Bankası T.A.O., and Türkiye Sınai Kalkınma Bankası A.Ş. (TSKB). Due to the absence of ESG observations for certain banks in specific years, the dataset has an unbalanced panel structure. In addition, since independent and control variables are included in the model with a one-year lag (t − 1), the number of observations used in the regressions is lower than the raw bank–year observations.

Although TSKB operates as a development and investment bank and thus differs institutionally from deposit banks, such structural differences are time-invariant and are controlled through bank fixed effects. Consequently, the estimated coefficients primarily reflect within-bank time variation rather than cross-sectional heterogeneity.

To quantify the representativeness of the sample, the annual shares of the sample banks in sector-wide totals are calculated (Appendix A 

Table A1. The share of sample banks in sector totals (2008–2023).

Indicator	Mean (%)	Min (%)	Max (%)	2008 (%)	2023 (%)
Total Asset Share	63.8	61.2	66.1	63.1	66.1
Total Loan Share	63.9	61.1	66.5	65.6	66.5
Total Deposit Share	65.1	62.4	68.9	63.4	68.9
Total Equity Share	61.9	57.7	66.4	57.7	66.2

Notes: Calculation: Made by taking the sector total in the “Total” row of the TBB annual tables.

). Sector totals are derived from the “Total” row in the BAT’s annual tables ranking banks by total assets. Although the sample includes only eight banks, these institutions account for a substantial share of the Turkish banking sector’s assets, loans, deposits, and equity, implying that the analysis captures a large economically weighted segment of the listed banking population. Over the 2008–2023 period, the sample banks account on average for 63.8% of total sector assets (min–max: 61.2–66.1%), 63.9% of total loans (min–max: 61.1–66.5%), 65.1% of total deposits (min–max: 62.4–68.9%), and 61.9% of total equity (min–max: 57.7–66.4%). At the beginning of the sample period (2008), the representation ratios for assets, loans, deposits, and equity are 63.1%, 65.6%, 63.4%, and 57.7%, respectively, increasing to 66.1%, 66.5%, 68.9%, and 66.2% by the end of the period (2023).

These figures indicate that, despite the relatively limited number of banks, the sample possesses substantial economic weight, and the empirical findings are driven by large-scale banks located at the core of the Turkish financial system.

3.2. Model

The analysis is conducted using panel data techniques, which are widely employed in the related empirical literature. All estimations are implemented in Python (version 3.12.13) using Google Colab. To mitigate potential simultaneity concerns and reduce reverse causality between ESG indicators and bank risk measures, all independent and control variables are included in the model with a one-year lag (t − 1). We include a COVID-19 dummy to capture the pandemic shock. Because crisis episodes can materially alter banks’ risk environment and lending behavior, we control for the COVID-19 period, consistent with evidence that banks operating under an expected credit loss framework tightened lending and reduced credit supply after the onset of COVID-19 (Chen & Huang, 2026).

$$Y_{i,t}={\alpha }_i+{\lambda }_t+\beta X_{i,t-1}+\gamma Z_{i,t-1}+\delta Covid\_dumm{y}_t+{\epsilon }_{i,t}$$

where:

• $i$ denotes the bank;
• $t$ denotes the year;
• $Y_{i,t}$ represents one of the four dependent variables (NPL, CAR, Liquidity_Risk, NBSP/E);
• $X_{i,t-1}$ denotes ESG indicators (ESG_Combined_Score, E_Score, S_Score, G_Score);
• $Z_{i,t-1}$ represents the set of control variables (Asset_Log, ROA, Equity_Ratio, InterestIER, ESG_Controversies_Score, Inflation, RGDP);
• ${\alpha }_i$ captures unobserved bank-specific fixed effects;
• ${\lambda }_t$ denotes the year’s fixed effects;
• $Covid\_dummy$ is a dummy variable equal to 1 for the year 2020 and 0 otherwise;
• ${\epsilon }_{i,t}$ is the error term.

In line with the objectives of the study, two empirical models are specified. Model A is designed to examine the effect of the aggregate ESG score, while Model B investigates the disaggregated effects of the environmental (E), social (S), and governance (G) scores on the four dependent variables. To avoid design-based multicollinearity, the composite ESG score and the E, S, and G pillar scores are not entered in the same regression. Model A uses the aggregate ESG score, whereas Model B replaces the composite measure with the disaggregated ESG pillars.

To determine the appropriate panel data specification, both fixed effects (FE) and random effects (RE) estimations were considered. The comparison between FE and RE models indicates the presence of a significant correlation between bank-specific effects and the explanatory variables. Moreover, the FE model provides higher explanatory power (Within $R^2=0.396$), and the main explanatory variable, ESG_Combined_Score_L1, is statistically significant only under the FE specification ($p=0.0175$). In addition, the poolability test (F = 5.11, $p=0.0001$) rejects the hypothesis of homogeneous slope coefficients across banks. Accordingly, all models are estimated using bank fixed effects. The fixed-effects framework is used to identify within-bank associations over time and should not be interpreted as establishing causal effects.

Given that banks may be simultaneously exposed to common macroeconomic and regulatory shocks, cross-sectional dependence is tested using the Pesaran (2004) CD test (see Appendix A 

Table A2. Cross-Sectional Dependence Tests (Pesaran CD).

Model	Dependent	CD_Stat	p_Value	Pairs_Used	N_Entities
A	NPL	0.976	0.329	28	8
A	CAR	0.573	0.567	28	8
A	Liquidity_Risk	0.027	0.979	28	8
A	NBSP/E	0.202	0.84	28	8
B	NPL	1.496	0.135	21	8
B	CAR	1.672	0.095	21	8
B	Liquidity_Risk	−0.31	0.756	21	8
B	NBSP/E	−0.712	0.476	21	8

). The test is applied to the residuals of the fixed-effects panel regressions for both Model A and Model B. For most specifications, the null hypothesis of cross-sectional independence cannot be rejected, suggesting that cross-sectional dependence is not a major concern in the data. Nevertheless, to ensure robust inference under potential weak dependence, Driscoll–Kraay standard errors are employed in all estimations.

Bank size is proxied by the logarithm of total assets, while ROA, Equity_Ratio, InterestIER, ESG_Controversies_Score, Inflation, RGDP, and a COVID-19 dummy for the year 2020 are included as control variables in the regression models.

3.3. Dependent Variables

To analyze the impact of ESG performance on banking risks within a multidimensional framework, this study employs four key indicators as dependent variables, each representing a distinct risk channel. These variables are selected to capture different dimensions of bank risk, allowing for a comprehensive assessment of how ESG performance is reflected across credit quality, capital adequacy, liquidity conditions, and balance sheet structure.

The selected dependent variables, together with the risk dimensions they represent and the rationale for their inclusion, are summarized in

Table 1. Dependent Variables Used in the Study, Their Definitions and Justifications for Selection.

Variable	Definitions	Dimension of Risk Represented	Justifications for Selection
NPL	The ratio of non-performing loans	Credit risk and asset quality	It is a fundamental indicator that directly reflects banks’ credit portfolio quality and risk-taking behavior (S. Liu et al., 2023; Maztoul, 2025). Accordingly, it is widely employed in the literature to assess the impact of ESG performance on credit allocation processes, borrower selection, and loan monitoring mechanisms (S. Liu et al., 2023; Ananta & Anwar, 2025; Maztoul, 2025; P. M. U. Nguyen & Pham, 2025).
Liquidity Risk	An indicator reflecting banks’ short-term liquidity position	Liquidity risk and funding flexibility	It measures banks’ capacity to meet their short-term obligations and reflects their exposure to liquidity pressures (J. Liu & Xie, 2024; Yang, 2024). This variable is included in the analysis based on the premise that ESG performance may indirectly influence funding conditions and liquidity stress through trust, transparency, and reputational channels (Zhao et al., 2021; Choi et al., 2023; Gangwani & Kashiramka, 2024; J. Liu & Xie, 2024).
CAR	Capital adequacy ratio	Capital resilience and risk resistance	The capital adequacy ratio represents the most fundamental regulatory indicator of banks’ resilience to unexpected losses and overall balance sheet strength (Tóth et al., 2021; Gangi et al., 2025). It is selected to test the hypothesis that stronger ESG performance is associated with enhanced risk management quality and, consequently, more robust capital buffers (Gangi et al., 2025; Magazzino et al., 2025; N. B. Nguyen, 2025).
Net Balance Sheet Position (NBSP/E)	Ratio of the difference between balance sheet assets and liabilities to equity	Balance sheet structural risk and total risk level	This indicator provides a comprehensive measure of banks’ balance sheet composition and structural risk profile (Baek & Kang, 2025). It allows for an assessment of the impact of ESG performance on banks’ balance sheet structure and risk-taking behavior without restricting the analysis to a single financial ratio (Azmi et al., 2021; Abdullah et al., 2023).

.

In the existing literature, the impact of ESG performance on bank stability is predominantly examined using aggregate indicators such as the Z-score or profitability ratios (ROA/ROE) (Abdullah et al., 2023; Cantero-Saiz et al., 2024; Choi et al., 2023; Magazzino et al., 2025). However, these measures may sometimes be insufficient to capture micro-level changes in banks’ operational processes and risk management practices (Abdullah et al., 2023).

In this context, the inclusion of NBSP/E (or alternatively the Net Interest Margin, NIM) offers a more granular and original perspective by linking ESG transparency to the efficiency of interest management between deposits and loans, which constitutes banks’ core intermediation function (Azmi et al., 2021; Abdullah et al., 2023). While prior studies document that ESG performance can reduce funding costs and improve credit quality, the direct balance sheet implications arising from the interaction of these two channels remain underexplored in the literature (Azmi et al., 2021; Cantero-Saiz et al., 2024; Baek & Kang, 2025).

3.4. Independent Variables

ESG performance is measured using a composite ESG score that jointly captures the environmental, social, and governance dimensions. This composite indicator reflects the overall level of banks’ sustainability practices and is widely employed in the literature examining the relationship between ESG performance and banking risks. In addition, the environmental (E), social (S), and governance (G) components of the ESG score are incorporated into the model to investigate the disaggregated effects of different ESG dimensions on banking risk outcomes.

3.5. Control Variables

In order to estimate more accurately the relationship between ESG performance and bank risk indicators, this study incorporates bank-specific control variables that are widely recognized in the literature as key determinants of banking risk and balance sheet behavior. Accordingly, indicators capturing bank size, profitability, financial leverage, and balance sheet structure are included as control variables in the empirical models.

The inclusion of these control variables allows the effects of ESG performance on bank risk measures to be isolated from bank-specific financial characteristics. This approach aims to ensure that the estimated results more reliably and consistently reflect the impacts attributable to ESG performance rather than confounding structural or financial factors.

4. Results

4.1. Descriptive Statistics

Table 2. Descriptive Statistics.

Variable	Mean	Std.Deviation	Minimum	Median	Maximum	Observation
NPL	3.61	1.46	1.27	3.4	7.6	105
CAR	16.81	2.68	13.08	16.14	24.6	105
Liquidity_Risk	21.81	9.08	3.35	21.07	45.18	105
NBSP/E	−42.32	44.65	−229.29	−37.98	39.69	105
ESG_Combined_Score	58.61	16.43	20.83	54.17	87.5	105
E_Score	67.42	25.31	12.5	70.83	95.83	105
S_Score	68.53	21.2	12.5	70.83	95.83	105
G_Score	48.06	17.88	12.5	45.83	79.17	105
Asset_Log	26.29	0.99	24.37	26.18	28.66	105
ROA	1.99	1.22	0.19	1.64	6.73	105
Equity_Ratio	10.82	2.31	4.83	11.31	15.52	105
InterestIER	190.37	35.88	115.7	186.87	300	105
ESG_Controversies_Score	88.85	16.63	29.17	95.83	95.83	105
Inflation	20.56	20.23	6.16	10.45	64.77	105
RGDP	4.96	3.99	−4.8	5.1	11.4	105
Covid_dummy	0.08	0.27	0	0	1	105

presents the descriptive statistics of the variables, including the mean, standard deviation, minimum, median, and maximum values. The ESG scores exhibit relatively high variance, indicating substantial heterogeneity across banks.

Table 3. Correlation Matrix.

Variables	NPL	CAR	Liquidity_Risk	NBSP/E	ESG_Combined	E_Score	S_Score	G_Score	ROA
NPL	1	0.18	−0.27	−0.11	−0.04	−0.17	−0.01	−0.05	−0.22
CAR		1	−0.03	−0.13	0.49	0.36	0.31	0.38	0.59
Liquidity_Risk			1	0.39	−0.34	−0.14	−0.49	0.01	−0.01
NBSP/E				1	−0.42	−0.31	−0.51	0.03	0.04
ESG_Combined_Score					1	0.77	0.84	0.55	0.3
E_Score						1	0.68	0.25	0.21
S_Score							1	0.17	0.15
G_Score								1	0.23
ROA									1

presents the correlation matrix for the variables used in the study. The results indicate statistically significant positive correlations between ESG scores and several financial indicators. Nevertheless, to address potential concerns regarding multicollinearity—particularly among the ESG sub-dimensions—a Variance Inflation Factor (VIF) analysis was conducted. As reported in Appendix A 

Table A3. Variance Inflation Factors (VIF)—Model A and Model B.

Variable	Model A—ESG Combined	Model B—E, S, G Separate
Constant	3207.524	3754.171
ESG_Combined_Score_L1	2.837	—
E_Score_L1	—	2.436
S_Score_L1	—	3.239
G_Score_L1	—	1.351
Asset_Log	3.946	4.729
ROA	4.463	4.768
Equity_Ratio	3.333	3.839
InterestIER	1.896	1.816
ESG_Controversies_Score	1.188	1.184
Inflation	7.851	8.446
RGDP	1.082	1.107

, the VIF values confirm that multicollinearity does not pose a serious concern among the explanatory variables.

4.2. Model A Regression Results

To empirically test the hypotheses developed in the previous section, Model A is first estimated to examine the relationship between the ESG Combined Score and the dependent variables.

Table 4. Fixed-effects panel regression results (Model A: ESG Combined Score)1.

Variables	NPL	CAR	Liquidity_Risk	NBSP/E
ESG Combined Score (t − 1)	0.040 ***	0.096 ***	−0.040	−0.271
	−0.011	−0.027	−0.073	−0.257
Size: log(Assets) (t − 1)	0.298	−0.765	−8.413 ***	−21.550 **
	−0.642	−0.869	−2.042	−9.254
ROA (t − 1)	0.260 **	0.874 ***	−0.315	12.320 **
	−0.13	−0.238	−0.567	−5.539
Equity ratio (t − 1)	0.107	0.099	−1.102 *	−1.07
	−0.143	−0.227	−0.649	−2.625
Interest income/expense ratio (t − 1)	0.018 **	−0.009	−0.042	−0.536 ***
	−0.008	−0.012	−0.03	−0.119
ESG controversies score (t − 1)	−0.002	−0.020 **	−0.010	−0.343 *
	−0.008	−0.008	−0.027	−0.202
Inflation (t − 1)	−0.042 **	−0.004	0.199 **	−0.515
	−0.02	−0.052	−0.078	−0.533
Real GDP growth (t − 1)	−0.047	0.074	0.639 ***	0.127
	−0.055	−0.066	−0.122	−0.778
Covid dummy	2.144 ***	1.644 *	−11.957 ***	−6.042
	−0.526	−0.943	−1.216	−6.399
Constant	−10.362	31.195	261.22 ***	675.83 ***
	−17.881	−22.263	−56.303	−227.29
Observations	97	97	97	97
Banks	8	8	8	8
Years	2008–2023	2008–2023	2008–2023	2008–2023
R-squared (within)	0.396	0.397	0.567	0.680
F-statistic (robust)	207.76	150.32	93.17	291.77

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01.

reports four panel regression models illustrating the association between the ESG Combined Score and the dependent variables—NPL, CAR, Liquidity Risk, and NBP Equity. In addition, the results of the cross-sectional dependence (CD) tests indicate that cross-sectional dependence does not pose a concern for the estimated models.

Table 4 shows that, after controlling for bank fixed effects and including lagged control variables, the ESG_Combined_Score (t − 1) has a positive and statistically significant effect on NPL (p < 0.01) and CAR (p < 0.01). In contrast, the effect of ESG_Combined_Score (t − 1) on Liquidity Risk and NBSP/E is not statistically significant. Standard errors are estimated using the Driscoll–Kraay procedure, and all specifications include bank fixed effects as well as the full set of control variables.

4.3. Model B Regression Results

Table 5. Fixed-effects panel regression results (Model B: E, S and G)2.

Variables	(1) NPL	(2) CAR	(3) Liquidity Risk	(4) NBSP/E
E_Score (t − 1)	0.004(0.586)	−0.044 *** (0.000)	−0.020 (0.731)	0.287 (0.124)
S_Score (t − 1)	0.007 (0.423)	0.012 (0.536)	−0.036 (0.378)	−0.575 ** (0.049)
G_Score (t − 1)	−0.012 * (0.085)	−0.001 (0.963)	−0.006 (0.895)	0.040 (0.782)
Size: log(Assets) (t − 1)	−0.221 (0.728)	1.507 (0.185)	−6.378 (0.159)	−31.628 ** (0.010)
ROA (t − 1)	−0.431 (0.377)	0.589 (0.360)	−0.033 (0.980)	9.539 ** (0.011)
Equity Ratio (t − 1)	0.240 (0.484)	0.060 (0.852)	−0.832 (0.472)	−1.475 (0.395)
Interest Income/Expense (t − 1)	−0.010 (0.457)	−0.015 (0.427)	0.037 (0.550)	−0.198 *** (0.003)
ESG Controversies Score (t − 1)	0.005 (0.444)	0.011 (0.364)	−0.048 * (0.091)	−0.051 (0.690)
Inflation (t − 1)	0.023 (0.592)	0.011 (0.874)	0.092 (0.582)	0.480 (0.267)
Real GDP Growth (t − 1)	−0.114 *** (0.001)	−0.056 (0.452)	−0.050 (0.709)	1.718 (0.155)
COVID-19 Dummy	0.621 (0.496)	0.946 (0.232)	−3.833 (0.319)	2.265 (0.785)
Constant	Yes	Yes	Yes	Yes
Observations	97	97	97	97
Banks	8	8	8	8
Years	2008–2023	2008–2023	2008–2023	2008–2023
R2 (Within)	0.313	0.316	0.471	0.615
F-statistic (robust)	295.67	328.6	1082	469.58

Notes: * p < 0.10, ** p < 0.05, *** p < 0.01.

presents the results of Model B regression analyses, which are conducted to examine the relationship between the ESG subcomponents—Environmental (E), Social (S), and Governance (G) scores—and the dependent variables.

Within the framework of Model B, the effects of lagged Environmental (E), Social (S), and Governance (G) scores on banks’ risk and performance indicators are analyzed using bank fixed effects. For the NPL dependent variable, neither the environmental nor the social score is statistically significant (E: β = 0.004, p = 0.586; S: β = 0.007, p = 0.423), while the governance score exhibits a weakly significant negative relationship (G: β = −0.012, p = 0.085). The results of the CAR model indicate that the environmental score has a statistically significant and negative effect (β = −0.044, p < 0.001). With respect to liquidity risk, none of the ESG subcomponents are statistically significant. In contrast, the NBSP/E model reveals that the social score has a negative and statistically significant effect (β = −0.575, p = 0.049), whereas the environmental and governance scores are not statistically significant. Across all specifications, bank fixed effects and the full set of control variables are included. The number of observations remains constant at 97, covering eight banks.

4.4. Robustness Results

The robustness checks conducted for Model A and Model B (Appendix A 

Table A4. Summary of Robustness Tests—Model A and Model B.

Model	Robustness	NPL	CAR	Liquidity Risk	NBP Equity
Model A (ESG Combined)	Baseline	0.040 ***	0.096 ***	−0.038	−0.271
		−0.011	−0.027	−0.073	−0.257
	R1: No COVID-19	0.049 ***	0.103 ***	−0.091	−0.298
		−0.01	−0.028	−0.084	−0.255
	R2: Balanced (2020–2022)	0.019	0.192 ***	0.135 ***	−3.432 ***
		−0.015	−0.046	−0.017	−0.41
	R3: Winsorized	0.040 ***	0.102 ***	−0.046	−0.237
		−0.012	−0.024	−0.076	−0.246
Model B (E, S and G Pillars)	Baseline	G: −0.012 *	E: −0.044 ***	–	S: −0.575 **
	R1: No COVID-19	G: −0.010	E: −0.051 ***	ESGC: −0.054 **	S: −0.573 **
	R2: Balanced (2020–2022)	G: −0.027 **	E: −0.375 ***	G: 0.132 **	G: 2.647
	R3: Winsorized	G: −0.011 *	E: −0.046 ***	–	S: −0.650 **

Notes: ESGC denotes the ESG Controversies Score (lagged), capturing downside ESG risk and adverse events. It is reported only when statistically significant. Standard errors are reported in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.10.

) comparatively assess the consistency of the effects of ESG indicators on bank risk measures under alternative model specifications.

In the baseline pillar model (Model B), E is negatively related to CAR and S is negatively related to NBSP/E, with G weakly related to NPL.

The robustness checks conducted for Model A and Model B (Appendix A Table A4) comparatively assess the consistency of the effects of ESG indicators on bank risk measures under alternative model specifications. For Model A, the positive and statistically significant effects of the lagged ESG_Combined_Score (ESG_Combined_Score_L1) on NPL and CAR are generally preserved when excluding the COVID-19 dummy, using a balanced panel, and applying winsorization to the sample. In contrast, for Liquidity Risk and NBP Equity, the sign and statistical significance of the ESG_Combined_Score_L1 coefficients vary across specifications, indicating that these results exhibit more limited robustness.

The robustness results for Model B suggest that the effects of the disaggregated ESG dimensions differ across risk channels. Across all robustness tests, the negative and statistically significant effect of the lagged environmental score (E_Score_L1) on CAR remains consistent. Similarly, the negative effect of the lagged social score (S_Score_L1) on NBP Equity retains its statistical significance in both the baseline model and the winsorized sample. By contrast, for NPL and Liquidity Risk, the coefficients of the E, S, and G scores are not statistically significant in most specifications.

In Model B, the disaggregated ESG analysis indicates a weak negative relationship between the governance score (G_Score_L1) and non-performing loans (NPL) in the baseline specification, which is statistically significant only at the 10% level. However, this relationship does not remain statistically significant when excluding the COVID-19 dummy, using balanced panel estimations, or applying winsorization. This finding suggests that the observed governance–NPL relationship is sensitive to sample composition and model specification.

In the balanced panel estimations, coefficient magnitudes and significance levels in both models appear more volatile. Given the relatively short time dimension and limited number of observations, these results are interpreted as supportive rather than decisive evidence for the main findings. When using the winsorized sample, the core results for the composite ESG score in Model A and the key findings for the environmental and social dimensions in Model B are largely preserved.

Overall, the comparative robustness analyses indicate that Model A provides relatively stable evidence on the effects of the composite ESG measure, while Model B highlights differentiated and channel-specific effects of the disaggregated ESG dimensions across bank risk indicators. On balance, the robustness checks provide stronger support for the ESG_Combined–CAR association, the E–CAR association, and the S–NBSP/E association, while the ESG_Combined–Liquidity Risk and G–NPL relationships seem less stable.

To further address concerns related to the limited number of banks in the sample, a leave-one-out (jackknife) analysis is conducted. The consistency summaries reported in

Table A5. Leave-One-Out (Jackknife) Robustness Results for Model A-ESG Combined Score.

Dropped_Bank	Dependent_Variable	Coef_ESG_Combined_L1	p_Value
bank1	NLP	0.051	0.000
bank1	CAR	0.090	0.018
bank1	Liquidity_Risk	−0.130	0.076
bank1	NBP_Equity	−0.419	0.217
bank2	NLP	0.053	0.000
bank2	CAR	0.111	0.000
bank2	Liquidity_Risk	−0.100	0.325
bank2	NBP_Equity	−0.350	0.329
bank3	NLP	0.048	0.000
bank3	CAR	0.100	0.001
bank3	Liquidity_Risk	−0.116	0.090
bank3	NBP_Equity	−0.355	0.272
bank4	NLP	0.047	0.003
bank4	CAR	0.112	0.001
bank4	Liquidity_Risk	−0.064	0.510
bank4	NBP_Equity	−0.268	0.325
bank5	NLP	0.054	0.000
bank5	CAR	0.106	0.005
bank5	Liquidity_Risk	−0.030	0.731
bank5	NBP_Equity	−0.221	0.460
bank6	NLP	0.048	0.000
bank6	CAR	0.102	0.001
bank6	Liquidity_Risk	−0.084	0.309
bank6	NBP_Equity	−0.265	0.258
bank7	NLP	0.050	0.000
bank7	CAR	0.082	0.000
bank7	Liquidity_Risk	−0.101	0.290
bank7	NBP_Equity	−0.465	0.043
bank8	NLP	0.046	0.001
bank8	CAR	0.122	0.000
bank8	Liquidity_Risk	−0.087	0.362
bank8	NBP_Equity	0.035	0.857

Notes: This table reports leave-one-out (jackknife) estimates. In each iteration, one bank is excluded from the sample, and the baseline fixed-effects model is re-estimated using the remaining banks. The reported coefficients correspond to the relevant ESG variables. Robust standard errors are computed using the Driscoll–Kraay method. All control variables are included in the estimations but are not reported for brevity.

and

Table A6. Leave-One-Out (Jackknife) Robustness Results for Model B-ESG Combined Score.

Dropped_Bank	Dependent_Variable	Coef_E_L1	P_E	Coef_S_L1	P_S	Coef_G_L1	P_G
bank1	NLP	−0.012	0.166	0.013	0.265	0.007	0.368
bank1	CAR	−0.048	0.081	−0.013	0.607	−0.003	0.839
bank1	Liquidity_Risk	0.064	0.094	−0.193	0.032	−0.142	0.001
bank1	NBP_Equity	0.053	0.691	−0.246	0.148	0.487	0.000
bank2	NLP	−0.004	0.764	−0.010	0.483	−0.002	0.792
bank2	CAR	−0.057	0.084	−0.004	0.894	0.008	0.480
bank2	Liquidity_Risk	0.103	0.011	0.006	0.914	−0.053	0.337
bank2	NBP_Equity	0.016	0.922	−0.488	0.009	0.291	0.027
bank3	NLP	−0.025	0.003	−0.009	0.229	−0.002	0.742
bank3	CAR	−0.076	0.001	0.030	0.287	−0.021	0.102
bank3	Liquidity_Risk	0.017	0.657	0.069	0.321	−0.065	0.146
bank3	NBP_Equity	0.270	0.089	−0.674	0.002	0.330	0.025
bank4	NLP	−0.011	0.215	−0.024	0.128	−0.004	0.490
bank4	CAR	−0.054	0.068	−0.025	0.379	0.006	0.573
bank4	Liquidity_Risk	0.016	0.557	0.076	0.055	−0.050	0.428
bank4	NBP_Equity	−0.117	0.560	−0.368	0.079	0.486	0.001
bank5	NLP	−0.017	0.194	0.004	0.816	0.001	0.932
bank5	CAR	−0.043	0.063	−0.043	0.376	0.001	0.944
bank5	Liquidity_Risk	0.050	0.296	−0.001	0.985	−0.065	0.196
bank5	NBP_Equity	0.173	0.184	−0.425	0.070	0.401	0.017
bank6	NLP	−0.017	0.088	−0.007	0.546	0.000	0.919
bank6	CAR	−0.057	0.033	−0.007	0.797	0.004	0.747
bank6	Liquidity_Risk	0.056	0.110	−0.003	0.952	−0.073	0.106
bank6	NBP_Equity	0.132	0.264	−0.506	0.005	0.327	0.005
bank7	NLP	−0.018	0.116	−0.011	0.389	−0.001	0.797
bank7	CAR	−0.041	0.068	0.018	0.322	0.008	0.522
bank7	Liquidity_Risk	0.054	0.191	−0.005	0.932	−0.076	0.137
bank7	NBP_Equity	0.129	0.248	−0.443	0.033	0.398	0.001
bank8	NLP	−0.024	0.023	−0.009	0.387	−0.003	0.603
bank8	CAR	−0.070	0.023	−0.020	0.442	−0.002	0.852
bank8	Liquidity_Risk	0.085	0.093	0.017	0.751	−0.082	0.145
bank8	NBP_Equity	0.153	0.307	−0.567	0.007	0.417	0.003

Notes: This table reports leave-one-out (jackknife) estimates. In each iteration, one bank is excluded from the sample, and the baseline fixed-effects model is re-estimated using the remaining banks. The reported coefficients correspond to the relevant ESG variables. Robust standard errors are computed using the Driscoll–Kraay method. All control variables are included in the estimations but are not reported for brevity.

show that the main results—particularly the effects of ESG_Combined_Score_L1 in Model A and the environmental and social dimensions in Model B—are not driven by any single bank. In contrast, the coefficients associated with G_Score_L1, especially for NPL, display weaker consistency, with statistical significance frequently disappearing across subsamples. This pattern suggests that governance-related effects are more sensitive to model specification and sample composition and may reflect a relatively fragile relationship within the structure of the Turkish banking system.

In this context, the following Discussion section compares the empirical findings with the existing literature and provides a detailed examination of potential mechanisms and policy implications.

5. Discussion

This study examines the impact of ESG (Environmental, Social, and Governance) performance on key banking risk and balance sheet indicators—namely non-performing loans (NPL), capital adequacy (CAR), liquidity risk, and net balance sheet position to equity (NBSP/E)—using both a composite ESG score and its disaggregated E, S, and G components. The findings partially align with the dominant views in the existing literature, while also revealing distinctive patterns, particularly in the context of emerging market dynamics and within the framework of the “skimping hypothesis.”

5.1. The Relationship Between ESG and Non-Performing Loans (NPL)

According to the baseline results of Model A, the composite ESG score exhibits a positive and statistically significant association with non-performing loans (NPL) (β = 0.0397, p < 0.01). This finding should be treated with caution, especially given the small sample and the sensitivity of some estimates to specification choice.

In contrast, the results of Model B indicate that the governance (G) score displays a weak but negative relationship with NPL (β = −0.012, p < 0.10), which is consistent with studies emphasizing that stronger corporate governance structures enhance risk monitoring and internal control mechanisms (P. M. U. Nguyen & Pham, 2025; Aevoae et al., 2023). Nevertheless, this relationship does not remain statistically significant across alternative model specifications, suggesting that the governance–NPL link is sensitive to model assumptions and sample composition. Governance-related results, particularly for NPL, should be viewed as tentative rather than conclusive.

5.2. The Relationship Between ESG and Liquidity Risk

The empirical results indicate that neither the composite ESG score nor its individual components exert a statistically significant effect on liquidity risk. This finding diverges from studies arguing that ESG performance serves as an “insurance mechanism” against liquidity shocks (Gangwani & Kashiramka, 2024; Zhao et al., 2021). A plausible explanation is that the banks in the sample already maintain relatively high liquidity ratios (with an average of 21.81%), which may limit the marginal role of ESG in further mitigating liquidity risk. Moreover, prior evidence suggests that the liquidity-related benefits of ESG tend to materialize primarily during periods of financial stress or in regions with heightened climate vulnerability (Choi et al., 2023), conditions that may not be fully captured within the sample period and institutional context considered in this study.

5.3. The Relationship Between ESG and Capital Adequacy (CAR)

The baseline Model A finding that the composite ESG score is positively and statistically significantly associated with the capital adequacy ratio (CAR) (β = 0.096, p < 0.01) is consistent with the argument that ESG engagement can affect capital buffers through information asymmetry and risk-management channels. In contrast, the negative and statistically significant effect of the Environmental (E) score on CAR observed in Model B (β = −0.044, p < 0.01) constitutes a noteworthy result. While Gangi et al. (2025) document a positive association between environmental responsibility and capital ratios, our findings suggest that the green transition may impose short-term pressure on capital adequacy through high adaptation costs and transition risks associated with divesting from “brown” assets. This outcome is consistent with the view that prudential regulation is still in a costly adjustment phase as environmental risks become internalized within risk-weighted asset frameworks (Esposito et al., 2021).

5.4. The Relationship Between ESG and Net Balance Sheet Position/Equity (NBSP/E)

In the NBSP/E specification, the negative and statistically significant effect of the Social (S) score (β = −0.575, p < 0.05) indicates that social performance may exert downward pressure on financial margins. While Azmi et al. (2021) argue that ESG engagement can improve net interest margins (NIM) by reducing funding costs, Ananta and Anwar (2025) as well as Jaiwani and Gopalkrishnan (2025) emphasize that socially oriented credit expansion (e.g., financial inclusion initiatives) and community-related projects may weaken operational efficiency and NBSP/E in the short run. Our findings are therefore more cautiously consistent with studies suggesting that the effects of ESG on bank performance vary across dimensions, with some social dimensions potentially weakening accounting-based (Menicucci & Paolucci, 2023).

5.5. Control Variables and the Macroeconomic Context

The estimated effects of the control variables indicate that the COVID-19 dummy has a positive and statistically significant impact on NPL and CAR, while exerting a negative effect on liquidity risk. These findings reflect the temporary balance sheet effects of crisis-period credit expansion and public support packages on banks during the pandemic (Danisman & Tarazi, 2024). This interpretation is also consistent with Aldousari et al. (2025), who show that crisis-period bank risk can vary with ownership structure, with private banks appearing more vulnerable to systemic shocks while publicly backed banks may display greater resilience. In addition, the result that inflation increases liquidity risk while reducing NPL ratios—through the nominal easing of debt repayment—confirms a characteristic feature of high-inflation economies such as Türkiye.

5.6. Limitations and Directions for Future Research

This study is subject to several limitations that should be taken into account when interpreting the findings. This study is subject to several limitations that should be considered when interpreting the findings. First, the analysis is limited to eight Borsa Istanbul-listed banks with available ESG data, and missing observations for some bank-year pairs reduce the effective sample size. As a result, the panel is unbalanced, which limits direct generalizability beyond Türkiye or to other bank types and weakens the statistical power of the estimates for the disaggregated E, S, and G dimensions.

To mitigate sensitivity to sample definition, a comprehensive set of robustness checks was conducted, including alternative sample constructions, exclusion of the COVID-19 year, winsorization of outliers, and leave-one-out (jackknife) analyses. These exercises indicate that the core findings are largely preserved across alternative specifications.

Beyond sample size, a second limitation concerns the way ESG performance is measured. ESG scores provided by LSEG Refinitiv offer a broad and comparable signal; however, they may not always capture the dimensions that are most material for banks in a given setting. In a bank-dominated emerging market, ESG-related risks may be concentrated in specific channels, such as the sectoral composition of the loan portfolio, exposures to transition-sensitive industries, or governance practices that shape risk appetite and provisioning discipline. Future research could therefore complement LSEG Refinitiv ESG scores with banking-relevant, materiality-oriented indicators (e.g., climate-sensitive portfolio exposures, green versus brown credit composition, or supervisory guidance-based ESG practices) to examine whether the ESG components that are most material are also those most strongly associated with risk outcomes.

Third, the study focuses on bank-level risk and balance sheet indicators and does not directly examine systemic risk measures that could provide additional insights into the macro-financial implications of ESG performance. Future research may extend this framework by incorporating systemic risk indicators to assess the economy-wide effects of ESG integration. Moreover, comparative designs covering different countries and banking systems, as well as alternative bank types, could further enhance the external validity of the findings.

Relatedly, although the empirical strategy uses lagged ESG variables and bank fixed effects to mitigate simultaneity concerns, it is not designed to support strong causal claims. Future research could build on difference-in-differences designs around regulatory milestones or on the staggered adoption of sustainability governance and reporting practices.

Finally, heterogeneity across banks is an important dimension that cannot be fully explored with a small listed-bank sample. Fixed effects absorb time-invariant differences, yet they do not answer whether the ESG–risk association differs systematically by ownership structure or business model (e.g., state-owned vs. private vs. foreign-owned banks; conventional vs. participation banking). With broader coverage, future research could test interaction effects (e.g., ESG × ownership type) and examine whether institutional features amplify or dampen the ESG–risk link, especially in emerging markets where public ownership and policy mandates may shape bank behavior.

It should be emphasized that the sample limitation does not stem from data availability but from the structure of the population itself. The study covers the vast majority of publicly listed banks in Türkiye for which ESG scores are reported, thereby offering a “near-population” panel of the Turkish banking sector.

6. Conclusions

This study contributes to the literature by examining the relationship between ESG performance on banks’ core risk and balance sheet indicators within a comprehensive framework, using the Turkish banking system as a case study. The findings suggest that the relationship between ESG performance on banks’ risk profiles is not unidimensional. Rather, different ESG components appear to operate through different risk channels, pointing to the importance to approaching ESG in the banking sector in a more holistic and context-sensitive way. Overall, the baseline results indicate that composite ESG performance and selected ESG dimensions are associated with some bank risk indicators, although governance-related findings appear weaker and more sensitive to empirical specifications in the Turkish banking sector.

Using panel data covering the 2008–2023 period, this study analyzes the relationship between banks’ ESG performance and asset quality as well as financial risk indicators in the Turkish banking sector. By employing two complementary model specifications—one based on a composite ESG score and the other on disaggregated environmental, social, and governance components—the analysis provides a broad assessment of how ESG dimensions relate to non-performing loans, capital adequacy, liquidity risk, and net banking profitability.

The empirical findings suggest that the composite ESG score is positively associated with asset quality and capital adequacy, whereas its relationship with liquidity risk and NBSP/E is less stable across specifications. When ESG dimensions are examined separately, substantial heterogeneity emerges across the E, S, and G components. In particular, the environmental score exhibits a negative relationship with capital adequacy, while the social score is negatively associated with NBSP/E. In contrast, the governance score displays weaker and less robust relationships with the examined risk indicators.

The robustness checks show that several of the baseline results remain broadly consistent, across alternative specifications, although some findings—particularly those related to liquidity risk and the governance–NPL relationship—are more sensitive to the specification and sample used. For this reason, the findings should be interpreted as associations rather than causal effects, with greater caution for the less stable results.

Overall, the findings suggest that ESG performance should not be treated as a uniform construct in the context of banking risk and asset quality. Instead, different ESG dimensions appear to operate through different channels, underscoring the value of disaggregated ESG analysis for regulators, investors, and bank managers. Given the limited availability of ESG data in emerging markets, this study contributes to the growing literature by providing evidence from the Turkish banking sector. At the same time, it highlights the need for future research to examine the effects of ESG performance on various types of financial risks and balance sheet indicators within a more comprehensive and causally oriented framework, using larger bank samples, alternative ESG measures, and different econometric approaches.

7. Theoretical and Policy Implications

7.1. Theoretical Implications

From a theoretical perspective, the findings of this study weaken the assumption that ESG affects bank risk in a unidirectional manner through a single mechanism. The results indicate that the ESG–risk relationship can operate through multiple channels and that its magnitude and direction may vary depending on banks’ business models, balance sheet structures, and the macro-financial conditions of the country in which they operate. In particular, the observation that improvements in ESG performance may coincide with short-term adverse movements in certain risk indicators suggests that ESG should not be regarded as an inherently or universally “risk-reducing” factor. This pattern implies that the effects of ESG may be time-varying and that risk may be redistributed across different risk categories rather than uniformly reduced.

Moreover, the differential associations observed between the environmental, social, and governance pillars and various outcome variables indicate that treating ESG as a single composite score may be theoretically incomplete. Instead, the underlying mechanisms appear to be pillar-specific, pointing to the need for more granular theoretical frameworks that explicitly model the distinct channels through which E, S, and G dimensions influence bank behavior and risk-taking. Finally, evidence derived from an emerging, bank-based economy operating in a volatile macroeconomic environment supports the view that findings from developed-country settings cannot be automatically generalized. This reinforces the importance of contextual factors in shaping theoretical interpretations of the ESG–risk nexus.

7.2. Policy Implications

The findings convey an important warning for regulators and supervisory authorities: ESG performance should not be treated as a standalone “risk-reducing label,” but rather integrated into a holistic supervisory framework through a disaggregated assessment of its effects across different types of risk. The fact that certain risk indicators may deteriorate even in environments where capital adequacy improves indicates that supervision based solely on capital metrics may be insufficient. Instead, ESG objectives should be monitored concurrently with indicators related to credit risk, liquidity risk, and balance sheet structure.

Accordingly, in the design of macroprudential policies, a more systematic use of stress tests and scenario analyses that explicitly assess the impact of ESG- and climate-related shocks on banks would be warranted. Finally, from the perspective of investors and market discipline, the results suggest that interpreting ESG performance as an unconditional signal of “lower risk” may be misleading. Rather, the ESG–risk relationship appears to vary across risk types, time horizons, and institutional contexts, implying a more nuanced and conditional interaction.