Do green innovation and governance limit CO₂ emissions: evidence from twelve polluting countries with panel data decision tree model

Abstract

We examine the effectiveness of green innovation on CO₂ emissions in the top twelve polluting nations—China, the US, India, Russia, Japan, South Korea, Canada, Mexico, Turkey, Italy, Poland, and the UK—from 1996 to 2020. Using panel data fixed and random effect model and decision tree analysis, we found that industrialization, urbanization, and economic growth increase CO₂ emissions, whereas green energy consumption and governance decrease CO₂ emissions. In the panel data tree-based model, governance is in the second and third positions in the decision tree fixed, and random effect model. Green innovation is not statistically significant despite the expected negative sign. The findings suggest that policymakers should encourage investment in green energy production and governance to combat environmental degradation. Investment in green energy should be escalated to ensure energy efficiency in the long term.

1 Introduction

Environmental issues are becoming increasingly critical. The burning of the Amazon forest and the melting of northern glaciers are major concerns. Global climate change and environmental degradation are serious hazards, threatening human life and jeopardizing the health of ecosystems, biodiversity, and natural resources. Consequently, the massive use of fossil fuels caused CO₂ emissions worldwide to rise to 36.8 gigatons (Gt) in 2022 [1]. Carbon dioxide emissions account for 75% of all greenhouse gas (GHG) emissions, significantly contributing to environmental degradation [2]. Global carbon emissions increased from 9.147 billion tonnes to 33.798 billion tonnes between 1960 and 2020. The world average temperature is predicted to rise by 1.2 °C by 2020, surpassing the European Union's three°C human tolerance limit [3]. According to Intergovernmental Panel on Climate Change (IPCC) projections, the Earth's average temperature will increase by more than 3 °C by the twenty-first century if global warming persists unchecked [4]. Furthermore, environmental pollution threatens human health and significantly influences psychological well-being, cognitive processes, and decision-making [5, 6]. Excessive CO₂ emissions have increased natural disasters, including extreme rainfall and droughts [7, 8]. The consequences of these actions include threats to vital resources like water and food supplies. Climate change poses significant threats to the environment, human population, and economy by disrupting supply chains, damaging infrastructure, and raising operational risks for businesses due to extreme weather events [9]. Consequently, these occurrences represent severe risks to social welfare, political stability, and economic growth [10, 11]. This problem affects not just the environment's sustainability but also humankind's future [12].

Climate change is primarily caused by greenhouse gas emissions, with fossil fuels and nonrenewable energies accounting for about three-quarters of global CO₂ emissions [12]. Businesses face significant challenges in adapting to climate change, developing eco-friendly ideas, and mitigating climate-related risks. Since climate change impacts environments, communities, and economies [13], businesses find it hard to include green innovations in their strategies. They encounter obstacles such as limited financial resources, technical limitations, and a shortage of qualified workforce. Moreover, innovation has an inherent element of risk. Moreover, innovation has an inherent element of risk. Corporations continue struggling to balance their short-term financial aims with their long-term sustainability ambitions, which has resulted in a notable decrease in green investment [12]. Due to climate change vulnerability, companies typically prioritize short-term challenges above long-term green innovation investments. In response, countries have modified their development policies to address climate change.

Globally, countries have implemented climate policies to regulate greenhouse gas emissions and advance sustainable development [14]. For instance, during the 27th United Nations Climate Change Conference in 2022, governments worldwide demonstrated their commitment to addressing climate change through laws, policies, and legal avenues. Furthermore, governments are implementing incentives to promote clean energy infrastructure and green technological development [15, 16]. The UN introduced the sustainable development goals in 2015, with 'SDG-13' emphasizing the urgent need to combat climate change and its impact. Every global economy aims to achieve a carbon–neutral target by reducing emissions by 2050. Green innovations are gaining traction globally as societies and industries respond to the urgent need for sustainability and environmental stewardship. Green innovations are multifaceted, encompassing technological, economic, and societal advancements to create a sustainable future for future generations. Green innovation promotes a transition to a green economy that is less reliant on volatile fossil fuel markets and more resilient to global economic shocks. Green innovation aims to mitigate the negative impacts of human activities on the environment, including climate change, biodiversity loss, deforestation, and pollution. Green innovation is crucial to combat climate change by reducing greenhouse gas emissions from fossil fuel burning, focusing on renewable energy, energy efficiency, and carbon capture technologies. Green innovation aims to reduce pollution and create healthier living environments to combat public health issues caused by environmental degradation and pollution. Global countries are committing to international agreements like the Paris Agreement to tackle environmental challenges, with green innovation playing a crucial role in achieving emissions reduction goals. Green innovation represents a comprehensive approach to addressing pressing global issues and creating a sustainable future for all. In this background, this paper analyses the effectiveness of green innovation in reducing CO₂ emissions in the top twelve polluting countries, investigates the contributing roles of industrialization, urbanization, economic growth, green energy consumption, foreign direct investment, and governance, and provides policy recommendations for encouraging renewable energy and governance to combat environmental degradation.

This study employs a robust methodological framework covering data from 1996 to 2020. The panel data model improves the reliability of results by addressing country-specific heterogeneity and providing more observations. In our study, we used panel fixed and random effect models and then compared whether the fixed or random effects model is more appropriate for empirical estimation. Additionally, decision tree models within the panel data structure are introduced to evaluate the significance of independent variables on CO₂ emissions, using mean square error for model performance evaluation.

Our investigation added three new insights to the body of literature. Firstly, the study investigates the impact of governance on CO₂ emissions, utilizing existing research to bridge the gap between the two. This study addresses a critical literature gap and provides insights for climate-related strategic decision-making. Exploring these systems presents an opportunity to design efficient solutions that combine energy saving, emission reduction, and green transformation, promoting sustainable development. The study's pivotal findings cater to policymakers and stakeholders in the top twelve polluting nations, highlighting the necessity of implementing proactive green energy consumption policies to diminish CO₂ emissions. This research adds to the growing work that looks at the effectiveness of green energy consumption and the achievement of emissions reduction. Second, the study uses principal component analysis to present the governance indicators. Third, we use a decision tree model in a panel data frame.

This study is organized into six sections. It starts with an introduction to the critical environmental issues and a review of the relevant literature (Sections 1 and 2), followed by the methodology used (Section 3). The data, variables used and their sources are explained in Section 4. Section 5 covers the presentation of results, encompassing empirical findings, robustness checks, and interpreting these outcomes. Lastly, Section 6 addresses the study's implications and limitations and provides policy recommendations for mitigating the impact of green innovation on CO₂ emissions.

2 Review of literature

Carbon dioxide emissions are the primary source of climate change; they pose serious risks to ecosystems, economies, and public health everywhere. For this reason, CO₂ emissions are a global issue. The greenhouse effect is strengthened by CO₂ accumulation in the atmosphere, which raises global temperatures and causes more frequent and severe weather events like hurricanes, droughts, and floods. CO₂ emissions contribute to air pollution, which harms public health by resulting in respiratory illnesses and early mortality. CO₂ emissions reduction is crucial for mitigating climate change's effects, protecting vulnerable populations, promoting sustainable development, and ensuring a livable planet for future generations.

Green innovation is increasingly recognized as a crucial strategy to combat CO₂ emissions and mitigate the impacts of climate change. Green innovation, often used interchangeably with environmental innovation, eco-innovation, and eco-efficiency, is a closely related concept in the literature. Green innovation is a process that promotes the development of green technologies, aiming to transition towards low-carbon energy consumption, energy saving, and pollution reduction. Green innovation focuses on developing innovative products, technologies, and business strategies to reduce environmental impact, conserve energy and resources, and improve production efficiency, promoting sustainable resource use [17]. It involves resource recycling, environmental protection, cleaner production, renewable energy use, energy-efficient technologies, and balancing environmental responsibility and economic objectives [18]. It encourages new ways to make things better for the environment by reducing pollution and other adverse effects of resource extraction [19]. Green innovation fosters economic growth, energy security, public health, and climate resilience, driving sustainable development and creating new opportunities for businesses and societies in a carbon-constrained world, thereby addressing the climate crisis and promoting sustainable growth. Green innovation is recognized as a powerful solution for tackling environmental challenges, offering advantages for both the environment and the economy [20,21,22].

2.1 Green innovation and environmental imperatives

The topic of green innovation has attracted considerable academic attention in recent years, primarily due to increasing environmental concerns and the depletion of natural resources. Research indicates that green innovation is crucial for addressing climate change and other environmental challenges, steering the world toward carbon neutrality [23, 24]. Consequently, many organizations are adopting green innovation as a strategic approach to promote environmental protection and economic growth. Therefore, as highlighted by [25], green innovation is essential for businesses to combat climate change and promote sustainable development. Green innovation aims to reduce environmental degradation, particularly climate change, by reducing resource consumption, pollution, and greenhouse gas emissions, driving motivation for such initiatives. Green innovation has demonstrated its effectiveness in various areas, such as reducing toxic emissions [26], lowering CO₂ emissions [27], boosting carbon productivity [28], enhancing urban eco-efficiency [16], supporting new urbanization efforts [29], and harmonizing ecological protection with economic growth [20]. The goals of green innovation include reducing waste, conserving energy, cutting pollution, and lessening a company's adverse environmental effects [30, 31]. Green innovation fosters job creation enhances public health by reducing pollution and enhances climate resilience.

Green innovation is critical for businesses to reduce environmental risks, establish new market niches, and increase competitiveness in sustainable industries, especially in light of rising global warming concerns [32, 33]. As environmental regulations become stricter, businesses must adopt green innovation as a crucial strategy. However, green innovation investment may increase costs and lead to uncertain returns, impacting a company's short-term financial performance due to more extended research and development cycles [34, 35]. The high risks associated with green innovation could result in substantial losses for firms, potentially impacting managers' careers through diminished reputations, reduced compensation, and possible dismissals [36]. Managers who prioritize financial targets often avoid green innovation, focusing instead on profitable projects during their tenure. Despite these challenges, green innovation helps businesses achieve long-term financial gains, competitive advantages, and compliance with strict environmental standards [37]. Due to the high levels of confidentiality, substantial investment requirements, and elevated risks associated with green innovation activities, significant information asymmetry can arise between companies and investors [38]. Green innovation's value is widely acknowledged, but corporate decision-makers and shareholders may face challenges in deciding whether to invest in it.

The role of green innovation in reducing CO₂ emissions has been examined in limited studies. To illustrate, using the DOLS approach [39], found that green innovation significantly reduced CO₂ emissions in Indonesia from 1990 to 2020 [40]. study uses the DOLS approach to examine the impact of green energy consumption on Colombia's CO₂ emissions from 1990 to 2020. The study indicates that implementing green energy practices significantly reduces CO₂ emissions. Furthermore, [41] investigated the influence of green energy consumption on CO₂ emissions in Ghana, spanning from 1990 to 2020. Employing the dynamic ARDL simulation method, the study illustrates that green energy consumption effectively mitigates environmental degradation by reducing CO₂ emissions. In addition, [42] examined the impact of green investment on carbon emissions and found that green investment decreases CO₂ emissions. Similarly, [43] found that increased green innovation is associated with reduced CO₂ emissions in the top 10 most polluted nations from 1991–2018. Meanwhile, [44] found a positive relationship between green technology innovation and economic growth in Singapore but a negative relationship with carbon emissions from 1990–2018.

2.2 Green innovation, technological change, and CO₂ emissions

Green innovation is a process that promotes environmental sustainability by developing less carbon-intensive technologies like renewable energy, energy-efficient processes, and sustainable manufacturing processes. It helps transition from fossil fuel dependency and reduces CO₂ emissions. Supportive policies, research incentives, and international collaboration accelerate the adoption of these technologies, contributing to a more sustainable future. Green innovation can reduce CO₂ emissions in polluting countries by promoting sustainable practices and technological advancements, including renewable energy technologies, energy-efficient processes, green transportation, and sustainable agriculture. Furthermore, innovation and large-scale adoption of green technology are crucial for reducing global greenhouse gas emissions and protecting environmental assets. Without these efforts, sustaining growth and addressing climate change will be costly. The study [45] found that technological advancements in BRICS countries can reduce CO₂ emissions and stimulate economic growth, based on their analysis from 1985 to 2014. Bergougui [46] found that technological innovation, fossil fuels, and renewable energy contribute to CO₂ emissions, with positive shocks decreasing emissions and negative shocks increasing them. Dunyo et al. [47] found that existing technologies seem to lower CO₂ emissions and that this effect is stronger during uncertain times. R&D operations, on the other hand, cause a rise in CO₂ emissions when there is uncertainty. Obobisa et al. [48] investigated the relationship between CO₂ emissions, institutional quality, and green technological innovation and found that green technology innovation reduces CO₂ emissions. Borgi et al. [49] investigated the impact of eco-innovation and renewable energy on CO₂ in G7 countries, finding that these strategies reduce CO₂ emissions when governance variables are present. Lin and Ma [50], found that Green technology innovations have varying impacts on CO₂ emissions in different cities and periods. They indirectly reduce emissions through industrial structure upgrades. The effect is more significant when a city's human capital level is high, with higher human capital cities showing better carbon emission reduction effects. [2], indicate that green technology innovation alone may not significantly reduce CO₂ emissions without supportive measures like renewable energy consumption and political stability. Thus, while green innovation is vital for sustainable development and combating climate change, a multifaceted approach is necessary to optimize its impact.

2.3 Green energy consumption and CO₂ emissions

Green energy consumption is crucial for reducing CO₂ emissions in countries with high pollution levels. These countries can significantly mitigate their environmental impact by transitioning from fossil fuels to renewable sources like solar, wind, biomass, and hydroelectric. Green energy reduces CO₂ emissions, is abundant and available, enhancing energy security, and reduces dependence on imported fossil fuels, making it a more sustainable alternative. Governments significantly accelerate green energy adoption through policies and incentives like tax subsidies, and regulatory frameworks. These measures attract private investments and stimulate green technology innovation by creating a favorable environment for renewable energy development. Green energy consumption diversifies energy, improves air quality, creates jobs, and positions countries strategically in the global economy, aligning with international sustainability goals and enhancing environmental stewardship reputation. Promoting green energy consumption is crucial for reducing CO₂ emissions, achieving climate targets, fostering economic growth, energy independence, and environmental resilience in polluting countries.

Carrión-Flores [26] study uses the bootstrapped autoregressive distributed lag (BARDL) model to analyze the effects of renewable energy and technological advancements on Malaysia's ecological footprint and carbon dioxide emissions, finding that renewable energy reduces environmental degradation. Behra et al. [2] found that renewable energy consumption, political stability, and fiscal decentralization can mitigate CO₂ emissions in seven OECD economies from 2000 to 2019, with green technology innovation having an insignificant effect. Amer et al. [51] examined the impact of renewable energy consumption on CO₂ emissions in GCC countries and found a negative relationship between renewable energy consumption and CO₂ emission. Zaman et al. [52] examined the nexus between education expenditure, female employment, renewable energy consumption, and CO₂ emissions in China and found a negative relationship between these variables and CO₂ emissions. Li et al. [53] explore the relationship between renewable energy, fossil fuel consumption, economic growth, urbanization, and CO₂ emissions in China. The study found an inverted U-shaped relationship between CO₂ emissions and GDP per capita, with fossil fuel use increasing emissions. Renewable energy consumption can reduce emissions by 0.259% in the long run, but rapid development increases emissions by 0.285 to 0.288% in the short term.

2.4 Governance, green innovation, and CO₂ emissions

Governance indicators are crucial for managing CO₂ emissions, involving transparent policies, accountable practices, and regulations. It influences industries and communities to reduce their carbon footprint, combat climate change effectively, and contribute to sustainable development, creating a positive externality for society. Zheng and Jin [54] emphasize the importance of governance in effectively implementing policies and regulations aimed at sustainable energy projects. Governance fosters economic growth by promoting transparency, accountability, and resource management. It minimizes corruption, ensures the rule of law, and promotes fair competition. This boosts investor confidence, attracts foreign investment, and facilitates economic diversification. This stimulates job creation, productivity, and overall living standards. Political stability can significantly reduce CO₂ emissions by enabling governments to implement long-term environmental policies, establish regulatory frameworks, and encourage industries to invest in cleaner technologies and renewable energy. This stability fosters collaboration with international partners, promoting innovation and a sustainable future. Green innovation is promoted through effective governance and supportive policies, including subsidies for R&D, tax incentives, carbon pricing, and environmental standards regulations. Green innovation drives economic development and quality of life, making it crucial for sustainable development and mitigating CO₂ emissions in polluting countries.

Sarpong [55] empirically investigated the nexus between CO₂ emissions and governance within both oil-producing and non-oil-producing countries in the Sub-Saharan Africa (SSA) region. The study found that governance exhibits a negative association with CO₂ emissions. Baloch and Wang [56] investigated the importance of governance in reducing CO₂ emissions in BRICS and found that Governance enhances environmental quality and reduces CO₂ emissions. Gani [57] explores the nexus between governance and carbon dioxide emissions from developing economies and finds that political stability, the rule of law, and control of corruption reduce CO₂ emissions per capita. Shabani et al. [58] examined the nexus between renewable energy, governance, and CO₂ emissions from 179 countries and found that governance lowers CO₂ emissions. According to [59], there is a positive relationship between governance and CO₂ emissions in high-income countries, while in middle-income and low-income countries, improved governance is associated with lower CO₂ emissions.

2.5 FDI and CO₂ emissions

The two primary theories that dominate the academic discussion regarding the relationship between energy consumption and foreign direct investment (FDI) are the pollution halo and pollution haven effects. According to the 'pollution halo' theory, when economies develop, FDI introduces cutting-edge technologies and creative management strategies that subsequently result in a significant decrease in carbon emissions [60, 61]. FDI significantly reduces CO₂ emissions by introducing advanced technologies and practices, and companies investing in countries with strict environmental regulations adopting cleaner production methods. FDI can significantly contribute to developing renewable energy infrastructure, displacing fossil fuels and reducing CO₂ emissions in host countries through investments in wind, solar, and hydroelectric power projects. FDI stimulates economic growth, leading to higher environmental standards and better regulation enforcement, often resulting in a shift towards cleaner technologies and practices. Governments play a crucial role in ensuring that FDI contributes positively to environmental sustainability through effective policies, incentives, and regulations.

According to the 'pollution haven' hypothesis, FDI inflows may encourage the adoption of outdated and ecologically unfavorable technologies in host nations, hence increasing carbon emissions and degrading the environment [62, 63]. Developing countries often lower environmental standards to boost economic growth, which can lead to greater environmental degradation. FDI can increase CO₂ emissions due to various mechanisms. First, FDI may not adhere to strict environmental regulations in host countries, leading to higher emissions due to outdated technologies and inadequate waste management. In order to attract FDI, host countries relax their environmental regulations. As a result, environmental quality in the host countries is deteriorating. Second, FDI can increase consumption and production activities, leading to higher energy demands and emissions. High carbon intensity industries, like heavy manufacturing or extractive industries, can further exacerbate CO₂ emissions. To mitigate these impacts, host countries need to implement rigorous environmental standards, incentivize green technologies, and align FDI with sustainable development goals.

2.6 Urbanization and CO₂ emissions

Urbanization significantly impacts CO₂ emissions due to increased energy consumption for transportation, heating, cooling, and industrial processes, primarily reliant on fossil fuels. People who live in urban areas are expected to consume more energy than those who live in rural areas because they must travel longer distances for employment and must build, run, and maintain urban infrastructure and services, such as housing, water supplies, roads, and bridges [64, 65]. Mitigating urban CO₂ emissions requires sustainable urban planning strategies, prioritizing public transport, energy-efficient buildings, green spaces, and renewable energy integration. Addressing these factors can help cities reduce CO₂ emissions and promote global environmental sustainability.

According to the International Energy Agency, urban areas contribute 70% of global greenhouse gases, and their environmental impact is expected to rise if not addressed properly. Urbanization may contribute to emissions by producing high energy consumption due to economic activities that sustain it. Urbanization, particularly non-renewable energy consumption, leads to a decrease in environmental quality through increased CO₂ emissions. [66] assess the impact of urbanization and the use of renewable energy on Europe's CO₂ emissions from 1995 to 2018. Results show that while increased urbanization has a negative impact on air quality, increased usage of renewable energy reduces CO₂ emissions. Grodzicki and Jankiewicz [67] identified the negative impact of urbanization as varying across different levels of urban development. However, some studies considered the relationship between urbanization and CO₂ emissions as being not very significant or inconclusive [68,69,70,71].

3 Data and methodology

Using a panel data model, PCA, and decision tree model, we investigated whether green innovation lowers CO₂ emissions in the top 12 polluting nations. The countries included in this study are China, the United States, India, Russia, Japan, South Korea, Canada, Mexico, Turkey, Italy, Poland, and the United Kingdom.

3.1 Panel data analysis

The panel data model is employed for several purposes. (i) to address country-specific heterogeneity, (ii) to increase the number of observations to obtain reliable results, (iii) less collinearity between the variables, and (iv) greater flexibility and effectiveness. Panel data enhance the efficiency of econometric estimations by providing greater sample variability and more degrees of freedom compared to cross-sectional data [72].

The panel data model appears as follows:

$$lnY_{it} = \alpha + X_{it} \beta^{\prime} + \mu_{i} + \delta_{t} + \varepsilon_{it}$$

(1)

$\alpha$ represent a constant term. $ln{Y}_{it}$ is the logarithm of CO₂ emissions of the country, i (where i, 1–12, countries are considered) at time period t ($ln{Y}_{it}$), ${X}_{it}$ is a set of the logarithm of independent variables that influence CO₂ emission except for the variables that are derived from Principal Component Analysis. ${\beta }{\prime}$ represents the slope coefficient vector associated with the independent variables, µ_i represents an unobserved country-specific effect, and δ_t is the time trend. The error term ε_it is uniformly distributed across nations and time periods.

$$ln{CO<Subscript>2</Subscript>}_{it}=\alpha +{\beta }_{1}{lnGI}_{it}+{\beta }_{2}{lnGE}_{it}+{\beta }_{3}{lnEG}_{it}+{\beta }_{4}{lnIND}_{it}+{\beta }_{5}{lnURB}_{it}+{\beta }_{6}{FDI}_{it}+{\beta }_{7}{PCA}_{it}+{\mu }_{i}+{\delta }_{t}+{\varepsilon }_{it}$$

(2)

All the variables are natural logarithmic forms except PCA, presented in Eq. 2. We created a principal component analysis using variables such as Voice and accountability, Political stability and absence of violence/Terrorism, Government effectiveness, Regulatory quality, Rule of Law, and Control of corruption. All the variables are represented in natural logarithm form except governance indicators variables. Natural logarithm is a statistical technique used to transform highly skewed values normally distributed and to address heteroscedasticity.

$ln{CO<Subscript>2</Subscript>}_{it}$ is the natural logarithm of CO₂ emissions, presented in Eq. (2). $lnGI$ is the natural logarithm of green innovation. $lnGE$ is the natural logarithm of green energy consumption. $lnEG\text{ is the}$ natural logarithm of economic growth. $lnIND$ is the natural logarithm of industrialization, $lnURB$ is the natural logarithm of urbanization. $lnFDI$ is the natural logarithm of foreign direct investment. $PCA$ is the principal component analysis comprising six variables. Here, the outcome variable is the CO₂ emissions, and the fixed and random effects are denoted by εi and εt, respectively. All of the β parameters are estimable using our panel frameworks.

The parameters of β, ranging from β1 to β7, correspond to several factors attributed to the CO₂ emissions over time t throughout the nation i. The model's error component, represented by εit, is thought to represent random noise with a normal distribution.

3.2 Principal component analysis

In our paper, we have created a variable through PCA that represents governance indicators. Principal Component Analysis (PCA) is a method used to reduce dimensionality in large correlated datasets by transforming them into uncorrelated variables. Which can be expressed in the below form.

$$\begin{gathered} Z_{1} = a_{1}.\,X = a_{11} X_{1} + a_{12} X_{2} + \ldots + a_{1n} X_{n} \hfill \\ Z_{12} = a_{2}.\, X = a_{21} X_{1} + a_{22} X_{2} + \ldots + a_{2n} X_{n} \hfill \\ Z_{n} = a_{n}.\,X = a_{n1} X_{1} + a_{n2} X_{2} + \ldots + a_{nn} X_{n} \hfill \\ \end{gathered}$$

(3)

where X_i is the original variable, Z_i transform variables are the principal components, and a_i is the coefficient vector, respectively. The coefficients are calculated by maximalizing the Var(Z_i) subject to the constraint that a^T_ia_i = 1 and cov(Z_iZ_j) = a^T_iƩa_i = 0, j = 1,2,3…., i-1, where Ʃ is the covariance matrix of X which is symmetric nonnegative definite matrix and it eigen values are λ_1, λ₂….. λ_n and it is corresponding vectors are e₁, e₂……. e_n. Through the method of Lagrange multiplier shown below.

$${\text{L}} = {\text{a}}_{{\text{i}}}^{{\text{T}}} \sum {\text{a}}_{{\text{i}}} - \, \lambda \left( {{\text{a}}_{{\text{i}}}^{{\text{T}}} {\text{a}}_{{\text{i}}} - {1}} \right)$$

(4)

We will be able to find the coefficient vector a_i, and calculate the principal components from that.

3.3 Decision tree regression

Most of the literature employs panel data models to investigate the impact of independent variables on CO₂ emissions. In this paper, we have used panel data models and the implementation of the Decision tree model in the Panel data structure. The decision tree regression algorithm is a tree-based machine learning model that is mostly preferred for solving classification problems but can also be used as a regression problem. It consists of three nodes. The Root node consists of the entire sample. The interior nodes represent the features/independent variables of the data set based on which the data set is splitting into subsets. Different measures are used to split tree-based models, such as Information Gain for classification tasks and Reduction in Variance for regression tasks, depending on the task at hand. Finally, the leaf nodes represent the output. The Decision tree model splits a dataset into trees to identify influential variables for predicting the dependent variable, eliminating the impact of outliers by splitting the data set, and does not consider the distance measure, which is used in other statistical methods.

We employ the panel unit root test developed by [73], which offers several advantages: (i) the IPS test accommodates unbalanced panels, and (ii) it relies on cross-sectional independence. The IPS test sets forth the null hypothesis of non-stationarity in contrast to the alternative hypothesis that the series is stationary. Next, we analyze our empirical equation by examining regression models with fixed effects followed by models with random effects. We have employed the fixed effects model, which examines whether intercepts vary over time or within a group. The random effect model investigates differences in error variance components between individuals or over time. The Hausman specification test is used to determine whether the fixed effects or random effects model is more appropriate for empirical estimation. The null hypothesis was accepted, indicating that random effect is preferred over a fixed effect model.

We proceed with a new application using a decision tree model in a panel data structure. The fixed and random effect models utilize the same transformed data set, and they are referred to as the decision tree fixed and random effect model. We are attempting to determine which model is better by using both models and the mean square error (MSE) in place of the Hausman test.

4 Variables used, country selection, and data source

CO₂ emission is considered as a dependent variable. Green innovation, Green energy consumption, Economic growth, Industrialization, Urbanization, Foreign direct investment, and governance indicators such as Voice and accountability, Political stability, absence of violence/terrorism, Government effectiveness, Regulatory quality, the rule of law, Control of corruption are treated as independent variables.

Statista's 2024 report reveals that the carbon dioxide emissions of the world's most polluting countries between 2010 and 2022 (measured in million metric tons) are China, United States, India, Russia, Japan, Indonesia, Iran, Germany, Saudi Arabia, South Korea, Canada, Mexico, Brazil, Turkey, South Africa, Australia, Vietnam, Italy, Poland and the United Kingdom. The countries were chosen based on the consistent data availability on all the variables used in our empirical analysis. Our empirical analysis highlights the twelve most polluting countries in the world from 1996 to 2022. Notably, we could not include these eight polluting countries due to data inconsistency: Indonesia, Iran, Germany, Saudi Arabia, Brazil, South Africa, Australia, and Vietnam.

According to research by the International Energy Agency (IEA), using renewable energy might result in a 60% reduction in CO₂ emissions [1]. Renewable energy technology innovation minimizes carbon intensity and diversifies non-fossil energy sources. Previous research indicates that industrialization promotes environmental degradation [40, 74]. The transition from agriculture to industry promotes economic progress but at the expense of the environment. According to [75], industries that contribute significantly to GDP consume over half of the world’s energy.

Urbanization refers to the people living in urban areas. Economic growth is represented as real GDP per capita. Industrialization refers to Industry (including construction) corresponding to ISIC divisions 05–43, encompassing manufacturing (divisions 10–33), mining, construction, electricity, water, and gas. It measures value added as the net output of these sectors, calculated by subtracting intermediate inputs without deductions for asset depreciation or resource depletion, based on ISIC revision 4. Table 1 provides a comprehensive overview of variables, detailing their definitions, measurement criteria, and expected signs.

Table 1 Overview of the variables used, their measurements, and the expected sign

5 Empirical results

The paper's empirical findings are discussed in this section. The variables utilized in this study, except the governance indicator, are expressed in the natural logarithmic form before moving further with empirical estimation. We took into account seven independent variables in our investigation. The six independent variables are Green innovation, green energy consumption, economic growth, industrialization, urbanization, and foreign direct investment. Principal component analysis (PCA) determines the seventh independent variable. Variables like voice and accountability, political stability and the absence of violence or terrorism, government effectiveness, regulatory quality, rule of law, and corruption control are independent variables. We first examined the correlation between these factors before moving on to a more in-depth investigation. These variables have a strong correlation, as we found. It is important to note that we used PCA, as indicated in Table 2, to prevent the multicollinearity issue.

Table 2 Correlation matrix

Using PCA has the advantages of (i) decreasing the dimensions from six to one variable and (ii) handling multicollinearity issues in the data. We can obtain a maximum of six PCAs as we have six variables. PCA 1 accounts for over 90% of the data's variability. Thus, PCA 1 has been taken into account in our analysis. Tables 3 and 4 display the Principal Component Analysis results and highlight the contributions of variables to PCA 1.

Table 3 Principal component analysis

Table 4 Contribution of variables on principal component 1

After determining PCA 1, we have identified our seventh independent variable. Next, we will conduct the unit root test to proceed with our research.

In our analysis, we utilize the [76] panel unit root test, which is considered superior to other tests in the literature. This test assesses the null hypothesis of non-stationarity against the alternative hypothesis of stationarity for the series under examination. Unit root test results in level and first differences are shown in Table 5. All variables are non-stationary at level I(0); upon differencing once, they become stationary (1).

Table 5 Unit root test

The Hausman test, random effect, and fixed effect are shown in Table 6. The majority of variables show statistical significance and show expected signs. The fixed effect model revealed that green innovation, green energy consumption, FDI, and governance have an expected negative sign. It is essential to highlight that although green innovation and FDI carry negative signs, they are not statistically significant. The coefficient of governance and green energy consumption indicates that the former has a more significant effect on lowering CO₂ emissions and improving environmental quality. A one percent increase in governance and green energy consumption will reduce CO₂ emissions by 0.20% and 0.18%, respectively. Hausman test says that the random effect is better than the fixed effect model. Examining the random effect model, we see that the results are comparable to those of the fixed effect model. Green innovation, green energy consumption, foreign direct investment, and governance all showed an expected negative sign in our random effect model. We found that the potential for reducing CO₂ emissions via governance is greater than that of green energy usage. One per cent increase in governance and green energy consumption will reduce CO₂ emissions by 0.19% and 0.18%, respectively. Green innovation and FDI have insignificant impacts on CO₂ emissions. We found that urbanization, industrialization, and economic expansion positively affect CO₂ emissions. To reduce CO₂ emissions, policymakers must increase the use of green energy and encourage strong governance standards.

Table 6 Fixed effect, random effect and Hausman test

In Table 7, we have done a cross-sectional dependency test using the Breusch-Pagan LM and Pesaran CD tests. Both tests reject the null hypothesis of cross-sectional dependency. The result indicates that there is a cross-sectional dependency among the countries. It may be said that an increase in pollution in one country will spill over into the remaining countries and create a negative externality.

Table 7 Cross-section dependence test results

We have added a new dimension to the panel data analysis by using a machine learning tree-based model on the panel data sets, following the fixed effect and random effect analyses. We have employed the same data format in fixed and random effect models. The transfermed data is deducted from the country-specific mean for each variable to obtain the fixed effect [75]. On the other hand, the data set is transferred for the random effect using the overall mean and the theta value, which in our case is 0.97. Following the data set transfer, we determined the feature importance for each data set, which is shown in Table 8. The amount of each independent variable helps to lower the uncertainty in the CO₂ emission and determines the feature relevance. This is typically determined by the reduction in Gini impurity or entropy achieved by splitting on specific independent variables. The effectiveness of a decision tree model lies in its ability to inherently rank and select variables based on their importance in predicting the dependent variable. Unlike linear models, where coefficients directly indicate the magnitude of influence, decision trees use feature importance metrics to highlight key predictors, offering a clear understanding of each variable's impact on the outcome. In other words, how much will the mean square error decline if we drop one independent variable? Or how much will the mean square error decline in the testing data sets if you drop one variable? The decision tree fixed effect model reveals economic growth, governance, and industrialization as the most crucial variables. The decision tree random effect model revealed that governance ranks third and industrialization ranks second. Both models show that economic growth ranks first in terms of its impact on CO₂. Both cases consider green innovation and foreign direct investment to be the least significant features. The results align with the panel data set, except for economic growth, which ranks third based on coefficient values. The results are consistent for panel data analysis with and without a decision tree compared to their least performances. The study reveals that green innovation and foreign direct investment (FDI) are insignificant and have low coefficient values.

Table 8 Independent variable importance to CO₂ emissions based on decision tree fixed effect and random effect model

6 Conclusion and policy implications

This study addresses a notable gap in the current literature by examining the influence of green innovation on CO₂ emissions. The study examines the impact of green innovation on CO₂ emissions in the top twelve polluting nations from 1996 to 2020. The countries included in the analysis are China, the US, India, Russia, Japan, South Korea, Canada, Mexico, Turkey, Italy, Poland, and the UK. Using panel data and decision tree analysis, we found that CO₂ emissions are increased by industrialization, urbanization, and economic growth, while they are decreased by green energy consumption and governance.

In the panel data tree-based model, governance ranks second and third in the Decision Tree fixed and random effects models. The study suggests that policymakers should encourage investment in green energy production and improve governance to combat environmental degradation effectively. The long-term energy efficiency should be ensured through increased investment in green energy production. Therefore, it is recommended that countries looking to lower their CO₂ emissions focus on green energy consumption and governance. Promoting governance can strengthen countries, making it challenging to generate economic outputs using unclean production factors. The nation's degree of governance is also crucial for long-term planning and investments in green energy. Therefore, in order to lower CO₂ emissions, it is suggested that countries address green energy and governance simultaneously. Our study highlights the necessity of a worldwide strategy for enhancing governance. The empirical analysis indicates that green innovation is not statistically significant in reducing emissions, indicating that it may need to be more effective. The influence of governance reform in one nation, how it spreads to other regions, and its consequences on CO₂ emissions should have been included in our analysis, which prompted the need for more research.