Identifying the Key Drivers in Energy Technology Fields: The Role of Spillovers and Public Policies

Abstract

This study investigates the salient roles of knowledge spillover and environmental policies on clean technology innovation. Employing a panel vector autoregressive model (PVAR) and connectedness network analysis with a comprehensive longitudinal dataset comprising 100 million patent documents across 26 countries, the study identifies clean technology fields that are most efficient in driving innovation and subsequently quantifies the spillover effects for each field. The impact of public environmental policies and regulations on clean technological innovations is also examined in depth. The results reveal that clean innovation is a complex and nuanced system, with significant knowledge spillovers occurring within and across energy and non-energy-related clean technology fields. The results also show that environmental policies significantly influence clean innovation, with technology support and adoption support policies having the most substantial impact. Furthermore, the results reveal that the impact of market-based policies on clean innovation is weaker than that of non-market-based policies, which is an important consideration for policymakers. The findings hold significance for policymakers in addressing sustainability goals and their implications.

1. Introduction

In order to meet the Paris Climate Agreement’s aim of confining global warming to 1.5 °C to 2 °C by the end of this century, significant investments in environmental technologies are paramount, and a drastic shift from high-carbon technologies is essential. As the world intensifies its efforts to mitigate climate change, the role of environmental policies in fostering innovation across clean technology fields has become increasingly pivotal [1,2,3]. Acemoglu et al. [1] highlighted the importance of appropriate policy measures in fostering environmentally friendly technologies and mitigating the negative impacts of industrialization and economic growth on the environment. Decarbonizing the economy necessitates a profound shift toward sustainable practices, prompting policymakers to craft regulations that stimulate cross-technology knowledge spillovers. These spillovers are essential as they enhance research productivity, especially for less advanced technologies, by transferring insights and breakthroughs across different clean tech sectors. This dynamic not only bridges the gap between emerging and mature technologies but also catalyzes greater overall innovation. Empirical evidence suggests that strategic environmental policies can significantly amplify these knowledge exchanges, fostering an ecosystem where cleaner technologies evolve more rapidly and synergistically. However, the challenge lies in designing policies that balance immediate economic growth with long-term sustainable advancements. Understanding the intricate impacts of these policies on cross-technology knowledge spillovers is crucial for accelerating the clean energy transition and achieving a sustainable future.

Clean energy innovation is essential for combating climate change and achieving net-zero CO₂ emissions by offering sustainable alternatives to fossil fuels. Advancements in clean technologies increase efficiency, affordability, and accessibility, enabling a global shift toward low-carbon energy. This transition reduces greenhouse gas emissions, curbs climate change impacts, and fosters economic growth and energy security [4,5,6]. Numerous countries have implemented diverse environmental policies and regulations in pursuit of achieving net-zero CO₂ emissions. As Stern [7] points out, the efficacy of these policies and regulations is anticipated to be significantly enhanced if they establish incentives that foster the advancement of clean energy technologies by stimulating innovation. Climate change policies and regulations play a vital role in guiding and accelerating the transition toward a low-carbon, sustainable future. They establish clear targets to reduce greenhouse gas emissions, promote clean energy, and enhance energy efficiency. By setting standards, incentivizing clean technologies, and penalizing polluters, these policies stimulate innovation, encourage behavioral changes, and ensure accountability [8]. Climate regulations also foster international cooperation, enabling collective action to address the global challenge of climate change, and thus, safeguard the environment for future generations [9].

Recognizing the role of clean innovations, this study focuses on cross-technology spillovers across various clean technology fields and the impact of public environmental policies and regulations on both innovations and their interactions. Empirical evidence highlights the crucial role of cleaner production innovations in improving environmental performance without hindering economic growth [2,3,10,11,12]. The primary challenge for this transition to clean technology is that today’s innovation decisions are shaped by both current policies and the historical development of inventions.

The aim of this study is to examine the dynamic interaction among 16 clean energy technology fields. These fields include seven directly related to clean energy—solar, wind, bioenergy, grid, hydrogen and fuel cells, nuclear, and renewables—and nine not directly related to energy: agriculture energy efficiency, air–rail–marine, building energy efficiency (including renewable integration), carbon capture and storage, e-mobility, energy efficiency, industry energy efficiency or substitution, storage (excluding e-mobility), and vehicle fuel efficiency. This study also assesses the impact of four types of environmental policies and regulations—market-based policies, non-market-based policies, technology support policies, and adoption support policies—on clean energy innovation and their interactions. This study focuses on clean technology innovations because they illustrate significant examples of deliberate differential treatment by government policies aiming to mitigate climate change. Public policies incentivize clean technologies in two fundamental ways: firstly, by enforcing carbon pricing policies, which have been demonstrated to shift innovation activities toward clean technologies and away from dirty ones [1,2,3,13,14,15,16]; and secondly, by providing direct support for clean innovation [1,2,3,13,17,18,19]. For example, the European Union’s (EU) Horizon 2020 program invested over 3 billion euros in research and innovation for clean energy projects in 2019. Additionally, many governments have implemented policies and incentives to encourage individuals and businesses to adopt these technologies, such as tax credits for purchasing electric cars or subsidies for installing solar panels.

Our study makes several significant contributions to the existing body of literature. First, unlike previous studies that relied on aggregate patent citations [2,16,20,21], we investigate cross-technology spillovers based on the direct innovation outcome measured by patents in 16 distinct clean sub-technology fields. Our data, sourced from the Worldwide Patent Statistical Database (PATSTAT) provided by the European Patent Office (EPO), encompasses over 100 million patent documents from more than 90 patent authorities worldwide, covering 44 countries from the period 2000 to 2001. This rich dataset, disaggregated into 16 fields over 2 decades and numerous countries, enables a nuanced assessment of spillover dynamics and policy effects based on the direct innovation outcome measured by patents. Second, to the best of our knowledge, we are the first to employ a dynamic panel data vector autoregressive (PVAR) model combined with the Diebold–Yilmaz (DY) spillover index [22,23] to evaluate knowledge spillovers across clean technology fields based on their innovation outcome measured by patents. This approach takes into account the temporal evolution of innovation in each field. We extend the DY index within the PVAR model framework, allowing us to assess cross-technology spillovers flexibly using comprehensive spillover metrics. Third, our study also examines the impact of environmental policies and regulations on clean innovation using four subcategories within the same dynamic PVAR and DY spillover index methodology. Although prior studies (see, e.g., refs. [3,10,15,16,17,18,19,20,21,22,24,25,26,27]) have examined the role of policies on clean innovations, our study employs a comprehensive set of policy variables that capture various dimensions of public policies and regulations across all fields of clean technologies. Additionally, we investigate the role of policy variables in interaction with clean innovations, accounting for dynamic (lagged) impacts and feedback effects. Fourth, to determine the key technology fields and policy types, such as market-based policies, non-market-based policies, technology support policies, and adoption support policies, we utilize network connectedness statistics (see, e.g., refs. [26,27]). This approach allows us to identify technology fields and policy types that play a central role in knowledge spillovers, those that act as significant intermediaries, those that are most connected to others for knowledge spillovers, and those that belong to the same cluster. In assessing these contributions, we take into account that while market-based policies are designed to create financial incentives for innovation, non-market-based policies directly support research, development, and the adoption of clean technologies. Our analysis, therefore, investigates the relative strength and efficiency of these policy types in stimulating technological advancements. By addressing these dimensions, our study provides a robust and nuanced understanding of cross-technology spillovers, the role of policies, and the dynamics of innovation in clean technology fields. To empirically evaluate these objectives, we adopt a PVAR model, which is well-suited for capturing the dynamic relationships between variables over time. This model allows us to account for the interconnectedness of innovation drivers.

The results of the empirical analysis in this study have significant implications. Cross-technology knowledge spillovers among clean energy technology fields help mitigate the risk of diminishing returns to innovation by enhancing the research productivity of less developed technologies. Meanwhile, the substitutability of technologies fosters increasing returns by enlarging the market for more advanced technologies. This dynamic means that advanced technologies offer greater rewards for innovation, thereby incentivizing further advancements that bolster their competitive advantage. The presence of knowledge spillovers from innovative activities strongly supports the case for government intervention, as private R&D investments under laissez-faire conditions are typically inadequate. Effective government policies can amplify these spillovers, ensuring that investments in R&D are sufficient to drive the necessary technological advancements. By fostering an environment where clean innovations can thrive, policymakers can help accelerate the transition to a sustainable economy, addressing environmental challenges while promoting economic growth.

The rest of the study is organized as follows. In Section 2, we provide a comprehensive review of the relevant literature. Section 3 outlines the methods used in our empirical analysis. Section 4 describes the data employed in the study. Section 5 presents the empirical analysis. Section 6 presents a discussion of the empirical results. Finally, Section 7 concludes the study.

2. Literature Review

The significance of energy technology in the contemporary world is pivotal, as it profoundly influences the economic, environmental, and socio-political aspects of human well-being. In the existing literature, researchers have extensively explored the dynamic relationship between clean and dirty energy markets. This section is focused on strands of literature with different methods and techniques for empirical analysis of key drivers of energy technology.

In earlier studies on energy technology innovation, Newell [28] contributes to the literature by focusing on different aspects of technological innovation in the production and use of energy. He mentioned driving innovation in lower-carbon energy technologies once regulatory constraints have been adopted and prices begin to capture the environmental externality associated with greenhouse gases. Furthermore, Acemoglu et al. [1] explained technological change regarding particular factors of production and why issues of directed technical change are essential for sustainable growth. In the theoretical framework, Dechezleprêtre et al. [29] investigated how Tobin’s Q is linked to ‘clean’ and ‘dirty’ innovation and innovation efficiency at the firm level, with their findings showing that the value of clean innovation and innovation efficiency to those firms provides successful clean research and development activities.

The first strand of literature comprehensively explores the dynamic relationship between clean and dirty energy markets in the empirical literature. The study of Cheon and Urpelainen [30] investigated R&D expenditures and patents for renewable energy technology in industrialized countries. They found strong support for the interactive effect of international oil prices and sectoral innovation systems on public policy and innovation. Xia [31] examined the relationship between fossil fuels and clean energy markets by applying the bibliographic mapping method from 1991 to 2022. The results of this study suggest that earlier investigations primarily concentrated on the oil and clean energy markets, with a scarcity of studies exploring market interconnections. Moreover, Dias et al. [32] explored the dynamic relationship between clean and dirty energy markets using five clean energy indexes and four dirty energy indexes from 2018 to 2023. The results indicated that indexes for both clean and dirty energy lack hedging characteristics and fail to act as safe havens during periods of economic uncertainty. Another study by Tiwari et al. [33] investigated the dynamic associations between clean and dirty energy markets by applying the time–frequency wavelet’s multiple cross-correlation daily returns from 2013 to 2020. Their empirical findings showed that the fluctuations between clean and dirty energy sources exhibit volatility over extended periods, both in the medium and long term. Later, Bian et al. [34] analyzed the linkages between energy industry development and technological efficiency by applying the PVAR model. The findings demonstrated that there is a bidirectional, dynamic association between the new energy industry and technological efficiency.

The majority of energy technology research in the literature has concentrated on the empirical consequences of adopting renewable energy sources, investigating the tangible outcomes of transitioning from conventional fossil fuels to cleaner alternatives. Ilyas et al. [35] conducted a study investigating the transition to renewable energy from the perspective of environmental degradation and economic growth by using the Generalized Method of Moments (GMM) estimation model. Their empirical findings indicated that renewable energy significantly reduces environmental pollution and contributes positively to environmental sustainability. Yu et al. [36] investigated the dynamic link between renewable energy and energy intensity, demonstrating that renewable energy development can reduce energy intensity. More importantly, various studies [37,38,39,40,41,42,43,44] have pointed out the outcomes of adopting renewable energy sources. Lastly, Jafri and Liu [45] investigated the effect of education, environmental law, and technology on renewable energy consumption (REC) in China, and the findings concluded that REC is enhanced alongside an increase in average years of schooling.

Over time, there have been several studies on clean and dirty energy spillovers and connectedness in the context of global environmental sustainability and a transition to more sustainable energy systems. Fuentes and Herrera [46] explored the dynamics of connectedness among the realized volatility indices of 16 clean energy stocks by applying an impulse–response analysis. Their empirical findings indicated a unidirectional connection between the implied volatility indices and clean energy stocks. Similarly, Saeed et al. [47] investigated the connectedness of clean energy stocks and crude oil markets by applying quantile-based estimators. The findings provided evidence that there is significant return connectedness that varies with time, but it is less volatile in the tails. Later, Mamkhezri and Khezri [48] investigated the spillovers of renewable energy, R&D expenditure, and CO₂ emissions by applying fixed-effects panel analysis from 2003 to 2017. Their empirical findings showed that there is a significant spillover effect from economic growth and renewable energy on the reduction in CO₂ emissions. Lastly, Chen et al. [49] explored the risk of transmissions among clean energy markets, green bonds, and other financial markets in China by using spillover network analysis. The findings showed that within clean energy markets, there is diversity in the way net risk spillover, the types of hedging assets used, and responses to market volatility are manifested, both in terms of direction and magnitude.

In recent years, a plethora of research has given increasing attention to the risk transmission of clean and dirty energy markets. Khalfaoui et al. [50] analyzed the spillover effect of US stock market returns on climate change-related risks, including the green index, carbon price, and general climate uncertainty, using network analysis. They concluded that there is a strong spillover connectedness network among climate change-related risks that act as net contributors. Furthermore, Deng et al. [51] investigated the link between clean energy and non-ferrous metals, and the risk of contagion between the two markets through the spillover index model and the minimum spanning tree (MST) method. They concluded that risk is mainly transmitted from clean energy to non-ferrous metals. Next, Farid et al. [52] examined the connectedness of dirty and clean energies pre- and post-COVID-19. Their empirical evidence suggested that weak linkages between clean energy equities and dirty energies exist in the short run. Conversely, there are limited instances of significant co-movements between dirty and clean energy markets in the long run.

Another study by Tiwari et al. [53] employed time-varying and frequency-domain spillover estimates to investigate the asymmetric relationship between clean and dirty energy markets from 2011 to 2020. The dispersion of positive and negative volatilities in global energy market indices varies over time with different intensities. Additionally, empirical evidence suggests that, in the long run, positive news contributes more to the integration of international energy markets compared to negative news. Their findings also indicated the presence of asymmetric risk effects in investment opportunities within the realms of clean and dirty energy. Some other studies [54,55,56,57] have confirmed the asymmetric relationship between clean and dirty energy markets with causal factors.

Additionally, Lorente et al. [58] employed the Quantile Vector Autoregressive and wavelet coherence to examine the connectedness among the climate change index, green financial assets, renewable energy markets, and geopolitical risk index. Their empirical results indicated that the climate change market can be a safe haven, and the climate change index, green financial assets, and clean energy are strong influencers in the financial markets and are vital to reducing geopolitical risk. On the other hand, Marra and Colantonio [59] employed the PVAR model to examine the effects of social, technical, political, and cultural determinants on the energy transition from 1996 to 2020. Their findings showed the significant role of causal determinants in energy transition and increased public awareness. Lastly, more studies [10,39,60,61,62,63,64] have concentrated on providing spillovers of clean and dirty technology employing PVAR methods in the context of sustainable development.

Despite the growing interest in the transition to clean energies, identifying clean and dirty technology fields has not been extensively studied as a driving factor in innovation. Conversely, both clean and dirty technologies play crucial roles in promoting sustainable development by addressing environmental, economic, and social challenges at both local and global levels. Therefore, this study investigates the salient roles of innovations within 18 unique clean technology fields. Unlike earlier research, we concentrate on identifying the technology domains that demonstrate the highest effectiveness in fostering innovation and subsequently measure the spillover effects for each of these domains.

3. Methodology

3.1. Panel Vector Autoregressive Model

We use a panel vector autoregressive (PVAR) model, which was first introduced by Holtz-Eakin et al. [65] and later extended by Binder et al. [66], to analyze the dynamic interactions across the clean energy fields and environmental policy variables. By estimating the PVAR model, we aim to comprehend the dynamic interactions among 16 clean technology innovation variables, which are measured by patents specific to each technology field, and four indicators of environmental policy and regulations. The PVAR model treats all 20 variables as endogenous, enabling us to analyze their interactions across multiple entities (countries) over time. In other words, the PVAR model is better equipped to estimate dynamic interactions among variables that are contemporaneously affected by common shocks to the innovation process. Therefore, the PVAR model provides the opportunity to capture short-, medium-, and long-term interactions and varying transmission speeds among the variables by allowing us to estimate dynamic feedback between the variables’ lags and contemporaneous terms. It also allows us to measure the spillover of innovation across the clan energy fields and public policy variables using the DY spillover index approach and connectedness network metrics. To the best of our knowledge, spillovers among clean-tech innovation fields and environmental variables have not been explored before using the PVAR models.

Given $i=1,2,\dots ,N$ countries, $t=1,2,\dots ,T$ years, and $k=1,2,\dots ,K$ variables, we define $X_{it}=\left(X_{1,it},X_{2,it},\dots ,X_{K,it}\right)\prime$ as the $K\times 1$ vector of variables. For our case, all variables, $X_{1,it},X_{2,it},\dots ,X_{K,it}$, are assumed to be second-order stationary; thus, we do not have to consider cointegration analysis or other issues related to nonstationary. Then, the dynamic relationship among the $K$ variables can be modeled using a PVAR model of order $p$, which can be written as:

$$X_{it}=A_{0,i}+A_1{X}_{it-1}+A_2{X}_{it-2}+\dots +A_p{X}_{it-p}+u_{it}=A_{0,i}+{\displaystyle \sum }_{l=1}^p{A}_l{X}_{it-l}+u_{it}$$

(1)

where $A_l$, $l=1,2,\dots ,p$ are $K\times K$ coefficient matrices, $A_0,$ represents country-specific intercepts (fixed effects), and $u_{it}$ is an identically and independently distributed zero-mean error term with $K\times K$ covariance matrix $\mathsf{\Sigma }$, $u_{it}~iid\left(0,\mathsf{\Sigma }\right)$. Consistently with the stationarity of all variables, we assume that all unit roots of the matrix of the characteristic equation relating to Equation (1) lie within the unit circle to ensure the stability of the PVAR model. The PVAR model also assumes parameter homogeneity across countries for the matrices $A_1, A_2, \dots , A_p$ relating to the lagged terms so that the same dynamics apply to all countries. The lag order $p$ of the PVAR model is usually selected based on the information criteria, such as the Akaike information criterion (AIC) or the Schwarz (Bayesian) information criterion (SIC). Holtz-Eakin et al. [65] proposed an equation-by-equation estimator for the PVAR model. Binder et al. [66] expanded on this by extending the estimator of Holtz-Eakin et al. [65] to accommodate a PVAR model that includes only endogenous variables lagged by one period. When $p>1$, further extensions are required by including additional lags of the endogenous variables, as well as predetermined and strictly exogenous variables. However, the presence of lagged dependent variables on the right-hand side of the system of equations can introduce bias into the estimates, particularly when $T$ (the time dimension) is small, even if $N$ (the cross-sectional dimension) is large. In such cases, the Generalized Method of Moments (GMM) provides more reliable estimates of the parameters of a PVAR model, especially under conditions of fixed $T$ and large $N$.

The steps of inference based on the PVAR model are given in Figure 1. The process of conducting a connectedness network analysis based on a PVAR model follows a structured methodology. First, data preparation is undertaken to ensure the dataset is ready for analysis. This is followed by model specification, which involves defining the functional form, selecting variables, transforming and differencing data as needed, and choosing appropriate instruments. Once the model is specified, parameter estimation is performed using methods such as GMM or fixed effects. After estimation, model diagnostics are conducted including checks for serial correlation, heterogeneity, and stationarity and tests for parameter significance, variable relevance, and instrument validity. Structural inference and forecasting are then carried out, incorporating impulse response analysis and forecast error variance decomposition (FEVD) to generate forecasts.

The final step, the connectedness network analysis, represents the novel contribution of this study to the literature. Building upon generalized forecast error variance decomposition, this analysis performs a detailed spillover examination using Diebold and Yilmaz spillover indices. Furthermore, the connectedness network analysis is conducted through network metrics, offering new insights into the interconnections and spillover effects across variables within the PVAR framework. The connectedness network analysis significantly advances the understanding of dynamic spillover and network connectedness in applied econometrics.

The selection of a dynamic panel vector autoregressive (PVAR) model for analyzing knowledge spillover across clean energy technology fields is justified by several critical considerations. Firstly, PVAR models effectively capture dynamic relationships among multiple variables over time, which is essential in the context of clean energy technologies where the effects of knowledge spillovers may evolve rather than manifest immediately. Additionally, PVAR models address endogeneity and simultaneous feedback effects between variables, allowing for a nuanced exploration of reciprocal influences among technology fields. The utilization of panel data in PVAR models provides both cross-sectional and time-series dimensions, facilitating a comprehensive analysis that accounts for both short- and long-term effects. Furthermore, the incorporation of lagged dependent variables within PVAR models enables the quantification of lagged effects of knowledge spillover, enhancing the understanding of how past innovations impact current technological advancements. This flexibility in model specification also allows for the inclusion of various exogenous factors and control for unobserved heterogeneity across different technology fields, thereby increasing the robustness of the findings. In contrast, static fixed-effects models, while useful for controlling time-invariant characteristics, may overlook important temporal dynamics and interactions. PVAR models additionally support impulse response analysis and variance decomposition, which elucidate the structural relationships between different technology fields and their responses to shocks or innovations. Collectively, these attributes position the PVAR model as a more powerful analytical tool for examining the complexities of knowledge spillover in the rapidly evolving clean energy sector.

3.2. Diebold–Yilmaz Spillover Index

Once the parameters of the PVAR model are estimated, the dynamic response of an endogenous variable, say $X_l$, to shocks in other variables, $X_q, q\ne l$, over different horizons, $h=1,2,\dots ,H,$ is typically analyzed using impulse response functions (IRFs) and forecast error variance decomposition (FEVD). Recently, many time series studies [22,23,67,68] have also utilized the Diebold–Yilmaz (DY) spillover index to quantify the extent of spillovers, or interconnectedness, among different time series variables.

The DY spillover index is based on the FEVD. One challenge in computing FEVD is identifying the shocks associated with each variable in the reduced-form PVAR model as specified in Equation (1). One common approach is to orthogonalize the shocks to create a recursive system, resulting in orthogonalized IRFs and FEVD. However, this method can introduce an order dependence issue for the orthogonalized FEVD. To mitigate this issue, we utilize the generalized FEVD proposed by Pesaran and Shin [69], which avoids the order dependence problem. The generalized forecast error decomposition (GFEVD) of the PVAR in Equation (1) is given by

$${\theta }_{kj}\left(H\right)={\displaystyle \frac{{\sigma }_{jj}^{-1}{{\displaystyle \sum }}_{h=0}^{H-1}{\left(e_k^{\prime }{C}_h\Sigma e_j\right)}^2}{{{\displaystyle \sum }}_{h=0}^{H-1}(e_k^{\prime }{C}_h\Sigma C_h^{\prime }{e}_k)}}, H=1,2,\dots ,\mathsf{{\rm Y}}$$

(2)

where $k,j=1,2,\dots ,K$, $H=1,2,\dots ,\mathsf{{\rm Y}}$ is the forecast horizon and $C_h$ is a $K\times K$ matrix obtained from the vector moving-average representation $X_{it}={\mu }_{0,i}+{\displaystyle \sum }_{j=0}^{\infty }{C}_j{u}_{it-j}$ of the PVAR in Equation (1). FEVD defined in Equation (2) quantifies how much of the forecast error variance of each variable in the system can be attributed to shocks from each variable, including itself. FEVD helps in interpreting the relative importance or contributions of each variable’s innovations over time, providing insights into the dynamic interactions and the degree of interconnectedness among the variables in the PVAR system.

The Diebold–Yilmaz spillover index is a quantitative measure that gauges the extent of spillover/interconnectedness and dynamic linkages across different variables (technology field and policy variables in our case).

An issue with the FEVD in Equation (3) is that the sum over all variables is not equal to 1, ${\displaystyle \sum }_{j=1}^K{\theta }_{kj}\left(H\right)\ne 1$, which makes interpretation of the DY spillover index difficult. After normalizing ${\theta }_{kj}\left(H\right)$ as ${\tilde{\theta }}_{kj}\left(H\right)={\theta }_{kj}\left(H\right)/{\displaystyle \sum }_{j=1}^K{\theta }_{kj}\left(H\right)$ so that ${\displaystyle \sum }_{j=1}^K{\tilde{\theta }}_{kj}\left(H\right)=1$, the DY total spillover index is defined as

$$TCI={\displaystyle \frac{{{\displaystyle \sum }}_{k,j=1,k\ne j}^K{\tilde{\theta }}_{kj}}{K}}\times 100$$

(3)

${\displaystyle \sum }_{j=1}^K{\tilde{\theta }}_{kj}\left(H\right)=1$, ${\tilde{\theta }}_{kj}\left(H\right)$ measures the proportion of the forecast error variance in a particular variable $k$ after $H$ periods that can be attributed to shocks from variable $j$, relative to the total variance due to shocks in all variables including its own shocks. In other words, it quantifies how much of the variability in the forecast of each variable is due to spillovers from other variables. The TCI measures the proportion of the forecast error variance in all variables that can be attributed to shocks from other variables in the system, relative to the total variance of all variables that is explained by their own shocks and those of other variables. In other words, it quantifies how much of the variability in the forecast of all variables in the system is due to spillovers from other variables, measuring the extent to which the variability is explained by these spillovers or interconnectedness. The index ranges from 0% to 100%. A value closer to 0% indicates low interconnectedness or spillover effects, meaning that variables are mostly driven by their own shocks. A value closer to 100% signifies high interconnectedness, where shocks in one variable significantly affect other variables in the system.

There are also two directional DY spillover indexes constructed in a similar way to TCI but considering only one variable. The Diebold–Yilmaz directional spillover index “from” is a measure that quantifies the extent to which a particular variable in the system is affected by shocks from all other variables. This is in contrast to the “to” index, which measures the influence of a particular variable on all other variables. The “from” index measures spillover to variable $k$ from all other variables $j$, which is computed as:

$$D{S}_{j\to k}={\displaystyle \frac{{{\displaystyle \sum }}_{k=1, k\ne j}^K{\tilde{\theta }}_{kj}}{K}}\times 100$$

(4)

Analogously, the “to” index measures spillover from variable $k$ to all other variables to $j$ and is computed as:

$$D{S}_{j\leftarrow k}={\displaystyle \frac{{{\displaystyle \sum }}_{j=1, k\ne j}^K{\tilde{\theta }}_{kj}}{K}}\times 100$$

(5)

Finally, the net spillover index is calculated as the difference between the directional spillover “to” and the directional spillover “from” for a specific variable. Essentially, it tells us whether a variable is a net transmitter or a net receiver of shocks within the system, which is computed as $N{S}_k=D{S}_{j\leftarrow k}-D{S}_{j\to k}$

$$N{S}_k=D{S}_{j\leftarrow k}-D{S}_{j\to k}$$

(6)

A positive net spillover index indicates that the variable is a net transmitter of shocks within the system. This means that it contributes more to the forecast error variances of other variables than it receives from them. A negative net spillover index indicates that the variable is a net receiver of shocks. This means that the variable is more influenced by shocks from other variables than it influences them.

The Diebold–Yilmaz spillover index methodology offers significant advantages for analyzing knowledge spillover across clean energy technology fields, rendering it particularly well-suited for this context. Firstly, this methodology provides a quantitative measure of spillover effects, enabling researchers to assess the extent to which knowledge in one technology field influences others, thereby identifying not only the presence of spillovers but also their magnitude and direction. Moreover, the dynamic framework inherent in the Diebold–Yilmaz index accounts for the temporal evolution of spillover effects, capturing how the nature and intensity of knowledge interactions can change over time as technology fields develop. Importantly, this approach facilitates the examination of bidirectional spillovers, essential for understanding the intricate network of interactions among different clean energy technologies, as innovations in one field can simultaneously affect and be influenced by advancements in others. The methodology also aids in identifying key technology fields that serve as major sources or recipients of spillovers, providing vital insights for policymakers and stakeholders aiming to promote innovation by highlighting areas for strategic investment or collaboration. Additionally, the robustness of the Diebold–Yilmaz index to model specification ensures greater confidence in the results, as spillover estimates are less likely to be influenced by arbitrary choices in model design. The insights derived from this analysis can inform strategic decisions by government entities and industry stakeholders, guiding investment priorities, collaboration initiatives, and research funding to maximize the impact of innovation efforts. Furthermore, the methodology enables comparative studies across different technology fields, regions, or time periods, offering a consistent framework for assessing knowledge spillovers and facilitating broader analyses of the dynamics of clean energy technologies on a global scale. Collectively, these attributes underscore the significance of the Diebold–Yilmaz spillover index for providing a nuanced, quantitative, and dynamic understanding of the interactions among clean energy technology fields, ultimately aiding in the formulation of effective strategies for fostering innovation and enhancing technological development.

3.3. Network Connectedness Metrics

In the empirical analysis of this study, the DY spillover indexes can be utilized to determine the extent of knowledge spillover across clean technology fields and to assess the influence of policy variables on innovations within these fields. Directional spillover indexes help quantify the significance of spillovers both from and to a specific technology field or policy variable. They also identify whether a field is a net transmitter or receiver of innovation. However, these measures alone are not sufficient to identify the key drivers of knowledge spillover or the centrality of a particular technology field or policy variable. In our context, centrality metrics are used to identify the most influential or central technology fields and environmental policy types within the innovation landscape, focusing on the interactions among these fields and policies. Various types of centrality metrics exist, each designed to capture different aspects of connectedness. Below, we briefly explain the centrality metrics used in our study.

The first metric the study uses is in-degree centrality, which measures the number of incoming links or connections a node has in a network. For the application in the study, a high in-degree centrality suggests that the technology field benefits significantly from the developments in other fields. In the context of innovation networks, a technology field or policy with high in-degree centrality is considered pivotal in attracting knowledge or influence from other fields or policies.

Out-degree centrality, the second metric used in the study, measures the number of outgoing links or connections a node has in a network, indicating how many other nodes it influences. In the context of innovation networks, a clean technology field with high out-degree centrality is a significant source of knowledge spillovers to other fields. High out-degree centrality for a policy variable indicates its strong influence on innovations across multiple clean technology fields, shaping the overall innovation landscape.

The third metric used in the study, closeness centrality, measures how close a node is to all other nodes in a network. In innovation networks, a clean technology field with high closeness centrality can quickly access and disseminate knowledge spillovers. A policy with high closeness centrality efficiently influences and integrates with multiple clean technology fields, acting as a central conduit for innovation and information flow.

Eigenvector centrality, the fourth metric used in the study, is a measure of the influence of a node in a network, accounting not only for the number of direct connections it has (degree centrality) but also the quality of those connections. In a network of clean technology innovation fields and environmental policies, eigenvector centrality can be interpreted to identify key innovation fields or policies that are most influential within the network.

The fifth centrality metric used in the study is betweenness centrality, which measures the extent to which a node lies on the shortest paths between other nodes, indicating its role as a bridge or intermediary. In innovation networks, a clean technology field with high betweenness centrality facilitates knowledge spillovers between other fields, acting as a critical connector. For a policy variable, a high betweenness centrality implies that the variable is essential for influencing and coordinating innovation across multiple clean technology fields, ensuring efficient information flow and integration within the network.

Page rank centrality, the sixth metric used in the study, measures a node’s importance based on both the quantity and quality of its incoming links, considering the influence of the nodes linking to it. A clean technology field with high page rank centrality is highly influential, receiving significant knowledge spillovers from other key fields and spreading it to other key fields with its outgoing links. For a policy variable with high page rank centrality, innovation is effectively driven by attracting influence from crucial clean technology fields, thereby playing a pivotal role in shaping the innovation landscape.

We also identify optimal clusters using a community detection algorithm that maximizes modularity. Modularity measures the degree to which a network can be divided into distinct modules, also referred to as groups, clusters, or communities. Thus, we group together technology fields and policy variables that are more densely connected within the same cluster compared to those outside, providing insights into how these components interact and influence one another within the innovation network. This approach helps uncover the underlying structure and key clusters that drive innovation in clean technologies and environmental policies.

4. Data

The study employs a panel dataset with an annual frequency, covering a period of 21 years from 2000 to 2021, for a total of 44 countries. Clean patent data came from the Worldwide Patent Statistical Database (PATSTAT) [70] provided by the European Patent Office (EPO). The PATSTAT is a comprehensive and extensive database that contains over 100 million patent documents from over 90 patent authorities worldwide. It is one of the largest, most reliable, and up-to-date patent databases available for patent analysis. PATSTAT, with its expansive and comprehensive database of patent documents from around the world for 44 countries and 47 distinct sectors, allows us to identify clean energy-related innovation. The particular dataset of PATSTAT data used in this study has been sourced from the Intellectual Property Database of the Science, Technology, and Innovation (STI) Micro-data Lab of the OECD (available at http://oe.cd/ipstats, accessed on 10 February 2023). The patent documents relevant to the study have been identified through the Y02 classification scheme developed by the EPO [71] and IEA [72]. We identify 18 clean technology fields for which the data are constructed by searching over 100 million patent documents. The evolution of the shares of each technology field over time is presented in Figure 2.

The clean technology fields in Figure 2 include bioenergy, grid, hydrogen and fuel cells, nuclear, renewables, solar, wind, other renewables, agriculture energy efficiency, air–rail–marine, building energy efficiency, carbon capture and storage, e-mobility, energy efficiency, industry energy efficiency or substitution, storage (excluding e-mobility), vehicle fuel efficiency, and renewable energy integration in buildings. As Figure 2 shows, the patent counts for renewable energy integration in buildings are quite small. Thus, this category is merged with the related main field of building energy efficiency. Similarly, other renewable energy fields with small patent counts are merged with the main category of renewables. As a result, 16 clean energy fields are included in the PVAR model. The policy variables used in the study cover the period from 2000 to 2020. One of the countries, Türkiye, has unbalanced data. Therefore, merging the patents and policy datasets results in a panel dataset with 26 countries and 21 years, covering the period from 2000 to 2020.

Table 1. Data information.

Variable Name	Code	Description	Category	Period
Clean energy patents
Bioenergy	BN	Bioenergy	Energy	2000–2021
Grid	GR	Grid	Energy	2000–2021
Hydrogen	HY	Hydrogen and fuel cells	Energy	2000–2021
Nuclear	NC	Nuclear	Energy	2000–2021
Renewables	RN	Renewables (all types)	Energy	2000–2021
Solar	SL	Solar	Energy	2000–2021
Wind	WN	Wind	Energy	2000–2021
Agriculture	AG	Agriculture energy efficiency	Non-energy	2000–2021
AirRailMarine	AR	Air—rail—marine	Non-energy	2000–2021
Building	BL	Building energy efficiency (including renewable integration)	Non-energy	2000–2021
CarbonCapture	CR	Carbon capture and storage	Non-energy	2000–2021
EMobility	EM	e-Mobility	Non-energy	2000–2021
EnrEfficiency	EN	Energy efficiency	Non-energy	2000–2021
Industry	IN	Industry energy efficiency or substitution	Non-energy	2000–2021
Storage	ST	Storage (excluding e-mobility)	Non-energy	2000–2021
Vehicle	VH	Vehicle fuel efficiency	Non-energy	2000–2021
Environmental policy
EPS	EP	Environmental policy stringency index	Policy	2000–2020
MarketPolicies	MP	Market-based policies	Policy	2000–2020
NonMarketPolicies	NP	Non-market-based policies	Policy	2000–2020
TechnologySupport	TS	Technology support policies	Policy	2000–2020
AdoptionSupport	AS	Adoption support policies	Policy	2000–2020
Countries
	Australia, Austria, Belgium, Brazil, Canada, Czech Republic, Denmark, Finland, France, Germany, Hungary, India, Ireland, Italy, Japan, Korea, Netherlands, Norway, People’s Republic of China, Poland, Russian Federation, Spain, Sweden, Switzerland, Türkiye, United Kingdom, United States

presents the variable names and other information, as well as the names of the 26 countries, after merging the data with policy variables and forming a balanced dataset.

As Table 1 shows, seven of the clean technology fields—which include bioenergy, grid, hydrogen and fuel cells, nuclear, renewables, solar, and wind—are energy fields. The remaining nine clean technology fields—which include agriculture energy efficiency, air/rail/marine, building energy efficiency (including renewable integration), carbon capture and storage, e-mobility, energy efficiency, industrial energy efficiency or substitution, storage (excluding e-mobility), and vehicle fuel efficiency—are non-energy fields related to clean energy or the clean energy transition.

Figure 2 indicates that storage-related clean energy technology patents hold the largest share of total clean energy patents. This share has shown an increasing trend, except for a decline during the COVID-19 period in 2021. The storage field is followed by industry energy efficiency or substitution, e-mobility, and building energy efficiency, respectively, in terms of their shares of total clean energy patents. Solar energy leads in innovation among the clean energy technology fields.

Figure 3 presents the density plot (Figure 3a) and a box plot (Figure 3b) of the logarithm of all clean energy patents across all years, technology fields, and countries. Additionally, it includes box plots of the logarithm of clean energy patents for each year across all countries and technology fields (Figure 3c). As the figure indicates, the patents exhibit a right-skewed distribution, reflecting an increasing trend in clean energy innovation. The yearly box plots reveal a flattening effect in the trend after 2017, followed by a decline during the COVID-19 period of 2020–2021.

One of the primary aims of this study is to examine how climate change policies and regulations affect clean innovations. Given that our study utilizes panel data from 27 countries, and that a multitude of policies and regulations have been introduced by these nations, it can be quite challenging to construct a small number of variables that effectively capture such a diverse array of policies and regulations. Consequently, we employ the four main categories of the environmental policy stringency (EPS) index developed by the OECD [70]. The EPS index is a measure developed to assess the rigor and comprehensiveness of environmental policies across different countries. It categorizes policies into four main subcategories. Market-based policies include financial instruments like taxes for CO₂, NOx, SOx, and diesel fuel, as well as CO₂ certificates and renewable energy certificates, which create economic incentives to reduce emissions. Non-market-based policies involve regulatory measures such as emissions limit values for NOx, SOx, and particulate matter, and sulfur content limits for diesel. Technology support policies focus on investments in research and development (R&D) to drive innovation in clean technologies. Lastly, adoption support policies for wind and solar encompass subsidies, tax credits, and infrastructure development to promote these renewable energy sources. Together, these components offer a comprehensive view of a country’s commitment to reducing its environmental impact and supporting sustainable development. The policy indexes range from zero, indicating the least stringent policies, to six, representing the most stringent policies, and cover the period from 1990 to 2020 for 40 countries.

The average EPS index across all countries is displayed in Figure 4b for each year from 2000 to 2020. As we observe from Figure 4b, the global trends and changes in climate change policies and regulations from 2000 to 2010 were characterized by a growing recognition of the need to address climate change and reduce greenhouse gas emissions. The global trends and changes in climate change policies and regulations since 2011 have been characterized by a mix of progress and setbacks. While there have been some positive developments, such as the adoption of the Paris Agreement in 2015, there has also been a slowdown in the pace of policy implementation and a weakening of some existing policies [9]. Figure 4a represents the trend in the number of patents related to clean innovations over the period from 2000 to 2020. The number of patents for clean innovations steadily increased until peaking at around 25,000 in 2019. By 2020, the number of patents dropped dramatically. This steep decline could be attributed to several factors, such as reduced government support or incentives for clean technology, changes in market dynamics, or even the effects of global economic events (e.g., the COVID-19 pandemic) that may have impacted innovation and patent activity.

5. Empirical Results

The dynamic nature of the PVAR model and the estimation methods we employ require a balanced dataset. After balancing the panel and combining all variables, we obtain a panel dataset with 546 observations covering 26 countries and 21 years, from 2000 to 2020.

Table 2. Descriptive statistics.

Variable	Obs.	Mean	Median	Min	Max	S.D.	Q(0.25)	Q(0.75)	CIPS Test a
Bioenergy	546	18.08	7.12	0.00	167.76	26.59	2.25	20.64	−4.35 ***
Grid	546	10.90	2.00	0.00	136.83	22.00	0.00	9.12	−2.81 **
Hydrogen	546	59.35	8.00	0.00	690.40	123.96	2.00	38.29	−2.94 ***
Nuclear	546	9.28	1.00	0.00	85.00	17.00	0.00	9.00	−3.40 ***
Renewables	546	12.57	5.00	0.00	112.00	19.08	0.00	16.00	−3.60 ***
Solar	546	69.69	10.29	0.00	935.73	146.49	3.00	44.67	−2.71 **
Wind	546	32.31	7.50	0.00	497.21	58.64	2.51	31.10	−3.08 ***
Agriculture	546	5.73	2.00	0.00	126.00	11.84	0.00	6.00	−3.69 ***
AirRailMarine	546	39.22	5.45	0.00	838.46	99.47	1.25	22.00	−2.76 **
Building	546	86.00	18.87	0.00	891.64	164.51	4.33	62.78	−2.83 **
CarbonCapture	546	6.15	1.00	0.00	75.77	11.97	0.00	6.00	−3.57 ***
EMobility	546	95.31	6.00	0.00	2924.01	268.40	1.00	36.11	−3.04 ***
EnrEfficiency	546	5.55	1.00	0.00	87.33	12.55	0.00	3.65	−2.86 **
Industry	546	86.43	10.67	0.00	1405.68	207.78	3.00	52.70	−2.88 **
Storage	546	118.07	8.44	0.00	2071.82	287.27	2.69	42.33	−2.66 *
Vehicle	546	13.23	1.33	0.00	263.00	30.06	0.00	7.25	−2.80 **
MarketPolicies	546	2.57	2.78	0.17	4.89	0.99	1.97	3.22	−2.72 **
NonMarketPolicies	546	1.41	1.17	0.00	4.17	0.91	0.83	1.83	−4.40 ***
TechSupport	546	4.33	4.75	0.50	6.00	1.44	3.50	5.50	−2.87 **
AdoptionSupport	546	1.99	1.75	0.00	6.00	1.32	1.00	3.00	−2.75 **

^a CIPS denotes cross-sectionally augmented Im–Pesaran–Shin test for unit roots. *, **, and *** denote significance at the 10%, 5%, and 1% levels, respectively.

presents the descriptive statistics for the dataset used to estimate the empirical specifications in the study. Policy variables are measured on a scale from 0 to 6, with the highest average observed for technology support policies, which have a mean of 4.33, followed by market-based policies with a mean of 2.57. Non-market policies exhibit the weakest stringency, with a mean of 1.41. The minimum value for all variables, except market-based policies, is uniformly zero, as there is at least one country in one or more periods with no environmental policy introduced and no clean patent registration. The 0.25th [Q(0.25)] and 0.75th [Q(0.75)] quantiles reported in Table 1 indicate moderate variability across all countries and years compared to the range of the variables measured, defined by the maximum minus the minimum value. Among the technology fields, storage leads in patents with a mean of 118.07, followed by e-mobility and industry energy efficiency or substitution, with means of 95.31 and 86.43, respectively. The smallest patent mean is observed for energy efficiency (5.55), followed by agriculture (5.73) and carbon capture (6.15). Among the direct energy-related technology fields, solar leads in average patent count with a mean of 69.69, followed by hydrogen (59.35) and wind (32.31). Table 2 also reports a cross-sectionally augmented Im–Pesaran–Shin (CIPS) test for unit roots. The CIPS test rejects the unit root null hypothesis at the 5% level for all series, except for storage, for which the null is rejected at the 10% level. Therefore, we conclude that all series are second-order stationary.

Figure 5 presents the Pearson correlation coefficient estimates for all variables used in the study. All technology field patents exhibit positive correlations, with most estimates ranging between 0.75 and 0.95. Environmental policy variables are moderately correlated with innovation variables, with the highest correlations observed for technology support and adoption support. Market-based policies show the weakest correlation with innovations in clean technology fields. Policy variables also display moderate to high correlations among themselves, with moderate correlations (0.30 to 0.55) across market-based and non-market-based support policies, and a high correlation (0.74) between technology and adoption support policies.

To estimate the DY spillover indexes and perform network analysis, we specify a PVAR model, selecting the lag order based on the Bayesian information criterion. The optimal lag order for the PVAR is determined to be 1, and the model parameters are estimated accordingly. Once the PVAR model is estimated, the FEVD and DY spillover indexes can be calculated straightforwardly using Equations (2)–(6). The resulting DY spillover index estimates are presented in

Table 3. Spillover estimates.

	BN	GR	HY	NC	RN	SL	WN	AG	AR	BL	CR	EM	EN	IN	ST	VH	MP	NP	TS	AS	From
BN	24.20	1.43	3.28	1.12	7.44	9.28	3.39	1.64	4.96	9.70	3.81	2.96	6.00	2.11	4.01	4.55	1.99	2.74	2.01	3.38	75.80
GR	3.78	17.45	3.91	1.87	5.18	5.78	6.12	0.91	0.79	16.24	2.07	3.88	6.44	4.25	3.81	4.73	3.02	2.75	2.86	4.18	82.55
HY	0.97	7.19	30.65	1.94	1.58	4.62	2.13	3.16	9.89	10.19	1.94	4.61	2.95	0.73	8.53	5.57	0.57	0.34	0.89	1.56	69.35
NC	2.57	2.33	3.18	23.75	2.97	6.46	5.80	1.24	0.94	7.82	1.48	2.91	6.59	2.25	2.19	2.17	5.46	3.94	7.32	8.63	76.25
RN	6.94	3.73	2.92	0.70	29.39	6.40	3.32	0.82	2.21	10.04	1.66	2.17	7.06	1.62	2.50	2.94	4.19	4.02	3.09	4.31	70.61
SL	4.21	1.76	3.06	1.87	4.74	16.67	4.74	3.30	2.07	12.25	1.51	4.86	11.18	3.29	4.78	4.03	3.05	3.14	3.84	5.64	83.33
WN	6.02	2.55	0.79	1.12	5.47	9.27	30.68	0.24	1.48	5.66	2.60	3.69	4.17	0.96	4.34	3.60	3.18	2.79	4.45	6.95	69.32
AG	2.33	5.03	2.06	2.26	2.25	4.72	2.18	20.85	2.45	14.39	0.79	3.01	10.64	12.81	3.06	1.80	2.37	2.37	2.10	2.53	79.15
AR	2.23	0.67	0.18	0.89	2.20	1.35	3.03	3.61	36.84	4.70	0.88	1.20	1.02	3.92	2.54	9.97	6.04	3.80	6.75	8.17	63.16
BL	3.68	4.31	3.84	2.51	3.55	8.66	5.32	2.14	0.89	20.92	1.55	5.81	8.07	5.31	4.53	5.76	3.14	3.21	2.95	3.87	79.08
CR	6.03	2.39	4.95	1.34	5.10	10.18	6.71	0.17	1.06	10.40	20.20	4.00	5.02	0.37	3.27	3.92	2.81	2.70	3.88	5.53	79.80
EM	2.96	2.58	4.28	2.19	2.99	9.12	13.55	0.56	0.63	9.08	2.09	11.37	4.92	1.24	8.22	8.23	3.04	2.44	4.50	6.03	88.63
EN	3.17	3.14	2.30	2.52	3.76	8.54	7.62	1.81	0.56	11.83	1.26	5.50	16.34	4.29	4.80	5.37	3.99	3.50	4.33	5.38	83.66
IN	3.52	5.85	1.81	1.84	2.56	7.23	3.70	3.29	2.52	12.84	1.43	4.17	6.04	22.62	5.01	4.17	1.91	2.88	2.68	3.95	77.38
ST	2.96	3.76	5.36	2.39	3.26	11.15	7.60	2.08	0.69	12.08	1.52	7.49	8.32	2.67	9.11	5.65	2.76	2.23	3.63	5.29	90.89
VH	3.51	2.46	4.18	2.46	3.71	7.96	8.81	0.21	1.54	10.11	2.42	8.55	5.04	0.69	6.53	15.27	3.75	3.00	4.23	5.55	84.73
MP	0.24	0.38	1.37	0.62	0.07	0.18	0.51	0.09	0.09	0.40	0.07	0.02	0.48	1.64	0.29	2.25	80.70	9.19	1.17	0.23	19.30
NP	0.41	1.95	0.58	0.37	0.45	0.41	2.20	0.26	1.67	2.40	0.53	0.18	0.82	8.60	0.44	0.23	2.33	62.24	6.37	7.56	37.76
TS	0.41	0.58	0.59	0.09	1.67	0.08	4.56	0.71	0.25	0.31	0.19	0.37	0.60	0.46	0.11	0.27	9.66	3.59	47.16	28.33	52.84
AS	1.30	1.85	0.94	0.11	2.81	1.00	8.47	0.43	0.11	1.79	0.63	1.28	1.71	0.78	0.52	0.30	4.39	1.74	23.27	46.58	53.42
To	57.24	53.93	49.57	28.18	61.74	112.41	99.75	26.66	34.81	162.22	28.43	66.65	97.07	57.99	69.47	75.49	67.64	60.36	90.34	117.07	70.85
Net	−18.56	−28.63	−19.79	−48.06	−8.88	29.08	30.43	−52.48	−28.35	83.15	−51.37	−21.98	13.41	−19.39	−21.42	−9.24	48.33	22.59	37.50	63.65

Notes: The table reports the generalized spillover measures estimated using the approach of Diebold and Yilmaz [22,23]. The lag order of the PVAR models is 1, which is selected using the BIC. Bold denotes the overall spillover index.

.

The Diebold–Yilmaz spillover estimates reported in Table 3 indicate that 70.85% of the fluctuations in all 20 assessed variables can be attributed to internal spillovers among these variables. Specifically, 12% of the internal spillover effects originate from environmental policy variables, including market-based policies (MPs), non-market-based policies (NPs), technology support policies (TSs), and adoption support policies (ASs). Among these policy categories, all act as net transmitters, meaning they exert a greater influence on clean technologies than they receive. Notably, technology support policies and adoption support policies are more influential compared to market-based and non-market-based policies, suggesting that the direct promotion of technological advancement and adoption has a more substantial impact on the spread and efficacy of clean technologies. Within the energy-related sector, solar and wind technologies are the largest net transmitters, significantly affecting other technologies. In terms of non-energy-related technologies, building energy efficiency (BL), energy efficiency (EN), and vehicle fuel efficiency (VH) technologies act as strong net transmitters. Conversely, agriculture (AG), carbon capture and storage (CR), and nuclear (NC) technologies are the most prominent net receivers, benefiting significantly from advancements in other technologies.

Figure 6 presents various representations of the innovation connectedness network, based on a weighted adjacency matrix formed by the spillover estimates reported in Table 1. The figure is divided into four parts: Parts (a) and (b) display the full and net connectedness networks, respectively. The full connectedness network is derived from spillover estimates obtained using Equation (2), while the net connectedness network is formed from pairwise net spillover estimates calculated using Equation (6). Parts (c) and (d) show the innovation connectedness networks after applying a 25% threshold, which involves replacing the adjacency matrix values that are less than the 0.75th quantile with zero. Part (c) reports pairwise spillover estimates, while part (d) is based on the pairwise net spillover estimates.

In Figure 6, thresholding helps to highlight more significant connections within the connectedness networks. The connectedness networks are constructed based on forecast error variance decomposition, which projects error variances 10 years into the future. This long-term perspective helps to better understand the lasting impact of the variables. Within these networks, the vertices (nodes) are distinctively colored to represent different categories, including clean energy fields, traditional energy fields, non-energy fields, and environmental policy variables. This color coding facilitates easier identification and comparison across categories. The shape of each node conveys additional information. Circular nodes denote net volatility transmitters—variables that predominantly send volatility to other variables. On the other hand, square nodes indicate net volatility receivers—variables that primarily absorb volatility from others. Furthermore, the size of each node is proportional to its degree, which is the total of its in-degree and out-degree connections. A higher degree signifies a larger number of connections, indicating a more central or influential role within the network. This sizing method aids in quickly identifying the most connected and potentially most influential nodes in the network.

The notable characteristics of the clean innovation connectedness network representation displayed in Figure 6 reveal a complex and nuanced interplay among various clean technology fields and environmental policy variables. From Figure 6a, it is apparent that both direct energy and non-energy-related clean technology fields exhibit strong internal connections, with significant knowledge spillovers occurring within and across these categories. The environmental policies also demonstrate considerable interaction, with the most substantial feedback observed between technology support and adoption support policies. This finding is not surprising, as renewable energy technologies are often seen as the cornerstone of clean innovation and have received significant attention and investment in recent years. Interestingly, this central position is also reflected in the strong connections between renewable energy technologies and environmental policies, particularly technology support policies. Figure 6a also highlights the role of government policies in shaping the clean innovation landscape, with strong connections between technology support policies and both direct energy and non-energy-related clean technology fields. This finding underscores the critical role of government support in promoting clean innovation and highlights the need for effective policy design and implementation to drive progress in this area.

Although all policy variables significantly influence clean technology fields, Figure 6c highlights that technology support and adoption support policies have a more widespread effect on clean innovation, as evidenced by their larger in-degree and out-degree compared to market-based and non-market-based policies. The fields of building energy efficiency (BL), solar (SL), energy efficiency (EN), and wind (WN) emerge as the largest knowledge-sharing entities and are identified as net knowledge transmitters. Conversely, vehicle fuel efficiency (VH), storage (ST), grid (GR), bioenergy (BN), and renewables (RNs) are net knowledge receivers.

The prominence of building energy efficiency (BL), solar (SL), energy efficiency (EN), and wind (WN) as the largest knowledge-sharing entities can be attributed to several factors. First, these fields have experienced significant technological advancements and policy support over the past decades, leading to the development of mature knowledge bases and established innovation ecosystems. For instance, solar and wind technologies have been at the forefront of clean energy transition efforts, supported by extensive research, public subsidies, and global deployment initiatives. Moreover, building energy efficiency and overall energy efficiency are critical across multiple sectors, making them central to the diffusion of innovations aimed at reducing energy consumption and improving sustainability in both industrial and residential settings. The combination of strong policy support, widespread adoption, and the maturity of these technologies enables these fields to act as net knowledge transmitters, facilitating the diffusion of best practices and innovations across the broader clean technology landscape.

Conversely, vehicle fuel efficiency (VH), storage (ST), grid (GR), bioenergy (BN), and renewables (RNs) are identified as net knowledge receivers due to several reasons. These fields, while crucial to the clean energy transition, are either still evolving or face significant technological and infrastructure challenges. For instance, storage technologies, such as batteries, are in the early stages of development and deployment, particularly in scaling for industrial or grid-level applications. Similarly, grid innovations and bioenergy have seen slower technological advances, making them more reliant on knowledge from more established sectors like solar, wind, and energy efficiency to drive their own development. Vehicle fuel efficiency, although progressing with the rise of electric vehicles, still depends heavily on innovations from energy storage and grid technologies. These factors contribute to these fields being net knowledge receivers, as they absorb innovations from the more developed clean technology fields to enhance their own growth and impact.

Connections based on net spillovers, as shown in Figure 6b,d, are represented by arrow lines sized proportionally to the net spillover magnitude. These figures underscore that all policy variables act as net transmitters, with adoption support having the largest impact. The net connectedness networks indicate that policy variables exert a relatively stronger influence on solar (SL), nuclear (NC), bioenergy (BN), grid (GR), agriculture (AG), air–rail–marine (AR), carbon capture and storage, energy efficiency (EN), and e-mobility (EM). Policy variables exert a relatively stronger influence on these technology fields due to the strategic importance and regulatory focus on these sectors in the global clean energy transition. These fields are either highly dependent on government support for research and development, such as nuclear energy and carbon capture, or are directly impacted by policy-driven adoption incentives, such as solar, bioenergy, and e-mobility. For instance, subsidies and feed-in tariffs have historically accelerated solar and bioenergy growth, while stringent emissions regulations promote advancements in nuclear, CCS, and transportation technologies like e-mobility and air–rail–marine [73]. Policies that encourage energy efficiency and grid modernization also align with national and international goals for reducing emissions and integrating renewable energy, further strengthening their influence in these fields.

When considering the innovation connectedness networks with 25% thresholding, as depicted in Figure 6c, a similar pattern emerges with some minor adjustments and clearer network links. Policy variables, including technology support, adoption support, and market-based policies, maintain their relative importance and net transmitter status. However, non-market-based policies shift to a net receiver position, indicating that policymakers are increasingly factoring in feedback from clean innovation when forming policies. The interaction within direct energy-related technology fields becomes weaker, while the strong interaction across the energy and non-energy clean technology fields remains robust. Policy variables continue to impact both categories of clean innovation, albeit with a reduced influence from market-based and non-market-based policies.

Examining the net connectedness network with thresholding in Figure 6d reveals that policies have the strongest impact on nuclear and air-rail-marine technologies. Only non-market-based policies are influenced by technology fields, specifically industry energy efficiency (IN), and they exhibit the weakest impact on clean technologies.

In analyzing homogeneous groups within the clean innovation connectedness network, we identify three distinct clusters. This indicates that interactions among clean technology fields with respect to knowledge sharing and the impact of public policies are modular but not fragmented, as they are confined to these three clusters. Significantly, four environmental policy variables—market-based policies, non-market-based policies, technology support policies, and adoption support policies—converge to form a separate, distinct cluster. When considering the pairwise spillover network illustrated in Figure 6a, we observe that agricultural energy efficiency and industrial energy efficiency or substitution form a unique cluster. In contrast, the remaining clean technology fields amalgamate into a large cluster comprising 14 fields. This result appears intuitive, as energy efficiency domains inherently exhibit unique characteristics that segregate them from other clean technology fields. The third cluster identified within the clean innovation connectedness network includes all renewable energy technologies, such as solar, wind, and hydro power. This cluster also includes energy storage and grid technologies, indicating a strong interconnectivity between these fields. This is not surprising, as renewable energy technologies are often interconnected and complementary in their use. For example, solar panels and wind turbines can be used in conjunction to generate electricity, and energy storage technologies are necessary to store excess energy generated by renewable sources.

To understand the key roles played by various clean technology fields and environmental policies,

Table 4. Metrics of network characteristics.

		In-Degree	Out-Degree	Closeness	Eigenvector Centrality	Betweenness	Page Rank
Energy	BN	75.80	57.24	0.03	0.59	0	0.06
	GR	82.55	53.93	0.04	0.61	0	0.06
	HY	69.35	49.57	0.05	0.52	167	0.06
	NC	76.25	28.18	0.04	0.46	16	0.05
	RN	70.61	61.74	0.04	0.58	0	0.05
	SL	83.33	112.41	0.04	0.86	0	0.06
	WN	69.32	99.75	0.03	0.74	0	0.05
	Group mean	75.32	66.12	0.04	0.62	26	0.05
Non-energy	AG	79.15	26.66	0.06	0.48	10	0.06
	AR	63.16	34.81	0.05	0.40	172	0.04
	BL	79.08	162.22	0.03	1.00	0	0.06
	CR	79.80	28.43	0.05	0.49	13	0.05
	EM	88.63	66.65	0.04	0.70	2	0.06
	EN	83.66	97.07	0.03	0.80	0	0.06
	IN	77.38	57.99	0.04	0.58	0	0.06
	ST	90.89	69.47	0.03	0.72	0	0.06
	VH	84.73	75.49	0.04	0.70	0	0.06
	Group mean	80.72	68.76	0.04	0.65	22	0.06
Policy	MP	19.30	67.64	0.03	0.34	83	0.02
	NP	37.76	60.36	0.04	0.39	66	0.03
	TS	52.84	90.34	0.03	0.58	11	0.03
	AS	53.42	117.07	0.03	0.68	2	0.03
	Group mean	40.83	83.85	0.03	0.50	41	0.03

Notes: The network statistics are based on the connectedness networks defined using the spillover table reported in Table 3.

presents network centrality metrics across several dimensions. In-degree metric shows that technology fields have equivalently structured inflows, with non-energy-related technologies receiving more inflows (benefiting more from other technologies) than energy-related technologies. The highest in-degree values are found in the fields of storage (ST), e-mobility (EM), energy efficiency (EN), vehicle fuel efficiency (VH), solar (SL), and grid (GR). This indicates that all these clean technology fields efficiently receive knowledge from other fields. Conversely, all four policy variables exhibit small in-degree values, indicating a limited response of policies to changes in clean technologies. The out-degree metric shows significant differences across the clean technology fields in terms of out-degree centrality. The highest out-degree values are noted for building energy efficiency (BL), solar (SL), wind (WN), and energy efficiency (EN), highlighting the crucial role of these fields in disseminating innovations to other fields. In stark contrast, all policy variables show high out-degree values, with the highest value observed for adoption support, signifying their critical role in spreading policy effects to other domains. Closeness centrality values are quite uniform across the fields, with agriculture standing out as having the closest location to all technology fields. This uniformity suggests that all fields are in an equally efficient position to share knowledge across the network. The most influential fields, as indicated by eigenvector centrality, include building energy efficiency (BL), solar (SL), wind (WN), energy efficiency (EN), storage (ST), e-mobility (EM), and vehicle fuel efficiency (VH). These fields or policies are key innovation domains within the network, suggesting that they play pivotal roles in influencing and driving overall network dynamics.

As indicated by the betweenness centrality, the fields of hydrogen and fuel cells (HY, energy-related) and air–rail–marine (AR, non-energy-related), along with market-based (MP) and non-market-based (NP) policies, hold substantial bridging roles in the transmission of technology. These elements act as crucial intermediaries, facilitating the flow of innovation and information between disparate parts of the network. Hydrogen and fuel cells, as energy-related technologies, have broad applications across energy generation, storage, and transportation sectors. They are increasingly viewed as critical to decarbonizing hard-to-abate sectors like industrial processes and heavy transport, while also complementing renewable energy sources by providing long-term energy storage solutions. Their versatility enables them to act as a key intermediary, connecting different technology fields, from energy production to vehicle propulsion, thereby facilitating the transfer of knowledge and innovation across sectors. Similarly, air–rail–marine technologies, though non-energy-related, are crucial in reducing emissions from transportation sectors that are traditionally hard to decarbonize. As they intersect with advancements in fuel efficiency, hydrogen, and biofuels, they play a significant role in spreading technological innovations that help mitigate emissions in heavy-duty transport modes. The bridging roles of these fields are essential to achieving cross-sectoral synergy in the clean energy transition.

Considering the page rank metric, the relative importance of all nodes appears similar, with non-energy technologies having a marginally higher role overall. This uniformity implies a balanced contribution across the network, though non-energy technologies slightly edge out in terms of importance.

6. Discussion

The findings from the connectedness network analysis using spillover indices reveal distinct patterns of knowledge transmission and reception among clean technology fields, particularly when examining the influence of different policy types. While all policy variables significantly impact clean innovation, our findings underscore that technology support and adoption support policies exert a more extensive influence compared to market-based and non-market-based policies. This is evidenced by the higher in-degree and out-degree of these policies, signifying their broader role in facilitating knowledge flows. Notably, the clean technology fields of building energy efficiency (BL), solar (SL), energy efficiency (EN), and wind (WN) emerge as the largest knowledge-sharing entities and act as net knowledge transmitters. These fields are not only pivotal in advancing clean innovation but also serve as key hubs in disseminating knowledge to other sectors. In contrast, vehicle fuel efficiency (VH), storage (ST), grid (GR), bioenergy (BN), and renewables (RNs) are primarily net knowledge receivers, relying more heavily on the knowledge shared by the aforementioned sectors. These results align with previous studies that emphasize the central role of specific technologies, such as solar and wind, in the diffusion of clean energy innovations [18,73]. However, the identification of building energy efficiency and energy efficiency as critical knowledge transmitters provides new insights into the interconnectedness of clean technology fields, expanding upon earlier findings by highlighting sectors previously underexplored in terms of their influence on broader clean innovation ecosystems [74].

The fields of building energy efficiency (BL), solar (SL), wind (WN), energy efficiency (EN), storage (ST), e-mobility (EM), and vehicle fuel efficiency (VH) emerge as the most influential sources of knowledge spillover in clean energy technologies due to their foundational role in the clean energy transition. These technologies have been subject to extensive research, policy support, and global innovation efforts, positioning them as key drivers of technological advancements across the clean energy sector. Building energy efficiency and energy efficiency have become increasingly central due to their cross-sectoral applicability. Innovations in these fields improve energy use in both residential and industrial settings, directly reducing energy demand and creating synergies with renewable energy adoption [75]. Solar and wind technologies have been among the most rapidly growing renewable energy sources, benefitting from substantial policy incentives, cost reductions, and deployment at scale [18]. Their technological advancements have generated spillovers to related fields, such as energy storage and grid management, which rely on the integration of intermittent renewable energy sources. Storage is a critical component of the clean energy transition as it enables the reliable integration of variable renewable energy into the grid, facilitating energy flexibility and resilience [76]. Likewise, e-mobility and vehicle fuel efficiency are crucial to reducing transportation emissions, a sector that accounts for a significant portion of global greenhouse gas emissions. Innovations in these areas, such as advancements in battery technology and electric vehicle adoption, create knowledge spillovers that extend into energy storage and grid technologies, driving further advancements in clean energy integration [18]. Overall, these fields represent interconnected hubs of innovation, with strong knowledge spillovers influencing other sectors of clean energy technology development. Their centrality within the innovation network reflects their role in advancing complementary technologies, thereby driving the broader diffusion of clean energy innovations.

In this study, the fields of building energy efficiency, solar, wind, energy efficiency, storage, e-mobility, and vehicle fuel efficiency are identified as the most influential within the innovation network, as indicated by their high eigenvector centrality scores. This finding aligns with the existing literature, where similar technological domains are highlighted as critical drivers of innovation in energy systems and sustainability. For instance, previous studies emphasize the central role of renewable energy technologies such as solar and wind in shaping future energy systems. According to Bogdanov et al. [77], solar and wind technologies have been identified as pivotal in transitioning to low-carbon energy systems due to their scalability and integration into energy grids. Similarly, the importance of building energy efficiency has been discussed in Eleftheriadis et al. [78], which stresses the growing innovation in sustainable building technologies and their substantial influence on energy consumption patterns. Shaukat and Khan [79] also underscore the role of e-mobility and vehicle fuel efficiency in reducing greenhouse gas emissions, highlighting their importance in both the automotive sector and broader energy systems. The current findings reinforce these previous observations by quantitatively demonstrating the prominence of these fields in network dynamics. Moreover, the inclusion of energy storage and energy efficiency technologies aligns with Bessa et al. [80], who emphasize that energy efficiency and storage are critical for managing energy demand and supporting renewable energy integration.

The results of this study identify hydrogen and fuel cells and air–rail–marine fields as key intermediaries in the transmission of technology, as evidenced by their high betweenness centrality. This indicates that these fields play substantial bridging roles, connecting various innovation domains and facilitating the flow of information and technological advancements across the network. These findings are consistent with previous studies that have highlighted the intermediary roles of both hydrogen technologies and the air–rail–marine sector in promoting technological diffusion and innovation across multiple sectors. Hydrogen and fuel cells have been widely recognized for their diverse applications across energy generation, storage, and transportation, serving as a flexible solution in the transition to low-carbon energy systems. McPherson et al. [81], for instance, emphasize the importance of hydrogen technologies in linking energy generation and storage with transportation, reinforcing their role as a bridge between various energy sectors. Similarly, Sovacool et al. [82] highlight the versatility of hydrogen in acting as a connector in decarbonization efforts, particularly in sectors like transportation and industrial processes. These studies align with the current findings, reinforcing the significance of hydrogen as a central node in energy innovation networks. The air–rail–marine sector’s bridging role in the transmission of technology, despite being non-energy-related, is also supported by the literature. Hesse and Rodrigue [83] demonstrate how innovations in transportation infrastructure, particularly in air, rail, and marine systems, serve as crucial intermediaries, linking different parts of the global economy and enabling the transfer of technological advancements across sectors. This finding is further supported by Geels [84], who argued that non-energy-related sectors like transportation play a pivotal role in integrating innovations that influence energy efficiency and system connectivity, contributing to overall network dynamics. The importance of betweenness centrality in identifying these bridging technologies aligns with Schilling and Phelps [85], who emphasize that technologies and sectors with high betweenness centrality often act as hubs for innovation, facilitating the diffusion of new technologies and ideas across otherwise disconnected parts of the network. These insights further underscore the pivotal role of hydrogen, fuel cells, and air–rail–marine sectors as innovation intermediaries.

The findings from the connectedness network analysis indicate that policy variables, particularly adoption support, play a critical role as net transmitters, significantly influencing key clean technology fields. This aligns with previous studies that highlight the importance of government policies in driving innovation and adoption in clean energy sectors. For instance, Popp et al. [18] noted that policy interventions such as subsidies, tax incentives, and emission regulations have been instrumental in the growth of solar, bioenergy, and e-mobility technologies. Similarly, Johnstone et al. [18] found that policy-driven support for nuclear and carbon capture and storage technologies is essential due to the high capital costs and long development timelines associated with these fields. The strong policy influence on energy efficiency and grid technologies is also consistent with the findings of Lafferty and Ruud [86], who highlighted the role of regulatory frameworks and incentives in advancing energy efficiency measures and grid modernization efforts. This study adds to the existing literature by demonstrating the disproportionate impact of adoption support policies on these sectors, reinforcing the need for targeted policy interventions to accelerate clean energy transitions.

7. Conclusions

The transition to a low-carbon and sustainable economy is a significant challenge that requires significant advances in clean technology innovation. The growing recognition of the critical role of clean technology innovation in achieving environmental sustainability has led to increased attention and investment in this area. However, to promote clean innovation effectively, policymakers need a comprehensive understanding of the knowledge-sharing dynamics among clean technology fields and environmental policies. This paper presents a novel approach to analyzing the knowledge-sharing dynamics among clean technology fields and environmental policies using a complex network analysis framework. The proposed approach provides a comprehensive overview of the knowledge-sharing dynamics among clean technology fields and environmental policies and can help policymakers identify the most influential clean technology fields and policies and inform the design of effective policies to promote clean innovation.

The results of the analysis reveal a complex and nuanced interplay among various clean technology fields and environmental policy variables. This study highlights the identification of strong knowledge spillovers within and across clean technology fields, especially between energy and non-energy sectors. It also reveals that technology support and adoption policies are influential in driving clean innovation, particularly in renewable energy and emerging technologies. The clean innovation connectedness network representation highlights the strong connections between renewable energy technologies and environmental policies, particularly technology support policies. The results also show that clean technology fields are highly interconnected, with significant knowledge spillovers occurring within and across categories. The analysis also highlights the critical role of government policies in shaping the clean innovation landscape, with technology support policies and adoption support policies exerting the most significant influence. The results also reveal that policy variables have a relatively stronger impact on solar, nuclear, bioenergy, grid, agriculture, air–rail–marine, energy efficiency, and e-mobility technologies. Moreover, the clean energy cluster is closely connected to the policy cluster, indicating that public policies play a critical role in promoting the development and adoption of renewable energy technologies. This finding highlights the importance of policy support in driving clean technology innovation. The close connection between the clean energy cluster and the policy cluster also suggests that policies targeting renewable energy technologies may have spillover effects, not only within the renewable energy domain but also across other clean technology fields.

Moreover, the presence of distinct clusters within the clean innovation connectedness network has important implications for policymakers and industry practitioners. First, it highlights the need for targeted policies that address the specific characteristics and needs of different clean technology fields. For example, policies promoting energy efficiency may not be as effective in promoting renewable energy technologies, as these fields exhibit different technological and market dynamics. Second, the close connection between the renewable energy cluster and the policy cluster suggests that policies targeting renewable energy technologies may have spillover effects, not only within the renewable energy domain but also across other clean technology fields. This highlights the potential for policy interventions to promote knowledge spillovers and cross-sectoral collaboration, which can lead to more efficient and effective clean technology innovation.

The network centrality metrics highlight the importance of certain clean technology fields and policies in driving innovation and knowledge diffusion in the clean technology network. The fields of building energy efficiency, solar, wind, energy efficiency, storage, e-mobility, and vehicle fuel efficiency are identified as key players in receiving and disseminating knowledge, while policies such as adoption support, market-based policies, and non-market-based policies play crucial roles in spreading policy effects to other domains. Additionally, the fields of hydrogen and fuel cells and air–rail–marine, along with market-based and non-market-based policies, are identified as important intermediaries in facilitating the flow of innovations and information between different parts of the network. These findings highlight the need for a balanced approach toward clean technology development and implementation, with a focus on both energy and non-energy technologies, as well as a mix of policies to drive innovation and diffusion.

The analysis of knowledge spillover across clean technology fields and the impact of environmental policies underscores several critical policy implications. First, targeted investment in key technologies is paramount. By focusing funding and resources on the most efficient technology fields and fostering public–private partnerships, we can maximize innovation outputs and accelerate advancements. Second, enhancing spillover effects through cross-sector collaboration and the establishment of innovation hubs will facilitate knowledge and technology transfer, amplifying the benefits across different clean technology fields. Furthermore, tailored public policies are necessary; sector-specific approaches should be developed to align regulatory support with the unique needs of each field, accompanied by clear and stable long-term policy signals to reduce uncertainty and encourage sustained investments. Support for emerging technologies is also crucial, requiring increased early-stage funding and backing for pilot and demonstration projects to validate and de-risk new innovations. Lastly, enhanced international cooperation in the form of harmonized global standards, best practices, and joint research initiatives can accelerate global progress in clean energy technologies, ensuring a collaborative approach to tackling environmental challenges.

Considering the empirical findings, this study proposes policy suggestions to distinguish the role of market-based policies such as carbon pricing and emissions trading in driving innovation in more mature technologies, such as renewable energy and emissions reduction technologies, where market signals can effectively guide investment and development. On the other hand, non-market-based policies such as R&D funding and regulatory standards are shown to be more effective in fostering innovation in earlier-stage or underperforming technologies, such as energy storage, waste-to-energy, and energy efficiency technologies, where direct support is crucial for overcoming technical and market barriers.

While market-based policies, like carbon pricing, were found to have a weaker impact, they still hold the potential for fostering a conducive innovation environment. From this point, governments should prioritize increased R&D funding and demonstration projects for technologies like energy storage while leveraging market-based incentives such as tax credits and subsidies for sectors like emissions reduction and renewable energy integration. Governments should refine these mechanisms to better encourage investment in clean technologies, particularly in energy-intensive sectors. Promoting public–private partnerships can leverage resources and expertise in emerging sectors, facilitating knowledge spillovers between energy and non-energy-related technologies. Finally, establishing a monitoring and evaluation framework will enable governments to assess the effectiveness of various policies, allowing for real-time adjustments and enhancements in their approach to advancing sustainability goals.

In conclusion, our analysis of the clean innovation connectedness network provides valuable insights into the interconnectivity and interdependence among different clean technology fields. By identifying distinct clusters within the network, we highlight the modular but not fragmented nature of interactions among clean technology fields. Moreover, our findings suggest that public policies play a critical role in promoting clean technology innovation and that policies targeting renewable energy technologies may have spillover effects across other clean technology fields. These insights have important implications for policymakers and industry practitioners seeking to promote clean technology innovation and transition to a more sustainable future.

Future research should explore the long-term effects of evolving environmental policies on clean technology innovation and provide comparative studies across different regions or sectors that could provide deeper insights into the varying impacts of policy frameworks on clean innovation.