The nexus between road transport intensity and road-related CO₂ emissions in G20 countries: an advanced panel estimation

Abstract

This study determines the dynamic linkages between road transport intensity, road transport passenger and road transport freight, and road carbon emissions in G20 countries in the presence of economic growth, urbanization, crude oil price, and trade openness for the period of 1990 to 2016, under the multivariate framework. This study employs the residual-based Kao and Westerlund cointegration technique to find long-run cointegration, and continuously updated bias-corrected (CUP-BC) and continuously updated fully modified (CUP-FM) methods to check the long-run elasticities between the variables. The long-run estimators’ findings suggest a positive and significant impact of road transport intensity, road passenger transport, road freight transport on road transport CO₂ emissions. Economic growth and urbanization are significant contributing factors in road transport CO₂ emissions, while trade openness and crude oil price significantly reduce road transport CO₂ emissions. The Dumitrescu and Hurlin causality test results disclose unidirectional causality from road transport intensity and road transport freight to the road transport CO₂ emissions. However, the causality between road passenger transport and road transport CO₂ emissions is bidirectional. Finally, comprehensive policy options like subsidizing environmental-friendly technologies, developing green transport infrastructure, and enacting decarbonizing regulations are suggested to address the G20 countries’ environmental challenges.

Introduction

Carbon dioxide is thought to be the most prevalent greenhouse gas (GHG), which is more responsible for global warming (Xu and Lin 2017; Paramati et al. 2017; Peng et al. 2018). Carbon is estimated to account for over 74% of all GHGs (IEA 2020a). A substantial portion of energy-related CO₂ emissions comes from the transport sector, contributing about 24% (8 billion tonnes) of global anthropogenic GHG emissions (IEA 2019a). On a global average, road transport contributes to about 75% (three-quarters) of overall transportation emissions in 2018. Within road transport, light-duty vehicles (passenger vehicles) contribute 45.1%, and the other 29.4% of transport CO₂ emissions come from heavy-duty vehicles (trucks carrying freight) (IEA 2019b; CIAT 2020). A major contributing factor to CO₂ emissions in all regions is transportation infrastructure and fossil fuel consumption.

The transport industry uses a significant amount of energy, which fuels economic development and stimulates urbanization, resulting in more private vehicles (Lu et al. 2007; Achour and Belloumi 2016a). This trend suggests that almost half of all potential travel growth will be caused by increased passenger vehicles and trucks (Dulac 2013; IEA 2020b). As a result, the transport sector is a substantial and growing source for CO₂ emissions, economic growth, and country development (Andrés and Padilla 2018; Sajid et al. 2019; Dong et al. 2020). While fossil fuels constitute 81% of global energy consumption in 2018 (IEA 2020c), the mounting CO₂ emissions generated by road freight transport are not expected to slow or cut down. No single solution will be adequate to challenge mitigating these emissions (OECD 2019). The number of passengers (roughly 2.4 billion Millennial passengers) projected to travel globally by 2050 is predicted to be approximately two times greater than it was in 2010. Although transport sector emissions have risen dramatically over the last 25 years, contributing to rising regional emissions, the transport sector’s share of the total has remained persistent, with approximately 10% of emissions across countries (Aggarwal and Jain 2016; Amin et al. 2020).

Transport infrastructure is a significant economic segment that fosters economic and social growth and enables effectual resource allocation and materials and mobility (Kuştepeli et al. 2012; Maparu and Mazumder 2017). Transportation facilities can, directly (efficiency and productivity channels) and indirectly (effects on trade, urbanization, fuel-energy consumption, and CO₂ emissions), significantly contribute to economic development (Beyzatlar et al. 2014). The demand for transport increased with high economic growth, swift pace in urban growth, growing disposable income, assortment in leisure pursuits, the imbalanced distribution of energy and material resources, and rapid growth in private cars’ numbers. However, the environmental effects of transport activities raise significant challenges. The transport sector contributes substantially and steadily to GHG emissions, and its share in environmental degradation is rising in all world regions. The transport sector’s share of CO₂ emissions and its continuing growth have drawn policymakers’ attention to economic growth, transport activities, and environmental stress.

The EKC hypothesis is an effective conceptual framework for addressing environmental problems associated with greenhouse gas emissions because it emphasizes the structural changes in the transportation sector, energy efficiencies, and scale effects in the economy (Al-Mulali and Ozturk 2016; Sarkodie and Strezov 2019a, 2019b). In the revised environmental Kuznets curve theory, the concept of scale-effect suggested by Grossman and Krüger could initially stimulate the demand for transportation and travel activities, leading to a remarkable increase in energy consumption for transportation and other sectors (Grossman and Krueger 1991; Sarkodie 2018). As a result, increased road transportation could exacerbate air pollution through the economy’s scale effect. Therefore, increasing environmental impacts should be considered in conjunction with increased road transport activities, operations, and energy usage. Energy consumption may increase carbon dioxide emissions as the transportation system enormously contributes to the energy-emission nexus (Franco and Mandla 2014; Erdogan et al. 2020). This means that the transport sector contributes more to greenhouse gas emissions than other sectors of the economy because emissions from the transport sector directly impact the environment (IEA 2009; Alshehry and Belloumi 2017; Andrés and Padilla 2018; Solaymani 2019). Therefore, more stringent environmental regulations and policies are needed to decouple the strong relationship between transportation intensity and carbon dioxide emissions (Ben Abdallah et al. 2013; Ouyang et al. 2019).

Given the above discussion, a clear and strong linkage exists between transport infrastructure and activity, economic development, and transport-related CO₂ emissions. The first empirical studies branch is primarily concerned with economic growth and energy consumption relationship. After the seminal work of Kraft and Kraft (1978), with the different econometric techniques, energy consumption, and economic growth causal relationship has been investigated in prior studies (Akarca and Long 1980; Yu and Hwang 1984; Belloumi 2009; Apergis and Payne 2010; Zhang and Ren 2011; Dagher and Yacoubian 2012; Kandemir Kocaaslan 2013; Mutascu 2016). The second group of literature looks at the relationship between economic growth and CO₂ emissions. This strand underlines the linkage between economic growth and environmental impacts, which has been explored in past empirical work (Shafik 1994; Cole et al. 1997; Cole 2004; Fodha and Zaghdoud 2010; Lotfalipour et al. 2010; Andreoni and Galmarini 2012; Kofi Adom et al. 2012; Abid 2015; Al-Mulali et al. 2016; Alam et al. 2016; Anastacio 2017; Awad and Abugamos 2017; Ahmad et al. 2017). The third group is primarily concerned with the connection between transportation infrastructure and activity and economic development. The basic conception is that enhancements in transport infrastructure and increased transport activity trigger economic growth, but economic growth can also enhance transport activity. Numerous studies have investigated and proven this relationship, including (Canning and Bennathan 2000; Kuştepeli et al. 2012; Pradhan and Bagchi 2013; Pradhan et al. 2013; Beyzatlar et al. 2014; Achour and Belloumi 2016b; Maparu and Mazumder 2017; Saidi et al. 2018).

The fourth strand appeared in recent literature, combining the previous three classes to examine the multifaceted link between the transport sector, environmental pollution and degradation, transport-energy use, and economic growth. Several empirical reviews exist in the literature. The most relevant studies that considered transport-energy consumption and gasoline demand are (Bentzen 1994) for Denmark; (Eltony and Al-Mutairi 1995) for Kuwait;(Ramanathan and Parikh 1999) for India. Also, Liddle (2009), Ramanathan (2001), Samimi (1995), and Xu and Lin (2015) identified the cointegrating relationship between transport demand and performance and macro-economic variables for the USA, India, Australia, and China, respectively. Equally, Azlina et al. (2014), Botzoris et al. (2015), Liddle and Lung (2013), and Saboori et al. (2014) scrutinized the dynamic long-run causal relationship between transport-related energy usage, economic growth, and environmental degradation. Consequently, there is a strong relationship between transport sector-related activities and transport CO₂ emissions with other macro-economic factors. However, recent studies neglect specific prestigious panels of countries such as G20 countries and road transport-related variables that affect pollution. Therefore, this study’s strong emphasis is to consider the impact of road transport intensity, road passenger transport, and road freight transport with additional macro-economic variables on road transport CO₂ emissions in the G20 countries.

For several reasons, the selection of G20 (group of twenty) countries in this study is justified. First, the G20 countries have a significant impact on global economic development, growth, and global emissions. Precisely, the G20 countries seized approximately 85% of the global GDP and also responsible for approximately 80% of GHG emissions, with 70% of the climate impacts. Because of compelling economic growth and higher energy demand and consumption, transport-related CO₂ emissions of G20 countries increased by 1.2% in 2018. Figure 1 shows the yearly uptrend for road transport CO₂ emissions for G20 countries in 1990–2016. Second, most of the G20 countries’ energy supply is from coal and oil has increased, and 82% of the energy mix is still based on fossil fuels. Furthermore, most G20 countries have similar transport-energy consumption trends, share of transport CO₂ emissions, and economic growth. Despite the critical role of the G20 countries in the world economy, the factors contributing to transport CO₂ emissions are worth investigating.

Fig. 1

Trend for road transport CO₂ emissions for G20 countries 1990–2016

This current research extends in several respects beyond the established literature. First, there is no established literature that focuses on the relationship between road transport intensity, road transport CO₂ emissions, and other macro-economic factors, to the best of our knowledge. Environmentally sustainable, effective, and economically productive, and efficient policies addressing the transportation sector are needed due to their potential impact on the environment. Second, this study uses a large sample of 19 G-20 countries and utilizes a long period (1990–2016), containing recent data. Third, our research model also develops an overall measure of road transport intensity comprised of road freight transport and road passenger transport because ecological issues are anthropogenic and a key role of the transport sector in the world’s economy. Fourth, as per methodological perspective, CADF and CIPS second-generation unit root tests, Kao and Westerlund panel cointegration tests, advanced panel long-run cointegrating regression Continuously Updated Fully Modified (CUP-FM) and Continuously Updated Bias-Corrected (CUP-BC) estimators, and panel granger Dumitrescu-Hurlin (D-H) causality test are employed to exhibits more reliable, accurate and robust results considering the problem of cross-sectional dependence, residual autocorrelation, heteroscedasticity, endogeneity, and slope heterogeneity.

The rest of the paper is organized in the following way: the “Data, model construction, and methodology” section provides data source information, model construction, and econometric methodology; the “Empirical results and discussion” section presents the empirical findings with discussion, and it concludes the study with some policy implications in the “Conclusion and policy implications” section.

Data, model construction, and methodology

Data

This research is intended to build a linkage between road transport intensity, road passenger transport, road freight transport, and CO₂ emissions from road transport, considering economic growth, urbanization, crude oil price, and trade openness as additional determinants of road transport carbon emissions. The road transport intensity is classified as road passenger transport and road freight transport or a combination of both measures using the concepts of the net and gross mass movement (Peake 1994; Scholl et al. 1996; Michaelis and Davidson 1996; SACTRA 1999; Arvin et al. 2015). In this study, we also disaggregate the combined measures of road transport intensity into road passenger transport and road freight transport for investigating the impact of both modules simultaneously because the environmental effect of road passenger and freight transport might be different.

The unique data on road transport CO₂ emissions (kt) is subscribed and compiled from the International Energy Agency online data services (IEA 2021). We have used the OECD (OECD 2020) database to gather data on road passenger transport (RDPT) and road freight transport (RDFT). The data for urbanization (%), economic growth (GDP), and trade openness (%) are amassed from the World Bank Indicators platform (WDI 2021), while the crude oil price data is taken from DataStream. West Texas Intermediate (WTI) is an index to measure crude oil price in US dollar per barrel (Sadorsky 2014; Khalfaoui et al. 2015; Basher and Sadorsky 2016; Sarwar et al. 2019; Nguyen et al. 2020; Habib et al. 2020), which aptly reflects the global oil demand and supply (Kao and Wan 2012; Cross and Nguyen 2017); many prior studies have used it as a significant determinant of carbon emissions (Zeng et al. 2017; Zou 2018; Mensah et al. 2019; Malik et al. 2020). The crude oil prices may have a diverse effect on the energy demand curve of each country (Kilian 2009; Mensah et al. 2019; Ahmed et al. 2020a). Therefore, it may influence the environment quality differently based on the structure of the economy and oil demand. For example, according to Boufateh (2019), the positive change in crude oil price harms environmental quality. In other words, positive shocks in crude oil prices cause an increase in the use of polluting energy.

This study uses annual data that covered the period of 1990–2016 for G20 (Group of Twenty) countries. The list of G20¹ countries covers Argentina, Australia, Brazil, Canada, People’s Republic of China, France, Germany, India, Indonesia, Italy, Japan, Republic of Korea, Mexico, Russia, Saudi Arabia, South Africa, Turkey, the UK, and the US. Yearly trends for road passenger transport and road freight transport are presented in Appendix. A comprehensive description, sources, and measurement of variables are tabulated in Table 1.

Table 1 Description of variables

Principal component analysis

Two variables, road passenger transport (RDPT) and road freight transport (RDFT), were utilized together by employing principal component analysis (PCA) to construct an air transport intensity index. Being a distinct form of factor analysis, to form an index, the PCA reduces variable’s variance (dimensionality) by melding the variables into a smaller and more compact linear combination based on their inherent variance (Gries et al. 2009; Menyah et al. 2014; Jollife and Cadima 2016; Latif et al. 2018). However, this study aims to formulate an index of road transport intensity (RDTI) for an in-depth and extensive analysis. This analysis uses the road transport intensity index, a weighted index of all road transport intensity indicators. This index also offers a single weighted relative measure, integrating most of the information on specific intensity parameters to comprehend the proper connection between road transport intensity and CO₂ emissions from road transport. It also contributes to our study because it is the first time for the RDTI index to be measured by PCA. The PCA process involves matrix construction for the dataset, standardized variables formation, computation of the correlation matrix, sorting of the eigenvalues and corresponding eigenvectors, and the panel component’s selection and reorienting (Jolliffe 2011; Hassan et al. 2011; Mirshojaeian Hosseini and Kaneko 2011). Table 2 demonstrates the PCA analysis for the RDTI index.

Table 2 PCA results for weighted RDTI index

Table 2 showed that the first factor’s highest eigenvalue is 1.851, while the second factor has the lowest eigen value, i.e., 0.149. Ensuing, the variance range of the first factor (0.925) and second factor (0.075) are given. The table also includes the eigenvectors that display the two main component factors’ loadings. RDTI index was developed with PC 1 as it has no negative value and contains most of the variable information relative to another component.

Economic modeling

Consistent with the prior studies (Stead 2001; Åhman 2004; Alises et al. 2014; Arvin et al. 2015; Adams et al. 2020), the current study adopted an empirical model; in our scenario, road transport CO₂ emissions is a dependent variable dictated by other independent variables like road transport intensity, road passenger transport, road freight transport, GDP per capita, urbanization, crude oil price, and trade openness are expressed as:

$$ {RDCO}_2=\mathrm{f}\left( RDTI, RDPT, RDFT, GDP, URB, COP, TROP\right) $$

(1)

In Eq. (1), RDCO₂ refers to road transport CO₂ emissions, RDTI denotes road transport intensity, RDPT indicates road passenger transport, RDFT is road freight transport, GDP is economic growth, URB shows urbanization, COP is the crude oil price, while TROP is trade openness. In this study, the log-linear enhanced function is used to transform the data into natural logarithmic form to remove data dispersion, reduce nonnormality and generate more reliable and consistent results than a standard linear augmented function (Vogelvang 2004; Charfeddine and Ben Khediri 2016; Kahia et al. 2017; Charfeddine and Kahia 2019). The specifications of the log-linear function for our empirical model can be seen in Eq (2).

$$ {\displaystyle \begin{array}{l}\ln {RDCO}_{2 it}={\varphi}_o+{\xi}_1\left(\ln {RDTI}_{it}\right)+{\xi}_2\left(\ln {RDPT}_{it}\right)+{\xi}_3\left(\ln {RDFT}_{it}\right)+{\xi}_4\left(\ln {GDP}_{it}\right)\\ {}+{\xi}_5\left(\ln {URB}_{it}\right)+{\xi}_6\left(\ln {COP}_{it}\right)+{\xi}_7\left(\ln {TROP}_{it}\right)+{\omega}_{it}\end{array}} $$

(2)

where i denotes the number of the countries (i = 1,2,3….. 19), t indicates the time dimension (from 1990 to 2016), φ_οis an intercept, and ω_itis the stochastic term, The coefficients of road transport intensity, road passenger transport, road freight transport, economic growth, urbanization, crude oil price, and trade openness are signified by ξ₁, ξ₂, ξ₃, ξ₄, ξ₅, ξ₆, and ξ₇respectively. Road Transport intensity is the degree to which road transport facilities are used, which can also be expressed as road freight transport and road passenger transport or a combination of both measures (Stead 2001). These intensity measures and economic growth have been originated from contributing to environmental degradation (Arvin et al. 2015; Wang et al. 2020a). By taking into account the impact of road transport intensity on road transport CO₂ emissions, Model 1 can be derived as follows:

$$ {\displaystyle \begin{array}{l}\ln {RDCO}_{2 it}={\varphi}_o+{\xi}_1\left(\ln {RDTI}_{it}\right)+{\xi}_2\left(\ln {GDP}_{it}\right),+{\xi}_3\left(\ln {URB}_{it}\right)+{\xi}_4\left(\ln {COP}_{it}\right)\\ {}+{\xi}_5\left(\ln {TROP}_{it}\right)+{\omega}_{it}\end{array}} $$

(3)

We replaced road transport intensity with road passenger transport (million passenger-kilometers) in model 2.

$$ {\displaystyle \begin{array}{l}\ln {RDCO}_{2 it}={\varphi}_o+{\xi}_1\left(\ln {RDPT}_{it}\right)+{\xi}_2\left(\ln {GDP}_{it}\right),+{\xi}_3\left(\ln {URB}_{it}\right)+{\xi}_4\left(\ln {COP}_{it}\right)\\ {}+{\xi}_5\left(\ln {TROP}_{it}\right)+{\omega}_{it}\end{array}} $$

(4)

We replaced road passenger transport component with road freight transport (million tonne-kilometers-goods) in model 3.

$$ {\displaystyle \begin{array}{l}\ln {RDCO}_{2 it}={\varphi}_o+{\xi}_1\left(\ln {RDFT}_{it}\right)+{\xi}_2\left(\ln {GDP}_{it}\right),+{\xi}_3\left(\ln {URB}_{it}\right)+{\xi}_4\left(\ln {COP}_{it}\right)\\ {}+{\xi}_5\left(\ln {TROP}_{it}\right)+{\omega}_{it}\end{array}} $$

(5)

The linkage between road transport CO₂ emissions and urbanization can explain the environmental impact of the urban population. Previous research identified urbanization as a significant environmental degradation determinant with positive and negative outcomes (Poumanyvong and Kaneko 2010; Ozturk et al. 2016; Wang et al. 2016; Charfeddine et al. 2018). The rigorous empirical works on the environmental effect of international trade are at best mixed. The following studies endorsing the pro-environmental repercussions of trade openness (Birdsall and Wheeler 1993; Frankel and Rose 2005; Ozturk and Acaravci 2013; Al-Mulali and Ozturk 2015), while (Dauda et al. 2021; Pata and Caglar 2021) exemplify a negative elasticity of CO₂ emissions in terms of trade openness. We have also included crude oil price as an explanatory variable because variations in fuel price control CO₂ emissions, boost energy efficiency, and promote vehicle fuel economy (He et al. 2005; Maghelal 2011; Shahbaz et al. 2015; Talbi 2017).

Econometric methodology

This study analyzes road transport intensity, road passenger transport, and road freight transport on road transport CO₂ emissions in G20 countries. This study follows econometric panel techniques suited to large T and N panels. We carry out CD tests, perform panel unit root and panel cointegration tests, and then move into long-run panel estimations and perform causality tests. Figure 2 shows the flowchart of econometric analysis used in this study.

Fig. 2

Flowchart of econometric analysis

Testing cross-sectional dependence

Our research instigates by examining the dependence in the empirical model between cross-sectional (units) countries. In the case where the cross-sectional units are dependent on one another, the cross-sectional dependence (CSD) problem arises (Nathaniel et al. 2021; Liu et al. 2021). Due to the high degree of globalization, international trade, economic and financial integration, and financial crisis spillover, one country is more sensitive to the economic shocks that can be widely shared with other countries (Munir et al. 2020). This interaction of nations has the potential to create an inappropriate dependency in panel data between cross-section countries. The presumption of cross-sectional independence is one of the drawbacks of traditional econometrics and analytical approaches (Andrews 2005). If cross-sectional dependence in a panel is ignored, the results obtained from such methods can be biased and misleading (Aydin 2019). Overlooking CSD precedes spurious and skewed elasticity estimations (Behera and Mishra 2020).

To that extent, Breusch and Pagan (1980) posited the Lagrange Multiplier (LM) simple test is used to investigate and counter cross-sectional dependency. By using the following panel data model, the LM statistic can be determined:

$$ {CSD}_{lm}={T}_{ij}{\sum}_{i=1}^{N-1}{\sum}_{j=i+1}^N{{\hat{\rho}}^2}_{ij}. $$

(6)

where T is the time dimension, N designates the number of cross-sectional countries (units), and ${\hat{\rho}}_{ij}$ signifies an estimation of the pair-wise correlation between residuals derived from estimates of Ordinary Least Squares (OLS) for each sequence.

The LM test is only valuable and effective for such cases where the T is amply large and the N is comparatively short (Chou 2013). Pesaran (2020) has suggested the following CD test based on the LM statistic as a solution to this problem:

$$ CSD=\sqrt{\frac{2T}{N\left(N-1\right)}}\left({\sum}_{i=1}^{N-1}{\sum}_{j=i+1}^N{T}_{ij}{{\hat{\rho}}^2}_{ij}\right)\Rightarrow N\left(0,1\right). $$

(7)

Under both tests’ null hypothesis, cross-section units’ independence is presumed and spread as a standard two-tailed normal distribution. In contrast, the alternative hypothesis postulates the dependency between countries (cross-section units).

Slope of homogeneity

Based on Hashem Pesaran and Yamagata (2008) slope homogeneity tests, we examined the slope coefficients’ homogeneity after checking the cross-sectional dependence. Earlier empirical approaches with an assumption of slope homogeneity ignored the country-specific characteristics (Breitung 2005; Bedir and Yilmaz 2016). Assuming homogeneity, in the case of large (N) and small (T), could yield misleading results. The problem of heterogeneity is critical to address because, due to differences in demographic, social, and economic structures of G20 countries, there is a possibility of heterogeneity in slope parameters, which could affect the consistency and accuracy of panel estimators. For this purpose, this study applied the robust slope homogeneity method proposed by Blomquist and Westerlund (2013). Blomquist and Westerlund (2013) suggested, based on Swamy’s model (Swamy 1970) and a robust version of the Hashem Pesaran and Yamagata (2008) test (denoted Δ), a generalized test to deal with both heteroskedasticity and serially correlated errors. This approach is very effective against more generalized cross-correlation constructs and considers the trivial size distortions in all assessments. To extend Δ, the data-generating process is provided as:

$$ {y}_{i,t}={\alpha}_i+{\xi}_i{\chi}_{i,t}+{\omega}_{i,t}, $$

(8)

where i = 1…N χ_{i, t}is a vector of regressors, with ξ_i the slope coefficients for the associated vector. The HAC version of Δ proposed by Blomquist and Westerlund (2013) is derived as :

$$ {\tilde{\Delta}}_{HAC}=\sqrt{N}\left(\frac{N^{-1}{St}_{HAC}-{k}_2}{\sqrt{2{k}_2}}\right) $$

(9)

$$ {St}_{HAC}=\sum \limits_{i=1}^N{T}_i{\left({\overset{\frown }{\varGamma}}_{2i}-{\overset{\frown }{\varGamma}}_{2 HAC}\right)}^{\prime}\left({\overset{\frown }{X}}_{i,{T}_i}{\overset{\frown }{V}}_{i,{T}_i}^{-1}{\overset{\frown }{X}}_{i,{T}_i}\right)\left({\overset{\frown }{\varGamma}}_{2i}-{\overset{\frown }{\varGamma}}_{2 HAC}\right) $$

(10)

where ${\overset{\frown }{\varGamma}}_{2 HAC}$ is a robust HAC estimator, and ${\overset{\frown }{X}}_{i,{T}_i}$is a trajectory matrix that partially eliminates the heterogeneous variables. ${\overset{\frown }{V}}_{i,T}$ implies variance estimator with kernel function k and bandwidth parameter B_{i, T}.

Panel unit root test

After checking the CSD and slope homogeneity, the next step in the analysis is to test the order of the cointegration of the various variables considered in this study. If the CSD is existent across cross-sections, then the first-generation panel unit root tests may offer misleading and worthless results (Dogan and Seker 2016a). Indeed, to address this issue, Khan et al. (2020) and Rauf et al. (2018) suggested non-parametric and parametric tests to avoid bias in findings. Except for Dickey and Fuller (1979) ADF test, Im et al. (2003) IPS test, Levin et al. (2002) LLC test, and Phillips and Perron (1988) PP test, the cross-sectional dependency and slope heterogeneity problems can also be countered by the both CADF (Cross-sectional augmented Dickey-Fuller) and CIPS (Cross-sectional Im, Pesaran and Shin) tests. The findings from both methods are also more robust and accurate. Pesaran (2007) has suggested both of these tests. CADF test statistic is stated as follows:

$$ \Delta {Y}_{i,t}={\Gamma}_i+{\Gamma}_i{Z}_{i,t-1}+{\Gamma}_i{\overline{Y}}_{t-1}+{\Gamma}_i\Delta \overline{Y_t}+{\varsigma}_{it} $$

(11)

where Δ displays change operator, Y denotes studied variable, and ς_itis a residual term. Based on the Eq.(11), CIPS equation is specified as:

$$ \Delta {Y}_{i,t}={\Gamma}_i+{\Gamma}_i{Z}_{i,t-1}+{\Gamma}_i{\overline{Y}}_{t-1}+{\sum}_{j=0}^p{\Gamma}_{ij}\Delta \overline{Y_{t-j}}+{\sum}_{j=1}^p{\Gamma}_{ij}\Delta {Y}_{i,t-j} $$

(12)

where $\overline{Y}$ is the average for cross-sectional units and is illustrated as:

$$ {\displaystyle \begin{array}{l}{Y}^{i,t}={\Gamma}^1{\overline{\log RDTI}}^{i,t}+{\Gamma}^2{\overline{\log RDPT}}^{i,t}+{\Gamma}^3{\overline{\log RDFT}}^{i,t}+{\Gamma}^4{\overline{\log GDP}}^{i,t}+{\Gamma}^5{\overline{\log URB}}^{i,t}\\ {}+{\Gamma}^6{\overline{\log TROP}}^{i,t}+{\Gamma}^7{\overline{\log COP}}^{i,t}\end{array}} $$

(13)

CIPS test statistics are labeled as:

$$ \overset{\frown }{CIPS}={N}^{-1}\sum \limits_{i=1}^N{CADF}_i $$

(14)

Panel cointegration tests

The current study is aimed at building a linkage between CO₂ emissions from road transport and road transport intensity, road passenger transport, and road freight transport, considering urbanization, economic growth, trade openness, and crude oil price as additional determinants for the G20 countries. It deploys two cointegration tests, i.e., Kao (1999) and CSD-robust Westerlund (2007) panel cointegration tests. To check the long-run non-heterogeneous connection, the residual-based Kao test is used. This test also uses the specific DF and ADF statistics to verify the null hypothesis of having no cointegration over the alternative hypothesis, namely, cointegration.

For estimation of residuals, we used the following regression:

$$ {u}_{i,t}=\varsigma {u}_{i,t-1}+\sum \limits_{j=1}^n{\Phi}_j\Delta {u}_{i,t-j}+{\omega}_{it} $$

(15)

For instance, Basile et al. (2011) determined ADF statistics as:

$$ {\displaystyle \begin{array}{l}\\ {} ADF=\frac{t_{ADF}+\frac{\sqrt{6N}{\overset{\frown }{\sigma}}_v}{\left(2{\overset{\frown }{\sigma}}_v\right)}}{\sqrt{\frac{{\overset{\frown }{\sigma}}_{0v}^2}{\left(2{\overset{\frown }{\sigma}}_v^2\right)}+\frac{3{\overset{\frown }{\sigma}}_v^2}{\left(10{\overset{\frown }{\sigma}}_{0v}^2\right)}}}\end{array}} $$

(16)

where ${\sigma}_v^2={\sum}_{ae}-{\sum}_{ae}{\sum}_e^{-1},{\sigma}_{0v}^2={\mathrm{\Re}}_a-{\mathrm{\Re}}_{ae}{\mathrm{\Re}}_e^{-1},\mathrm{\Re}$exhibits the long-run covariance matrix. Next, we also utilize Westerlund (2007) cointegration technique. It provides robust and accurate results and helps to handle cross-sectional error term dependence (Kapetanios et al. 2011). Besides, the test has no limitation for the common factor (Khan et al. 2020). The null hypothesis, in this case, implies that cointegration between cross-section units does not exist. Besides that, the alternative hypothesis indicates the presence of cointegration between considered variables.

Hence, the baseline equation for Westerlund (2007) test can be simplified as:

$$ {\gamma}_i(L)\Delta {y}_{it}={\xi}_{1i}+{\xi}_{2i}t+{\gamma}_i\Big({y}_{it-1}-{\Psi}_i^{\prime }{x}_{it-1}+{\lambda}_i{(L)}^{\prime }{\nu}_{it}+{\varepsilon}_{it} $$

(17)

where γ_i holds the cointegration vector between studied variables x and y. γ_i signifies the coefficient for the rectification of errors. The four group and panel test statistics can be specified as:

$$ {G}_{\tau }={N}^{-1}\sum \limits_{i=1}^N\frac{\theta_i}{SE\left({\hat{\theta}}_i\right)} $$

(18)

$$ {G}_{\alpha }={N}^{-1}\sum \limits_{i=1}^N\frac{T{\theta}_i}{\theta {\prime}_i(1)} $$

(19)

$$ {P}_{\tau }=\frac{{\hat{\theta}}_i}{SE\left({\hat{\theta}}_i\right)} $$

(20)

$$ {P}_{\alpha }=T{\alpha}^{\prime } $$

(21)

G_τ and G_α characterize the group means statistics, while P_τand P_α denote the panel statistics. The pre-requisite for conducting the regression analysis is fulfilled by statistical evidence of cointegrating links between the variables.

Long run estimations

Prior studies have used various first-generation econometric methods to estimate long-run effects but neglect the issue of cross-sectional dependency (Ulucak and Bilgili 2018; Zafar et al. 2019; Ahmed et al. 2020b). In order to overcome this problem, we calculate the long-run parameters with the CUP-BC and CUP-FM estimators, developed by Bai et al. (2009) and Bai and Kao (2006), following the recent studies (Fang and Chen 2017; Koçak et al. 2020; Wang et al. 2020b; Ahmed et al. 2020b; Ahmed and Le 2021). Both methods have certain benefits: i) the issue of cross-sectional dependency and unobserved non-linearity is being considered; (ii) these approaches are preferable over other estimators because they are capable of generating accurate and robust results even within sight of residual autocorrelation, endogeneity, and heteroscedasticity (Bai et al. 2009; Camarero et al. 2014; Ahmed et al. 2021); iii) consistent results are also achieved even with factors and regressors having a mixed integration order, i.e., (I(1) and I (0)) (Bai et al. 2009; Tamarit et al. 2011). Owing to these advantages, these estimators for our sample have the appropriate size and power estimates. The CUP-FM is the most suitable tool for this analysis because it is ideal for a small data sample. Additionally, these estimators are extensively used to estimate the long-run parameters (Fang and Chang 2016; Fang and Chen 2017; Ulucak and Bilgili 2018; Koçak et al. 2020).

For CUP-FM and CUP-BC estimators, the following equation is employed:

$$ \left({\hat{B}}_{CUP},{\hat{F}}_{CUP}\right)=\mathrm{argmin}\frac{1}{nT^2}\sum \limits_{i=1}^n{\left({w}_i-{y}_i\beta \right)}^{\prime }{M}_F\left({w}_i-{y}_i\beta \right). $$

(22)

where M_F = X_T − T⁻²VV^′, the identity matrix for dimension T is X_T. V assumes a vector of common latent factors.

Moreover, heterogeneous FMOLS, DOLS, and DSUR were used to validate CUP-BC and CUP-FM estimators’ findings.

Heterogeneous panel causality test

We analyze the directional flow and causal relationships between interest variables by utilizing the heterogeneous Dumitrescu and Hurlin (2012) (D-H) panel granger causality test to provide additional details to the policymakers. This approach addresses the question of heterogeneity and CSD and also has no constraint T >N (Dogan and Seker 2016b; Koçak and Şarkgüneşi 2017). In this case, the null hypothesis of the D-H causality is presumed to reflect that no causal direction was found between variables contrary to the alternative hypothesis, which directs the causal associations among considered variables. The model can be formulated as:

$$ {y}_{i,t}={\varphi}_i+\sum \limits_{k=1}^p{\upxi}_i^k{y}_{i,t-k}+\sum \limits_{k=1}^p{\chi}_i^j{T}_{i,t-k}+{\omega}_{i,t} $$

(23)

k signifies the lag length, whereas ξⁱ(j)represents the autoregressive parameters.

The Wald statistics of all panel are computed to test the null hypothesis by averaging the values for each cross-section of the individual Wald statistics:

$$ {Wbar}_{N,T}^{Hnc}=\frac{1}{N}\sum \limits_{i=1}^N{W}_{i,T} $$

(24)

Empirical results and discussion

This study examines the nexus between road transport intensity, road passenger transport, road freight transport, and CO₂ emissions from road transport for a panel of G20 countries. In Table 3, descriptive statistics disclose the average road transport CO₂ emissions of 178095.418 kt of CO₂ with the maximum value of 1544553 kt of CO₂.

Table 3 Summary of descriptive statistics (before natural logarithms)

The average urban population level is 70.838%, with the maximum value approaching 91%. The average GDP (million 2010 US dollars) is 22121.566 US$, which directs the high economic growth in G20 countries. The Pearson correlation results in Table 4 validate that CO₂ emissions from road transport are positively correlated with road transport intensity. As with this point, all the other independent variables RDFT, RDPT, URB, GDP, and COP, have a strong positive association with CO₂ emissions except trade openness, which negatively correlates. Road passenger transport and road freight transport are positively and substantially linked to each other. The correlation of RDTI, RDFT, and RDPT with urbanization and TROP are negative, while positive relationship with GDP and crude oil price is observed.

Table 4 Correlation analysis

The estimation of cross-sectional dependency (CSD) has become the main focal point in recent literature. The failure to contain CSD may lead to biased results and efficacy loss. Thus, the Breusch-Pagan LM and Pesaran Scaled LM tests were used to assess the existence of CSD.

Table 5 documents the results of cross-sectional dependency. Based on respective p-values of LM and CD statistics, we reject the null hypothesis of no cross-sectional independence at the 1% statistical significance level for road transport CO₂ emissions, road transport intensity, road passenger transport, road freight transport, urbanization, economic growth, crude oil price, and trade openness. These findings indicate the prevalence of an unobserved common factor as shocks in road transport intensity occurring in one country may cause variations in other countries’ related factors. In other words, all variables in this analysis have a cross-sectional dependency.

Table 5 Cross-sectional dependence

Concerning the CSD, the issue of slope heterogeneity is disclosed in Table 6 by both tests of the slope’s homogeneity. Both tests ($\tilde{\Delta}$ and ${\tilde{\Delta}}_{adj}$) have p-values less than 1%, so the alternative hypothesis of slope heterogeneity is accepted for G-20 countries and affirming the existence of heterogeneity. Table 6 shows that the problem of slope heterogeneity persists in the panel based on the significance values of the given tests.

Table 6 Slope homogeneity analysis

Table 7 tabulates the outcomes of CIPS and CADF panel unit root tests. The empirical findings of both CIPS and CADF tests propose that all variables have unit root and showing stationary behavior when their first differences are made. Put differently, these results suggested that all variables are integrated at 1(1), and also, there must be an indication of a long-run cointegration nexus between the analyzed variables.

Table 7 Unit root tests

Next, we deploy the panel cointegration tests of Kao (1999) and Westerlund (2007). The results of Kao (1999) are shown in Table 8, which endorses the cointegration relationship among selected variables. The ADF statistic displays a cointegration relationship and rejects the null hypothesis at a 1% or 5% significance level. Furthermore, the Kao test traces the long-term equilibrium relationship between road transport CO₂ emissions, road transport intensity, road passenger transport, road freight transport, economic growth, urbanization, crude oil price, and trade openness.

Table 8 Kao panel cointegration tests

Likewise, the Westerlund (2007) test approximations in Table 9 validate that road transport CO₂ emissions, road transport intensity, road passenger transport, road freight transport, urbanization, economic growth, crude oil price, and trade openness have a long-run equilibrium relationship. Using group (G_τ) and panel (P_τ) statistics, the null hypothesis, depending on corresponding p-values, has been refuted at 5% and 10% significance levels for all models. The results of both tests imply the cointegration existence among analyzed variables in the selected panel.

Table 9 Westerlund (2007) cointegration tests

After establishing cointegrating nexus among considered variables, we determine the long-run effects of variables. For this purpose, we employ the CUP-BC and CUP-FM approaches. Figure 3 shows the graphical summary of long-run relationships. Table 10 offers key estimates of this analysis.

Fig. 3

Long-run relationships

Table 10 Long-run estimation of CUP-FM and CUP-BC estimator

Our study’s primary purpose is to explore the effect of road transport intensity on road transport emissions. Empirical evidence shows that road transport intensity positively and significantly affects road transport CO₂ emissions at a 1% critical level in model 1. Assuming other factors unchanged, a 1% increase in road transport intensity will increase road transport CO₂ emissions by 0.015% (according to CUP-FM) and 0.008 % (according to CUP-BC). This result implies that road transport intensity increases the road transport CO₂ emissions in G20 countries and suggests that more economic activity leads to higher energy demand and deter the environment. Immense energy demand inevitably leads to fossil fuels such as diesel and gasoline consumption, thereby emitting large amounts of CO₂ emissions. This finding is aligned with the prior studies (Wang et al. 2018; Song et al. 2019).

In model 2, the coefficient of road passenger transport has a positive and significant effect on road transport CO₂ emissions under CUP-FM and CUP-BC. The results indicate that a 1% increase in road passenger transport raises road transport CO₂ emissions by 0.021 and 0.017% based on CUP-BC and CUP-FM long-run estimation methods. The estimated coefficient for road passenger transport is positive and statistically significant. These findings confirm the argument that road passenger transport escalates transportation demands, fuel consumption, and associated environmental degradation. The main reason for the increase in road transport CO₂ emissions is passenger transport, roughly half of all transport emissions. Passenger transport produces more emissions than other transportation modes because of the comparatively high demand for road transportation. In effect, road passenger transport burns more gasoline, resulting in higher emissions. This result is synchronized with the previous findings of (Ribeiro and Balassiano 1997; Li and Yu 2019) that road passenger transport increases road transport CO₂ emissions.

Regarding road freight transport and road transport CO₂ emissions, we again verify the positive relationship in model 3. It shows that a 1% increase in RDFT raises RDCO₂ by 0.010% (CUP-FM) and 0.012% (CUP-BC). These outcomes are plausible because road freight transport establishes a significant share in road transport CO₂ emissions. Additionally, the prior literature documents that road freight transport degrades the environment. Besides, future environmental transport issues cannot be ignored because the rise in vehicle numbers upsurges road expansion, so the rising demand for energy and the growing usage of low-quality fuels has increased environmental pollution. The major sources of energy use for most public transports are electricity and coal.

In contrast, fossil fuels like oil and natural gas are used as the key source of energy for private vehicles (Saboori et al. 2014). The findings are congruent with (McKinnon and Piecyk 2009; Xu and Lin 2018; Anwar et al. 2020) and differ with findings of (Danish et al. 2020). All the three measures of road transport intensity, based on passenger and freight transport, in our empirical results reveal the pollution-enhancing role of road transport activities. These findings imply that the higher level of road transport intensity stimulates environmental degradation via related CO₂ emissions. Road transportation operates through the scale effect because G20 countries have tremendous growth patterns and account for approximately 85% of global GDP and around 75% of international trade (OECD 2021). Therefore, these countries are faced with the challenges of balancing economic growth and environmental quality to achieve sustainable development. Though G20 countries have comparatively more advanced and efficient transportation system than other countries, yet the road transport activities are intensifying the road emissions in these countries. Figure 1 also supports these arguments, which exhibits an overall rising trend of road-related carbon emissions in G20 panel over the time. The scale effect of road transportation mainly stimulates the energy demand especially for fossil fuel resources in these countries. Moreover, the recent figures show that transport-related CO₂ emissions surge by approximately 1.2% because international trade and economic growth have significantly enhanced freight and passenger transport (Climate Transparency 2021). These rising trends of road-related environmental degradation require the urgent attention to further improve the energy efficiencies in the road transport sector and implement the renwable energy options.

Urbanization also has a positive effect on road transport CO₂ emissions in G20 countries. A 1% rise in urbanization may increase RDCO₂ by 0.0323% (CUP-FM), and 0.027% (CUP-BC) in model 1, 0.0549% (CUP-FM) and 0.0555% (CUP-BC) in model 2, and 0.015% (CUP-FM) and 0.015% (CUP-BC) in model 3. From the results, urbanization significantly influences road transport CO₂ emissions, the urban population in G20 countries deteriorates environmental quality by rising energy demand and expansion of the road transport infrastructure. Urbanization is the population migration from rural to urban areas, which has shown remarkable growth for the private transport industry as urban private transport networks expand in urban areas for long-distance travel. As cities expand, public infrastructure projects, including highways and roads, are being pursued. The empirical outcomes are aligned with the Poumanyvong and Kaneko (2010); Martínez-Zarzoso and Maruotti (2011); Wang et al. (2017); Song et al. (2018); Fan et al. (2020); and Hashmi et al. (2021) and contradict the result of Adams et al. (2020). These findings reveal that the urban population in G20 countries has caused a substantial increase in road-related carbon emissions by stimulating travel and energy demand. Therefore, the national and municipal authorities should mitigate the steady decline in environmental quality due to the tremendous growth of urbanization in these economies by designing and launching environmental-friendly and energy-efficient road transport vehicles and green transportation systems.

The economic development coefficient (GDP) has a positive and significant effect (under both CUP-FM and CUP-BC) at a 1% level in all three models. When economic growth increases by 1%, RDCO₂ also accelerates by 0.041% and 0.036% in the case of CUP-FM and CUP-BC, respectively. This acceleration may be due to the scale effect, which is dominant in G-20 countries as compared to technique and composition effects. As income level rises, people purchase and consume more goods, move to cities for better health care facilities and educational services, enhance transport demand, and expand the road network. This empirical outcome is aligned with the findings of Danish et al. (2020); and Godil et al. (2021). As stated above, G20 countries hold more than 80% of global GDP, which has intensified the environmental degradation by aggregating the demand for both passengers and freight transport to meet the pace of economic development. Therefore, economic growth has a scale effect on road transport activities, which are mainly dependent on fossil fuel and other energy resources for transport vehicles leading to more air pollution in G20 countries.

Considering crude oil price and road transport CO₂ emissions, we again found evidence of a strong negative linkage between them. It shows that a 1% increase in COP reduces RDCO₂ by 0.025% and 0.020% under CUP-FM and CUP-BC, respectively. As crude oil prices rise, the road transport CO₂ emissions decrease because the people reduce their cars’ use. This result is consistent with the previous research findings (Maghelal 2011; Rasool et al. 2019; Abumunshar et al. 2020). Thus, the overall findings in all three models unveil that oil price improves the environmental quality in G20 countries by reducing the demand for oil resources. As road vehicles mainly consume fossil fuel energy resources, the rising oil prices significantly reduce the demand for non-renewable energy resources, causing the decline in road-related emissions in the long run.

Finally, the relationship between TROP and RDCO₂ is negative, implying that TROP increases environmental quality. A 1% increase in TROP lessens RDCO₂ by 0.008 (CUP-FM) and 0.015 (CUP-BC) in G-20 countries. This suggests that with increasing economic level, the impact of trade openness on carbon emissions also changes and notably enhances the environment of affluent and rich G20 countries. The recent trend of augmented trade would stimulate the transfer of high emission-intensive industrial units from developed to developing countries, causing developed countries to accomplish emission reduction at the expense of developing countries. This is consistent with the acknowledged carbon transfer phenomenon in the international trade process (Essandoh et al. 2020). This result confirms the outcomes of Acheampong (2018), which indicate that trade openness in sub-Saharan African countries cuts carbon emissions and opposes the findings of Ahmed et al. (2017); and Chen et al. (2021). The empirical findings confirm the positive role of trade openness in effect with CO₂ emissions in G20 countries because the higher level of exports and imports creates healthy competition among the countries to improve the environmental quality. Moreover, these countries are highly developed and have introduced comparatively stringent environmental regulations about foreign trade, state of the art transportation system, and environmental-friendly trade policies.

In addition, all three models are estimated using heterogeneous and robust FMOLS, DOLS and DSUR, and the results are presented in Table 11. The results of FMOLS, DOLS, and DSUR are aligned with the CUP-FM and CUP-BC estimates. These robust results show that this study’s findings are accurate and can be extended for policy consequences.

Table 11 Results of HFMOLS, DOLS, and DSUR

We utilize the method of Dumitrescu and Hurlin (2012) (DH) to examine and find the causal paths of relationships in the short run between variables after assessing the long-term coefficients and the existence of CSD. Table 12 shows the findings of the heterogeneous DH panel causality test. The most exciting outcome from this causality analysis documents the bidirectional causality and feedback relation between road transport passenger and road transport CO₂ emissions. That means the road transport passenger changes lead to variations in road transport CO₂ emissions, and vice versa. However, our results designate a unidirectional relationship from road transport freight to road transport CO₂ emissions, signifying that road transport freight Granger-causes road transport CO₂ emissions, but not the other way around. Referring to the above finding, unidirectional causality extending from road transport intensity to road transport CO₂ emissions has been found. Besides, our results also confirm feedback (bidirectional) causal link between the GDP and road transport CO₂ emissions, between urban population and road transport CO₂ emissions. A strong interdependence between economic growth and road transport CO₂ emissions is demonstrated. This outcome is consistent with the findings of Aslan et al. (2021) for Mediterranean countries but is not aligned with (Zhang et al. 2014; Arvin et al. 2015).

Table 12 Dumitrescu-Hurlin panel causality tests

Similarly, the bidirectional relationship between urbanization and road transport CO₂ emissions implies that urbanization leads to road transport CO₂ emissions, while high road transportation emissions motivate policymakers and legislators to formulate and implement current urban policies obstructing rural to urban migration and relocation. The finding of this feedback relationship is also supported by (Al-Mulali et al. 2013). Moreover, empirical findings also posit that there is a two-way causality between crude oil price and road transport CO₂ emissions. Finally, both trade openness and road transport CO₂ emissions affect each other in the Granger causality sense and found a bidirectional link.

Conclusion and policy implications

This study scrutinized the influence of road transport intensity, road passenger transport, and road freight transport on road transport CO₂ emissions in G20 countries in a multivariate framework for the period of 1990–2016 while incorporating economic growth, urbanization, trade openness, and crude oil price. Numerous econometric approaches were employed to achieve the objectives of this analysis. For instance, CSD, robust unit root tests such as CADF and CIPS are used to verify unit root properties, and the residual-based Kao’s (1999) and Westerlund’s (2007) panel cointegration tests are used to check the existence of cointegration among the studied variables. The long-run CUP-FM and CUP-BC estimators are used to measure long-term elasticities. Finally, panel causality is checked through Dumitrescu-Hurlin (DH) heterogeneous panel causality test. These latest econometrics techniques help to address the problems of slope homogeneity and cross-sectional dependency in panel data.

The results of the cointegration tests suggest the long-term equilibrium between variables. The long-run estimation unfolds a positive association between road transport intensity, road passenger transport, road freight transport, and road transport CO₂ emissions. Simply speaking, road transport intensity, road passenger transport, road freight transport contribute to enhancing road transport CO₂ emissions. Economic growth and urbanization are significant factors in promoting road transport CO₂ emissions, while trade openness and crude oil price significantly reduce road transport CO₂ emissions. The causality test estimates indicate that a unidirectional relationship extends from road freight transport and road transport intensity to road transport CO₂ emissions, signifying that road transport freight and road transport intensity Granger-cause road transport CO₂ emissions. We also found that bidirectional causality and feedback relation exists between road passenger transport and road transport CO₂ emissions.

Based on the empirical results, several relevant policy implications for G20 countries are proposed. Road transport intensity degrades the environment as a whole, increases transport-related CO₂ emissions, and enhances economic activity. However, G20 countries should take two major steps to remove barriers to the transport sector’s decarbonization. Incomplete commitments of international agreements (Kyoto Protocol and Paris Agreement) and the high cost for transport-related clean technologies have precluded significant reductions in overall GHG emissions and the transport sector in particular. All Parties of The Paris Agreement, especially the G20 countries as the largest emitters of carbon emissions, uphold their pledges towards emission reduction targets and intensify these efforts in the future. To deliver nationally determined contributions (NDCs), the governments should use taxes and subsidies as a tool to change the clean energy’s relative costs in the transport sector. They should provide subsidies to promote environmental-friendly technologies. They should also provide incentives to foster the research and development (R&D) of public and private clean and renewable technologies, capitalize and develop green infrastructure like urban road transport systems, and enact regulations, which will gradually decarbonize all economic sectors, including the transport sector. Movings towards sustainable transportation require strategies at individual levels like awareness regarding sustainable transport benefits, an adaptation of sustainable lifestyle, and car-sharing.

This research has some limitations because it focuses on G20 countries and country-specific analysis is not included. Based on this limitation, a time-series estimation at the individual and country-level would help to understand the link between road transport intensity and the environment. Also, this relationship with other transport modes like air transport and railways would be more vital for understanding this relationship.

Appendix

ESM 1

(DOCX 20 kb)