Green Bonds and Energy Markets Under Climate Risk Shock: A Spillover Perspective

Abstract

In the context of escalating climate change, it is imperative to understand its multifaceted impacts on financial markets, as climate risks not only affect the low-order moments (mean and variance) but also the high-order moments (skew and kurtosis) of the energy market and the bond market. This study employs a quantile vector autoregressive framework, a combination of time-domain and frequency-domain analyses, and quantile-to-quantile regression to assess the dynamic spillover effects under varying market conditions. The results reveal that spillover effects are particularly pronounced during extreme events, both high positive shocks (above the 80th percentile) or high negative changes (below the 20th percentile). Furthermore, during periods of high climate risks, the dynamic interaction between the energy market and green bonds intensifies, strengthening their roles in the context of spillover effects and altering their respective positions. Our findings also exhibit that the coal markets and green bonds act as net recipients of spillovers, highlighting their potential as effective hedging instruments. Finally, climate risks contribute to an increasing spillover of risk in the new energy sector, with the long-term trend showing the most significant growth in spillover intensity.

1. Introduction

The accelerating degradation of the natural environment has led to a marked increase in the frequency and intensity of unusual weather patterns and climatic anomalies. Phenomena such as sustained temperature rises, sea-level expansion, intensified storm systems, and widespread wildfire events now pose formidable threats to both energy infrastructure and the broader fabric of societal functioning. These challenges are becoming increasingly significant [1]. The systemic implications of climate-related hazards have therefore emerged as a central priority within financial market analysis, drawing considerable scrutiny from both practitioners and policymakers [2,3]. As documented by the Intergovernmental Panel on Climate Change (IPCC), the consequences of shifting climatic conditions have been observed across virtually every scientific discipline and throughout all components of the climate system. The term “climate change” encompasses the ongoing upward trajectory of mean global temperatures, the progressive warming of the planet, the escalation of severe weather systems, and a rising incidence of catastrophic episodes—notably coastal inundation driven by higher sea levels, prolonged agricultural dry spells, riverine and surface flooding, and hazardous thermal extremes that disrupt ecosystems across the globe. The interdependence between economic and financial networks and environmental equilibrium is profound, leaving these systems exposed to the compounding repercussions of sustained climatic shifts. As a result, a broad cross-section of industries has directed considerable research effort toward understanding the ramifications of climate-related hazards [4]. Climate risks, whether through physical harm, evolving market conditions, policy constraints, or reputational challenges, present significant difficulties for businesses and can lead to substantial financial losses.

Within this context, the energy industry stands out as especially exposed to both the tangible physical damages and the structural transition pressures brought about by shifting climatic patterns. Escalating heat levels, prolonged water scarcity, and catastrophic flooding events routinely interfere with energy generation capacity and distribution logistics. At the same time, regulatory frameworks designed to curb carbon emissions are propelling an accelerated worldwide pivot away from fossil-based energy carriers. Such transformations carry far-reaching consequences not merely for commodity pricing within energy markets but also for the broader financial architecture that underpins energy-related investment, production financing, and risk valuation. Furthermore, frequent climate disasters have introduced inevitable risks to energy supply, potentially impacting the energy market [5,6]. Meanwhile, various greenhouse gas emission reduction measures have been adopted to mitigate climate risks [7]. The energy structure has been optimized, and the proportion of renewable energy consumption has increased [8,9]. Measures addressing climate change have notably influenced the dynamics of the energy market, reshaping both supply structures and demand patterns [10].

To counter these threats, China has pursued an ambitious agenda of expanding clean energy capacity and developing sustainable financial instruments. Official statistics reveal that the proportion of non-fossil energy in China’s total primary energy consumption climbed from 15.5% in 2013 to 25.9% by 2023, signaling a decisive structural pivot toward greener energy sources. This trajectory reflects sustained improvements in the alignment between energy generation practices and ecological preservation goals, which has facilitated the growth of an energy-efficient economy. Such developments have simultaneously enhanced the synergy between energy production and environmental stewardship, while also transforming the operational landscape of both conventional energy and green finance markets vulnerabilities. Statistical data indicate that the share of new energy consumption in China has experienced significant growth, rising from 15.5% of total energy use in 2013 to 25.9% by 2023. This energy consumption shift reflects a structural transformation of the national energy mix toward cleaner and low-carbon sources. The compatibility between energy production and ecological sustainability has improved substantially, fostering the emergence of an energy-conserving society and further optimizing the overall energy consumption structure. These changes have improved the compatibility between energy production and ecological sustainability, while also reshaping the dynamics of energy and green financial markets.

During the last ten years, heightened recognition of planetary warming trends and the growing frequency of climate-driven catastrophes have refocused scholarly and practitioner attention on how climate-associated hazards propagate through financial and commodity markets [11,12]. Consequently, climate risk exposure has become an increasing concern for financial institutions, including asset managers and banks, particularly regarding asset allocation strategies and the management of loan portfolios [13]. Concerns over both physical climate risks and transition-related uncertainties are increasingly influential in shaping how financial institutions assess investment opportunities and monitor corporate behavior [14]. Simultaneously, the financial consequences of how green investment channels respond to shifting climate conditions have drawn considerable scholarly scrutiny. A notable regulatory insight from Hong et al. highlights that financial market participants’ limited familiarity with climate hazards may result in suboptimal adaptive responses [15]. Regulators worry that this knowledge gap could result in inadequate risk assessment and management practices when addressing climate-related financial challenges. Moreover, the correlation between energy prices and extreme events is of paramount importance in energy finance studies. Events such as droughts and reduced rainfall have profoundly impacted energy price risks [16]. It is anticipated that climate risks will also increase the frequency of extreme climate events [17]. Beyond the uncertainties inherent in climate dynamics, the sample period has also been shaped by notable geopolitical upheavals—notably the military confrontation between Russia and Ukraine and the escalating tensions in the Middle East—which have amplified turbulence across global energy markets and heightened financial instability. While these episodes are acknowledged here as relevant macroeconomic context for the investigation window, they are not modeled as explicit control variables or dummies within the baseline econometric specification. Rather, their influence is captured implicitly via the observed market dynamics and volatility patterns embedded in the baseline econometric model as separate control or dummy variables. Instead, their potential effects are reflected indirectly through observed market dynamics and volatility captured in the empirical framework.

Although the scholarly literature on the interplay among climate risk, sustainable finance, and energy commodities has expanded substantially, several notable lacunae persist. Prior investigations tend to concentrate on aggregate-level effects while largely ignoring the differential behavior across distinct energy sub-sectors (such as fossil-based versus renewable sources) and heterogeneous categories of green bonds. Furthermore, scant attention has been devoted to the temporal evolution of these interconnections—specifically how they manifest across varying investment horizons or under conditions of severe market stress. These deficiencies constrain our capacity to fully comprehend the channels through which climate risk propagates across interlinked market structures. This study seeks to address several key questions: Are there differences in the risk transmission mechanisms between climate risks arising from various sources and international energy and green bond prices? Do income spillover effects between these variables differ in intensity when comparing short-term versus long-term horizons? While climate risks affect energy prices, do fluctuations in energy prices further exacerbate climate risks?

To bridge these identified research gaps, the present investigation adopts the generalized forecast error variance decomposition methodology advanced by Diebold and Yilmaz, applying it to evaluate the cross-market spillover dynamics among energy commodities, green bond instruments, and indicators of climate risk [18]. Based on the empirical analysis, the approach developed by Baruník–Křehlík (BK) decomposes connectivity into long-, medium-, and short-term components, enabling a more nuanced evaluation of systemic risk during periods of market turbulence. Křehlik was employed to examine the changes in spillover-influencing factors across high, medium, and low frequencies [19]. An empirical study was conducted to investigate the relationship between energy prices, green bonds, and climate risks during different periods. Additionally, we employed the quantile vector autoregression (QVAR) method to explore spillover effects under various quantile conditions, thereby revealing the time-domain and frequency-domain characteristics among these variables. This approach enriches the research methods for such issues and enhances our ability to respond to and mitigate climate risks more effectively.

The principal contributions of this paper are threefold. In the first instance, it illuminates how cross-market linkages operate across varying spectral bands, thereby furnishing a more granular picture of inter-market dynamics. Specifically, the Diebold–Yilmaz (DY) framework is applied to gauge static spillover transmission among climate risk variables, energy sector indices, and green bond instruments. The analysis confirms that episodes of pronounced climate risk coincide with a notable elevation in the aggregate connectivity measure relative to tranquil market conditions. Complementing this, the Baruník–Křehlík (BK) method is employed to examine long-, medium-, and short-term connectivity, allowing for a detailed assessment of risks during market turbulence. Second, we use a quantile vector autoregression (QVAR) model to quantify spillover effects among climate risk, green bonds, and energy markets. Our analysis finds that the new energy market consistently exhibits relatively high risk spillover, highlighting its role as a major contributor to overall market risk. Furthermore, under median and extreme quantile transformations, the natural gas and coal markets demonstrate the most pronounced changes, with the net spillover index shifting from positive to negative, thereby acting as risk recipients. Third, our findings indicate substantially higher risk spillover effects in new energy markets. Notably, geopolitical events, such as the Palestine–Israel conflict, have influenced spillover dynamics across different time horizons, with short-term versus medium- to long-term responses varying among commodities. These heterogeneous responses reflect differences in supply chains, demand patterns, policy interventions, and degrees of financialization. Monitoring such changes is essential for investors to develop differentiated investment portfolios and optimize returns through frequency-based connectivity analysis.

The paper is organized as follows: Section 2 reviews relevant literature, Section 3 presents the research methodology, Section 4 discusses empirical findings and Section 5 concludes with policy implications.

2. Literature Review

2.1. Climate Risk and Green Bond Markets

With mounting societal recognition of climate-related threats, the green finance sector has evolved into an essential conduit for channeling investment capital toward environmentally sustainable endeavors [20]. Among green financial instruments, green bonds (GBs) have attracted considerable academic and investor attention due to their potential for financing environmentally responsible projects while hedging against climate risk [21]. For instance, in line with the International Capital Market Association’s (ICMA) Green Bond Principles, the green bond market has seen rapid expansion [22]. According to the Climate Bonds Initiative (CBI), global green bond issuance reached around $259 billion in 2019, reflecting a 51% year-on-year growth from 2018 and setting a new benchmark for the green finance market [23]. To this end, recent literature has explored the links between climate risk and green bond performance. Pastor et al. established a theoretical model demonstrating that climate risk creates return performance heterogeneity between sustainable and non-sustainable industries through two distinct channels [24]. This finding is supported by Ardia et al. [25] and Bua et al. [26], who classify climate risk into two primary subtypes: physical climate risk and transitional climate risk. These risks are believed to positively influence the stock market [27,28] and the commodity market [29,30].

Additionally, policy uncertainty, particularly in leading financial systems like the United States, can significantly influence the adoption and performance of green bonds [31,32]. The significance of studying green bonds has been amplified due to the uncertainty and fluctuations in environmental policies. This uncertainty significantly influences both the adoption and efficacy of green bonds as financial instruments, particularly their capacity to mitigate climate risks amid volatility [33]. Wang et al. analyzed corporate climate risk perceptions following green bond issuance, demonstrating enhanced climate risk awareness among most firms post-issuance [34]. Sartzetakis further examined the role of green bonds in facilitating low-carbon transitions while identifying strategies to overcome barriers to market expansion [35]. Despite this progress, current research often treats green bonds as isolated instruments, with limited attention to their dynamic interactions with energy markets and climate shocks.

2.2. Climate Risk and Energy Markets

Shifting climatic conditions have increasingly become a dominant source of price instability and systemic vulnerability within energy trading arenas. A substantial body of empirical evidence demonstrates that severe meteorological disturbances, sustained temperature increases, and large-scale natural catastrophes are capable of disrupting the equilibrium of energy supply networks and commodity pricing. For example, Wen et al. [16] demonstrate that environmental shocks and pandemics increase volatility in crude oil markets, while Lee [36] found that natural disasters significantly disrupt the supply–demand equilibrium of both conventional and renewable energy sources. Research by Liu and Chen [37] and Liang [38] further confirm the pivotal role of extreme climate conditions in driving price fluctuations in traditional energy markets. Moreover, the operational efficiency of power systems, especially in thermal, hydroelectric, and wind generation, is closely tied to climatic variables [39]. Abnormal climate phenomena may also damage power transmission infrastructure, thereby compromising the security and reliability of electricity supply systems [40].

In the renewable energy sector, hydropower and biofuels are particularly vulnerable to environmental conditions [41,42]. Wang et al. [43] specifically highlight the market risks arising from the inherent intermittency of solar and wind power generation, while Sarker’s research team found that clean energy pricing mechanisms cannot be insulated from the influence of conventional energy supply–demand dynamics [44]. Additionally, climate mitigation policies, such as carbon pricing and fossil fuel extraction limits, have introduced new uncertainties into traditional energy pricing models [45]. Of particular note is how extreme climate events or major policy announcements tend to intensify the interaction mechanisms between conventional and green energy markets. This situation underscores the urgent need for the international community to recognize the critical value of climate policy regulation in facilitating energy structure optimization and maintaining market stability [46].

2.3. Spillover Analysis

To adequately represent the intricate web of risk dependencies linking financial and energy market segments, scholars have deployed a diverse toolkit of quantitative techniques. Early contributions relied on models including DCC-GARCH, BEKK-GARCH, and copula-based approaches to characterize volatility transmission. Nevertheless, these traditional frameworks exhibit notable limitations in modeling the nonlinearity, asymmetry, and time-conditional nature of risk propagation—shortcomings that become especially apparent during periods of acute market dislocation.

To this end, the variance decomposition method of prediction error proposed by Diebold and Yilmaz transcends the single-market perspective; however, its static attributes have prompted the innovative application of frequency domain analysis techniques. Spectral variance decomposition accurately captures the time-varying characteristics of risk transmission through frequency band division [18,47,48]. Additionally, the frequency-domain overflow index framework has become the technical benchmark for subsequent research [19]. This method breaks down risk spillover into various frequency bands, allowing for the evaluation of connectivity frequency response variables across different types and enabling the analysis of risk spillover effects over distinct time periods. Recent research has constructed a complex network system that includes energy commodities, financial variables, and the stock market. For instance, the frequency-domain correlation network established by the Ferrer team using the BK method not only confirmed cross-market risk transfer but also revealed differences in risk-driven mechanisms across various time scales [49]. In studies of extreme events such as COVID-19, time-frequency analysis methods have demonstrated unique value. The empirical research conducted by Jiang quantified the dynamic risk transmission between the financial and energy markets during the pandemic [50]. Meanwhile, scholars such as Umar found that the correlation fluctuations between energy markets during the crisis exhibited time-varying heterogeneity, providing a new explanation for the risk amplification mechanism [51].

The DY and BK methods primarily estimate conditional expectations to analyze the propagation of market shocks. However, the impact of these shocks varies under different market conditions. The research by Briss somewhat underestimates the effects of unexpected events in extreme scenarios [52]. To address this issue, some studies have enhanced the conditional mean VAR model by incorporating conditional quantile analysis. Khalfaoui employed quantile regression to study spillover effects in the energy market under extreme conditions. This study utilizes the quantile vector autoregression (QVAR) model to quantify spillover effects at the 0.10 and 0.90 conditional quantiles, offering a novel approach to study spillover effects at different shock intensities [53]. Ando further extended the QVAR model based on the DY framework, providing a deeper analysis of network structure and cross-quantile spillover heterogeneity, going beyond the reliance on mean-based methods [54]. Additionally, Chatziantoniou proposed a quantity-driven time-frequency spillover framework that merges the QVAR model with the frequency-domain analysis of the BK method. This framework highlights the cyclical behavior of risk transmission, focusing on specific quantiles and frequency bands in the energy market. It further explores volatility spillovers at various quantiles across distinct frequencies, thereby improving the model’s robustness and adaptability [55]. Moreover, Ando emphasized that, compared to traditional conditional mean estimators like OLS, this quantile-based approach is less susceptible to outliers, enhancing estimation precision and reliability [54]. Similarly, Saeed’s study also tended to underestimate the effects of rare events under extreme conditions [52]. In response, recent research has advanced VAR modeling by incorporating conditional quantile mechanisms. Khalfaoui’s quantile regression approach to energy market dynamics under extreme scenarios supports the relevance of Ando’s quantile-based DY model in capturing network structure and spillover diversity across quantiles.

3. Statistical Analysis and Methodology

3.1. Data

This study utilizes a comprehensive dataset covering the period from 1 January 2018 to 29 December 2023, which enables the investigation of the dynamic interactions among China’s green bond market, energy markets, and climate-related uncertainty. This period is marked by significant developments in China’s green finance sector, energy transition policies, and heightened climate risk awareness, making it particularly suitable for empirical analysis.

To represent the green bond market, this study employs the China Green Bond Index, jointly compiled by the Shanghai, Chinabond Financial Valuation Center Co., Ltd., and the Beijing, China Energy Conservation Standardization Technical Service Co., Ltd. As the first officially recognized benchmark index aligned with the Green Bond Endorsed Project Catalogue, it offers a comprehensive and authoritative reflection of the domestic green bond market. It captures both the development trends and pricing behavior of China’s green bond instruments and has been extensively cited in academic literature as a proxy for evaluating the performance and systemic characteristics of green debt instruments in China [56,57]. All green bond data are sourced from the Wind Financial Terminal (https://www.wind.com.cn/).

The energy market is examined through a selection of five indices that reflect both renewable and conventional energy dynamics. Specifically, the Hydropower Index (Water) and the Wind Power Index (Wind) serve as proxies for the renewable energy sector, reflecting the price movements of green energy in China. In contrast, the ICE Crude Oil Futures Price Index (Oil), the NYMEX Natural Gas Price Index (Gas), and the Coal Index (Coal) capture the dynamics of the traditional fossil fuel market, illustrating the price volatility of high-carbon energy sources. All energy-related data are sourced from the Wind database. The use of futures prices, rather than spot prices, is justified by their superior liquidity and trading volume, and ability to incorporate market expectations regarding global supply–demand dynamics and price expectations [58]. In addition, energy futures play an important role in financial markets as safe-haven assets and risk management tools, particularly during periods of heightened market volatility [59,60]. Consequently, energy futures prices are frequently employed as key indicators by policymakers in the development of strategies and market stabilization efforts.

Furthermore, given that climate risk is characterized by significant uncertainty and externality, this paper introduces the global climate risk glossary developed by scholars such as Rognone, Bua, Kapp, and Ramella in their research titled “The Transformation of European Financial Markets and Physical Climate Risk Pricing: A Text-based Approach. It is constructed using text mining techniques and authoritative information sources, systematically identifying key terms related to Transition Risk (TRI) and providing correlation rankings at the phrase level. This vocabulary list has been widely utilized to quantify the degree of climate risk exposure in the financial market. The data are publicly accessible at http://www.policyuncertainty.com/china_epu.html (accessed on 15 January 2026).

Table 1. Descriptive statistics.

	TRI	GB	Gas	Oil	Coal	Wind	Water
Mean	−0.0010	0.0373	−0.007	−0.2955	0.0143	0.2516	−0.0810
Maximum	0.1007	1.9738	0.8330	144.71	127.00	275.67	54.030
Minimum	−0.0692	−0.7755	−1.4390	−249.15	−78.45	−213.667	−65.837
Std. Dev.	0.0228	0.1034	0.1688	28.102	6.4999	41.700	12.338
Skewness	0.2558	1.7071	−0.8654	−1.3961	2.5297	0.3490	−0.5484
Kurtosis	0.7275	27.929	10.229	10.229	141.73	5.3591	3.2478
J-B	46.8765	46,558.9	6333.51	6615.51	1,182,676	1719.83	692.476
ADF	−8.1 ***	−8.4 ***	−9.1 ***	−11.0 ***	−10.2 ***	−10.7 ***	−10.4 ***

Note: J-B represents the Jarque-Bratest test statistic; *** represents significance at the 1% level.

shows the descriptive statistics for each variable. The Jarque–Bera (J-B) test indicates that none of the variables follow a normal distribution. Additionally, the results from the unit root test confirm the stationarity of all series, thereby satisfying the prerequisites for the application of the QVAR model.

3.2. Methodology

This study extends the analytical framework pioneered by Chatziantoniou et al. [55], employing the generalized forecast error variance decomposition approach within the quantile vector autoregressive (QVAR) model to measure the tail-risk spillover effect caused by climate risks. This method can effectively identify the risk transmission paths of the energy and green bond markets under different shock intensities, time dimensions and cycle characteristics. In terms of model setting, the stable state of the market is represented by the median of conditions. Meanwhile, the special situations of sharp market decline and rapid market rise are captured, respectively, through the two extreme quantiles of 0.10 and 0.90.

To construct a complete connectivity measurement system, we first establish the quantile vector autoregression model QVAR(p), and its mathematical expression is as follows:

$$x_t=\mu \left(\tau \right)+{\varphi }_1\left(\tau \right)x_{t-1}+{\varphi }_2\left(\tau \right)x_{t-2}+\cdots +{\varphi }_p\left(\tau \right)x_{t-p}+\mu \left(\tau \right)$$

(1)

where $x_t$ and $x_{t-i}$, $i=1,\dots p$ are $N\times 1$ endogenous variable vectors, $\tau$ takes values in $\left[0, 1\right]$, representing quantiles, $p$ is the number of lags in the QVAR model, $\mu \left(\tau \right)$ is the $N\times 1$ conditional mean vector, and $N\times N$ is the QVAR coefficient matrix. ${\mu }_i\left(\tau \right)$ proves that the error vector $N\times 1$ has an error variance–covariance matrix of $N\times N$, $\sum \left(\tau \right)$. The QVAR (p) model is transformed into the quantile vector moving average representation of QVAR (∞) using the Wold theorem:

$$x_t=\mu \left(\tau \right)+\sum _{j=1}^p{\varphi }_j\left(\tau \right)x_{t-j}+\mu \left(\tau \right)=\mu \left(\tau \right)+\sum _{i=0}^{\infty }{\Psi }_i\left(\tau \right)u_{t-i}$$

(2)

Subsequently, the generalized forecast error variance decomposition (GFEVD) is computed to apply the connectivity method [61,62]. The proportion of forecast error variance, as shown by the GFEVD, represents the impact of disturbances on the series. This metric quantifies the effect imposed by the disturbances and can be formulated as follows:

$${\theta }_{ij}\left(H\right)={\displaystyle \frac{{\left(\sum \left(\tau \right)\right)}_{jj}^{-1}{\sum }_{h=0}^H\left({\left({\Psi }_h\left(\tau \right)\sum \left(\tau \right)\right)}_{ij}\right)}{{{\sum }_{h=0}^H\left({\Psi }_h\left(\tau \right)\sum \left(\tau \right){\Psi }^{\prime }\left(\tau \right)\right)}_{ii}}}$$

(3)

$${\tilde{\theta }}_{ij}\left(H\right)={\displaystyle \frac{{\theta }_{ij}\left(H\right)}{{\sum }_{k=1}^N{\theta }_{ij}\left(H\right)}}$$

(4)

To ensure the rows of ${\tilde{\theta }}_{ij}\left(H\right)$ sum to 1, each row must be normalized by its sum to get ${\tilde{\theta }}_{ij}$. By normalizing, this paper derives the following identities: ${\sum }_{i=1}^N{\tilde{\theta }}_{ij}\left(H\right)=1$ and ${\sum }_{j=1}^N{\sum }_{i=1}^N{\tilde{\theta }}_{ij}\left(H\right)=N$. As a result, the sum of each row in the matrix equals 1, meaning that a shock in one sequence affects not only that sequence but also the other sequences to some extent $j$.

In the next stage, we compute various connectivity metrics, starting with the network pairwise connectivity (NPDC), which is derived through the following process:

$$NPDC\left(H\right)={\tilde{\theta }}_{ij}\left(H\right)-{\tilde{\theta }}_{ji}\left(H\right)$$

(5)

If $NPD{C}_{ij}\left(H\right)>0\left(NPD{C}_{ij}\left(H\right)<0\right)$, it indicates that the influence of series $j$ on series $i$ is stronger (or weaker) than that of series $i$ on series $j$. Thus, if $NPD{C}_{ij}\left(H\right)>0$ and series $j$ dominates series $i$, the reverse is also true.

Unlike other overall directional connectivity measures (TO), this metric quantifies how shocks in sequence $i$ affect all other sequences $j$:

$$T{O}_i\left(H\right)=\sum _{i=1,i\ne j}^N{\tilde{\theta }}_{ji}\left(H\right)$$

(6)

Unlike other overall directional connectivity measures (FROM), this metric quantifies how shocks in sequence $j$ affect all other sequences $i$:

$$FRO{M}_i\left(H\right)=\sum _{i=1,i\ne j}^N{\tilde{\theta }}_{ij}\left(H\right)$$

(7)

Net total directional connectivity reflects the difference between the overall directional connectivity to and from other sequences, representing the net effect of sequence $i$ on the network.

$$NE{T}_i\left(H\right)=T{O}_i\left(H\right)-FRO{M}_i\left(H\right)$$

(8)

If $NE{T}_i>0\left(NE{T}_i<0\right)$, series $i$ has a greater impact on all other series $j$ than the impact it receives, making it a net shock emitter. On the other hand, if the received impact exceeds the impact it exerts on others, it is a net shock receiver.

The network connectivity level, as determined by the total connectivity index (TCI), can be computed using the following formula:

$$TCI\left(H\right)=N^{-1}\sum _{i=1}^{N}T{O}_i\left(H\right)=N^{-1}\sum _{i=1}^{N}FRO{M}_i\left(H\right)$$

(9)

Simply put, this metric reflects the average impact of shocks in one series on the others. A higher TCI value indicates greater market risk, while a lower value suggests lower market risk.

Up until now, we have been focusing on assessing connectivity in the time domain. Similarly, we now shift our focus to evaluating connectivity in the frequency domain. The Fourier transform of QVMA (∞) defines the spectral density of $x_t$ at frequency $\omega$.

$$S_x\left(\omega \right)=\sum _{h=-\infty }^{\infty }E\left(x_t{x}_{t-h}^{\prime }\right)e^{-i\omega h}=\Psi \left(e^{-i\omega h}\right)\sum t{\Psi }^{\prime }\left(e^{+i\omega h}\right)$$

(10)

By combining spectral density with generalized forecast error variance decomposition, the frequency generalized forecast error variance decomposition is obtained. Here, ${\theta }_{ij}\left(\omega \right)$ represents the conditional quantile at time $\tau$, showing the part of variable $i$ spectrum influenced by variable $j$.

$${\theta }_{ij}\left(\omega \right)={\displaystyle \frac{{\left(\sum \left(\tau \right)\right)}_{jj}^{-1}{\left|{\sum }_{h=0}^{\infty }{\left(\Psi \left(\tau \right)\left(e^{-i\omega h}\right)\sum \left(\tau \right)\right)}_{ij}\right|}^2}{{\sum }_{h=0}^{\infty }{\left(\Psi \left(e^{-i\omega h}\right)\sum \left(\tau \right)\left(e^{i\omega h}\right)\right)}_{ij}}}$$

(11)

$${\tilde{\theta }}_{ij}\left(\omega \right)={\displaystyle \frac{{\theta }_{ij}\left(\omega \right)}{{\sum }_{k=1}^N{\theta }_{ij}\left(\omega \right)}}$$

(12)

To examine the spillover effects of tail risk across different frequency domains, this paper defines a frequency band: $d=\left(a,b\right):a,b\in \left(-\pi ,\pi \right),a<b$. The spillover from variable $j$ to variable $i$ within this band is measured by ${\tilde{\theta }}_{ij}\left(d\right)={\int }_a^b{\tilde{\theta }}_{ij}\left(\omega \right)d\omega$. This enables the calculation of directional and total spillover indices across different quantiles and frequency domains.

$$T{O}_i\left(d\right)=\sum _{i=1,i\ne j}^N{\tilde{\theta }}_{ji}\left(d\right),FRO{M}_i\left(d\right)=\sum _{i=1,i\ne j}^N{\tilde{\theta }}_{ij}\left(d\right)$$

(13)

$$TCI\left(d\right)=N^{-1}\sum _{i=1,i\ne j}^{N}T{O}_i\left(d\right)=N^{-1}\sum _{i=1,i\ne j}^{N}FRO{M}_i\left(d\right)$$

(14)

Using Equation (13), the net spillover index (NET), $NE{T}_i\left(d\right)=T{O}_i\left(d\right)-FRO{M}_i\left(d\right)$ is defined in the quantile frequency domain. This index helps assess each energy market’s contribution to tail-risk contagion across different time periods.

4. Empirical Results and Discussion

4.1. Time-Domain and Frequency-Domain Average Connectedness

4.1.1. Time-Domain Average Spillover Effects

Using the Hannan–Quinn Criterion (HQ), Akaike Information Criterion (AIC), final prediction error (FPE), and Schwarz Criterion (SC), a vector autoregression (VAR) model with an optimal lag order of one was established using the DY methodology.

Table 2. Static spillover effects.

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	98.74	0.25	0.02	0.53	0.18	0.21	0.06	0.18
GB	0.22	98.77	0.07	0.26	0.46	0.06	0.17	0.18
Gas	0.05	0.13	97.33	1.94	0.31	0.17	0.07	0.38
Oil	0.20	0.18	1.65	96.88	0.31	0.29	0.49	0.45
Coal	0.08	0.10	0.27	0.28	99.24	0.00	0.03	0.11
Wind	0.16	0.12	0.10	0.21	0.00	62.93	36.48	5.30
Water	0.03	0.19	0.02	0.60	0.06	36.11	63.00	5.29
TO	0.11	0.14	0.30	0.54	0.19	5.26	5.33	TCI
NET	−0.07	−0.04	−0.08	0.09	0.08	−0.04	0.04	11.87

presents the static spillover matrices, which capture the spillover effects of climate risk on green bonds and various segments of the energy market. The values along the main diagonal of the matrix indicate the proportion of the forecast error variance attributable to each variable’s own lag, thereby reflecting its self-impact or intrinsic shock. In contrast, the off-diagonal elements represent the magnitude of interactions between variables, illustrating the interconnectedness within the spillover network. Additionally, the “TO” row denotes the total spillover transmitted by a given variable to others, while the “FROM” column indicates the total spillover received by a specific market from external shocks. The “NET” row captures the net spillover—calculated as the difference between the “TO” and “FROM” values—thus representing the direction and magnitude of tail-risk transfer for each market.

The results in Table 2 indicate that the overall connectedness is 11.87%, suggesting that, on average, 11.87% of the forecast error variance in each variable can be attributed to shocks from other variables within the system. Notably, the wind energy index is the largest recipient of spillover effects (5.30%), while the hydropower index serves as the largest source of spillover (5.33%), indicating that new energy sources are more susceptible to significant fluctuations within this system. In contrast, coal prices exhibit the weakest correlation among the three energy sources (0.11%), which aligns with the proportion of coal in traditional energy. Green bonds show negative risk spillover and will be affected by changes in other energy prices. As shown in the table, the spillover effect of Transition Risk Indicators (TRIs) on price changes in energy and green bonds is 0.11%, while the impact received from these markets is 0.18%, resulting in a notable negative net spillover of −0.07%. By energy type, TRI has the greatest spillover effect on oil prices (0.25%), followed by natural gas and coal prices. This suggests that the intense competition among major players in the international oil market, along with sharp fluctuations in oil prices, often serves as a barometer for international geopolitical dynamics.

As illustrated in Figure 1, the overall connectedness index experienced a significant increase during the outbreak of the COVID-19 pandemic and the Russia–Ukraine conflict. This indicates that during these periods, the interlinkages between the climate risk index, green bonds, and the energy market became markedly stronger, reflecting heightened systemic interactions amid global uncertainty.

4.1.2. Time-Frequency Average Spillover Effects

This study uses the BK frequency domain method to decompose the spillover effects into three parts: short-term, medium-term, and long-term.

Table 3. Static spillover effects.

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	84.29	0.23	0.02	0.53	0.13	0.17	0.04	1.11
GB	0.11	60.39	0.04	0.13	0.22	0.05	0.12	0.68
Gas	0.04	0.13	80.52	1.33	0.22	0.16	0.07	1.95
Oil	0.17	0.15	1.17	80.21	0.29	0.23	0.39	2.41
Coal	0.06	0.08	0.21	0.21	82.71	0.00	0.02	0.59
Wind	0.11	0.06	0.09	0.20	0.00	51.81	31.05	31.51
Water	0.02	0.10	0.01	0.39	0.05	29.43	51.60	30.01
TO	0.52	0.75	1.55	2.79	0.90	30.05	31.70	TCI
NET	−0.59	0.07	−0.40	0.39	0.31	−1.46	1.69	9.75
The spillover table for band 3.14 to 0.63. Roughly corresponds to 1 day to 5 days.
	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	12.20	0.02	0.00	0.00	0.04	0.03	0.02	0.12
GB	0.09	31.68	0.02	0.11	0.20	0.00	0.04	0.45
Gas	0.01	0.00	14.17	0.51	0.08	0.01	0.00	0.60
Oil	0.03	0.02	0.41	14.04	0.02	0.05	0.08	0.60
Coal	0.01	0.02	0.05	0.05	13.94	0.00	0.01	0.14
Wind	0.04	0.05	0.01	0.01	0.00	9.37	4.59	4.70
Water	0.01	0.07	0.01	0.17	0.01	5.62	9.60	5.89
TO	0.19	0.19	0.49	0.86	0.34	5.72	4.73	TCI
NET	0.07	−0.27	−0.11	0.25	0.20	1.02	−1.16	1.79
The spillover table for band 0.63 to 0.10. Roughly corresponds to 5 days to 30 days.
	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	2.25	0.00	0.00	0.00	0.01	0.01	0.00	0.02
GB	0.02	6.71	0.00	0.02	0.04	0.00	0.01	0.10
Gas	0.00	0.00	2.64	0.10	0.01	0.00	0.00	0.11
Oil	0.01	0.00	0.08	2.62	0.00	0.01	0.02	0.11
Coal	0.00	0.00	0.01	0.01	2.59	0.00	0.00	0.03
Wind	0.01	0.01	0.00	0.00	0.00	1.75	0.84	0.86
Water	0.00	0.02	0.00	0.03	0.00	1.05	1.80	1.11
TO	0.04	0.04	0.09	0.17	0.07	1.07	0.87	TCI
NET	0.02	−0.06	−0.02	0.05	0.04	0.20	−0.24	0.33

Note: The spillover table for band 0.10 to 0.00. Roughly corresponds to 30 days to Inf days.

presents the static total spillover effect of returns between climate risk indicators, green bonds, and energy prices across short-, medium-, and long-term frequency bands. The overall connectedness is measured at 11.87%, comprising short-term connectivity (9.75%), medium-term connectivity (1.79%), and long-term connectivity (0.33%). Notably, short-term connectedness makes up the largest share, suggesting that spillover effects among the four variables are most significant in the short term. In other words, shocks to one variable tend to be rapidly transmitted to others within a 1- to 5-day window, while such linkages are relatively weak over medium- and long-term horizons. Consequently, analyzing short-term spillovers is particularly critical, as it reflects the speed at which financial markets adjust to new information and external shocks [63]. The variation in its overflow is nearly equivalent to the total connectivity.

Notably, the returns of both natural gas and coal show similar responses to Transition Risk Indicators (TRIs). This pattern implies that natural gas prices possess diversification potential within the traditional energy sector. The World Health Organization also emphasizes the hedging and offsetting potential of natural gas prices within the context of global energy dynamics [64,65]. This may be attributed to the rapid expansion of the liquefied natural gas (LNG) market, which has enhanced global market liquidity and reduced dependency on major gas-producing and pipeline transit countries. The diversification of supply channels has, to some extent, diluted the geopolitical leverage of these countries. As a result, climate risks exert limited influence on the volatility of natural gas prices.

It is noteworthy that the direction of net spillover for green bond prices shifts when transitioning from short- to medium- and long-term horizons. Specifically, the positive net spillover effect observed in the short term gradually diminishes as the frequency decreases, ultimately turning negative at longer horizons. In contrast, wind energy demonstrates an opposite trend, exerting a greater influence on other variables in the medium and long term. Moreover, among the three energy sources, natural gas and coal exhibit the least susceptibility to external shocks during these periods, suggesting that natural gas, in particular, remains the least affected by other variables. Overall, short-term spillover effects are most significant during periods of heightened climate risk, while medium- and long-term spillovers, though generally weaker, may still exhibit some activity during these events. Future research could further combine high-resolution climate event indicators with frequency decomposition methods to systematically investigate the dynamic mechanisms of medium- and long-term risk transmission, thereby deepening the understanding of interactions between green bonds and energy markets.

Secondly, this paper categorizes the overall dynamic connectedness into three distinct frequency ranges, as illustrated in Figure 2. The black line represents the total connectedness, while the red line indicates short-term (high-frequency) connectedness covering 1 to 5 days, the green line denotes medium-term (mid-frequency) connectivity for 5 to 30 days, and the blue line reflects the long-term (low-frequency) connectivity for periods exceeding 30 days.

It is evident that the magnitude of dynamic connectedness across frequencies mirrors the static results, with the highest levels observed in the short-term domain, followed by medium- and long-term horizons. The pattern and amplitude of short-term connectedness closely track those of total connectivity, suggesting that the system’s overall spillover dynamics are predominantly driven by short-term interactions. However, the figure reveals that during the aforementioned period of heightened climate risks, medium- and long-term connectedness remain significantly higher than during calmer periods. Further analysis indicates that the timing of these peaks in connectedness coincides with surges in the Transition Risk Index (TRI), suggesting that major climate events played a pivotal role in amplifying medium- and long-term spillovers among the variables.

As illustrated in Figure 3, the dynamic net spillover estimates exhibit notable temporal variation and distinct patterns across variables. The figure reveals that the net spillover values for each variable fluctuate between negative and positive, with greater volatility observed during periods of high TRI levels. Notably, the prices of green bonds display clear peaks accompanied by sudden fluctuations. In contrast, the net spillover of coal remains relatively stable, consistent with the potential hedging effect of coal discussed earlier.

A closer analysis indicates that, for the majority of the time, the net spillover of green bonds and traditional energy prices is negative, suggesting their vulnerability to the influence of other variables. From the net spillover chart of the Climate Risk Index (TRI), it is evident that, for the majority of the observed period, the net spillover of TRI is negative. However, it turned positive in early 2022 and 2023, indicating that during these times, the climate risk index had a significant influence on the green bond and energy markets. Overall, the information transmission effect within energy prices influences climate risks and green bonds.

Figure 4 shows the net spillover effects across different time intervals and frequency bands. Compared to other periods, the short-term frequency band displays the widest range of net spillover values, indicating stronger spillover effects, which aligns with the results in Table 3. The net spillover value of TRI in the short term fluctuates between negative and positive. In contrast, in the medium to long term, the net spillover of TRI remains positive, indicating that it acts as an information transmitter, amplifying the price fluctuations of other variables. Therefore, the medium to long-term impact of TRI on green bonds and energy prices is more significant. The notable peaks in the net spillover of hydropower and wind power prices do not appear in the medium to long term; instead, they are more pronounced in the context of positive net spillover, further confirming the medium to long-term impact of climate risks on the new energy market.

4.2. Dynamic Quantile Connectedness

This paper employs the QVAR model to analyze the spillover effects between climate risks and green bonds and energy markets under extreme scenarios.

Table 4. Tail net risk spillover table (median of conditions).

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	98.74	0.25	0.02	0.53	0.18	0.21	0.06	1.26
GB	0.22	98.77	0.07	0.26	0.46	0.06	0.17	1.23
Gas	0.05	0.13	97.33	1.94	0.31	0.17	0.07	2.67
Oil	0.20	0.18	1.65	96.88	0.31	0.29	0.49	3.12
Coal	0.08	0.10	0.27	0.28	99.24	0.00	0.03	0.76
Wind	0.16	0.12	0.10	0.21	0.00	62.93	36.48	37.07
Water	0.03	0.19	0.02	0.60	0.06	36.11	63.00	37.00
TO	0.75	0.97	2.13	3.81	1.32	36.83	37.29	TCI
NET	−0.51	−0.25	−0.54	0.69	0.56	−0.24	0.29	11.87

,

Table 5. Tail-risk net spillover table (0.90 quantiles).

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	35.56	10.90	11.70	11.05	7.01	11.84	11.94	64.44
GB	11.72	40.21	11.00	9.41	6.20	10.76	10.70	59.79
Gas	11.89	10.90	34.01	12.67	7.01	11.91	11.61	65.99
Oil	11.02	8.96	12.33	35.00	7.49	12.46	12.73	65.00
Coal	9.09	7.78	9.16	9.88	46.33	8.85	8.91	53.67
Wind	10.87	9.19	9.94	11.10	6.58	28.41	23.90	71.59
Water	10.33	9.03	9.96	11.02	6.13	24.04	29.69	70.31
TO	64.92	56.75	63.90	65.14	40.42	79.86	79.80	TCI
NET	0.48	−3.04	−2.09	0.14	−13.25	8.27	9.49	64.40

and

Table 6. Tail-risk net spillover table (0.10 quantiles).

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	35.15	11.42	11.17	10.91	7.51	11.67	12.16	64.85
GB	12.10	37.33	11.09	10.36	7.15	11.20	10.77	62.67
Gas	11.01	10.97	35.17	12.63	7.08	11.64	11.50	64.83
Oil	11.16	10.48	12.57	33.41	7.31	12.69	12.38	66.59
Coal	9.30	8.72	9.19	9.66	45.22	8.83	9.08	54.78
Wind	10.51	9.66	10.06	11.25	6.61	28.79	23.12	71.21
Water	10.46	9.11	10.17	10.95	6.51	23.68	29.12	70.88
TO	64.55	60.35	64.25	65.76	42.16	79.71	79.01	TCI
NET	−0.29	−2.32	−0.58	−0.83	−12.62	8.51	8.13	65.11

illustrate the dynamic and symmetrical quantile features of the total connectivity display. In extreme cases, the risk spillover among these markets is more pronounced, with total connectivity fluctuations exceeding 60%.

These results corroborate the conclusions of recent empirical work, which demonstrates that cross-market risk transmission intensifies markedly when the economy experiences adverse shocks or deteriorating conditions [66]. The structural fluctuations in energy prices have altered the spillover effect of climate risks. Consequently, the competition among traditional energy sources has shifted to a contest between green bonds and new energy sources. Notably, fluctuations in energy commodity prices often lead to an increase in climate risks. However, the role of the Total Risk Index (TRI) in the table varies under extreme upward and downward market conditions. This variation may be attributed to the fact that as climate risk escalates due to changes in traditional energy prices, the financial nature of energy commodities further drives fluctuations in wind energy and hydropower prices, particularly green bond prices, through an indirect transmission mechanism that influences energy price changes.

From the perspective of the time-domain, the previously mentioned static spillover findings indicate that, under extreme circumstances, the tail-risk spillover intensity between green bonds and energy markets rises significantly, primarily as a result of heightened climate risk. However, since the spillover index is centered around the median quantile, it fails to adequately capture the impact of tail-risk spillovers in such extreme scenarios. Furthermore, the renewable energy market demonstrates the highest sensitivity to broader market dynamics, exhibiting a marked increase in targeted spillover levels. Nevertheless, conventional energy markets continue to act as the primary transmitters of risk, exerting a considerable influence on the overall spillover structure. These dynamics are further complicated by frequency-dependent responses, which reflect the influence of both macroeconomic cycles and heterogeneous investor behavior. Consequently, time-domain analysis alone is insufficient for comprehensively capturing the transmission of tail risk across varying time scales. This highlights the necessity of employing frequency-domain methods that take into account distinct spectral elements, particularly associated with low-frequency (long-term) and high-frequency (short-term) bands, when investigating tail-risk spillovers in global energy markets.

4.3. Time-Frequency Quantile Connectivity

Table 7. Time-frequency quantile connectivity (0.10 quantiles).

	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	35.15	11.42	11.17	10.91	7.51	11.67	12.16	64.85
GB	12.10	37.33	11.09	10.36	7.15	11.20	10.77	62.67
Gas	11.01	10.97	35.17	12.63	7.08	11.64	11.50	64.83
Oil	11.16	10.48	12.57	33.41	7.31	12.69	12.38	66.59
Coal	9.30	8.72	9.19	9.66	45.22	8.83	9.08	54.78
Wind	10.51	9.66	10.06	11.25	6.61	28.79	23.12	71.21
Water	10.46	9.11	10.17	10.95	6.51	23.68	29.12	70.88
TO	64.55	60.35	64.25	65.76	42.16	79.71	79.01	TCI
NET	−0.29	−2.32	−0.58	−0.83	−12.62	8.51	8.13	65.11
1–5
	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	30.60	9.51	9.42	9.40	6.62	10.06	10.72	55.73
GB	7.03	21.95	6.82	6.20	4.27	6.88	6.93	38.12
Gas	8.91	8.78	28.52	9.82	5.75	9.24	9.20	51.70
Oil	8.11	7.35	8.74	25.35	5.65	9.38	9.22	48.45
Coal	7.37	6.32	7.15	7.54	36.45	7.03	7.41	42.81
Wind	7.99	7.13	7.56	8.20	5.21	22.23	18.21	54.29
Water	8.22	7.20	7.86	8.16	5.10	18.23	22.85	54.77
TO	47.63	46.28	47.54	49.32	32.60	60.81	61.68	TCI
NET	−8.10	8.16	−4.16	0.87	−10.21	6.52	6.91	49.41
5–30
	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	3.75	1.37	1.35	1.19	0.67	1.26	1.18	7.02
GB	3.98	12.05	3.32	3.27	2.22	3.41	3.10	19.29
Gas	1.72	1.72	5.53	2.30	1.08	1.96	1.91	10.70
Oil	2.43	2.33	3.02	6.58	1.30	2.63	2.57	14.28
Coal	1.52	1.75	1.57	1.69	7.26	1.40	1.36	9.29
Wind	1.99	1.80	1.90	2.41	1.05	5.33	4.04	13.19
Water	1.83	1.45	1.83	2.26	1.11	4.48	5.22	12.97
TO	13.47	10.43	13.00	13.11	7.43	15.15	14.16	TCI
NET	6.45	−8.87	2.30	−1.17	−1.86	1.96	1.19	12.39
30–inf
	TRI	GB	Gas	Oil	Coal	Wind	Water	FROM
TRI	0.80	0.53	0.40	0.32	0.22	0.36	0.26	2.09
GB	1.09	3.33	0.96	0.89	0.65	0.92	0.75	5.26
Gas	0.38	0.47	1.13	0.50	0.25	0.45	0.39	2.43
Oil	0.62	0.80	0.81	1.48	0.36	0.67	0.59	3.86
Coal	0.41	0.65	0.47	0.44	1.51	0.40	0.31	2.68
Wind	0.53	0.73	0.60	0.65	0.35	1.24	0.87	3.73
Water	0.41	0.46	0.48	0.53	0.30	0.96	1.06	3.14
TO	3.45	3.65	3.71	3.33	2.14	3.75	3.16	TCI
NET	1.36	−1.61	1.28	−0.53	−0.54	0.02	0.02	3.31

illustrates the spillover effects among climate risks, green bonds, and the energy market, with a particular emphasis on the 0.10 quantile within the frequency domain. At this extreme lower quantile, the tail-risk spillover index is estimated at 49.41% in the short term, 12.39% in the medium term, and 3.31% in the long term. The findings indicate that, in the short term, risk spillover is primarily influenced by the overall tail risk within the energy market. When examining spillover effects across various time horizons, it becomes evident that the new energy sector, particularly the wind energy segment, exhibits a relatively higher level of risk transmission.

The interaction between green bonds and the energy market has emerged as a significant source of systemic risk. The net spillover effect of climate risk reveals initial heterogeneity; in the short term, natural gas and coal primarily act as risk recipients, while in the medium and long term, they evolve into sources of risk spillover. From the perspective of risk spillover, the new energy market, especially the wind energy sector, has demonstrated the most pronounced characteristics of risk spillover across short, medium, and long-term horizons.

Figure 4 illustrates the overall analysis outcomes of the dynamic connectedness index under a specific quantile configuration, where warm-colored regions denote areas of pronounced correlation features. The research identifies a significant bidirectional correlation among climate risk, green bonds, and the energy market, manifested as both positive coordinated changes and negative hedging relationships. Notably, the dynamic total connectedness index exhibited a significant upward trend during 2020. This empirical finding suggests that the interlinkages between the energy and green bond markets have intensified in response to climate-related shocks, thereby elevating the aggregate level of systemic market risk. This discovery elucidates the mechanism by which climate factors amplify the risk transmission within financial markets from a time-varying correlation perspective.

This study further investigates the dynamic network correlation characteristics under various quantile conditions (see Figure 5). In Figure 5, the blue area represents the negative spillover effect of the network, while the red area reflects the positive spillover effect. The empirical results indicate that the net spillover effect among different markets exhibits significant time-varying characteristics, fluctuating between positive and negative values over the sample period. This finding underscores the importance of dynamic correlation research. Notably, climate risk underwent a transformation from being a risk receiver to a risk exporter in 2021, reaching its peak spillover intensity in 2022. During normal market periods (with the exception of the stock market crash in 2023), the green bond market primarily functions as a risk absorber. It is important to highlight the clear risk-hedging relationship between traditional energy and new energy markets: coal and natural gas demonstrate significant role transformation characteristics under varying market conditions (both normal and extreme). This dynamic balance mechanism effectively serves as a risk buffer.

Figure 5 further reveals the evolutionary patterns of net spillover effects for each variable across various temporal dimensions and under different quantile scenarios. These dynamics are primarily influenced by the responsiveness of both the energy and bond markets to external disturbances. In the short term, the new energy sector primarily functions as a net risk transmitter, while conventional energy and green bond markets exhibit divergent pathways of risk transmission. Specifically, during periods of heightened climate risks, traditional energy demonstrates a significant short-term risk-hedging capacity. Under extreme market conditions, the risk transmission role of the energy market displays pronounced time-varying characteristics. In the medium- and long-term dimensions, various energy types undergo periodic transitions between being net risk exporters and net risk receivers. However, it is important to note that the risk-hedging effect of traditional energy shows a gradual attenuation trend over time. The green bond market exhibits distinct sensitivity characteristics, maintaining a negative network correlation throughout the time and frequency domains, which reflects its unique response mechanism to shocks from other markets. Reason: Improved clarity, vocabulary, and technical accuracy while correcting grammatical and punctuation errors.

4.4. Connectedness Network Results

The present study draws on network-theoretic methods to conduct an in-depth examination of how extreme tail risk propagates among climate risk variables, green bond instruments, and energy market indices. Leveraging the generalized forecast error variance decomposition technique embedded in the QVAR framework, a network of tail-risk contagion pathways is established, comprising seven major index variables. Each index serves as a node in the network, while directed edges illustrate the direction and magnitude of risk transmission among markets. In the time-domain context, Figure 6 illustrates the topological layout of the tail-risk transmission network under typical market conditions, while Figure 7 and Figure 8 specifically examine the contagion pathways of tail risk within the energy sector. To gain a comprehensive understanding of risk transmission characteristics, the study also incorporates a frequency-domain perspective analysis, examining the evolution of the network structure across different quantiles and conditions.

The results of the network-based analysis indicate that tail-risk spillover among markets exhibits significant structural characteristics (as shown in Figure 6, Figure 7 and Figure 8). Overall, the nodes within the risk spillover network demonstrate stronger connectivity, underscoring the robust nature of the risk transmission mechanism under extreme conditions. Specifically, the coal market and green bonds have established the closest two-way risk spillover relationship, reflecting a substantial risk linkage effect between the two. In contrast, the degree of risk correlation between the new energy market and the traditional energy sector is relatively limited, with the intensity of risk transmission being significantly weaker. This characteristic suggests that fluctuations in the global energy market are primarily influenced by the internal correlation networks of traditional energy markets, which serve as the main channels for the dissemination of systemic risks. The research findings confirm that when faced with the shock of extreme climate risks, the new energy market and green bonds can offer effective hedging options against the risks associated with the traditional energy market through their unique risk transmission pathways.

5. Conclusions and Policy Implications

Situated within the context of worldwide economic efforts to confront climate-related threats, this paper probes how climate risk transmits to green bond and energy market variables under extreme market states. The analytical framework integrates a quantile vector autoregressive (QVAR) model with spectral decomposition and network-based methods. The main empirical findings are as follows: (1) combining time-frequency spillover analysis with quantile regression reveals that extreme tail-risk transmission among climate risk, green bonds, and energy markets is highly state-dependent and temporally variable, becoming especially acute during extreme episodes; (2) the evidence points to a predominance of short-horizon spillover effects, which highlights the outsized influence of investor behavioral factors in driving short-lived market dislocations; (3) when climate-related risks intensify, the interdependence between green bonds and the energy sector strengthens considerably, leading to shifts in each market’s position within the broader risk transmission network; and (4) coal prices predominantly absorb rather than transmit spillover risk, suggesting their utility as a potential hedging vehicle.

On the policy front, our results imply that risk management frameworks should be calibrated to the prevailing market regime and investment horizon. Short-horizon market participants must remain alert to abrupt changes in spillover intensity, whereas regulatory authorities ought to reinforce climate risk surveillance and establish safeguards against cross-asset contagion. The directional decomposition of spillover flows identifies which markets function primarily as risk emitters versus risk absorbers, thereby furnishing policymakers with evidence-based guidance for targeted regulatory action. In addition, the amplified long-horizon spillover effects observed in the renewable energy sector deserve dedicated policy attention: by refining price discovery mechanisms and acknowledging the sector’s intermediary function linking conventional energy and green finance, regulators can bolster overall market resilience and improve crisis response capabilities. In sum, an effective policy architecture should incorporate a thorough understanding of cross-market risk transmission pathways and deploy strategies that reflect the quantile-specific and frequency-dependent dynamics documented in this study.