Risk Spillovers and Network Connectedness between Clean Energy Stocks, Green Bonds, and Other Financial Assets: Evidence from China

Abstract

As climate change impacts energy consumption, investments in clean energy are now associated with increased levels of risk and uncertainty. Consequently, the management of risk for clean energy investors has garnered significant academic attention. This study was designed to explore the risk transfers among clean energy markets, how they respond to market volatility, and how exceptional events impact the risk spillover. This was performed by examining the risk spillover of and asymmetric connectedness between clean energy markets, green bonds, and other financial markets in China, in line with the connectedness framework and minimum spanning tree technique. The findings revealed that clean energy markets exhibit heterogeneity in terms of the direction and magnitude of net risk spillover, the types of hedging assets involved, and their response to market volatility. Exceptional events, such as the Russian–Ukrainian conflict and COVID-19 pandemic, have an impact on the spillover relationships. During stable market conditions, green bonds experience fewer spillovers from clean energy markets, whereas, in times of volatility, gold markets are subjected to fewer spillovers. In the time domain, the overall long-term spillover is stronger compared to the short and medium terms. In the frequency domain, there is a significant risk of low-frequency transmission. These findings hold practical implications for energy investors in portfolio construction and for policymakers in pursuing sustainability objectives.

1. Introduction

Traditional energies have played a significant role as an energy source in recent decades. However, their utilization has inevitably contributed to a range of climate change issues. With the gradual lifting of COVID-19 pandemic-related restrictions imposed by many countries [1], global carbon emissions are expected to increase in the foreseeable future, heightening the criticality of global climate governance [2]. However, clean energy can alleviate the adverse impact of industrialization on low-carbon development and more effectively address climate-related problems [3]. Consequently, numerous countries, including China, have begun optimizing their energy mix and promoting the adoption of clean energy [4].

The Chinese government has recognized the importance of promoting the development of natural and social capital as a crucial part of its development strategy for the future [5]. However, China’s clean energy market is in its early stage of development and requires significant investments and technological innovation [6]. Therefore, it is necessary to study whether and how the clean energy sector can provide investors with sensible asset allocation and portfolio strategies.

Investors and policymakers with an environmental consciousness have a keen interest in understanding the heterogeneity across different segments of the clean energy sector, mitigating financial risks, and examining the role and transmission pathways between clean energy markets and other financial markets. However, to the best of our knowledge, there are limited studies that systematically analyze the spillover effects between clean energy markets, emerging green bonds, technology stocks, gold, and traditional energy assets. Moreover, it is valuable and practical to consider extreme volatility, as research has shown that extreme volatility deserves special attention when examining market risk contagion characteristics for investors [7]. For instance, events such as the Russian-Ukrainian conflict have led to shifts in the correlation of financial assets [8]. The COVID-19 pandemic has also had a substantial impact on stock market performance, surpassing the effects of past crises, including the 2008 financial crisis [9]. Companies, including many in the energy sector in China, have suffered significant losses [10], and there may be potential changes in the risk aversion of underlying assets [11]. It has been observed that, during the COVID-19 pandemic, the correlation between clean energy indices and conventional energy increased, but clean energy assets can still act as risk diversifiers [12]. However, there is a lack of research on the specific performance and risk management measures under major contingencies such as COVID-19 and the Russia–Ukraine conflict.

To address the aforementioned research gap, this study aimed to investigate the movements of asset portfolios involving clean energy and other financial assets before and after the COVID-19 pandemic and, thus, provides valuable insights into the performance and risk management strategies of clean energy investments in the context of major contingencies. Additionally, due to the stricter COVID-19 protection measures implemented by the Chinese government compared to other countries, studies utilizing Chinese data may yield distinct findings compared to those relying on data from other nations.

The main objectives of this study were to provide insights into addressing some key questions in the field of clean energy investment and risk management: (1) Can other financial markets diversify risk for clean energy markets and identify the pathways through which risk is transmitted? (2) What are the major differences, compared to other assets, in the effectiveness of green financial markets, particularly green bonds, in hedging risk? (3) What is the heterogeneity of the hedging effect of different financial assets on various types of clean energy? (4) How is the role of risk diversification evolving over time?

The rest of this paper is organized as follows. Section 2 reviews the literature. Section 3 outlines the methodologies. Section 4 presents the data and preliminary analysis. Section 5 reports the empirical results. Section 6 concludes the paper.

2. Literature Review

In current research on clean energy, portfolio construction with other financial assets has been explored as a means to hedge risks and secure returns. These assets include oil prices [13], natural gas [14], the VIX index [15], the VXXLE index [16], carbon assets [17], and gold [18]. By diversifying the portfolio, investors aim to mitigate the risk associated with clean energy investments. Furthermore, the introduction of the green bonds market has brought about additional opportunities for risk management. Studies suggest a low or even negative correlation between green bonds and clean energy markets, indicating that green bonds can provide protection against price volatility in clean energy markets [19]. Moreover, green bonds have demonstrated higher average returns in comparison to traditional bonds [20], making them increasingly popular among investors. This trend presents new opportunities for clean energy investors to improve their risk management practices.

Research has already explored the relationship between the clean energy market and other financial markets. For instance, a study [21] found that green bonds serve the purpose of promoting portfolio diversification and act as a shock absorber for clean energy assets. Another study [22] observed that a strong positive correlation between technology and clean energy stocks and that higher technology stock returns generate interest in clean energy stocks. The research in [18] compared the hedge function of gold and crude oil for clean energy, concluding that gold is relatively weaker than crude oil in mitigating risks during extreme market volatility. While the United States primarily relies on oil as its main traditional energy source, China depends on coal power to fulfill more than two-thirds of its energy demand. The research conducted by [23] indicated that there is a significant two-way volatility spillover effect between traditional energy markets such as the coal power market and clean energy stocks in China. However, the existing literature primarily focuses on examining the relationship between a single financial asset and the overall clean energy index, with limited studies exploring the connections between multiple financial markets and clean energy markets.

Currently, the green bonds market and clean energy markets have experienced rapid growth in terms of scope and depth. Exploring the interaction between green bonds and clean energy markets holds significant implications for investors and policymakers with environmental preferences. Scholars have studied the spillover effects within this system and found that green bonds exhibit fewer spillover effects compared to other factors. For example, the authors of [24] revealed that green bonds serve as the primary recipients of shocks, as opposed to global solar, global wind, and carbon prices. A study [21] discovered that both the green bond index and the global clean energy index act as net recipients of short-term and long-term shocks. Furthermore, research indicates that the impact of green bonds on the clean energy market is greater than the reverse influence. For instance, with causal market-related data, the research in [25] found that the upward spillover from green bonds to the clean energy market is more significant than the downward spillover. However, in either the Granger causality of risk or the MVMQ-CAViaR model, the excessive spillovers from clean energy to the green bonds market are not significant, which implies that the green bonds market can serve as an emerging option to diversify clean energy portfolio risk.

The increasing interconnectedness among energy sub-sectors, markets, and industries has made the global new energy sector susceptible to the rapid diffusion of market information and associated risks during severe unexpected events, such as the COVID-19 pandemic. Consequently, these events can disrupt pre-existing network structures [26]. Therefore, it is crucial for clean energy investors and policymakers to comprehend the risk–return profile of clean energy investments, as well as the potential time-varying characteristics and risk aversion properties [27]. Under extreme market conditions, green bonds led to short-term growth in the clean energy sector and exerted an increasingly positive influence following the outbreak of the COVID-19 pandemic [28]. The research in [22] argued that the relationship between technology stocks and clean energy may differ during extreme market conditions, suggesting that technology stocks could provide opportunities for portfolio diversification. During a pandemic, a study [29] observed a significant difference in the correlation between market returns in the traditional energy sector and the corresponding producer stock indices. Furthermore, it was found that gold prices had a significant positive impact on clean energy stocks in both the short-term and long-term during the COVID-19 pandemic [30]. However, there is a scarcity of systematic studies examining the dynamic correlation and network structures between the clean energy market and the major financial markets. Furthermore, China’s stringent pandemic control strategies have resulted in divergent impacts compared to other countries, creating a research gap in this field.

Additionally, most of the existing studies on clean energy utilize composite indices and fail to account for the heterogeneity among different clean energy assets. Furthermore, there is a dearth of research examining the correlation between clean energy and financial assets, while simultaneously comparing various financial assets. Additionally, given China’s unique pandemic control policies and clean energy landscape, conducting relevant studies assumes greater significance.

The study contributes to the existing literature in three main ways: (1) It uncovers the heterogeneity in the size and direction of spillover effects across different clean energy markets. This provides valuable insights for asset allocation strategies under different investment conditions. (2) The study confirms the weak spillover effects between the clean energy market and the green bonds market, indicating that green bonds and the gold market serve similar functions as risk-averse assets. (3) In addition to time–frequency domain analysis, the study employed asymmetric analysis and dynamic analysis. By using the COVID-19 outbreak as a reference point, it investigated the risk spillover effects across different economic environments and time periods, shedding light on the risk transmission pathways in the spillover network.

This research helps to enhance portfolio construction for investors, increase social capital investment in the clean energy sector, and promote the development of green industries. Moreover, it provides valuable insights for policymakers to improve the effectiveness of climate policies by appropriately funding climate and sustainability projects.

3. Methodology

This study examined the symmetric and asymmetric return spillovers between the green bonds and clean energy markets, conducting a comprehensive spillover analysis in both the time and frequency domains. To enhance visibility and facilitate understanding, the results of network connectedness are also presented. This approach draws upon the spillover index methods introduced by Diebold and Yilmaz (2012) [31] (DY) and Baruník and Křehlík (2018) [32] (BK).

3.1. The DY Spillover Index

This study employed the DY connectedness framework to analyze the directional risk spillovers between the Clean Energy Equity Index and various financial markets over time. This framework is based on the original design of the generalized vector autoregressive (VAR) model, which calculates the forecast error variance decomposition (FEVD). One advantage of this framework is that the results are independent of the variable order, allowing for consistent and coherent tracking of different levels of connectedness.

The structural VAR(p) illustrates the n-variate process $x_{t,1}$,…, $x_{t,n}$ within the time of $t=1,\dots ,T$. The expression formula is as follows:

$$x_t={\sum }_{i=1}^p{\Phi }_{t-i}+{\epsilon }_t,$$

(1)

where $x_t$ denotes the n-dimensional column vector of each market price return, $\epsilon \sim (0,\sum )$ represents the vectors of independent identically distributed perturbations, and the autoregressive coefficient matrix ${\Phi }_i$ on time ($t=1,\dots ,T$) is assumed to capture the smooth covariance. Formula (1) can be further expressed in a moving average form as (2):

$$x_t={\sum }_{i=0}^{\infty }{A}_i{\epsilon }_{t-i},$$

(2)

where ${\epsilon }_t$ represents a vector of zero-mean errors with ∑ as the covariance matrix and the coefficient matrix $A_i$ of order $n\times n$ follows the recursive process of Equation (3):

$$A_i={\Phi }_1{A}_{i-1}+{\Phi }_2{A}_{i-2}+\dots {\Phi }_p{A}_{i-p}.$$

(3)

It can be observed that the formula incorporates a unit matrix of size $n\times n$ and $A_i=0$ for $i<0$. The variance decomposition enables the estimation of the proportion of the H-step-head error variance in $x_i$, which can be attributed to the shock in $x_j$.

Diebold and Yilmaz (2012) [31] utilized the generalized variance decomposition introduced by Pesaran and Shin (1998) [33] to examine these correlations. Within this framework, the proportion of innovations in variable k contributing to the variance of the H-step ahead prediction error for variable j is given by Equation (4):

$${\theta }_{ij}^g=\frac{{\delta }_{ii}^{-1}{\sum }_{h=0}^{H-1}{(e_i^{{}^{\prime }}{A}_h\sum e_j)}^2}{{\sum }_{h=0}^{H-1}(e_i^{{}^{\prime }}{A}_h\sum A_h^{{}^{\prime }}{e}_i)},$$

(4)

where ${\delta }_{ii}$ represents the standard deviation of the error term $\epsilon$, $A_h$ denotes the coefficient matrix in the infinite moving average process corresponding to prediction horizon H in the VAR model, H is the selected prediction horizon, ∑ is the covariance matrix of the error vector, and $e_i$ is the selected vector. The selected vector, denoted as $e_i$, takes a value of 1 in the j-th element and 0 otherwise, such that ${\sum }_{j=1}^N{\theta }_{ij}^g(H)\ne 1$. To normalize each entry of the variance decomposition matrix, it is necessary to divide each element by the sum of the elements in the corresponding row of the variance decomposition table. This normalization can be expressed as in Equation (5):

$${{\theta }_{ij}}^{{\sim }_g}=\frac{{\theta }_{ij}^g(H)}{{\sum }_{j=1}^N{\theta }_{ij}^g(H)}.$$

(5)

By following the construction process, we obtain Equations (6) and (7):

$${\sum }_{j=1}^N{{\theta }_{ij}}^{\sim g}(H)=1,$$

(6)

$${\sum }_{j=1}^N{{\theta }_{ij}}^{\sim g}(H)=N.$$

(7)

The total spillover index is utilized to quantify the contribution of shock spillover among variables to the overall forecast error variance. It is defined as in Equations (8) and (9):

$$S^g(H)=\frac{{\sum }_{i,j=1,i\ne j}^N{{\theta }_{ij}}^{\sim g}(H)}{{\sum }_{i,j=1}^N{{\theta }_{ij}}^{\sim g}(H)}·100=\frac{{\sum }_{i,j=1,i\ne j}^N{{\theta }_{ij}}^{\sim g}(H)}{N}·100,$$

(8)

$$S_{•i}^g(H)=\frac{{\sum }_{j=1,i\ne j}^N{{\theta }_{ij}}^{\sim g}(H)}{{\sum }_{j=1}^N{{\theta }_{ij}}^{\sim g}(H)}·100.$$

(9)

Based on these directional spillovers, the net volatility spillover from market i to all markets j can be computed as the disparity between the total volatility shock transmitted to all other markets and the total volatility shock received from all other markets, as shown in Equation (10):

$$S_i^g(H)=S_{•i}^g(H)-S_{i•}^g(H)$$

(10)

Then, we are able to construct the net directional connectedness metric for each pair using Equation (11):

$$S_{ij}^g(H)=(\frac{{{\theta }_{ij}}^{\sim g}(H)}{{\sum }_{k=1}^N{{\theta }_{ik}}^{\sim g}(H)}-\frac{{{\theta }_{ij}}^{\sim g}(H)}{{\sum }_{k=1}^N{{\theta }_{jk}}^{\sim g}(H)})·100.$$

(11)

3.2. The BK Spillover Index

To capture the connectedness in the frequency domain, Baruník and Křehlík (2018) enhanced the spillover framework of DY through the structure of their spillover approach [32]. They introduced the spectral variance decomposition of the VAR, which quantifies the proportion of uncertainty in a particular variable at a specific frequency that can be attributed to shocks in other variables within the VAR model. By setting $i=\sqrt{-1}$, the generalized causation spectrum can be expressed as in Formula (12):

$${(f(\omega ))}_{j,k}\equiv \frac{{\sigma }_{kk}^{-1}{\left|\Psi (e^{-iw}){\sum }_{j,k}\right|}^2}{\Psi (e^{-iw})\sum {\Psi }^{{}^{\prime }}{(e^{+iw})}_{j,j}},$$

(12)

of which ${(f(\omega ))}_{j,k}$ is interpreted as an intra-frequency causal measure representing the proportion of the spectrum of variable j at frequency $\omega$ that is caused by the shock in the variable k and $\Psi (e^{-iw})$ represents the Fourier transform of the impulse response. Baruník and Křehlík (2018) [32] proposed Expression (13) for the share of shocks to variable k in the fluctuations of variable j within the frequency band:

$${({\theta }_d)}_{j,k}=\frac{1}{2\pi }\underset{}{\overset{d}{\int }}{\Gamma }_j{(f(\omega ))}_{j,k}d\omega ,$$

(13)

where ${({\theta }_d)}_{j,k}$ represents the power of variable j at the frequency $\omega$. The frequency-based connectedness within the frequency band can be defined in Equation (14) using the spectral representation of the generalized forecast error variance decomposition (GFEVD):

$$C_d^F=100(\frac{{\sum }_{j\ne k}^{}{({\overline{\theta }}_d)}_{j,k}}{\sum {({\overline{\theta }}_{\infty })}_{j,k}}-\frac{Tr\left\{{\overline{\theta }}_d\right\}}{\sum {({\overline{\theta }}_{\infty })}_{j,k}}).$$

(14)

Hence, the total overflow can be calculated in Equation (15):

$$C_d^W=100(1-\frac{Tr\left\{{\overline{\theta }}_d\right\}}{\sum {({\overline{\theta }}_d)}_{j,k}}).$$

(15)

Similar to the time domain, directional overflows can also be calculated at different frequencies. The “FROM” and “TO” overflows can then be calculated using Equations (16) and (17). The directional frequency connectedness from others to variable j and from variable j to the others can be defined and estimated as:

$$C_{j\leftarrow •}^d=(\sum _{k=1,k\ne j}^{}{{\theta }_{j,k}}^{\sim d})\frac{\sum {\theta }^{\sim d}}{\sum {\theta }^{\sim \infty }}$$

(16)

and

$$C_{j⟶•}^d=(\sum _{k=1,k\ne j}^{}{{\theta }_{k,j}}^{\sim d})\frac{\sum {\theta }^{\sim d}}{\sum {\theta }^{\sim \infty }}.$$

(17)

Then, the net spillover index, which quantifies the net spillover of market j to all other markets, can be expressed in Equation (18):

$$C_j^d=C_{j⟶•}^d-C_{j\leftarrow •}^d.$$

(18)

3.3. Network Connectedness

To visualize the spillover network between the clean energy market and other financial markets and explore the risk transmission paths across the markets, this study constructed a weighted directed network based on the works of Diebold and Yilmaz (2012) [31] and Diebold and Yilmaz (2015) [34]. In the network, the clean energy market and other financial markets are regarded as network nodes, and the spillover relationship between markets is regarded as the connection of the network. In this directed network model, the elements of the adjacency matrix represent pairwise directional connectedness. The row sums of the adjacency matrix represent the total “FROM” directional connectedness, and the column sums represent the total “TO” directional connectedness. These “FROM” and “TO” directional connectedness values are adopted as weights for the network’s edges, which can be visualized using different colors and sizes. Each market is represented as a node in the network, with the color indicating whether it serves as a transmitter or recipient in the system and the size of the nodes revealing the strength of the net risk spillover.

3.4. Minimum Spanning Tree Approach

Currently, most studies on risk transmission paths are conducted within correlation networks. However, it is important to eliminate unnecessary edges from the market network in order to obtain pivotal information. Various methods can be employed for information filtering, such as the threshold method, minimum spanning tree (MST) method [35,36], planar maximum filter graph method, asset graph, and planar maximally filtered graph. Among them, the MST approach is particularly valuable as it not only highlights the core relationships between variables in a system, but also reveals potential risk transmission paths [37].

A spanning tree of a graph consists of all nodes and some edges of the graph, so that there is a path between any two nodes. While there are several methods to construct a spanning tree, the MST approach seeks the smallest sum of the edge weights. This approach is widely applied in the design of communication networks, power and leased-line telephone networks, wiring connections, links in a transportation network, etc. [38]. To filter out unnecessary information and obtain a simplified and robust structure, the MST is constructed, which represents the shortest path that connects all nodes in the network and serves as the most-efficient transmission path when risks occur.

4. Data and Descriptive Analysis

According to the Statistical Classification of Energy Conservation and Environmental Protection Clean Industries issued by the National Bureau of Statistics of China, the clean energy industry consists of eight sectors: nuclear power industry, wind energy industry, solar energy industry, biomass energy industry, hydropower industry, smart grid industry, other clean energy industries, and traditional energy clean and efficient utilization industry. Additionally, the traditional energy clean and efficient utilization industry are represented by waste-to-energy. The energy efficiency business of contract energy management is used to represent other clean energy sectors. To assess the overall performance of clean energy, the WIND series index (the data source is https://www.wind.com.cn/ accessed on 7 July 2022) was adopted considering the time span, index authority, and recognition. The green financial market is represented by green bonds, which are primarily represented by the CSI-China Green Bond Index. Other markets are represented as follows: the high technology sector is represented by the CSI Technology 50 Strategy Index, the traditional energy market by the CSI Energy Futures Index, and the gold market by the Shanghai Gold Index (refer to

Table 1. Description of the framework for the clean energy markets and financial markets. The clean energy market is represented by eight industries with the WIND series index, and financial markets are represented by four industries with the CSI series index.

Market Type	Primary Industry	Abbreviation	Index Name
Clean Energy Market	nuclear power	Nuc	WIND nuclear power concept index
	wind energy	Wind	WIND wind power concept index
	solar energy	Pho	WIND photovoltaic concept index
	biomass energy	Geo	WIND geothermal energy concept index
	hydropower	Hyd	WIND water and hydropower construction concept index
	smart grid	Elec	WIND smart grid concept index
	other clean energy industry	EM	WIND contract energy management concept index
	traditional energy clean and efficient utilization industry	Was	WIND waste power concept index
Financial Markets	green bonds	Grb	CSI-China Green Bond Index
	high technology	Tech	CSI Technology 50 Strategy Index
	traditional energy	Ener	CSI Energy Futures Index
	gold	Gold	Shanghai Gold Index

). All price data used in the analysis are daily closing prices, spanning from 5 January 2015 to 30 June 2022, resulting in 1822 observations for each market and 21,864 observations for the 12 markets. In the empirical analysis, the return series is computed with Equation (19):

$$R_t=log(P_t)-log(P_{t-1}).$$

(19)

Figure 1 depicts the descriptive statistical analysis of each market return (details are shown in

Table A1. Descriptive statistics of realized volatility in clean energy assets and financial markets.

	Grb	Gold	Ener	Tech	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM
Mean	0.0002	0.0003	0.0006	0.0005	0.0002	0.0004	0.0008	0.0003	0.0004	0.0004	−0.0001	0.0005
Median	0.0002	0.0002	0.0005	0.0005	0.0013	0.0012	0.0017	0.0012	0.0013	0.0016	0.0005	0.0016
Maximum	0.008	0.054	0.0817	0.0731	0.0807	0.0933	0.2568	0.0743	0.0881	0.0699	0.0741	0.0845
Minimum	−0.0082	−0.0481	−0.086	−0.0945	−0.1048	−0.1036	−0.1026	−0.1029	−0.1026	−0.1053	−0.0911	−0.1056
Std. Dev.	0.0009	0.0085	0.0165	0.02	0.0203	0.0212	0.0222	0.0188	0.0207	0.0205	0.0187	0.0211
Skewness	−0.1798	0.0158	−0.2624	−0.6126	−0.8913	−0.6139	0.1785	−0.887	−0.7725	−0.9064	−0.7903	−1.0031
Kurtosis	19.0495	7.1112	5.6796	5.8749	7.408	6.2935	14.9431	6.9604	6.1996	6.5959	7.2678	7.0532
Jarque-Bera	19,564.86 ***	1283.24 ***	566.02 ***	741.44 ***	1716.34 ***	937.90 ***	10,838.27 ***	1429.64 ***	958.42 ***	1231.15 ***	1572.43 ***	1552.73 ***
Sum	0.3721	0.4888	1.1576	0.866	0.3874	0.7231	1.3812	0.583	0.7976	0.6598	−0.1406	0.9662
Sum Sq. Dev.	0.0013	0.1313	0.4965	0.7312	0.7533	0.821	0.8946	0.6412	0.7792	0.7656	0.6363	0.8095
ADF	−27.66 ***	−42.03 ***	−42.34 ***	−42.07 ***	−39.98 ***	−40.70 ***	−39.40 ***	−39.16 ***	−40.06 ***	−39.34 ***	−38.12 ***	−40.06 ***

Notes: *** denote the rejection of the null hypotheses of normality and no autocorrelation at the significance levels of 1%.

in Appendix). With the exception of Was, all markets show positive means close to zero. The standard deviations of Pho and Wind are 0.0222 and 0.0212, respectively, indicating strong fluctuations in these two major variables. Although gold, the traditional safe-haven asset, exhibits a low standard deviation of 0.008491, green bonds have a significantly lower standard deviation of 0.0009. The kurtosis values of all series are at a high level, implying the existence of a spiky distribution. Regarding skewness, except for Gold and Pho, most market returns are negatively skewed. Furthermore, Jarque–Bera’s test rejects the null hypothesis that none of the variables follow a normal distribution. The Dickey–Fuller (ADF) test indicates that all variables are stationary series at the $1\%$ level of significance.

5. Empirical Results

In this section, static, asymmetric, and dynamic analyses are employed to examine the spillover effects among different markets under varying conditions. Specifically, spillover network analysis is conducted to illustrate the dynamic characteristics of spillovers, and an MST is conducted to draw more visually interpretable conclusions.

5.1. Static Spillover Analysis

5.1.1. Time Domain Spillover Analysis

Table 2. Static connectedness matrix over the full sample period. The significant total spillover index indicates a close interconnection among the markets. It suggests that risks are transmitted within the clean energy markets. Elec has the highest net spillover effect. Pho and Wind have negative net spillovers. Grb has the largest net spillover.

	Grb	Tech	Gold	Ener	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	72.26	2.93	0.45	0.55	3.16	2.39	0.92	4.19	3.72	3.95	2.45	3.00	27.74
Tech	0.34	17.76	0.03	0.09	9.73	9.05	7.00	11.06	12.92	10.23	9.98	11.82	82.24
Gold	0.31	0.99	91.55	1.09	0.67	0.30	2.28	0.78	0.85	0.21	0.79	0.17	8.45
Ener	0.24	0.42	2.48	89.76	1.09	1.01	0.94	0.80	0.71	1.49	0.56	0.49	10.24
Nuc	0.29	9.39	0.04	0.18	16.21	10.81	6.89	12.04	11.83	9.94	10.81	11.58	83.79
Wind	0.21	8.83	0.01	0.15	10.74	16.42	8.72	11.13	12.41	9.20	10.60	11.58	83.58
Pho	0.32	8.56	0.85	0.28	8.61	10.18	22.97	10.02	10.80	8.05	8.62	10.73	77.03
Hyd	0.36	9.28	0.07	0.18	11.12	9.83	6.64	17.49	11.69	10.33	11.68	11.33	82.51
Elec	0.27	10.45	0.03	0.10	9.99	9.95	7.18	11.27	15.71	10.58	11.31	13.15	84.29
Geo	0.33	9.79	0.01	0.18	10.57	9.27	6.24	11.46	12.77	15.77	11.12	12.48	84.23
Was	0.21	9.36	0.07	0.26	10.44	9.37	5.57	12.16	11.93	9.79	18.42	12.41	81.58
EM	0.23	9.75	0.01	0.10	10.10	9.55	7.21	11.30	13.28	10.82	11.84	15.81	84.19
TO	3.11	79.75	4.07	3.16	86.21	81.72	59.59	96.22	102.92	84.60	89.78	98.74	789.87
NET	−24.63	−2.49	−4.39	−7.08	2.42	−1.85	−17.44	13.71	18.64	0.37	8.20	14.55	65.82

Notes: Based on the AIC criterion, a 5-order lag term was selected. In the table, the “FROM” column represents the internal risk spillover of a market from other markets, and the “TO” row represents the risk spillover of a market to other markets. The “NET” row indicates the difference between FROM-spillovers and TO-spillovers in each market. In the bottom right corner, the total spillover percentage is calculated as the ratio of the sum of external spillover (“TO-spillover” or “FROM-spillover”) to the sum of total spillover (“ALL-spillover”).

presents the estimates of the total static spillover matrix between clean energy, green bonds, and other financial markets. The total static spillover index for the return system is 65.82%, indicating a close interconnection among the markets. The significant spillovers observed in clean energy indexes suggest that most of the changes in the system are driven by interactions within the clean energy markets rather than financial markets. Additionally, the information transmission between clean energy stock returns is highly correlated.

Among the markets, Elec exhibits the highest net spillover effect, followed by EM. Both of these clean energy assets, which are associated with addressing energy challenges, have attracted considerable investment attention. The development of smart grids holds a pivotal strategic position in the Chinese economy, while contract energy management is an essential tool for enhancing energy efficiency. Hyd, ranking third in net spillovers, shows promising prospects due to the declining trend in thermal power installations and the increasing focus on hydropower construction and investment. Hence, the top three transmitting markets represent industries that have consistently garnered investor interest. Furthermore, both Pho and Wind have negative net spillovers, indicating that the photovoltaic and wind energy markets are significantly influenced by other markets. This observation aligns with the fact that the costs of new energy generation methods such as photovoltaics and wind have been rapidly declining, resulting in higher volatility compared to other clean energy assets. Notably, the green bonds market exhibits the highest net spillover as a receiver, suggesting it having the least impact on the other markets. This characteristic endows the green bonds market with value in portfolio risk diversification.

Furthermore, a weak spillover effect was identified in the directional link between clean energy markets and the green bonds market [39]. Specifically, the risk spillover from the solar energy market to the green bonds market is 0.92%, the lowest among the spillover indices, which suggests that the green bonds market exhibits relatively low correlation with the solar energy market. On the other hand, the solar energy market was identified as the primary receiver with the smallest negative total spillover indices among the clean energy markets.

Moreover, the correlation between clean energy assets and other financial assets was examined. In addition to the green bonds market, the spillover effect from the clean energy market to the gold market was also found to be at a low level. Gold exhibited the lowest correlation among various clean energy assets, such as nuclear, wind, hydropower, geothermal, and contract energy management. Similarly, traditional energy demonstrated the least correlation with clean energy assets, including smart grids and waste-to-energy.

To provide a more-comprehensive understanding of the magnitude and direction of static spillover effects across different markets under varying market conditions, the study period from 2015 to 2022 was divided into two sub-periods, with the year 2019, marked by the COVID-19 pandemic, serving as the breakpoint. The results are presented in

Table A2. Static connectedness matrix for the subsample period of 2015–2019.

	Grb	Tech	Gold	Ener	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	96.54	0.51	0.65	0.59	0.12	0.22	0.3	0.14	0.17	0.21	0.19	0.37	3.46
Tech	0.03	14.08	0.18	0.12	9.96	11.06	10.14	10.55	12.08	10.45	10.13	11.23	85.92
Gold	0.74	1.09	90.01	0.01	1.23	0.96	0.28	1.18	1.04	1.28	1	1.19	9.99
Ener	0.47	0.63	0.02	91.15	0.87	0.9	1.13	1.13	0.7	0.96	1.02	1.02	8.85
Nuc	0.02	10.02	0.15	0.16	13.81	11.63	9.87	11.06	11.43	10.37	10.49	11	86.19
Wind	0.02	10.38	0.12	0.15	10.87	13.07	10.35	11.16	11.64	10.46	10.58	11.2	86.93
Pho	0.03	10.45	0.05	0.21	10.03	11.33	14.39	10.53	11.4	10.24	10.24	11.11	85.61
Hyd	0.01	10.17	0.16	0.18	10.57	11.46	9.93	13.4	11.42	10.63	10.95	11.11	86.6
Elec	0.02	11.18	0.15	0.12	10.46	11.43	10.25	10.92	12.87	10.7	10.43	11.48	87.13
Geo	0.01	10.38	0.2	0.18	10.18	11.03	9.88	10.89	11.53	13.8	10.45	11.46	86.2
Was	0.01	10.09	0.16	0.19	10.29	11.18	9.88	11.23	11.23	10.5	13.93	11.31	86.07
EM	0.02	10.6	0.15	0.16	10.28	11.25	10.19	10.87	11.73	10.9	10.76	13.1	86.9
TO	1.37	85.5	1.99	2.05	84.87	92.46	82.18	89.64	94.37	86.69	86.23	92.49	799.86
NET	−2.09	−0.41	−8	−6.8	−1.33	5.53	−3.43	3.04	7.24	0.5	0.16	5.59	66.65

and

Table A3. Static connectedness matrix for the subsample period of 2020–2022.

	Grb	Tech	Gold	Ener	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	83.32	2.25	2.58	2.73	1.47	0.72	0.96	1.51	1.16	1.19	1.1	1.03	16.68
Tech	0.65	23.57	0.21	0.82	10	8.02	11.98	6.98	11.83	9.89	6.26	9.8	76.43
Gold	2.26	1.03	88.68	1.68	0.48	1.1	0.96	0.71	0.71	0.43	0.81	1.16	11.32
Ener	1.48	2.21	1.41	71.42	4.87	1.64	2.2	3.76	2.26	2.75	2.64	3.35	28.58
Nuc	0.44	7.19	0.11	0.93	16.86	10.43	9.54	10.45	11.99	10.67	9.51	11.87	83.14
Wind	0.31	6.85	0.08	0.44	12.43	20.09	12.42	7.82	12.46	8.42	7.5	11.17	79.91
Pho	0.3	9.87	0.17	0.57	10.88	11.98	19.31	7.35	12.61	8.8	5.97	12.18	80.69
Hyd	0.38	5.91	0.06	0.99	12.33	7.73	7.5	20.09	11.1	11.23	11.7	11	79.91
Elec	0.31	8.34	0.12	0.45	11.85	10.33	10.85	9.29	16.65	10.12	8.99	12.69	83.35
Geo	0.4	7.95	0.08	0.72	11.85	7.89	8.7	10.59	11.45	18.75	9.39	12.22	81.25
Was	0.35	5.62	0.04	0.57	11.87	7.85	6.43	12.39	11.29	10.48	21	12.1	79
EM	0.45	7.06	0.08	0.66	11.87	9.41	10.66	9.36	12.86	10.96	9.75	16.89	83.11
TO	7.33	64.29	4.94	10.56	99.89	77.1	82.2	80.21	99.73	84.94	73.62	98.56	783.38
NET	−9.36	−12.14	−6.37	−18.02	16.75	−2.81	1.5	0.3	16.38	3.69	−5.38	15.45	65.28

.

The role of green bonds in risk diversification varies depending on the market environment. In a positive market environment, green bonds demonstrated a stronger ability to spread risk and exhibit minimal spillover effects with all clean energy assets, except for solar energy. Conversely, in a negative market environment, gold emerged as the asset with the smallest spillover effects with most clean energy assets, including nuclear, photovoltaics, hydropower, smart grid, geothermal energy, and waste-to-energy. Additionally, in a stable market environment, the spillover effects of clean energy assets are more influenced by the spillover effects of financial assets. However, in a volatile market environment, financial assets play a more significant role as recipients of shocks. Notably, the nuclear market (Nuc) and photovoltaic market (Pho) exhibited a shift in their spillover effects. When the market environment worsens, they transition from negative spillover effects to positive spillover effects and become transmitters of shocks. This suggests that they have a more-pronounced impact on other markets when the financial market deteriorates.

5.1.2. Frequency Domain Spillover Analysis

To delve deeper into the extent of risk spillover at different frequency bands, this study followed the approach of Liang et al. [40] and decomposed the risk into three frequency ranges: high frequency (1–5 days), medium frequency (5–30 days), and low frequency (more than 30 days). The frequency connectedness, as defined in Equations (12)–(15), was then calculated for each of these frequency ranges. This allowed for an examination of the spillover dynamics across different time horizons.

Table A4. Spillovers for the frequency band 3.14 to 0.63 corresponding to a time period of 1–5 days.

	Grb	Gold	Ener	Tech	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	1.59	0.02	0	0.03	0.03	0.05	0.04	0.04	0.04	0.03	0.03	0.03	0.34
Gold	0.01	1.32	0.04	0.01	0.01	0	0.01	0.01	0	0	0	0	0.09
Ener	0.02	0.01	0.79	0.07	0.03	0	0.02	0.02	0.01	0.02	0.03	0.02	0.25
Tech	0	0.01	0.02	0.19	0.09	0.13	0.06	0.09	0.15	0.12	0.09	0.12	0.88
Nuc	0	0.01	0.02	0.09	0.21	0.15	0.05	0.11	0.15	0.13	0.11	0.13	0.95
Wind	0.01	0	0.01	0.16	0.19	0.36	0.11	0.17	0.24	0.19	0.16	0.2	1.44
Pho	0	0.02	0.02	0.08	0.08	0.13	0.35	0.07	0.13	0.1	0.07	0.1	0.8
Hyd	0	0.01	0.02	0.13	0.16	0.18	0.08	0.28	0.2	0.18	0.17	0.18	1.51
Elec	0.01	0.01	0.01	0.15	0.14	0.19	0.09	0.14	0.28	0.19	0.15	0.2	1.28
Geo	0.01	0	0.01	0.15	0.16	0.18	0.08	0.17	0.23	0.33	0.16	0.21	1.36
Was	0.01	0.01	0.02	0.22	0.27	0.28	0.13	0.3	0.34	0.29	0.52	0.32	2.19
EM	0.01	0	0.02	0.16	0.17	0.2	0.09	0.17	0.26	0.22	0.18	0.3	1.48
TO	0.08	0.1	0.19	1.25	1.33	1.49	0.76	1.29	1.75	1.47	1.15	1.51	12.57
All	1.67	1.42	0.98	1.44	1.54	1.85	1.11	1.57	2.03	1.8	1.67	1.81	65.57

Notes: Based on the AIC criterion, a 5-order lag term was selected. The “FROM” column represents the internal risk spillover of a market from other markets, and the “TO” row represents the risk spillover of a market to other markets. In the bottom right corner, the total spillover percentage is calculated as the quotient of the sum of external spillover (“TO-spillover” or “FROM-spillover”) and the sum of total spillover (“ALL-spillover”).

,

Table A5. Spillovers for the frequency band 0.63 to 0.10 corresponding to a time period of 5–30 days.

	Grb	Gold	Ener	Tech	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	9.32	0.08	0.01	0.27	0.28	0.4	0.28	0.44	0.39	0.32	0.29	0.33	3.07
Gold	0.05	7.26	0.22	0.03	0.04	0.02	0.04	0.07	0.03	0.02	0.01	0.03	0.55
Ener	0.08	0.05	3.68	0.38	0.17	0.03	0.07	0.11	0.07	0.12	0.12	0.11	1.28
Tech	0.01	0.04	0.13	0.9	0.41	0.65	0.35	0.49	0.81	0.6	0.48	0.6	4.54
Nuc	0.01	0.03	0.09	0.5	1.08	0.82	0.37	0.67	0.85	0.68	0.64	0.71	5.34
Wind	0.04	0.02	0.07	0.81	0.96	1.81	0.71	0.92	1.31	0.95	0.92	1.06	7.72
Pho	0.03	0.15	0.11	0.42	0.41	0.69	1.63	0.47	0.71	0.5	0.42	0.59	4.46
Hyd	0.02	0.04	0.09	0.64	0.81	0.92	0.48	1.42	1.05	0.89	0.92	0.9	6.71
Elec	0.02	0.02	0.07	0.75	0.72	0.95	0.52	0.79	1.5	0.94	0.87	1.08	6.7
Geo	0.02	0.02	0.07	0.8	0.86	0.92	0.47	0.9	1.27	1.64	0.94	1.1	7.32
Was	0.03	0.04	0.15	1.07	1.26	1.34	0.61	1.45	1.7	1.35	2.52	1.59	10.52
EM	0.03	0.02	0.09	0.78	0.84	1	0.57	0.92	1.4	1.11	1.03	1.52	7.73
TO	0.33	0.51	1.12	6.41	6.72	7.69	4.43	7.19	9.52	7.41	6.58	8.04	66.41
All	9.66	7.77	4.78	7.35	7.84	9.55	6.1	8.65	11.09	9.12	9.16	9.62	65.95

and

Table A6. Spillovers for the frequency band 0.10 to 0.00 corresponding to a time period of more than 30 days.

	Grb	Gold	Ener	Tech	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	53.62	0.18	1.98	3.7	4.42	0.98	0.34	4.89	3.94	6.12	2.29	3.2	33.04
Gold	0.15	54.82	0.43	5.98	5.2	2.33	3.37	6.68	5.2	1.6	3.44	1.56	35.94
Ener	0.14	3.25	82.63	2.05	0.53	1.73	0.61	0.42	0.71	1.61	0.11	0.2	11.36
Tech	0.48	0.33	1	15.42	10.9	5.83	7.57	12.53	10	8.13	10.29	10.98	78.04
Nuc	0.24	0.16	0.27	9.51	13.2	8.02	8.6	13.22	9.47	8.15	10.91	10.66	79.21
Wind	0.04	0.11	0.18	7.21	9.87	11.01	10.84	12.08	9.39	6.73	10.87	10.3	77.62
Pho	0.08	0.24	0.3	8.54	10.02	8.1	15.69	12.85	8.89	6.75	10.42	10.84	77.03
Hyd	0.46	0.11	0.25	9.2	11.37	7.26	7.75	15.53	9.12	8.26	10.94	9.98	74.7
Elec	0.23	0.09	0.28	9.83	10.11	7.03	8.29	12.46	11.25	8.1	11.23	11.31	78.96
Geo	0.3	0.18	0.17	8.94	10.46	7.16	7.56	12.25	9.81	10.38	11.08	11.01	78.22
Was	0.11	0.11	0.34	8.84	9.62	6.73	6.43	11.8	8.57	7.15	14.03	10.44	73.73
EM	0.08	0.13	0.41	9.17	9.85	6.9	8.93	11.96	9.5	7.75	11.78	12.47	76.46
TO	2.31	4.89	5.61	82.97	92.35	64.98	70.29	111.14	84.6	70.35	93.36	90.48	777.33
All	55.93	59.71	88.24	98.39	105.55	73.08	85.98	106.67	95.85	80.73	107.39	102.95	71.3

present the total connectedness and its components for the high-frequency range (short-term component with a time period of 1–5 days), the medium-frequency range (medium-term component with a time period of 5–30 days), and the low-frequency range (long-term component with a time period of more than 30 days), respectively. The results indicated that the overall spillover effect is stronger in the long term compared to the short term. Additionally, during periods of crises, the connectedness tends to increase, suggesting a higher risk of transmission in the low-frequency range. In terms of specific assets, gold consistently exhibited the least connectedness with all clean energy assets in both the short term and medium term. However, this relationship changes in the long term, as green bonds replace gold as the financial asset with the least connectedness to wind and photovoltaics, while other clean energy assets become more correlated with traditional energy represented by energy futures.

5.2. Asymmetric Spillover Analysis

In this subsection, an asymmetric spillover analysis was conducted by decomposing return data into positive and negative parts. As shown in

Table 3. Full-sample positive return connectedness matrix.

	Grb	Gold	Ener	Tech	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	97.19	1.65	0.02	0.09	0.09	0.25	0.19	0.06	0.03	0.06	0.17	0.2	2.81
Gold	1.18	94.24	1.21	0.11	0.28	0.07	1.88	0.1	0.1	0.09	0.54	0.2	5.76
Ener	0.94	2.78	93.45	0.13	0.5	0.27	0.83	0.36	0.1	0.31	0.22	0.11	6.55
Tech	0.58	0.06	0.04	23.25	10.44	8.45	4.63	9.88	12.95	9.26	8.03	12.44	76.75
Nuc	0.25	0.05	0.22	9.86	21.03	10.26	3.99	12.04	11.85	9.58	9.27	11.6	78.97
Wind	0.05	0.01	0.04	8.59	10.36	25.04	6.54	10.34	12.84	7.48	7.22	11.49	74.96
Pho	0.12	2.4	0.51	8.11	6.33	9.98	35.44	8.09	8.29	6.07	5.31	9.35	64.56
Hyd	0.17	0.04	0.32	9.59	11.41	9.32	4.03	22.52	10.78	10.75	10	11.06	77.48
Elec	0.31	0.02	0.19	11.79	10.01	11.29	4.75	11	19.07	9.36	8.44	13.76	80.93
Geo	0.19	0	0.08	10.63	10.3	9.05	3.51	11.77	11.94	21.29	8.46	12.78	78.71
Was	0.13	0.06	0.35	9.33	10.66	8.33	2.58	12.62	10.1	8.69	25.18	11.99	74.82
EM	0.13	0.01	0.15	11.5	9.96	10.12	4.63	11.32	13.14	10.9	9.7	18.45	81.55
TO	4.04	7.08	3.14	79.71	80.35	77.38	37.56	87.56	92.12	72.56	67.36	94.98	703.85
NET	1.23	1.32	−3.41	2.96	1.38	2.42	−27	10.09	11.19	−6.15	−7.45	13.43	58.65

and

Table 4. Full-sample negative return connectedness matrix.

	Grb	Tech	Gold	Ener	Nuc	Wind	Pho	Hyd	Elec	Geo	Was	EM	FROM
Grb	99.01	0.2	0.15	0.01	0.02	0.02	0.04	0.14	0.06	0.25	0.09	0.02	0.99
Gold	0.03	99.38	0.06	0.24	0.01	0.03	0.03	0.01	0.11	0.02	0.05	0.02	0.62
Ener	0.07	0.54	85.67	1	1.24	2.03	1.66	1.12	1.53	2.05	1.36	1.73	14.33
Tech	0.01	0.02	0.06	15.1	10.42	10.44	11.42	10.84	11.58	9.85	9.94	10.32	84.9
Nuc	0.01	0.04	0.17	9.79	13.32	10.86	10.69	11.2	11.25	10.86	11.1	10.71	86.68
Wind	0.02	0.02	0.18	10.09	11.04	12.92	11.79	11	11.48	10.13	10.55	10.77	87.08
Pho	0.02	0.02	0.17	10.85	10.49	11.25	13.67	10.69	11.44	10.11	10.34	10.96	86.33
Hyd	0.01	0.04	0.09	10.19	11.02	10.53	10.81	13.5	11.28	10.56	11.37	10.61	86.5
Elec	0.01	0.02	0.11	10.96	10.75	10.72	11.31	10.88	12.49	10.59	10.85	11.32	87.51
Geo	0.01	0.04	0.17	10.14	11.14	10.18	10.63	10.85	11.51	13.26	11	11.07	86.74
Was	0.01	0.04	0.19	9.57	11.19	10.48	10.43	11.24	11.12	10.52	14.3	10.91	85.7
EM	0.02	0.05	0.13	10.29	10.58	10.57	11.29	10.81	11.74	10.46	11.06	13.01	86.99
TO	0.22	1.03	1.47	83.15	87.9	87.1	90.09	88.78	93.09	85.4	87.71	88.44	794.37
NET	−0.77	0.41	−12.86	−1.75	1.22	0.02	3.75	2.27	5.59	−1.34	2.01	1.45	66.2

, the total spillover between the clean energy market and other financial markets is more significant when the market environment is negative (66.2%) compared to when it is positive (58.65%), indicating an asymmetry in clean energy markets. Consistent with the analysis of split periods, it shows that the negative spillover effect of financial assets increases during periods of market volatility. Specifically, the spillovers from Pho and Was change from negative to positive when the market environment worsens. When the market environment is positive, green bonds are more effective in diversifying risk. However, as the market environment deteriorates, gold exhibits the smallest spillover effect from most clean energy assets. These findings differ from the previous analysis, as various clean energy assets, including nuclear, wind, hydropower, smart grids, and contract energy management, exhibited positive spillover effects when the market environment was negative. Additionally, in terms of financial assets, green bonds and technology markets change their roles from being net risk transmitters to risk recipients.

5.3. Dynamic Spillover Analysis

5.3.1. Rolling Time Window Analysis

In this subsection, time-varying return correlations were investigated by using a rolling window approach with a window length of 150 days. Figure 2 depicts the strong fluctuation of correlation levels in the system over time, emphasizing the significance of considering time variation when assessing the total return correlation of the system.

According to Figure 2, the overall spillover effect can be divided into four phases. In the first phase, which extends from early 2015 to late 2019, the spillover effect exhibited moderate fluctuations at around 72%. In the second phase, from early 2020 to early 2021, it showed a sharp increase and reached its apogee. This can be attributed to the perception of spillover effects in international financial markets following the outbreak of the COVID-19 pandemic, when investors perceived a gloomy investment outlook and sought assets to hedge against the risks associated with vulnerable assets. This finding aligns with the research of [41,42], which suggested that, during major outbreaks, the connectivity between financial markets intensifies, leading to a sentiment-driven investor behavior characterized by a pessimistic investment outlook. The third phase, spanning from 2021 to early 2022, witnessed a sharp decline in the overall spillover effect as the COVID-19 pandemic became more normalized and the control strategies became more effective. This alleviated investors’ fears and reduced the spillover effect. The fourth phase, after early 2022, saw a decline in the spillover effect to a bottom level of 52% before experiencing an increase, probably due to the outbreak of the Russia–Ukraine conflict.

Figure A1, Figure A2 and Figure A3 reveals interesting dynamics in the role of different assets over time. Here are the key observations:

(a)

Gold market: The spillover effect in the gold market was predominantly negative for most of the study period, but became positive in 2022. This shift may be attributed to increased demand for gold in the short term, driven by higher risk aversion following the outbreak of the Russia–Ukraine conflict. As the conflict situation normalizes and risk aversion subsides, the spillover effect of gold declines accordingly.

(b)

Technology stocks: The spillover effect of the technology stocks market shifted from negative to positive at the end of 2019, indicating a change in its relationship with other markets. This shift may be attributed to various factors such as market conditions, economic developments, or investor sentiment.

(c)

Risk diversification assets: Assets such as gold, green bonds, and energy futures generally experienced smaller shocks from and to other clean energy assets. This indicates their effectiveness in mitigating risk and their lower susceptibility to external influences. However, the degree of shocks from other assets to these financial assets increased at the end of 2019, suggesting a changing dynamic in their relationship with other markets.

(d)

Clean energy market: The spillover effect of the clean energy market has significantly increased after the COVID-19 pandemic. This suggests a higher degree of interconnectedness and influence between the clean energy market and other markets during this period. The reasons may include changes in government policies, market trends, or global economic conditions.

(e)

Nuclear energy: The spillover effect of nuclear energy increased sharply to an unprecedented level, primarily due to sanctions imposed on Russia. These sanctions had significant spillover effects and resulted in increased energy prices. This highlights the influence of geopolitical factors and policy decisions on the dynamics of the energy market.

Overall, these findings demonstrate the evolving nature of spillover effects among different assets and the impact of various events and factors on their interrelationships.

5.3.2. Spillover Network Analysis

Figure 3 illustrates the net volatility spillover network over the full sample based on pairwise directional connectedness in three periods (2015–2022, 2015–2019, and 2020–2022). The colors of the arrows represent the magnitude of risk spillover between the respective markets, ranging from purple (strongest) to pink and to white (weakest). Wider arrows indicate higher connectedness. Furthermore, the nodes are colored blue on the receptor side and yellow on the transmitter side. The node’s diameter corresponds to the market’s net spillover level.

Figure 3a depicts the directional linkages among pairs of markets during the full sample period, confirming the findings in Table 2. It reveals a lower correlation between financial markets, such as green bonds, gold, energy futures, and other markets. However, when dividing the sample period into two phases, distinct patterns emerge. In Figure 3b, representing a relatively stable market environment without major crises, the spillover effect is more regular. The connection edges between green bonds, gold, energy futures, and other markets become thinner and lighter in color, indicating a further decrease in correlation. Additionally, the connection edges among clean energy markets exhibit a consistent width and color, suggesting a relatively even distribution of correlation. Figure 3c illustrates the network connectedness after the outbreak of the COVID-19 pandemic. The correlations between different markets underwent varying changes. The risk spillover effect intensified across the financial system due to the pandemic, leading to enhanced information transmission. Notably, energy futures experienced the highest increase in spillover, while gold and green bonds exhibited relatively weaker cross-market spillovers, consistent with the findings of [43,44,45], suggesting their potential as protective measures against inter-market spillover risk. However, the previously high correlation between the technology market and other markets appeared to decrease. The distribution of correlations across clean energy markets became uneven, with the correlation decreasing to varying degrees. With the exception of smart grids, nuclear, and contract energy management, the remaining clean energy markets continued to exhibit strong correlations with other markets.

5.4. MST Analysis of Full Sample

Figure 4 presents the MST for the full sample period from 2015 to 2022. The MST shows that certain clean energy markets and various financial markets form clusters or groups. This suggests that these markets share strong interconnectedness and exhibit similar patterns of risk transmission. Of particular interest is the node “Elec”, which is connected to the largest number of other nodes in the MST. This indicates that the electricity market plays a crucial role in the risk transmission process and has a significant influence on the overall network structure. Changes or shocks in the electricity market can potentially impact a wide range of other markets, making it an important focal point for risk transmission analysis.

The community structure analysis revealed that green bonds, gold, and energy futures consistently belong to the same community, irrespective of the occurrence of the COVID-19 pandemic. This finding aligns with the results obtained from the previous time-phased study, where these assets exhibited lower correlations with other clean energy assets. The DY and BK models also indicated that green bonds, gold, and energy futures had weaker correlations with other clean energy assets. This implies that these assets can serve as effective hedging tools in an investment portfolio. Therefore, investors seeking to mitigate the risk associated with clean energy assets may consider incorporating green bonds, gold, and energy futures into their portfolios.

6. Conclusions

6.1. Summary of This Research

The main objective of this paper was to examine the spillover effects and risk transmission dynamics between clean energy markets, the green bonds market, and other financial markets. The analysis took into account the impact of the COVID-19 pandemic and the Russian–Ukrainian conflict and investigated both static and dynamic spillover effects. To achieve this, the paper utilized the Diebold and Yilmaz spillover index model to analyze the static spillover effects among different markets. The model allows for an assessment of the interdependencies and contagion risks between clean energy equities, green bonds, and other financial assets. Furthermore, the study employed rolling time windows to capture the dynamic nature of spillovers. This approach enabled us to observe how the spillover effects evolve over time and potentially identify periods of heightened risk transmission. In addition, the Baruník and Křehlík spillover model, based on the frequency domain, was employed to investigate the spillover effects in the short, medium, and long terms. This analysis provides insights into the different time horizons over which risk is transmitted between the markets under consideration. Lastly, the paper employed spillover network analysis to compare the direction and strength of risk transmission changes during three split periods. This approach allows for a more-detailed examination of the risk transmission channels and can help identify key markets and assets that play significant roles in the spillover network.

6.2. Key Findings

(a)

Static spillover effects: The analysis revealed that there is significant heterogeneity in the direction and magnitude of the net spillover effects among clean energy assets. Among them, smart grids exhibit the largest net spillover effect (18.64), indicating the strong influence on other markets and susceptibility to external shocks. The analysis highlighted the roles of the photovoltaic (−17.44) and green bonds markets (−24.63) as major receivers of spillover effects, implying that they tend to absorb risk from other markets rather than transmit it. Furthermore, the net spillover of the green bonds market is lower than other financial markets, indicating its potential contribution to risk diversification within the system.

(b)

Frequency domain spillover analysis: This analysis revealed that the spillover effect of long-term volatility (71.3) is significantly greater than that of short- (65.57) and medium-term (65.95) spillovers. In the short and medium terms, gold exhibited the lowest connectedness with other clean energy assets. This implies that gold may have a lower degree of correlation and dependence on the performance of clean energy markets during these time periods. As a result, choosing gold as a hedging asset in extreme cases could be a more-sensible strategy compared to green bonds.

(c)

Dynamic spillover effects: This research found that the spillover effects between the markets fluctuated over time, indicating changing levels of risk transmission. The results suggested that combining clean energy assets with other financial assets in a portfolio may not be practical during periods of heightened risk and market volatility. This is because the negative spillover effects of financial assets tend to increase under such conditions (from −2.09, −0.04, −8, −6.8 to −9.36, −12.14, −6.37, −18.02, respectively). In particular, the analysis highlighted the roles of green bonds in spreading the risk from clean energy assets (0.02, 0.02, 0.03, 0.01, 0.02, 0.01, 0.01, 0.02) when the market environment is positive and gold in spreading the risk from clean energy assets (0.11, 0.08, 0.17, 0.06, 0.12, 0.08, 0.04, 0.08) when the market environment is negative.

(d)

Special events shock: Since the COVID-19 pandemic, financial markets have become more exposed to shocks, leading to a notable increase in risk spillovers compared to clean energy markets. Specifically, the net spillover from nuclear energy has increased since 2020, becoming the largest transmitter of risk in the system. The directions of risk spillovers from gold and photovoltaics shifted from negative to positive, indicating a change in their relationship with other markets. Similarly, wind energy transitioned from being a transmitter of risk to a recipient. These findings highlight the importance of considering the changing dynamics and direction of risk spillovers in different markets, especially during periods of significant events.

(e)

Spillover network analysis: This analysis indicated that financial assets generally exhibit low correlations with clean energy assets. It was observed that the correlations were more evenly distributed before the COVID-19 pandemic. However, after the pandemic, there were significant variations in the band color and width across markets, indicating that the correlation structure of the system became more uneven and heterogeneous. The MST analysis suggested that gold, green bonds, and traditional energy shared a relatively lower level of correlation with clean energy assets. This finding supports the notion that these financial assets can act as diversification tools in an investment portfolio, providing a hedge against the risks associated with clean energy assets.

6.3. Recommendations

For investors, it is advised not to include an excessive number of clean energy assets in portfolios simultaneously due to the strong co-movements within the clean energy markets in both normal and extreme circumstances. It is crucial to carefully consider the spillover effects of clean energy markets when constructing portfolios. Green bonds can be a suitable choice for diversification purposes within the clean energy sector. During periods of heightened risk, gold can serve as a prudent strategy due to its risk-averse nature and potential to act as a hedge.

For policymakers, photovoltaics and green bonds represent a viable instrument for achieving environmental sustainability objectives due to their strong ability of absorbing risk from other markets. Therefore, policymakers should prioritize the development and promotion of green financial markets, taking into account the uncertainties associated with extreme market conditions. Appropriate environmental policies should be proposed to strengthen the smart grids market due to its strong influence on other markets and its susceptibility to external shocks. Additionally, policymakers should be flexible to make corresponding adjustments due to the changing dynamics of risk spillovers in the markets, especially in the face of special events.

6.4. Limitations and Future Research

Firstly, this study focused on markets in China. It will provide extensive inspiration with evidence from other states’ markets. Secondly, this study investigated the impact of the COVID-19 pandemic and the Russian–Ukrainian conflict on the risk spillovers. In future research, as time passes, it will derive robust and practical insights by including more special events. Thirdly, for comparison purposes, it will be valuable to conduct similar analysis using other techniques, such as planar maximum filter graph method and planar maximally filtered graph method.