Climate Policy Uncertainty and Sovereign Credit Risk: A Multivariate Quantile on Quantile Regression Analysis

Abstract

This study investigates the nonlinear and regime-dependent relationship between climate policy uncertainty (CPU) and sovereign credit default swap (CDS) spreads across a panel of developed and emerging economies from February 2010 to March 2025. Utilizing the Quantile-on-Quantile Regression (QQR) and Multivariate QQR (MQQR) frameworks, we capture the heterogeneous effects of CPU under varying market states and assess the marginal role of global risk factors, including geopolitical risk (GPR), economic policy uncertainty (EPU), and market volatility (VIX). The findings indicate that in developed markets, CPU exerts a nonlinear impact that intensifies during periods of heightened sovereign risk, while in low-risk regimes, its effect is often muted or reversed. In contrast, emerging economies exhibit more volatile and state-contingent responses, with CPU exerting stronger effects in calm conditions but diminishing in explanatory power once global risks are taken into account. These dynamics highlight the importance of institutional credibility and financial integration in moderating CPU-driven credit risk.

1. Introduction

The accelerating pace of climate change and the increasing frequency of extreme weather events have transformed environmental risks from peripheral concerns into central determinants of macroeconomic and financial stability. Climate-related disasters now pose systemic threats to sovereign balance sheets, debt sustainability, and long-term creditworthiness, particularly in countries with limited adaptive capacity or high exposure to physical climate risks (Zenios 2022; Mallucci 2022). Vulnerability to climate change, both in its chronic and acute forms, exerts downward pressure on sovereign ratings (Sun et al. 2023; Klusak et al. 2023) and raises sovereign borrowing costs, especially in emerging and climate-vulnerable economies (Beirne et al. 2021; Boitan and Marchewka-Bartkowiak 2022). Simultaneously, rising transition risks tied to decarbonization efforts, geopolitical tensions, and climate policy uncertainty (Guo et al. 2025; Collender et al. 2023) have introduced new layers of complexity into the pricing of sovereign risk. This growing risk landscape has prompted a growing body of research exploring the financial repercussions of climate vulnerability, yet several critical gaps remain. Existing studies often rely on linear models or average effects, which can obscure important asymmetries, tail dependencies, and regime-specific responses (Naifar 2023; De Wet 2023). However, sovereign risk is inherently nonlinear, often responding more sharply to climate shocks during episodes of market distress or elevated uncertainty (Bratis et al. 2024; Subramaniam 2022). For example, the sovereign CDS spreads of climate-vulnerable countries may exhibit sharp sensitivity to climate-related disruptions during adverse financial conditions or periods of high geopolitical risk, reflecting nonlinear amplification mechanisms (Afonso et al. 2024; Yang and Hamori 2023).

Motivated by this complex interaction, our study seeks to deepen the understanding of how climate vulnerability influences sovereign credit risk by moving beyond average effects and embracing a fully distributional approach. We employ Quantile-on-Quantile Regression (QQR) to capture the asymmetric and state-contingent effects of climate risk on sovereign credit default swap (CDS) spreads, a widely recognized forward-looking measure of sovereign risk (Yang and Hamori 2023; Naifar 2023). While prior work has shown that geopolitical and economic policy uncertainty significantly shapes sovereign CDS pricing (Naifar and Aljarba 2023; Demiralay et al. 2024; Inoguchi 2025), our study adds to this literature by focusing explicitly on climate risk as a driver of sovereign credit stress across different quantiles of market conditions. We further expand this framework by incorporating Climate Policy Uncertainty (CPU) and global financial risk indicators, including the Geopolitical Risk (GPR), Economic Policy Uncertainty (EPU), and the VIX volatility index, within a novel Multivariate Quantile-on-Quantile Regression (MQQR) approach. By doing so, we extend the insights of Naifar (2023), who links climate readiness and vulnerability to sovereign CDS, by focusing specifically on policy-related uncertainty as a critical determinant of sovereign risk transmission. This extension enables us to examine how climate vulnerability interacts with broader global risk factors in shaping sovereign credit default swap (CDS) spreads. Our contributions are threefold: (i) we introduce CPU into the sovereign CDS literature, providing the first systematic evidence on how climate policy uncertainty shapes sovereign credit risk; (ii) we advance the econometric framework by developing a multivariate QQR approach that captures nonlinear interactions between climate and global risk factors; and (iii) we provide comparative evidence for both advanced and emerging economies, highlighting heterogeneities in how climate-related uncertainty transmits into sovereign risk. Together, these contributions clarify the rationale for our work and position it within the frontier of sovereign risk research. Our methodology is aligned with recent empirical advances that advocate flexible and nonlinear approaches better to capture the complex determinants of sovereign credit markets (Pan et al. 2024; Gilchrist et al. 2022; Bajaj et al. 2023).

Theoretically, the CPU can influence sovereign credit risk through several channels. First, heightened uncertainty around the timing, scope, and stringency of climate policies may increase investors’ risk premia, as governments face unclear fiscal burdens related to transition costs, green investments, or stranded assets. Second, the CPU may amplify perceptions of policy inconsistency or credibility gaps, which are particularly relevant for sovereign creditworthiness where long-term fiscal stability is essential. Third, the CPU interacts with global financial risk by exacerbating volatility in capital flows and raising refinancing costs, especially in economies dependent on external financing. These mechanisms imply that sovereign CDS spreads may respond not only to realized climate shocks but also to uncertainty about the future policy path, with effects varying by institutional strength and market regime. This perspective is consistent with sovereign risk theory (Eaton and Gersovitz 1981; Arellano 2008), which emphasizes that sovereign borrowing costs depend critically on default risk under different states of the world, and with the literature on policy uncertainty (Pástor and Veronesi 2013; Julio and Yook 2012), which shows that uncertainty shocks can generate asymmetric and state-dependent effects on financial markets.

The central research questions guiding this study are as follows: (i) Does climate vulnerability significantly influence sovereign credit risk, and if so, how does this relationship vary across different quantiles of market conditions? (2) To what extent do global risk factors moderate or amplify this relationship? (3) Are the effects symmetric across advanced and emerging economies? Building on these questions, we formulate the following hypotheses: (H1) CPU exerts a positive and nonlinear impact on sovereign CDS spreads, with more potent effects in high-risk regimes; (H2) The influence of CPU is amplified by global risk factors (GPR, EPU, VIX), particularly during periods of financial stress; (H3) The impact of CPU is asymmetric across advanced and emerging economies, with emerging markets displaying greater sensitivity and volatility due to weaker institutional frameworks.

By addressing these questions, this paper makes several key contributions. First, we provide a comprehensive empirical assessment of the climate-sovereign risk nexus using a rich quantile-based framework that accounts for both nonlinearity and interaction effects. Second, we present new evidence on how climate risk transmission differs between advanced and emerging economies, particularly under conditions of financial stress, uncertainty, or geopolitical tension (Afonso et al. 2024; Aljarba et al. 2024; Naifar 2024). Finally, our study informs rating agencies, sovereign investors, and policymakers by identifying when and how climate vulnerability materially affects sovereign risk, thereby providing a more detailed basis for integrating climate metrics into sovereign credit assessments.

Our empirical analysis indicates several important insights into the dynamic relationship between CPU and sovereign CDS spreads. First, the QQR results demonstrate that the effect of CPU on sovereign risk is highly nonlinear and regime-dependent across countries. In developed economies, CPU tends to exert pronounced upward pressure on CDS spreads primarily during high-risk regimes while exhibiting muted or even negative effects in more stable market conditions, suggesting a degree of policy credibility and institutional resilience. In contrast, emerging markets show more volatile and asymmetric responses: in some cases, even moderate CPU shocks under tranquil conditions trigger a rise in sovereign CDS, highlighting investor sensitivity to climate uncertainty in weaker institutional contexts. Second, the MQQR approach confirms that global risk factors (GPR, EPU, VIX) significantly shape the transmission of CPU into sovereign risk, often amplifying effects during periods of joint financial stress and dampening standalone CPU impacts in calmer periods.

The remainder of this paper is organized as follows. Section 2 reviews the relevant literature. Section 3 describes the data and research methodology. Section 4 presents the empirical results and provides an in-depth discussion of the findings. Finally, Section 5 concludes the paper with a summary of key insights, policy implications, and suggestions for future research.

2. Literature Review

Understanding the multifaceted relationship between climate vulnerability and sovereign credit risk has gained momentum in recent years, driven by the escalating effects of climate change, rising policy uncertainty, and an increase in instances of geopolitical instability. A broad range of empirical studies have examined the determinants of sovereign credit risk, proxied by sovereign credit default swap (CDS) spreads, across diverse methodological and thematic dimensions.

A growing body of research highlights the impact of geopolitical risk on sovereign credit default swap (CDS) markets. Demiralay et al. (2024) find that country-specific geopolitical risks significantly elevate sovereign CDS spreads, particularly during periods of heightened market volatility and weak economic performance. Similarly, Naifar and Aljarba (2023) document asymmetric and heterogeneous effects of GPR on sovereign CDS across nineteen countries using QQR analysis, with more potent effects in upper quantiles. The influence of geopolitical events on sovereign risk is also evident in the Eurozone, as noted by Bratis et al. (2024), who identify volatility spillovers between GPR and sovereign CDS during the 2009–2012 crisis. Afonso et al. (2024) extend this by showing that geopolitical tensions, particularly in border regions, elevate sovereign risk in Europe, especially during turbulent periods. In country-specific studies, De Wet (2023) and Subramaniam (2022) demonstrate the asymmetric impact of GPR on Australian sovereign yields, emphasizing more potent effects in medium- and long-term maturities.

The role of global and macro-financial uncertainty is further explored in studies such as Gilchrist et al. (2022), who demonstrate that global financial risk has a strong influence on sovereign bond spreads, and Inoguchi (2025), who finds dynamic conditional correlations between the VIX and CDS spreads, particularly for emerging market economies (EMEs). Bajaj et al. (2023) highlight how oil prices and global risk indicators, such as the VIX, shape CDS spread dynamics in G20 economies. In a similar context, Demir and Danisman (2021) find that while economic uncertainty depresses bank credit growth, geopolitical risks have no significant aggregate effects. Other studies focus on the interconnectedness of sovereign risk markets. Aljarba et al. (2024) uncover significant volatility spillovers among the CDS spreads of emerging markets, influenced by global factors including the VIX, EPU, and GPR. Huang (2025) and Nagy and Neszveda (2025) provide evidence that geopolitical conflicts, particularly the Russo-Ukrainian War, significantly widen CDS spreads and act as early signals of financial resilience. The broader impact of economic and political uncertainty on CDS markets is supported by Pan et al. (2024), Blommestein et al. (2016), and Galariotis et al. (2016), who find that global or EMU-wide factors, as well as investor sentiment, are significant drivers of CDS spreads.

Concurrently, a separate yet increasingly integrated body of research investigates the relationship between climate risk and sovereign credit. Zenios (2022) argues that climate risks are priced in sovereign credit markets, proposing the integration of Integrated Assessment Models (IAMs) with debt sustainability analysis. Klusak et al. (2023) simulate climate-adjusted sovereign credit ratings and show that stringent climate policies could mitigate climate-induced downgrades. Sun et al. (2023) confirm the relevance of climate vulnerability and readiness in sovereign rating models, while Boitan and Marchewka-Bartkowiak (2022) find that countries with low climate resilience pay higher risk premia.

Climate uncertainty has also emerged as a key determinant of sovereign credit risk. Guo et al. (2025) demonstrate that a rise in CPU raises sovereign borrowing costs, particularly in vulnerable and less resilient economies. Naifar (2024) confirms this by showing that CPU drives spillovers in G20 CDS spreads. Additionally, Naifar (2023) demonstrates that climate vulnerability increases the spread of CDS, while climate readiness reduces it. Beirne et al. (2021) corroborate these findings, concluding that climate vulnerability has a substantial influence on sovereign bond yields in both advanced economies (AEs) and emerging market economies (EMEs). Capasso et al. (2020) and Collender et al. (2023) extend this to firm-level and transition risk indicators, linking carbon emissions and natural resource dependence to elevated credit risk.

From a methodological perspective, recent works have employed a variety of quantile-based and non-linear models. Boumparis et al. (2017) and Ramírez-Rondán et al. (2023) utilize panel quantile regressions to capture heterogeneous effects of uncertainty and political institutions on credit ratings. Naifar et al. (2020) employs quantile regression to distinguish between the roles of global and local drivers in Gulf Cooperation Council (GCC) countries. Pan et al. (2021) and Tekin (2025) emphasize the importance of differentiating between short-term and long-term dynamics when analyzing sovereign CDS spreads in response to economic and political shocks. Despite these advancements, a crucial gap remains in understanding how the joint distributional characteristics of climate vulnerability and sovereign credit risk change across different quantiles and interact with macro-financial and policy-related uncertainty factors. While several studies employ quantile regression (QR) or panel quantile regression to explore the heterogeneity in the sovereign credit risk nexus, only a few have adopted the more flexible QQR framework to uncover the nonlinear and quantile-dependent interactions. However, to the best of our knowledge, no existing study has employed the extended Multivariate Quantile-on-Quantile Regression (MQQR) model to analyze sovereign CDS spreads in the context of climate risk. In particular, the integration of CPU and global financial risk factors (e.g., GPR, EPU, VIX) within a unified, quantile-sensitive empirical framework remains largely absent in the extant literature. This theoretical link is consistent with the literature on uncertainty and sovereign credit risk more broadly. The EPU has been shown to elevate sovereign borrowing costs through channels of fiscal risk, investor sentiment, and capital market volatility (Demiralay et al. 2024). By analogy, CPU represents an emerging source of policy uncertainty with potentially similar but climate-specific implications for sovereign risk. Sovereigns perceived as slow, inconsistent, or unpredictable in climate policy may face higher refinancing costs as investors demand compensation for regulatory ambiguity and transition risk. Conversely, credible and predictable climate policy frameworks may mitigate these effects by reducing long-term fiscal and economic uncertainty.

This study fills this gap by employing a robust MQQR approach to examine how climate risk affects sovereign credit risk across the distribution of sovereign CDS spreads, and how global uncertainty indicators condition this relationship. By applying this advanced methodology to a comprehensive dataset of developed and emerging economies from 2010 to 2025, this study contributes new insights into the distributional and conditional effects of climate and policy risks on sovereign credit markets.

3. Data Description and Research Methodology

3.1. Data Description

This study utilizes a monthly panel dataset covering the period from February 2010 to March 2025 for ten countries: Australia, France, Germany, Japan, South Korea, the United Kingdom, the United States, Brazil, Indonesia, and South Africa. These countries represent a mix of developed and emerging economies, allowing us to explore heterogeneities in the climate risk–sovereign credit risk nexus across varying levels of economic resilience and fiscal capacity. All sovereign CDS spreads were obtained from the Investing.com platform and serve as a forward-looking measure of sovereign credit risk (Yang and Hamori 2023; Gilchrist et al. 2022). Methodological and data considerations primarily guide our choice of countries. First, from an initial sample of sixteen countries, we applied the Brock–Dechert–Scheinkman (BDS) test to identify the presence of nonlinear dependence in CDS returns. Only the ten retained countries exhibited statistically significant BDS test results, thereby justifying the use of nonlinear, quantile-based econometric methods1. Second, the selected countries ensure continuous and complete monthly data availability over the full sample period, which is crucial for preserving the comparability and robustness of our empirical estimations. Third, the classification of advanced versus emerging economies follows the International Monetary Fund (2023) standard: Australia, France, Germany, Japan, South Korea, the United Kingdom, and the United States are classified as advanced economies, while Brazil, Indonesia, and South Africa are classified as emerging economies. This provides an internationally recognized and replicable benchmark for our categorization. Finally, regime classification into high-risk and low-risk states is determined in a data-driven manner using quantile thresholds of the sovereign CDS distribution, which is fully consistent with the quantile-on-quantile methodological framework.

To capture climate-related uncertainty, we employ the CPU Index as a proxy for climate risk. The CPU index reflects investor concerns over regulatory ambiguity and long-term policy unpredictability associated with climate change. Its relevance is grounded in recent studies that document its significant role in shaping sovereign risk pricing (Guo et al. 2025; Naifar 2024). For instance, Guo et al. (2025) show that rising climate policy uncertainty amplifies sovereign borrowing costs, particularly for countries with lower resilience or higher vulnerability. In addition to the CPU, we control for three global risk factors commonly linked to sovereign risk spillovers. EPU, obtained from the PolicyUncertainty.com database, proxies for macroeconomic ambiguity and has been widely used to explain fluctuations in sovereign CDS spreads (Boumparis et al. 2017; Ramírez-Rondán et al. 2023). Studies by Pan et al. (2024) and Demir and Danisman (2021) confirm that elevated EPU tends to widen sovereign CDS spreads and restrict access to credit markets, especially during periods of financial fragility. Although the VIX originates from U.S. equity markets, extensive empirical evidence demonstrates that it functions as a global “fear gauge.” Sarwar and Khan (2017) show that increases in VIX significantly depress emerging market returns, especially during crisis periods. Smales (2022) finds that U.S. uncertainty shocks are transmitted to both developed and emerging financial markets, spreading “fear” globally. Likewise, Choi (2018) documents that VIX-driven financial uncertainty shocks substantially reduce output in emerging economies. Thus, the VIX serves as a globally valid proxy for market volatility and ensures comparability across the diverse set of countries in our sample, where region-specific volatility indices are often unavailable or inconsistent. The Market Volatility Index (VIX), collected from Investing.com, captures investors’ global risk aversion. Numerous studies highlight the VIX’s dominant role in sovereign CDS pricing (Stolbov 2017; Inoguchi 2025; Aljarba et al. 2024), with empirical findings indicating that spikes in VIX lead to cross-market spillovers and increased sovereign risk sensitivity, particularly in emerging economies. GPR captures country-level and global political instability. GPR is a well-established determinant of sovereign CDS spreads, particularly under conditions of conflict or crisis. Demiralay et al. (2024) find that country-specific geopolitical shocks significantly elevate sovereign risk premiums, especially during periods of heightened tension. Naifar and Aljarba (2023) further highlight the asymmetric effects of GPR across different quantiles and countries, emphasizing its nonlinear role in the transmission of sovereign risk.

The combination of sovereign CDS as a dependent variable and CPU, GPR, EPU, and VIX as independent or control variables allows us to explore how climate risk and global financial stressors jointly shape credit risk dynamics across quantiles. This selection is informed by an extensive body of literature emphasizing the state-dependent and nonlinear effects of these variables (Naifar and Aljarba 2023; Bajaj et al. 2023; Huang 2025). Furthermore, using monthly data helps smooth short-term noise while retaining sufficient temporal granularity for quantile-based estimation.

Table 1. Descriptive statistics of sovereign CDS spreads and global risk factors.

	Australia	France	Germany	Japan	South Korea	UK	USA
Mean	−0.007	−0.002	−0.006	−0.008	−0.006	−0.008	0.001
Median	−0.003	−0.001	−0.024	−0.01	−0.021	−0.017	−0.003
Min	−0.462	−0.507	−0.717	−0.429	−0.451	−0.611	−0.541
Max	0.515	0.753	0.787	0.555	0.581	0.82	0.832
Std_Dev	0.143	0.165	0.163	0.142	0.154	0.161	0.204
Skewness	0.396	0.73	0.778	0.382	0.418	1.054	1.052
Kurtosis	4.536	6.018	8.315	4.266	4.071	9.169	6.87
Jarque_Bera	22.524 ***	84.773 ***	231.276 ***	16.475 ***	13.93 ***	320.53 ***	146.332 ***
	Brazil	Indonesia	South Africa	CPU	GPR	EPU	VIX
Mean	0.002	−0.003	0.002	0.012	0.004	0.009	0.001
Median	−0.009	−0.011	0.002	−0.008	−0.009	−0.001	−0.009
Min	−0.309	−0.487	−0.242	−0.944	−0.6	−0.5	−0.614
Max	0.676	0.797	0.647	1.233	0.696	0.628	0.853
Std_Dev	0.141	0.163	0.125	0.366	0.203	0.18	0.238
Skewness	0.84	0.784	0.818	0.094	0.448	0.378	0.306
Kurtosis	5.072	6.035	5.78	3.213	4.176	4.083	3.657
Jarque_Bera	53.644 ***	88.027 ***	78.477 ***	0.61	16.495 ***	13.162 ***	6.08 **

Note: This table reports summary statistics for monthly log returns of sovereign CDS spreads for 10 selected countries, along with global risk indicators: Climate Policy Uncertainty (CPU), Geopolitical Risk (GPR), Economic Policy Uncertainty (EPU), and the Volatility Index (VIX), covering the period from February 2010 to March 2025. Jarque-Bera tests assess the normality of the return distributions, while the Augmented Dickey–Fuller (ADF) and Phillips–Perron (PP) tests examine the presence of unit roots. Asterisks ***, ** denote significance at the 1% and 5% levels, respectively.

reports the descriptive statistics of sovereign CDS spreads and the four control variables (CPU, GPR, EPU, and VIX) across the sample period.

Table 1 indicates considerable cross-country variation in the mean, standard deviation, and distributional characteristics of CDS returns. The United States exhibits the highest volatility (Std. Dev = 0.204), while South Africa displays relatively lower dispersion (Std. Dev = 0.125). Most CDS series are negatively skewed, except for those in the UK and USA, which exhibit strong positive skewness, suggesting asymmetry in the response of sovereign risk to shocks. The Jarque–Bera test statistics confirm that the majority of return distributions are non-normal at the 1% level, reinforcing the presence of fat tails and excess kurtosis. Turning to the explanatory and control variables, CPU shows the highest volatility (Std. Dev = 0.366) and wide range (–0.944 to 1.233), reflecting the episodic nature of climate-related policy shocks. GPR, EPU, and VIX also display significant excess kurtosis and strong Jarque–Bera statistics, indicating that uncertainty and volatility measures are prone to extreme realizations rather than normal variation. Among them, VIX demonstrates relatively high dispersion (Std. Dev = 0.238), consistent with its role as a barometer of global financial stress. These findings justify the use of nonlinear, quantile-based methods to capture distributional asymmetries, particularly under extreme market conditions.

3.2. Research Methodology

3.2.1. BDS Nonlinearity Test

To complement standard unit root diagnostics, we apply the Brock–Dechert–Scheinkman (BDS) test to assess whether the return series and global risk variables exhibit nonlinear dependence. The BDS test evaluates the null hypothesis that the data are independently and identically distributed (i.i.d.). Rejection of the null indicates the presence of nonlinear structure, thereby motivating the use of nonlinear quantile-based estimation techniques. The test is conducted across embedding dimensions M2–M6, where higher embedding dimensions capture more complex dependence patterns. M1 is excluded since it is equivalent to the trivial one-dimensional case.

3.2.2. Quantile-on-Quantile Regression (QQR)

To capture the nonlinear and state-contingent relationship between CPU and sovereign credit risk across different market regimes, we adopt a two-stage quantile-based econometric approach. First, we implement the univariate QQR to examine how the quantiles of sovereign CDS spreads respond to varying levels of CPU across its distribution. This approach allows us to detect heterogeneous effects and asymmetric transmission patterns that are not observable under linear or mean-based models. Before estimation, both the sovereign CDS spreads and explanatory variables are transformed into log returns to ensure stationarity and capture short-run changes. Therefore, although the equations below are expressed in levels for notational simplicity, the empirical implementation is conducted on stationary log-return series. Formally, the QQR specification is denoted as:

$${CDS}_t^{\tau }=f\left({CPU}_t^{\theta }\right)$$

(1)

This relationship is formalized as:

$${CDS}_t^{\tau }={\varnothing }_0+{\varnothing }_1{CPU}_t^{\theta }+{\epsilon }_t$$

(2)

where, ${CDS}_t^{\tau }$ represents the τ-th quantile of the sovereign CDS spreads at time t, ${CPU}_t^{\theta }$ denotes the $\theta$-th quantile of the climate policy uncertainty index, ${\varnothing }_1$ captures the quantile-specific dependence, and ${\epsilon }_t$ is the disturbance term.

3.2.3. Multivariate Quantile-on-Quantile Regression (MQQR)

To model the multidimensional influence of global risk factors on sovereign credit risk, we employ the MQQR framework, where the quantiles of sovereign CDS spreads (τ) are regressed on local values of CPU, while controlling for GPR, EPU, and VIX. Unlike standard quantile regression, this approach flexibly captures nonlinear and state-contingent dynamics in the CDS-CPU relationship, while maintaining tractability concerning control variables. The MQQR model is specified as:

$${CDS}_t^{\tau }=f\left({CPU}_t,{GPR}_t,{EPU}_t,{VIX}_t\right)$$

(3)

This relationship is formalized as:

$${CDS}_t^{\tau }={\varnothing }_0+{\varnothing }_1{CPU}_t^{\theta }+{\varnothing }_2{GPR}_t+{\varnothing }_3{EPU}_t+{\varnothing }_4{VIX}_t+{\epsilon }_t$$

(4)

where ${CDS}_t^{\tau }$ is the τ-th conditional quantile of sovereign CDS spreads. ${CPU}_t^{\theta }$ the local θ-th quantile level of CPU, i.e., captures nonlinear effects of CPU across its distribution. ${\varnothing }_1$ is a coefficient that varies across quantiles of both CPU and CDS (this is the essence of QQR). ${GPR}_t$, ${EPU}_t$, ${VIX}_t$ standard level variables (not quantile-transformed), hence modeled with constant coefficients ${\varnothing }_2$ to ${\varnothing }_4$.

4. Empirical Results

4.1. Pre-Test Diagnostics

Before proceeding with the main quantile-based estimations, we provide simple unconditional correlations as a descriptive benchmark. Figure A1 (Appendix A) reports Pearson correlation coefficients between sovereign CDS spreads and the four global risk indicators (CPU, GPR, EPU, VIX) over the sample period (February 2010–March 2025). As expected, these unconditional linear associations are modest and vary across countries, offering limited guidance on the underlying relationships. Given their weakness and instability, we do not interpret them further. Instead, they serve only as a preliminary check and motivate the use of nonlinear, state-dependent techniques such as QQR and MQQR, which are better suited to capture potential asymmetric and regime-contingent dynamics.

While the correlation analysis provides initial insights into the linear relationships among variables, it falls short in detecting the patterns that may characterize the data. To address this, the Brock–Dechert–Scheinkman (BDS) test serves as a robust statistical tool for identifying nonlinear dependencies in time series. Specifically, the BDS test examines whether a series (or its residuals) is independently and identically distributed (i.i.d.) by evaluating correlation integrals across different embedding dimensions. A rejection of the null hypothesis implies the presence of nonlinear dependence. By examining the residuals for violations of the independent and identically distributed (i.i.d.) assumption across various embedding dimensions, the BDS test offers compelling evidence on whether nonlinear modeling approaches are warranted. This diagnostic step is essential for preventing model misspecification and supports the validity of applying quantile-on-quantile regression techniques. In

Table 2. BDS Nonlinearity Test Results.

	Australia	France	Germany	Japan	South Korea	UK	USA
M2	2.8389 ***	4.0052 ***	3.0982 ***	3.0031 ***	3.1542 ***	1.8368 *	2.747 ***
M3	3.5574 ***	4.3108 ***	3.6419 ***	2.853 ***	3.4137 ***	1.8813 *	2.6168 ***
M4	3.4454 ***	4.3246 ***	3.6858 ***	3.3994 ***	3.7681 ***	2.1246 **	2.606 ***
M5	3.3793 ***	4.1365 ***	3.4465 ***	3.2688 ***	3.6561 ***	1.9652 **	2.5566 **
M6	3.1959 ***	3.9491 ***	3.2794 ***	3.0761 ***	3.8943 ***	2.1738 **	2.3485 **
	Brazil	Indonesia	South Africa	CPU	GPR	EPU	VIX
M2	1.9366 *	3.4551 ***	2.8046 ***	3.9663 ***	2.6965 ***	1.2001	6.5066 ***
M3	2.1715 **	4.0212 ***	3.6088 ***	3.8913 ***	2.7102 ***	1.6935 *	6.6413 ***
M4	1.8147 *	3.9345 ***	3.344 ***	3.9721 ***	2.6063 ***	1.8022 *	6.8474 ***
M5	1.491	3.9036 ***	2.8865 ***	3.4972 ***	2.6753 ***	1.6439	6.5223 ***
M6	2.099 **	3.7814 ***	2.8933 ***	3.1047 ***	2.8727 ***	1.3255	6.1866 ***

Note: This table reports the BDS (Brock-Dechert-Scheinkman) test statistics for sovereign CDS spreads and global risk variables (CPU, GPR, EPU, VIX) across embedding dimensions M2 to M6. The BDS test examines whether a time series is independently and identically distributed (i.i.d.) by evaluating correlation integrals across higher-dimensional embeddings. M1 corresponds to the trivial one-dimensional case and is therefore not reported, while M2–M6 represent embedding dimensions of 2 to 6, respectively. Rejection of the null hypothesis indicates the presence of nonlinear dependence. Asterisks denote significance levels: *** p < 0.01, ** p < 0.05, * p < 0.1.

, M2–M6 denote the embedding dimensions used in the test, where higher dimensions capture increasingly dependent structures. M1 is not reported because it represents the trivial one-dimensional case and does not provide helpful diagnostic information. Table 2 presents the results of the BDS test applied to sovereign CDS spreads for ten countries and global risk factors.

The test statistics for most countries and variables are statistically significant at conventional levels, particularly for Australia, France, Germany, Japan, South Korea, Indonesia, and South Africa, across all embedding dimensions, confirming the robustness of nonlinear structures. This provides strong justification for employing nonlinear methodologies such as MQQR. However, in some instances, such as Brazil (especially at higher embedding dimensions M4–M5), the UK, and EPU, exhibit weaker or marginal significance. For instance, Brazil loses statistical significance at M5, and the UK shows relatively lower BDS statistics, suggesting limited nonlinear dependence. Overall, the BDS test results support the appropriateness of a nonlinear quantile modeling strategy for most of the retained series, validating the study’s econometric approach and justifying the exclusion of countries or variables where nonlinear dependence is not confirmed.

4.2. Multivariate Quantile-on-Quantile Regression

To address potential omitted variable bias and capture the joint influence of global uncertainty on sovereign credit risk, we extend the univariate QQR framework by employing the MQQR approach. This methodology enables us to investigate how the effect of CPU on sovereign CDS spreads varies across different market states, while simultaneously controlling for major global risk factors. Figure 1 illustrates the nonlinear and state-dependent relationship between CPU and sovereign CDS spreads across all countries in the sample. Each country-specific panel presents the bivariate QQR surface on the left, capturing the partial effect of CPU across varying quantiles of both CPU and CDS without accounting for other risk drivers. The right panel in each country plot displays the MQQR surface, which incorporates additional global risk factors to isolate the marginal impact of CPU.

Across the sample of advanced and emerging economies, the relationship between CPU and sovereign CDS exhibits substantial nonlinearity and state dependence. In Australia, the CPU has asymmetric effects, exerting a negative or muted influence under stable credit conditions but significantly amplifying sovereign risk in high CDS regimes. This sensitivity persists, though it diminishes, once broader global risk factors (GPR, EPU, VIX) are controlled for. France similarly displays intensified responsiveness to CPU during joint stress states, with positive effects concentrated in the upper quantiles of both CPU and CDS. At the same time, more neutral or negative responses prevail during periods of low risk. Germany’s exposure is more conditional, with CPU increasing CDS spreads primarily in tranquil regimes, and only jointly contributing to risk when global uncertainties are elevated. Japan exhibits sharp nonlinearity, where mild CPU shocks can drive substantial CDS spikes during stress, yet moderate CPU is often associated with risk reduction, likely reflecting policy credibility. These dynamics are reinforced but more focused when controlling global risk factors. South Korea exhibits limited standalone sensitivity to CPU in stable periods, while significant positive effects emerge only under simultaneous spikes in CPU and CDS, especially when global risks are considered. In the United Kingdom, CPU consistently exerts adverse or negligible effects on sovereign spreads across most quantiles, implying resilience or institutional confidence, with only mild risk amplification observed under severe global stress. The United States presents a more systemic profile: CPU has limited impact under normal conditions but generates robust and consistent positive spillovers on CDS during joint tail events, reflecting its centrality in transmitting global climate-financial risks. Brazil’s CDS spreads increase in response to moderate CPU even under calm markets. However, stronger climate shocks under stress have a subdued or negative effect, possibly due to policy offsets or external buffers. Multivariate results confirm that global factors often dilute CPU’s role, except under specific stress regimes. Indonesia’s sovereign risk rises moderately in low-risk environments but declines in relevance during high-stress states, as global uncertainties dominate, as highlighted by the dampened CPU effects in the MQQR framework. Finally, South Africa demonstrates sensitivity to CPU mainly in lower CDS quantiles, suggesting that climate uncertainty elevates perceived risk under stable conditions but becomes less impactful when systemic volatility prevails. MQQR further shows that global factors absorb a significant portion of the CPU’s explanatory power. Collectively, these findings underscore the nonlinear, regime-specific nature of CPU’s influence on sovereign credit risk, with advanced economies generally showing more structured responses and emerging markets displaying more heterogeneous, state-contingent effects.

To enhance the interpretability and transparency of the results, we have supplemented the three-dimensional surface plots in Figure 1 with two-dimensional heatmaps presented in Figure 2. While Figure 1 illustrates the nonlinear and state-dependent relationship between CPU and sovereign CDS spreads for each country, both in the bivariate QQR specification (left panel) and in the MQQR specification with global risk controls (right panel), these plots alone make it difficult for the reader to gauge the precise magnitude and direction of the estimated partial effects. Figure 2 addresses this limitation by displaying the estimated ϕ1 coefficients across all (θ,τ) combinations in a heatmap format, incorporating numerical coefficient values.

Across countries, the results consistently show that climate policy uncertainty (CPU) exerts asymmetric and state-dependent effects on sovereign credit risk, with the magnitude and direction of influence varying by credit market conditions and interaction with global risk factors. In Australia, CPU amplifies sovereign risk primarily in moderate-to-upper CDS quantiles when uncertainty is high, though this effect weakens and partially overlaps with broader global risks once controls are introduced. France displays stronger standalone CPU effects in extreme states, with volatility across both tails, but after accounting for global uncertainty, the influence becomes more moderate, stable, and generally positive. For Germany, heightened CPU significantly raises risk in lower-to-mid CDS quantiles, and while global factors temper the impact, CPU retains an independent and persistent role in shaping credit spreads. Japan’s results highlight that CPU effects are mixed in the standalone setting but become more robustly positive under high CPU regimes when global risks are controlled, confirming its independent amplifying role in stressed environments.

In South Korea, CPU generates strong positive effects in elevated uncertainty states but negative ones under low CPU–high CDS conditions, with the MQQR results revealing a more balanced influence that persists mainly in high-uncertainty regimes. The UK shows relatively weak and often negative standalone effects, yet once global risks are controlled for, CPU’s positive role in driving sovereign risk becomes clearer, especially under high CPU conditions. In the U.S., CPU has strongly asymmetric standalone impacts, sharply raising risk in tranquil credit conditions while easing pressures under stress; with controls, these effects moderate but remain consistently positive in elevated uncertainty states. Emerging markets show similar conditional dynamics: in Brazil, CPU amplifies risk in stable markets and eases it in stressed ones, with the MQQR results confirming a persistent but more moderate influence; in Indonesia, CPU shocks produce both amplifying and dampening effects depending on prevailing market conditions, though after accounting for global risks, positive effects dominate at high CPU levels. Finally, in South Africa, CPU strongly raises risk in calmer conditions but reduces it under stress, with global factors tempering but not eliminating its independent effect, highlighting the nonlinear and conditional role of climate policy uncertainty across different sovereign credit environments.

These cross-country dynamics align with and extend insights from recent literature on sovereign risk under uncertainty. In particular, Guo et al. (2025) demonstrate that the effect of CPU on sovereign borrowing costs is more pronounced in less developed economies, where lower resilience and greater climate vulnerability amplify credit risk, which aligns with our findings for Brazil, Indonesia, and South Africa. Conversely, in advanced economies, where fiscal space and institutional strength often buffer external shocks, Zenios (2022) and Klusak et al. (2023) demonstrate that the pricing of climate risk becomes more conditional, materializing predominantly during periods of elevated stress. This mirrors our observation that the CPU significantly influences CDS spreads in developed countries, primarily in the upper quantiles, i.e., high-risk regimes. Additionally, the dilution of the CPU’s standalone role once global uncertainties are controlled for (MQQR results) is consistent with Gilchrist et al. (2022) and Inoguchi (2025), who find that global risk indicators, such as the VIX, substantially drive sovereign spreads, especially in emerging markets. More broadly, our findings align with Naifar (2024), who emphasizes that climate uncertainty-induced spillovers are highly contingent on country integration and regime type, with more integrated markets exhibiting stronger co-movement in response to global shocks. This study contributes to a growing body of evidence that sovereign credit risk under climate uncertainty is nonlinear, context-dependent, and moderated by both macro-financial resilience and exposure to global volatility.

5. Conclusions

This paper explores the nonlinear and state-dependent effects of CPU on sovereign CDS across developed and emerging economies using QQR and MQQR approaches. By incorporating global risk factors (GPR, EPU, VIX) in the MQQR framework, we aim to capture the marginal impact of CPU under varying credit market conditions and global uncertainty regimes. Our findings reveal that the CPU has a significant, asymmetric influence on sovereign CDS spreads, with its impact varying substantially across quantiles and countries. In developed economies, the CPU tends to exert pronounced positive effects under high-risk regimes, suggesting that climate uncertainty becomes particularly relevant during periods of financial stress. However, this influence often dampens in stable environments, reflecting stronger institutions, more credible climate policies, and deeper financial markets. In contrast, emerging economies display more volatile and contradictory responses, with CPU affecting sovereign spreads even in tranquil states. Yet, once global systemic factors are accounted for via MQQR, CPU’s role often diminishes, highlighting these countries’ greater exposure to external financial shocks and weaker climate resilience infrastructure.

These findings carry several important policy implications. For sovereign policymakers, the evidence suggests the need for more transparent and forward-looking climate strategies, particularly in emerging markets where climate-related policy effects are more erratic and pronounced. In advanced economies, strengthening policy credibility and consistency can help contain risk premiums during crisis episodes. In contrast, in emerging economies, institutional reforms and capacity building are essential to prevent the CPU from amplifying sovereign vulnerabilities even under normal conditions. Robust climate governance can mitigate risk premiums during financial turbulence.

For credit rating agencies, the results support incorporating CPU into sovereign risk models, moving beyond static ESG indicators to account for dynamic, state-contingent interactions between climate and finance. Explicit recognition of climate policy credibility as a rating factor would create more substantial incentives for governments to adopt predictable and long-term transition strategies. Global financial institutions, such as the IMF and the World Bank, should recognize the disproportionate climate vulnerability in emerging markets and develop tailored tools, including climate buffers, sovereign insurance, and infrastructure support, to build resilience. For advanced economies, multilateral support should prioritize global spillover management, ensuring that disorderly policy transitions do not destabilize emerging markets through capital flow reversals or rising spreads. Finally, investors are encouraged to integrate CPU indices into sovereign risk pricing and hedging models, particularly under joint stress conditions, as highlighted by the regime-dependent spillovers captured through QQR and MQQR. In practice, this means adopting differentiated pricing frameworks: penalizing policy unpredictability in emerging markets while rewarding credible transition pathways in advanced ones.

While the results offer valuable insights, future research could enhance this framework by incorporating country-specific measures of climate policy uncertainty that better reflect national regulatory developments. Additionally, examining the interactive effects between CPU and climate vulnerability, or fiscal space, could reveal important mediating channels in sovereign risk pricing. Further extensions may also assess the implications of climate-related risks within sovereign bond markets or investigate the dynamic role of climate-financial spillovers across asset classes and regions. Such efforts would enrich the understanding of how the CPU influences sovereign risk under varying structural and institutional conditions.