Non-Linear Dynamics of ESG Integration and Credit Default Swap on Bank Profitability: Evidence from the Bank in Turkiye

Abstract

This paper investigates the effect of Environmental, Social and Governance (ESG) scores and Credit Default Swap (CDS) spreads on the profitability of Halkbank, one of the biggest state-owned banks in Türkiye, an emerging economy. To this end, we employ Non-linear Autoregressive Distributed Lag (NARDL) and Markov Switching Regression (MSR) methods, taking into account non-linear market risks, using Halkbank’s quarterly data consisting of 63 observations for the period 2009Q1–2024Q3. Moreover, to prevent multicollinearity, we aggregate banking-specific and macroeconomic indicators into a single composite index using Principal Component Analysis (PCA). Our MSR findings suggest that ESG scores and CDS spreads negatively affect bank profitability and that these effects are particularly pronounced during periods of high market volatility. Similarly, NARDL findings suggest that ESG scores have asymmetric effects on bank performance, with both positive and negative changes in ESG performance having a negative impact on profitability, and moreover, negative changes have a more negative impact on profitability. This means that the bank’s sustainability initiatives may be costly and negatively affect profitability in the short run, but these effects will be more negative if initiatives deteriorate. Our findings emphasize the need for banks to adopt a gradual ESG approach that enables them to increase their capacity without compromising financial stability and for regulatory structures to have a flexible and sophisticated risk management framework capable of rapidly adapting to different market conditions. Therefore, our study provides valuable insights to sector managers and policymakers regarding the financial implications of sustainability approaches.

1. Introduction

The banking industry in emerging economies is experiencing significant structural change as the global emphasis shifts towards sustainability and evolving market dynamics (Demirgüç-Kunt et al., 2021). Profitability in this industry is increasingly challenged by external factors such as heightened regulatory oversight and volatile global economic conditions (Claessens et al., 2018). In addition, the role of Environmental, Social and Governance (ESG) activities in the banking sector has become increasingly central, as financial institutions are expected to integrate sustainability considerations into their strategic and operational decisions (Buallay, 2019; Birindelli et al., 2018). Consequently, the banking industry today is required to pursue sustainable and socially responsible practices alongside traditional profit-oriented objectives. On the other hand, Credit Default Swap (CDS) spreads are widely regarded as key market-based indicators of financial risk, reflecting investor sentiment and the perceived creditworthiness of financial institutions (Longstaff et al., 2005). All things considered, the Turkish banking sector is in a special position to be in the spotlight.

As one of the most important state-owned banks in Türkiye, Halkbank plays a dual role by focusing on profitability while striving to achieve development goals. These roles emphasize the need to examine the interaction of sustainability and market risk indicators with profitability indicators.

Since its establishment in 1938, Halkbank has been one of the key public banks in Türkiye, focusing on social banking and SME financing. Accordingly, it is one of the most important public banks in the country’s economic development process, with total assets reaching TL 2.6 trillion in 2024. In addition to its role in the development process, Halkbank has recently accelerated its ESG activities, and this has been an important milestone for Halkbank. In this context, both its dual role and its ESG initiatives make Halkbank an appropriate area to analyze.

The existing empirical literature emphasizes the effects of ESG initiatives and CDS spreads on bank profitability (Avino et al., 2019; Azmi et al., 2021; Menicucci & Paolucci, 2023; Liu et al., 2025). However, there is a lack of research on these effects from the perspective of a state-owned bank in an emerging economy. This gap is compounded by the fact that studies focus on profitability measures such as Return on Assets (ROA) and Return on Equity (ROE), both of which represent different financial performance structures. Moreover, while linear modeling is frequently used in past studies, asymmetric and non-linear structures that arise due to the nature of market behavior are ignored. Addressing these research shortcomings and complexities is important for creating a more nuanced perspective on the determinants of bank profitability.

In this respect, this study has several important contributions to the related literature by addressing four interrelated gaps. First, it investigates the relationship between ESG scores and both ROA and ROE separately to examine the non-linear effects of ESG activities on bank performance measures. Secondly, in order to determine the relationship between market risks and bank profitability, the effect of CDS spreads on bank profitability measures is analyzed. Thirdly, the study considers bank-specific factors and macroeconomic factors on profitability and analyzes these variables together. Finally, different dynamics under different market conditions are taken into account. Thus, the study provides important policy implications for state-owned banks in emerging economies and provides a roadmap for risk management of stabilization measures and sustainability activities against financial difficulties.

We focused on these gaps by analyzing the following: how do both ESG performance and CDS spreads influence the profitability of a state-owned bank in an emerging market, and do these relationships vary across different market regimes? We investigate Halkbank, one of Türkiye’s largest state-owned banks (TL 2.6 trillion assets), by employing quarterly data between 2009 and 2024.

For this purpose, PCA is used to reduce both banking-specific and macroeconomic factors into a single index. NARDL and MSR models, which have a robust methodological framework, are used to identify non-linear relationships between ESG scores and profitability factors. The presence of an asymmetric nature between the variables was analyzed with the BDS pretest. In addition to these methods, examining ROA and ROE separately as performance measures further extends the scope of the analysis.

The findings of the study show that there is a negative relationship between ESG scores and bank profitability indicators. This supports the notion that favorable sustainability movements may negatively affect profitability indicators, especially in the short term. Similarly, CDS spreads are also found to negatively affect profitability indicators. This supports the view that profitability may be negatively affected during periods of increased market risk. Moreover, these relationships between ESG scores, CDS spreads, and bank profitability indicators show a non-linear structure, which underlines the complexity of the interactions between variables.

The rest of the paper is designed as follows. Section 2 reviews the existing literature on the determinants of bank profitability, ESG integration, and the role of CDS spreads. Section 3 details the methodological background of the NARDL and MSR models used to examine the non-linear structure, together with the PCA used for dimensionality reduction along with the data sources. Section 4 presents the empirical results as well as the diagnostic tests, and Section 5 discusses the policy implications of the findings for emerging public sector banking. Finally, Section 6 concludes with a broad summary of the paper’s contributions to the literature and roadmap suggestions for future work.

2. Background and Literature Review

The literature on bank profitability has evolved significantly, with studies examining both internal and external determinants. Internal factors, including impaired loans, efficiency, and gross interest margins, significantly influence profitability metrics (Lamothe et al., 2024). Non-traditional banking activities have been found to enhance ROA and ROE, though they may lead to reduced net interest revenue (Saif-Alyousfi, 2020). Fernandes et al. (2021) document a non-monotonic relationship between cash holdings and profitability, noting that emerging market banks maintain higher cash levels than their developed market counterparts.

2.1. ESG and Bank Performance

Recent research reveals complex relationships between ESG practices and bank performance. In emerging markets, Gangwani and Kashiramka (2024) find that higher ESG disclosure scores correlate with reduced insolvency and liquidity risks. This finding aligns with studies by Useche et al. (2024), who document positive relationships between ESG strategies and profitability in Latin American contexts. However, the relationship exhibits regional variations. In Southeast Asia, Gutiérrez-Ponce and Wibowo (2023) observe negative short-term impacts of ESG practices on financial performance, while individual ESG pillars show varying effects.

The role of ESG in banking performance is further complicated by market conditions and institutional frameworks. Luo et al. (2024) demonstrate that country governance quality moderates the ESG-performance relationship, while Bouattour et al. (2024) identify three distinct ESG performance regimes using regime-switching models. In China, Wang et al. (2024) find that ESG performance stimulates green innovation in commercial banks, particularly when supported by effective remuneration incentives.

2.2. CDS and Bank Performance

The literature provides evidence on how bank performance is affected by CDS spreads through various channels. Ahnert et al. (2020) find that banks that pass regulatory stress tests have tighter CDS spreads. While Drago et al. (2017) point out that firm-specific factors (leverage, asset quality, etc.) are the main determinants of CDS spreads, Benbouzid et al. (2017) emphasize country-level financial structure as an important determinant of CDS spreads.

On the other hand, there is also evidence in the literature that the effect of CDS spreads on bank performance varies depending on time and regime. Du et al. (2023) find significant regime changes in treasury markets after the global financial crisis. Hainaut et al. (2018), using continuous-time regime switching models, find evidence on how regime changes affect bank capital structure.

When recent studies in this field are analyzed, it is noteworthy that more complex models are used. For example, Joaqui-Barandica et al. (2024) combined Singular Spectrum Analysis and PCA in examining the effects of key macroeconomic factors on profitability. On the other hand, Dao and Nguyen (2024) use Bayesian Quantile Regression for variable selection in macroeconomic stress tests, while Gemar et al. (2019) use Partial Least Squares Structural Equation Modeling to assess the sustainability of bank profitability.

The evolution of analytical approaches reflects growing recognition of nonlinear relationships in banking performance. Bouslimi et al. (2024) demonstrate complex relationships between credit and liquidity risks using non-linear analysis, while Vithessonthi et al. (2024) compare linear and non-linear modeling approaches in analyzing credit and investment relationships.

The literature increasingly recognizes the role of technological advancement in banking performance. Pilatin (2024) examines how big data usage moderates the relationship between intellectual capital and innovation performance in Turkish banks. Nabiyev and Ovenc (2023) analyze the symbiotic relationship between commercial banks and fintech firms, highlighting the importance of technological integration in modern banking.

Research on emerging markets reveals unique challenges and opportunities. Athari et al. (2023) demonstrate how country governance moderates the impact of various risks on banking sustainability in BRICS economies. Yan et al. (2024) highlight the role of state-owned capital in promoting green innovation, while Mirza et al. (2022) document positive effects of green lending on banking performance in BRIC countries.

Several methodological gaps persist in the literature. Saviano et al. (2024) identify challenges in ESG factor integration, particularly regarding data quality and knowledge management. Curcio et al. (2024) highlight the need for more comprehensive analysis of how ESG scores affect systemic risk, while Palmieri et al. (2023) emphasize the importance of considering industry and stock index membership in assessing ESG impacts.

The existing literature on the determinants of bank profitability provides valuable insights. First, although studies like Gangwani and Kashiramka (2024) and Lu and Yang (2024) examine ESG effects on bank performance, and others like Ahnert et al. (2020) and Drago et al. (2017) analyze CDS spread impacts, there is a notable absence of research examining their combined effects on bank profitability, particularly in emerging markets. Our study fills an important gap in the literature by analyzing how ESG scores and CDS spreads affect the profitability of a state-owned bank in Türkiye.

Second, although non-linear models were used by Bouattour et al. (2024) and Vithessonthi et al. (2024), these studies only examined one aspect of bank profitability. In our study, we examine the relationship between bank profitability and ESG scores and CDS spreads using NARDL and MSR models. The combination of the two models allows us to understand both asymmetric effects and regime-dependent behavior in this relationship.

Third, the construction of a composite index of banking-specific and macroeconomic factors using PCA in our study provides a simpler yet comprehensive analytical framework that can be adopted in future banking studies. In this respect, our study differs from Joaqui-Barandica et al. (2024) and Dao and Nguyen (2024), who use sophisticated techniques to deal with multiple indicators.

Moreover, what distinguishes our study from studies such as Yan et al. (2024) and Mirza et al. (2022) is that we specifically examine the performance of a state-owned bank in Türkiye, an emerging market. This aspect of our study allows us to provide unique insights into how ESG manages the complex relationship between implementation costs and profitability targets. Indeed, the negative relationships between ESG scores and profitability measures in our findings provide an important contribution to the debate on the short-term costs and long-term benefits of sustainability initiatives in banking.

Finally, our study’s examination of the joint effects of ESG and market risk indicators on bank performance addresses the methodological gaps identified by Saviano et al. (2024) and Curcio et al. Our findings contribute to understanding how these relationships change under different market conditions and regimes and address the need for more sophisticated modeling to understand the relationships in depth.

3. Data and Methodology

3.1. Data

This study empirically analyzes the impact of ESG scores and CDS spreads on the bank profitability of Halkbank, one of the largest state-owned banks in Türkiye, using quarterly data consisting of 63 observations between 2009Q1 and 2024Q3.

In Türkiye, banks are legally required by Banking Regulation and Supervision Agency regulations to publish interim (quarterly) financial statements under the ‘Regulation on Accounting Applications for Banks and Safeguarding of Documents’. Interim reports are subject to independent auditor review, typically a limited review rather than a full audit, which ensures that the financial information is fairly presented in all material respects.

Halkbank was selected for this study on the basis of several criteria. Firstly, as a state-owned bank with a significant presence in the market, Halkbank’s active sustainability initiatives, in addition to its development banking role, provide an important research area on the impact on profitability. Secondly, Halkbank’s listed (publicly traded) status makes its financial data easily accessible and enables market-based risk assessment through CDS spreads. Finally, Halkbank’s large size and systemic importance in the Turkish banking industry make it representative of state-owned banks in other developing countries. These characteristics make it an ideal case to analyze how ESG initiatives and CDS spreads affect Halkbank’s profitability.

We employ two widely used profitability measures as dependent variables: Return on Assets (ROA), which measures the efficiency of asset utilization in generating profits, and Return on Equity (ROE), which indicates the return generated on shareholders’ investment.

Our explanatory variables encompass four main categories. First, we include ESG performance measures, decomposed into positive (ESG_POS) and negative (ESG_NEG) changes to capture potential asymmetric effects. Second, we incorporate Credit Default Swap (CDS) spreads as a measure of market-perceived risk. The third category comprises seven bank-specific variables: the ratio of liquid assets to total assets, liquid assets to short-term liabilities, total assets, non-performing loans ratio, equity to total assets ratio, capital adequacy ratio, and net interest income after provisions to total assets. The fourth category includes macroeconomic indicators: the Central Bank Late Liquidity Window Rate, the general budget non-interest balance to GDP ratio, the economic growth rate, and the consumer price index. All variables for Halkbank were obtained from the Refinitiv Eikon database. Descriptive statistics are presented in Table A1 of Appendix A.

To address multicollinearity and reduce dimensionality, we first perform PCA to create two composite indices: PC_BANK for banking-specific indicators and PC_MACRO for macroeconomic variables. This approach allows us to capture the comprehensive effects of multiple related variables while maintaining model parsimony. PCA ensures that variables exhibiting high correlation are grouped into a single index, thereby preventing multiple linear dependencies. By preventing the subjective determination of weight matrices for variables, it derives the index to be created according to the structure of the data itself. For this reason, this method is frequently used in macro-finance and banking studies (Kapinos & Mitnik, 2016; Asongu, 2018; Destek, 2018).

3.2. Methodology

For empirical modeling, we firstly investigate linearity properties of the variables by employing Broock et al. (1996) BDS test. After we found all investigated variables exhibit non-linear properties, we employed Non-linear Autoregressive Distributed Lag (NARDL) approach to examine long-run relationships and asymmetric effects.

The non-linear bound test approach, which is an augmented version of Pesaran et al. (2001) developed by Shin et al. (2014), is employed to test for cointegration, with the null hypothesis of no cointegration tested against critical values. Superior properties of the NARDL bound test approach over the conventional models include the following: (i) NARDL model gives effective results irrespective of stationarity properties the variables, (ii) NARDL approach provides valid and efficient results when the sample is small, (iii) NARDL model presents superior results because it estimates cointegration and asymmetries jointly (Katrakilidis & Trachanas, 2012; Chishti et al., 2020).

ESG variable is separated into negative and positive parts parallel with the Shin et al. (2014) approach. Equation (1) indicates decomposed ESG variable.

$${ESG}_t^{+}={\sum }_{m=1}^t∆{ESG}_t^{+}={\sum }_{m=1}^t\mathrm{m}\mathrm{a}\mathrm{x}(∆{ESG}_t^{+},0)$$

$${ESG}_t^{-}={\sum }_{m=1}^t∆{ESG}_t^{-}={\sum }_{m=1}^t\mathrm{m}\mathrm{i}\mathrm{n}(∆{ESG}_t^{-},0)$$

(1)

Equations (2) and (3) presents augmented UECM model with negative ${(ESG}^{-}$) and positive ${(ESG}^{+})$ variables as proposed by Shin et al. (2014) for ROA and ROE, respectively.

$$\begin{gathered} {∆LROA}_t={\alpha }_0+{\alpha }_{1,i}{{\sum }_{i=1}^m{∆LROA}_{t-i}+\alpha }_{2,i}{{\sum }_{i=0}^m{{∆ESG}^{+}}_{t-i}+{\alpha }_{3,i}{\sum }_{i=0}^m{{∆ESG}^{-}}_{t-i}+{\alpha }_{4,i}{\sum }_{i=0}^m{∆CDS}_{t-i}+ \\ {\alpha }_{5,i}{\sum }_{i=0}^m{∆PC\_BANK}_{t-i}+{{\alpha }_{6,i}{\sum }_{i=0}^m{∆PC\_MACRO}_{t-i}+\alpha }_7{ROA}_{t-1}+{\alpha }_8{{ESG}^{+}}_{t-1}+{\alpha }_9{{ESG}^{-}}_{t-1}+{\alpha }_{10}{CDS}_{t-1}+ \\ {\alpha }_{11}{PC\_BANK}_{t-1}+{\alpha }_{12}{PC\_MACRO}_{t-1}+\epsilon }_t \end{gathered}$$

(2)

$$\begin{gathered} {∆LROE}_t={\alpha }_0+{\alpha }_{1,i}{{\sum }_{i=1}^m{∆LROE}_{t-i}+\alpha }_{2,i}{{\sum }_{i=0}^m{{∆ESG}^{+}}_{t-i}+{\alpha }_{3,i}{\sum }_{i=0}^m{{∆ESG}^{-}}_{t-i}+{\alpha }_{4,i}{\sum }_{i=0}^m{∆CDS}_{t-i}+ \\ {\alpha }_{5,i}{\sum }_{i=0}^m{∆PC\_BANK}_{t-i}+{{\alpha }_{6,i}{\sum }_{i=0}^m{∆PC\_MACRO}_{t-i}+\alpha }_7{ROA}_{t-1}+{\alpha }_8{{ESG}^{+}}_{t-1}+{\alpha }_9{{ESG}^{-}}_{t-1}+{\alpha }_{10}{CDS}_{t-1}+ \\ {\alpha }_{11}{PC\_BANK}_{t-1}+{\alpha }_{12}{PC\_MACRO}_{t-1}+\epsilon }_t \end{gathered}$$

(3)

Following confirmation of cointegration, we estimate long-run coefficients to quantify the relationships between our variables. The NARDL model is specified to include both positive and negative ESG variables, allowing for differential impacts of increases and decreases in ESG on bank profitability.

To capture regime-dependent behavior, we complement the NARDL analysis1 with a two-state MSR model. This approach allows parameters to vary across different regimes, providing insights into how relationships might change under different market conditions. The regime-switching model is estimated using maximum likelihood, with regime probabilities determined endogenously from the data. The MSR model representation for our study is indicated in Equations (4) and (5) for ROA and Equations (6) and (7) for ROE.

Low Volatility Regime for ROA:

$${ROA}_t={\alpha }_{\mathrm{1,0}}+{\alpha }_{\mathrm{1,1}}{ESG}_{t-i}{+{\alpha }_{\mathrm{1,2}}{CDS}_{t-i}+{\alpha }_{\mathrm{1,3}}{PC\_BANK}_{t-1}+{\alpha }_{\mathrm{1,4}}{PC\_MACRO}_{t-1}+\epsilon }_{1,t}$$

(4)

High Volatility Regime for ROA:

$${ROA}_t={\alpha }_{\mathrm{2,0}}+{\alpha }_{\mathrm{2,1}}{ESG}_{t-i}{+{\alpha }_{\mathrm{2,2}}{CDS}_{t-i}+{\alpha }_{\mathrm{2,3}}{PC\_BANK}_{t-1}+{\alpha }_{\mathrm{2,4}}{PC\_MACRO}_{t-1}+\epsilon }_{2,t}$$

(5)

Low Volatility Regime for ROE:

$${ROE}_t={\alpha }_{\mathrm{1,0}}+{\alpha }_{\mathrm{1,1}}{ESG}_{t-i}{+{\alpha }_{\mathrm{1,2}}{CDS}_{t-i}+{\alpha }_{\mathrm{1,3}}{PC\_BANK}_{t-1}+{\alpha }_{\mathrm{1,4}}{PC\_MACRO}_{t-1}+\epsilon }_{1,t}$$

(6)

High Volatility Regime:

$${ROE}_t={\alpha }_{\mathrm{2,0}}+{\alpha }_{\mathrm{2,1}}{ESG}_{t-i}{+{\alpha }_{\mathrm{2,2}}{CDS}_{t-i}+{\alpha }_{\mathrm{2,3}}{PC\_BANK}_{t-1}+{\alpha }_{\mathrm{2,4}}{PC\_MACRO}_{t-1}+\epsilon }_{2,t}$$

(7)

where ${\alpha }_{10}$ and ${\alpha }_{20}$ indicates constants and ${\alpha }_{11}$ and ${\alpha }_{21}$ denote the autoregressive coefficients. ${\epsilon }_{1t}$ and ${\epsilon }_{2t}$ are the white-noise errors, respectively.

This methodological framework allows us to examine non-linear relationships between bank profitability and its determinants while accounting for potential regime changes and asymmetric effects. The combination of NARDL and MSR approaches provides a robust analysis of how ESG performance and market risk influence Halkbank’s profitability under different market conditions.

4. Results

In the empirical analysis, we first investigated non-linear properties of the investigated variables by employing the BDS test. BDS (1996) test results are presented in Table 1.

Table 1. BDS test results.

Dimensions/Variables	2	3	4	5	6
ROA	0.184 [0.000]	0.305 [0.000]	0.384 [0.000]	0.443 [0.000]	0.482 [0.000]
ROE	0.174 [0.000]	0.290 [0.000]	0.368 [0.000]	0.419 [0.000]	0.455 [0.000]
CDS	0.120 [0.000]	0.211 [0.000]	0.266 [0.000]	0.295 [0.000]	0.308 [0.000]
ESG	0.181 [0.000]	0.298 [0.000]	0.375 [0.000]	0.429 [0.000]	0.464 [0.000]
PC_BANK	0.136 [0.000]	0.236 [0.000]	0.303 [0.000]	0.357 [0.000]	0.396 [0.000]
PC_MACRO	0.167 [0.000]	0.267 [0.000]	0.323 [0.000]	0.350 [0.000]	0.352 [0.000]

The BDS test results indicate the rejection of the null hypothesis of linearity for all investigated variables and show the non-linearity of the investigated variables.

After the non-linearity check, we investigated the co-integration relationship between the variables by employing the asymmetric cointegration test developed by Shin et al. (2014). Asymmetric co-integration model results are presented in Table 2.

Table 2. Asymmetric co-integration test results.

Variable	F Statistics	Critical Value (%1 Significance)
		Bottom Bound	Upper Bound
ROA	8.21	3.06	4.15
ROE	6.99	3.06	4.15

Table 2 reveals strong evidence of cointegration between bank profitability measures and their determinants. The bound test results demonstrate F-statistics of 8.21 and 6.99 for ROA and ROE models, respectively, both exceeding the upper critical values at all significance levels. This finding confirms the presence of long-run relationships among our variables, supporting the appropriateness of our modeling approach.

After the co-integration check, long-run estimation for ROA and ROE is presented in Table 3.

Table 3. Long-term coefficients and ECM coefficients from the NARDL model.

ROA			ROE
Long Term Coefficients from NARDL (3,1,1,0,0,0) Model			Long Term Coefficients from NARDL (2,0,0,0,0,1) Model
Variables	Coefficient	T Statistics	Variables	Coefficient	T Statistics
PC_BANK	0.300	2.165	PC_BANK	1.164	2.847
PC_MACRO	0.150	2.346	PC_MACRO	0.842	2.785
ESG_POS	−0.016	−2.217	ESG_POS	−0.094	−2.370
ESG_NEG	−0.053	−2.078	ESG_NEG	−0.748	−2.191
CDS	−0.016	−2.396	CDS	−0.029	−2.177
C	2.033	4.378	C	18.503	2.773
Error Correction Coefficients from NARDL (3,1,1,0,0,0) Model			Error Correction Coefficients from NARDL (2,0,0,0,0,1) Model
ECT	−0.260	−4.245	ECT	−0.192	−4.094
Diagnostic Checks			Diagnostic Checks
Breusch-Godfrey LM Test	0.726 [0.488]		Breusch-Godfrey LM Test	1.047 [0.358]
White Heteroskedasticity Test	0.298 [0.586]		White Heteroskedasticity Test	0.273 [0.602]
Ramsey’s RESET Test	1.424 [0.160]		Ramsey’s RESET Test	2.083 [0.114]
Jarqua-Bera Normality Test	0.274 [0.871]		Jarqua-Bera Normality Test	0.655 [0.720]

Notes: The values in brackets represent the p-value.

The long-run estimation results for the ROA model reveal significant relationships between bank profitability and all explanatory variables. The banking sector composite index exhibits a positive coefficient of 0.300, suggesting that improvements in banking sector conditions substantially enhance profitability. Similarly, the macroeconomic composite index shows a positive coefficient of 0.150, indicating that favorable macroeconomic conditions contribute to improved profitability, albeit with a smaller magnitude than banking-specific factors.

Notably, our analysis reveals asymmetric effects of ESG performance on ROA. Both positive and negative changes in ESG scores demonstrate significant negative effects on profitability, with coefficients of −0.016 and −0.053, respectively. This finding suggests that ESG initiatives, regardless of their direction, may impose short-term costs on bank profitability. The CDS spread also exhibits a negative relationship with ROA, with a coefficient of −0.016, indicating that increased market-perceived risk adversely affects profitability.

The ROE model results demonstrate similar patterns but with larger magnitudes, suggesting that shareholders’ returns are more sensitive to these factors. The banking sector index shows a stronger positive effect with a coefficient of 1.164, while the macroeconomic index demonstrates a coefficient of 0.842. The negative effects of ESG performance are more pronounced in the ROE model, with coefficients of −0.094 and −0.748 for positive and negative changes, respectively. The CDS spread maintains its negative influence with a coefficient of −0.029.

The error correction terms for both models are negative and significant (−0.260 for ROA and −0.192 for ROE), indicating that approximately 20% of any deviation from long-run equilibrium is corrected within one quarter. This moderate speed of adjustment suggests that while market forces work to restore equilibrium, the process is gradual rather than immediate.

The MSR model results are presented in Table 4.

Table 4. MSR model results.

ROA		ROE
Variable/Model	MSR Model	Variable/Model	MSR Model
Regime 1: Low Volatility		Regime 1: Low Volatility
PC1_BANKA	(0.404) *	PC1_BANKA	(0.819) *
PC1_MAKRO	(0.122) **	PC1_MAKRO	(1.320) **
ESG	(−0.013) *	ESG	(−0.155) **
CDS	(−0.001) *	CDS	(−0.019) *
C	(1.633) *	C	(18.209) *
Regime 2: High Volatility		Regime 2: High Volatility
PC1_BANKA	(0.744) ***	PC1_BANKA	(3.287) **
PC1_MAKRO	(0.716) ***	PC1_MAKRO	(3.039) *
ESG	(−0.038) *	ESG	(−0.205) *
CDS	(−0.011) *	CDS	(−0.024) *
C	(5.492) *	C	(42.048) *

Notes: *,** and *** indicates 1%, 5% and 10% significance level, respectively.

According to Table 4, the MSR model results identify two distinct regimes with significantly different parameter estimates. Under Regime 1 (low volatility) for the ROA model, the banking sector and macroeconomic indices show moderate positive effects (0.404 and 0.122, respectively), while ESG and CDS maintain negative but relatively smaller impacts (−0.013 and −0.001). In contrast, Regime 2 (high volatility) exhibits stronger effects across all variables, with notably larger negative impacts from ESG (−0.038) and CDS (−0.011).

The ROE model similarly demonstrates regime-dependent behavior, with Regime 2 showing substantially larger coefficients than Regime 1. The banking sector index effect increases from 0.819 to 3.287, while the ESG impact becomes more negative, moving from −0.155 to −0.205. These findings suggest that the relationships between profitability and its determinants become more pronounced during periods of high market volatility.

Our models pass all diagnostic checks, with no evidence of serial correlation, heteroskedasticity, or functional form misspecification. The Breusch-Godfrey LM test, ARCH test, and Ramsey RESET test all confirm the adequacy of our model specifications.

5. Discussion

Our empirical findings provide novel insights into the complex relationship between ESG performance, market risk, and bank profitability in the context of state-owned banks in emerging markets. The results reveal several important patterns that contribute to both theoretical and practical implications for banking sector sustainability.

The documented negative relationship between ESG scores and bank profitability represents a nuanced departure from conventional wisdom. While Gangwani and Kashiramka (2024) found positive ESG effects in emerging markets, our regime-switching analysis reveals that the relationship is more complex and state-dependent. The stronger negative effect during high volatility periods (coefficient increasing from −0.013 to −0.038 for ROA) suggests that ESG initiatives may create additional financial strain during market stress.

This finding can be interpreted through several theoretical lenses. First, it aligns with the “transition cost hypothesis” suggested by Gutiérrez-Ponce and Wibowo (2023), where initial ESG investments create short-term profitability pressures. Gutiérrez-Ponce and Wibowo (2023) report that overall ESG scores are negatively related to financial performance indicators such as ROA, ROE, and Tobin’s Q. Similarly, Yuen et al. (2022) and Buallay et al. (2021), within the scope of selected banks in developed and developing economies, and Maama (2021), within the scope of banks in Ghana, provide robust evidence supporting the “trade-off hypothesis” that ESG activities may reduce bank profitability. Second, the regime-dependent effects support Bouattour et al. (2024) argument about the non-linear nature of ESG-performance relationships. The larger negative coefficients during high volatility periods suggest that the cost burden of sustainability initiatives becomes more pronounced when banks face market stress, potentially due to reduced operational flexibility and increased compliance costs.

The negative relationship between CDS spreads and profitability, particularly its regime-dependent nature, provides important insights into market risk transmission mechanisms. Our findings extend Ahnert et al. (2020) by showing that the impact of market risk varies across different market conditions. The increase in negative effects during periods of high volatility (from −0.001 to −0.011 for ROA) highlights a non-linear risk–return relationship that is particularly pronounced during periods of stress.

These findings shed light on some important implications for understanding the risk-return dynamics of state-owned banks. In this regard, while non-state-owned banks may be more flexible in terms of managing market risks, state-owned banks are more likely to face constraints due to the impact of public policies. The stronger negative effect in periods of high volatility supports the view that state-owned banks may be more vulnerable to adverse conditions due to tighter policies.

Our findings suggest the need for more comprehensive analyses of ESG activities of state-owned banks, especially in developing countries. Firstly, the negative impact of ESG initiatives in the short run emphasizes the need for policymakers to take a gradual approach when setting sustainability targets. This would make it more feasible for state-owned banks to build capacity and undertake operational initiatives before their financial stability deteriorates. This is in line with Hui et al.’s (2024) recommendations for ESG and extends this framework by taking into account the regime change factor.

Second, the relationship between ESG performance and CDS spreads emphasizes the need for a simultaneous risk management structure that takes into account sustainability and market risk factors. Accordingly, as suggested by Sáiz et al. (2024), banks should strike a balance between profitability and sustainability initiatives, but our findings suggest that this balance is even more critical during periods of market turmoil. This requires the development of risk management strategies that can integrate changes in market conditions while remaining aligned with sustainability objectives.

Our findings provide important contributions to the determinants of bank profitability in several aspects. First, our results challenge linearity assumptions by showing significant regime-dependent effects. Second, we empirically support the switching cost hypotheses for a state-owned bank and conclude that these costs are not the same for different market conditions.

The regime-switching results also raise important questions about the nature of ESG implementation in banking. The stronger negative effects during high volatility periods suggest potential interaction effects between market conditions and sustainability initiatives that warrant further investigation. Future research could explore the mechanisms through which market volatility amplifies the cost burden of ESG initiatives and identify potential strategies for mitigating these effects.

6. Conclusions

This study analyzes the relationship between Halkbank’s ESG performance, CDS spreads, and profitability using quarterly data covering the period from the first quarter of 2009 to the third quarter of 2024. The MSR and NARDL methods are employed to identify asymmetries under different market regimes. Accordingly, the study provides valuable insights into the dynamics between sustainability performance and profitability for a state-owned bank in Türkiye over a period that includes episodes of financial turbulence.

Our findings reveal important insights in several respects. First, based on the NARDL analysis, both positive and negative decompositions of ESG scores were found to negatively affect the bank’s profitability, with this adverse effect being particularly stronger under high market volatility regimes. This indicates that the bank’s sustainability investments reduce profitability through cost-effectiveness. The increased expenses associated with meeting sustainability obligations and adapting operational processes may be key factors eroding profitability.

Secondly, our NARDL findings also show that CDS spreads negatively affect bank profitability, with this effect again being stronger during periods of market stress. This stronger impact suggests that factors influencing profitability become more pronounced under turbulent conditions. In such periods, rising risks increase funding costs, which in turn constrain the supply of credit.

Thirdly, both the bank composite index and the macroeconomic composite index contribute positively to bank profitability. Moreover, our results indicate that profitability behaves differently across volatility regimes. In particular, according to the MSR results, the higher coefficient in high-volatility regimes relative to low-volatility regimes suggests that management must exercise greater caution during periods of elevated market risk.

The findings of this study have important implications for both policy and practice. For policymakers, our results highlight the need to consider prevailing market conditions when designing and implementing sustainability requirements for state-owned banks. For bank managers, the findings underscore the importance of carefully managing the timing and execution of ESG initiatives, particularly during periods of heightened market risk.

Although our findings make significant contributions to the existing literature, it is important to acknowledge the limitations of this study. First, we examine the effect of cumulative ESG scores on bank profitability. However, analyzing the impact of each component (E, S, and G) separately would provide clearer insights into their individual effects, as these may be obscured when aggregated into a single ESG score. Second, because our study focuses on a single state-owned bank, the results may have been influenced by that institution’s specific governance structure, operational principles, or policy environment. Future studies involving multiple state-owned banks or comparative analyses between public and private banks could help determine whether the relationships identified here are bank-specific or indicative of broader sectoral patterns.

Future research may also explore how cross-country or public–private bank comparisons shape the dynamics we have highlighted. In addition, longer-term analyses could examine whether the negative ESG effects documented in this study persist as banks adopt more mature and efficient sustainability practices. These avenues represent valuable opportunities for further scholarly inquiry.