Electric vehicles and carbon emission intensity: cross-country evidence using extended STIRPAT and System GMM frameworks

The rapid growth of greenhouse gas emissions and increasing climate-related risks have intensified global efforts to transition toward low-carbon development pathways. The transport sector remains one of the largest contributors to global carbon emissions due to its heavy dependence on fossil fuels, making transport decarbonisation an increasingly important policy priority International Energy Agency (IEA, 2024). Carbon emission intensity (CEI), which measures carbon emissions per unit of economic output, has become an important indicator for evaluating sustainable development performance and decarbonisation efficiency across countries. Unlike aggregate emissions, CEI reflects the carbon efficiency of economic activity and provides a more suitable measure for assessing whether sustainable economic growth is becoming environmentally sustainable. Nevertheless, substantial cross-country differences in CEI persist due to variations in economic structure, energy systems, industrial composition, urbanisation and technological readiness. Figure A1 ^[1] illustrates the variation in CEI across selected economies, highlighting the uneven progress of global decarbonisation efforts and the continuing challenge of balancing economic development with environmental sustainability.

In response to growing climate concerns and transport-related emissions, electric vehicles (EVs) have emerged as a key technological solution to support sustainable mobility and accelerate the energy transition. Governments worldwide have promoted EV adoption through subsidies, tax incentives, investment in charging infrastructure and industrial transition policies aligned with the United Nations' (UN) Sustainable Development Goals (SDGs), particularly SDGs 7, 9, 11, 12 and 13 (UNDP, 2025; IEA, 2024; IPCC, 2023). The rapid expansion of EV adoption has been particularly evident in China, the USA and several European economies, reflecting a growing global commitment to transport electrification. Figure A2 ^[1] demonstrates the substantial growth in global EV stock over the past decade, indicating that EVs are becoming an increasingly important component of a nation's long-term decarbonisation strategies.

Despite rapid EV adoption, the environmental effectiveness of EVs remains widely debated. While EVs eliminate direct tailpipe emissions, their overall carbon mitigation impact depends heavily on the electricity generation structure, energy efficiency, urbanisation patterns and supporting infrastructure conditions (Burchart-Korol et al., 2018; Owais Khan et al., 2024). In economies where electricity generation remains highly dependent on fossil fuels, EV adoption may generate weaker environmental benefits due to indirect emissions from electricity production (Miotti et al., 2016). Similarly, rapid urbanisation may offset part of the environmental gains from EV diffusion by increasing transport demand, congestion and energy consumption (EC) (Wang et al., 2022b). These structural differences suggest that the environmental impact of EV penetration is unlikely to be uniform across countries and development stages.

Existing studies on EVs and environmental sustainability have primarily focused on total carbon emissions, transport-sector emissions, or lifecycle assessment approaches (Hao et al., 2015; Miotti et al., 2016; Burchart-Korol et al., 2018). Although these studies provide valuable insights into the environmental implications of transport electrification, relatively limited cross-country evidence examines whether EV penetration reduces CEI, which captures the carbon efficiency of economic activity rather than aggregate emission levels alone. In addition, many previous studies rely on EV sales as a proxy for penetration, even though short-term sales growth may not adequately reflect long-term fleet transformation and sustained decarbonisation outcomes. The distinction between EV stock and EV sales, therefore, remains underexplored in the empirical literature.

Another limitation is the limited attention given to heterogeneity across development stages and structural conditions. Because EV effectiveness depends on renewable electricity availability, urbanisation, infrastructure readiness and progress in the energy transition, the findings from advanced economies may not fully generalise to middle-income countries (MIC). To address these gaps, this study examines the relationship between EV penetration and carbon-emission intensity using panel data from 32 major EV-adopting economies from 2010 to 2023. An extended Stochastic Impacts by Regression on Population, Affluence and Technology (STIRPAT) framework and system generalised method of moments (System GMM) estimation are employed to address endogeneity, persistence and unobserved heterogeneity. By incorporating renewable electricity share, urbanisation and income-group heterogeneity, the study seeks to provide a more comprehensive understanding of the structural conditions shaping EV-driven decarbonisation outcomes across countries.

This study contributes to the literature in three main ways. First, it directly links EV penetration to carbon-emissions intensity rather than aggregate emissions, providing a clearer assessment of carbon-efficiency outcomes. Second, it compares EV stock and EV sales to distinguish between sustained fleet transformation and short-term market expansion. Third, it incorporates interaction effects involving renewable electricity share and urbanisation to examine whether the environmental effectiveness of EV adoption depends on broader structural and development conditions. The study, therefore, aims to provide policy-relevant insights for countries pursuing transport electrification as part of broader climate transition and sustainable development strategies. Beyond the environmental sustainability literature, this study contributes to the business and economic policy literature by examining how transport electrification interacts with economic development, energy transition and structural conditions across countries. The findings provide insights into the conditions under which green technology adoption can support sustainable growth and carbon-efficient development, particularly in emerging and developing economies undergoing energy and transport transitions.

The growing urgency of climate change and carbon reduction commitments has increased global interest in EVs as a key strategy for promoting sustainable transport and reducing dependence on fossil fuels. The transport sector remains one of the largest contributors to greenhouse gas emissions, making transport electrification an important component of long-term decarbonisation strategies (IEA, 2024; IPCC, 2023). Compared with conventional internal combustion engine vehicles, EVs produce no direct tailpipe emissions and are generally considered more energy efficient, particularly when supported by cleaner electricity supply systems (Hao et al., 2015; Miotti et al., 2016).

Existing studies generally suggest that EV adoption contributes to lower transport-related carbon emissions and improved environmental sustainability, although the magnitude of its environmental benefits varies across countries and energy systems (Burchart-Korol et al., 2018; Owais Khan et al., 2024; Chenayah et al., 2024). Lifecycle assessments show that EV-related emissions depend not only on vehicle operation but also on battery production, electricity generation structure and energy efficiency throughout the supply chain (Hao et al., 2015; Miotti et al., 2016). Consequently, EV decarbonisation benefits tend to be weaker in fossil-fuel-dependent electricity systems and stronger in systems with higher renewable energy penetration (Miotti et al., 2016; Burchart-Korol et al., 2018).

The environmental effectiveness of EV adoption also depends on broader structural and policy conditions, including charging infrastructure availability, technological readiness, renewable electricity generation and urban transport systems (Owais Khan et al., 2024). Rapid urbanisation may offset some of the environmental gains from electrification by increasing transport demand, congestion and EC (Wang et al., 2022b). Accordingly, existing evidence remains mixed, with mitigation outcomes varying across electricity systems, infrastructure conditions and development stages (Miotti et al., 2016; Owais Khan et al., 2024). Much of this literature focuses on total carbon emissions, transport-sector emissions, or lifecycle assessments rather than CEI, which reflects the carbon efficiency of economic activity. Compared with aggregate emissions, CEI provides a more suitable indicator for assessing whether sustainable economic growth is becoming environmentally sustainable.

Another important issue concerns the measurement of EV penetration. Existing studies frequently use EV sales as a proxy for EV diffusion because sales data are more readily available across countries. However, sales primarily capture current market expansion and may not fully reflect accumulated fleet transformation and long-term decarbonisation effects. By contrast, EV stock reflects the total number of EVs operating within the transport system and may better capture sustained environmental outcomes. Nevertheless, relatively few studies directly compare EV stock and EV sales when evaluating environmental performance, leaving uncertainty regarding which indicator better reflects the long-run impact of transport electrification.

The environmental effectiveness of EV diffusion is also likely to differ across countries due to variations in economic development, infrastructure readiness, renewable electricity availability, urbanisation patterns and progress in the energy transition. High-income economies generally have cleaner electricity systems, stronger charging infrastructure and more comprehensive policy support for transport electrification, potentially enabling EV adoption to yield greater mitigation benefits (IEA, 2024). In contrast, middle-income economies often rely more heavily on fossil-fuel-based electricity generation while facing affordability and infrastructure constraints, which may weaken the effectiveness of EVs. Consequently, findings derived primarily from advanced economies may not be fully generalisable to countries at earlier stages of the transport and energy transition.

Although existing studies provide important insights into EV adoption and environmental sustainability, several gaps remain. Relatively few cross-country studies directly examine the relationship between EV penetration and CEI, despite CEI being a more appropriate indicator of carbon efficiency than aggregate emissions. In addition, previous studies have often relied on EV sales as the primary measure of penetration, even though EV stock may better reflect sustained fleet transformation. Limited attention has also been given to heterogeneity across development stages and structural conditions, particularly renewable electricity availability and urbanisation pressure. These limitations highlight the need for broader cross-country evidence on the structural conditions shaping EV-driven decarbonisation outcomes.

This section presents the theoretical framework, empirical model specification and robustness strategy used to examine the effect of EV penetration on CEI.

Figure A3 ^[1] illustrates the analytical framework applied in this study. It extends the Impact–Population–Affluence–Technology (IPAT) identity by incorporating key components of population, affluence and technology to explain changes in CEI. The extended STIRPAT framework incorporates EV penetration, gross domestic product (GDP) per capita, urbanisation, population density (PD), research and development (R&D) investment, primary EC and a nation's share of renewable electricity. Primary EC captures energy demand, while renewable electricity share reflects electricity generation structure, which is important because EV decarbonisation outcomes depend on grid carbon intensity (Franzò and Nasca, 2021; Sacchi et al., 2022; Singh et al., 2024). The framework, therefore, facilitates analysis of how EV adoption and structural conditions jointly influence CEI (Xing et al., 2023).

This study examines the relationship between EV penetration and CEI across countries with different income levels. Consistent with the Environmental Kuznets Curve (EKC), environmental outcomes are closely linked to economic development (Grossman and Krueger, 1995). PD, urbanisation and technological progress may also influence EV adoption patterns and environmental performance.

To examine these mechanisms, this study introduces the IPAT model (Ehrlich and Holdren, 1971) and extends it into an empirically estimable framework. The IPAT theory assumes that environmental impact results from the combined effects of population, affluence and technology. By adjusting these factors, environmental pressure can be mitigated.

Where I represents environmental impact, while P, A and T denote population, affluence and technology.

Because the IPAT environmental formula assumes unit elasticities, this present empirical analysis adopts the STIRPAT framework developed by Dietz and Rosa (1997) and extended by York et al. (2003).

The extended model takes the following form:

Where I, P, A and T follow the definitions in Equation (1), a represents the intercept and the IPAT model becomes a special case when elasticities equal one. Since Equation (2) is nonlinear and difficult to estimate directly, this study applies a logarithmic transformation to derive a computable linear form.

Thus, the refined STIRPAT model is expressed as Equation (3):

In this study, the environmental impact variable is CEI, which is calculated as the proportion of carbon emissions to GDP. The urbanisation rate (UR) and PD are included to represent the population component. GDP per capita (constant US$2015) represents a measure of affluence. Technological progress is captured by research and development investment. This situation reflects the argument that technological advancement occurs through both production innovation and green innovation (Jiang et al., 2023; Long et al., 2018). Green innovation investment supports sustainability and reduces carbon emissions (Fang, 2023), while production-oriented innovation drives the automotive industry's development.

Primary EC captures energy demand, while the renewable electricity share (RE) captures the structure of electricity generation. Together, these variables represent the energy system dimension and help avoid attributing cross-country differences in grid carbon intensity to EV penetration. Individual fixed effects, time fixed effects and error terms are included to control for unobserved heterogeneity.

EV stock (ST) is the primary indicator of EV penetration. To address persistence, endogeneity and unobserved heterogeneity in panel data, this study employs the System Generalised Method of Moments estimator (Blundell et al., 2000; Fukase, 2010; Hsiao, 2022; Teixeira and Queirós, 2016). The dynamic model is specified as follows:

Where re denotes the renewable electricity share, measured as the percentage of total electricity output generated from renewable sources, the other variable definitions remain consistent with Equation (3).

The System GMM estimator uses the second to fourth lags of CEI as GMM-style instruments with the collapse option to limit instrument count, while the remaining explanatory variables enter as Instrumental Variable (IV)-style instruments (Blundell and Bond, 1998; Roodman, 2009). The total number of instruments is 24 in Model 1 and 26 in Model 2, both below the number of groups (32), satisfying the rule of thumb that instruments should not exceed groups. The exclusion restriction assumes that past CEI values influence current CEI only through the modelled socio-economic pathways. The structural persistence of carbon intensity supports this assumption but cannot be directly tested. A remaining concern is that changes in CEI may trigger policy responses affecting EV adoption or R&D investment. Individual fixed effects and lagged instruments reduce simultaneity bias, but time-varying policy feedback cannot be ruled out entirely.

This study conducts robustness checks through two approaches: replacing the key EV variable and testing for regional heterogeneity.

First, EV sales are used as an alternative indicator of EV penetration because they capture contemporaneous market diffusion and policy-driven adoption. Replacing EV stock with EV sales provides a robustness check for the baseline results. The robustness model is specified as follows:

Where sa refers to EV sales, the other variables and parameters remain the same as in Equation (4).

Second, this study tests regional heterogeneity because income levels strongly influence vehicle adoption. Since cars are expensive and not always essential, the demand for automobiles tends to be more elastic than that for basic goods. EV adoption also depends on infrastructure readiness, including charging facilities and supportive road systems. Moreover, EV battery manufacturing relies on rare metals, which increase production costs and limit diffusion in less developed economies.

Because EV adoption remains concentrated in relatively wealthier economies (IEA, 2025; World Bank, 2023), the sample is divided into 21 high-income countries (HIC) and 11 MIC based on World Bank classifications. Israel is classified as an HIC, while Vietnam, Indonesia and Türkiye are classified as MICs.

In addition, an interaction term analysis was conducted to empirically test the mechanisms underlying the heterogeneity results. The interaction between EV stock and the UR (lnst × lnur) tests whether urbanisation-driven carbon pressure moderates the mitigation effect of EV penetration. The interaction between EV stock and the share of renewable electricity (lnst × lnre) tests whether the electricity generation structure conditions the effectiveness of EV adoption. These tests provide direct empirical evidence on the channels through which structural factors influence EV decarbonisation outcomes rather than relying solely on income group comparisons.

Several potentially relevant variables are not included in the model due to data limitations. The availability of charging infrastructure, which may affect the utilisation rate and the practical emissions benefits of EVs, is not directly measured because comparable cross-country indicators do not exist for the full sample period (Fang et al., 2020). Policy instruments such as purchase subsidies, carbon taxes and fuel economy standards also differ across countries and over time. However, they are difficult to operationalise in a panel regression framework due to their heterogeneous design and implementation. The omission of these variables means that the estimated coefficients should be interpreted as conditional associations rather than strictly causal effects. Individual fixed effects absorb time-invariant cross-country differences in institutional and infrastructure environments, partially mitigating omitted-variable concerns, but time-varying policy changes are not directly controlled for.

This paper evaluates the relationship between EV penetration and CEI using panel data from 32 countries covering the period from 2010 to 2023. EV stock and sales data are obtained from the IEA Global EV Data Explorer (IEA, 2025). The sample covers the period from 2010 to 2023, capturing the full trajectory of EV adoption from early adoption to mass-market diffusion. During this period, the global EV stock grew from approximately 17,400 units to over 40 million (IEA, 2025), providing sufficient within-panel variation to support dynamic panel estimation. All socio-economic control variables are consistently available for the 32 sampled countries through 2023 from the World Bank and IEA databases.

The sample comprises 32 countries with consistent EV stock and sales data available from the IEA Global EV Data Explorer. Four additional economies (Vietnam, Indonesia, Israel and Türkiye) are included following expanded IEA coverage. Because the sample focuses on countries with active or emerging EV markets, the findings should be interpreted primarily as evidence from economies with measurable EV adoption.

Socio-economic and environmental variables, including GDP per capita, research and development expenditure, primary EC, renewable electricity share, UR, PD and CO₂ emissions, are sourced from the World Bank database (World Bank, 2023) and Climate Watch (Climate Watch, 2023). Renewable electricity share is measured as the percentage of total electricity output generated from renewable sources obtained from the World Bank (indicator EG.ELC.RNEW.ZS). Missing observations are addressed using interpolation. For the four newly added countries, EV stock and sales data are available from the IEA from 2012 onwards for Israel and Türkiye and from 2014 onwards for Vietnam and Indonesia. Missing observations in the early years of the panel were interpolated, consistent with the treatment applied to the original 28-country sample. To estimate the extended STIRPAT model in a computable form, all variables are transformed into logarithms. Tables A1 ^[1] and A2 ^[1] provide the descriptive statistics and correlation patterns, while Tables A3 ^[1] to A5 ^[1] present the regression results and robustness tests.

Table A1 ^[1] reports the descriptive statistics. The sample exhibits substantial variation in CEI, EV penetration and the renewable electricity share, reflecting differences in development stages, energy structures and industrial profiles across countries (Doluweera et al., 2020; Shen et al., 2019). This variation provides a suitable basis for examining the heterogeneous outcomes of EV decarbonisation.

Table A2 ^[1] presents the correlation test results. CEI is negatively correlated with EV stocks and GDP per capita but positively correlated with primary EC. These correlations are broadly consistent with the arguments in the literature review. For instance, EV adoption is expected to support decarbonisation only when the energy system becomes cleaner and when structural transitions support lower emission intensity (Franzò and Nasca, 2021; Sacchi et al., 2022). However, correlation does not establish causality. EV penetration may be negatively correlated with CEI, partly because countries with higher income levels and better governance also invest more in cleaner energy and technologies. Therefore, dynamic estimation is required to isolate the contribution of EV penetration while controlling for persistence and socio-economic drivers. CEI is also negatively correlated with the share of renewable electricity (r = −0.492), indicating that countries with cleaner electricity systems tend to have lower carbon intensity. To assess potential multicollinearity, variance inflation factors were computed across all models, with a mean variance inflation factor (VIF) of 2.41 and no individual VIF exceeding 10.

Before estimating the System GMM model, a benchmark comparison is conducted. According to Roodman (2009), the dynamic GMM estimator is valid only if the coefficient of the first-order lag of the dependent variable lies between the ordinary least squares (OLS) and fixed-effects estimates. Following this approach, OLS, least squares dummy variable (LSDV) and fixed-effect regressions are first estimated, and the results are presented in Table A3 ^[1].

The lagged CEI coefficient lies within the valid range for both OLS and fixed-effects estimations, indicating that the dynamic setting is appropriate. The benchmark results support the dynamic specification and confirm that CEI exhibits persistence over time, consistent with the path-dependent nature of carbon intensity (Tao et al., 2023).

Table 1 reports the main System GMM results. Model 1 uses EV stock as the penetration indicator, while Model 2 replaces it with EV sales to improve robustness. The diagnostic tests confirm model validity: AR(1) is significant, AR(2) is insignificant and the Hansen test supports the validity of the instrument set. Lagged CEI also remains significantly positive, indicating strong persistence in CEI (Tao et al., 2023).

Most importantly, EV penetration was significantly associated with lower CEI. As shown in Table 1, a 1% increase in EV stock reduces CEI by 0.045% in Model 1, while a 1% increase in EV sales reduces CEI by 0.042% in Model 2. These findings provide cross-country evidence that EV penetration contributes to decarbonisation, and they support earlier studies showing that EV mitigation effects depend on enabling structural and energy conditions (Hofmann et al., 2016; Nuez et al., 2022; Yu et al., 2022).

Importantly, EV stock exhibits a slightly larger estimated mitigation effect than EV sales, suggesting that cumulative penetration may generate somewhat more stable carbon benefits than short-term market expansion. This finding is consistent with concerns that rapid turnover in EV technology may shorten vehicle lifetimes and weaken the efficiency of lifecycle mitigation (Yang et al., 2020). The results, therefore, provide tentative evidence that sustained fleet transformation may yield greater carbon benefits than temporary market diffusion.

Primary EC is significantly and positively associated with CEI, indicating that energy demand remains a major constraint for decarbonisation. This finding is consistent with Wang et al. (2017, 2022a) and reinforces the argument that EV penetration cannot be evaluated independently of the electricity generation structure and energy supply conditions (Franzò and Nasca, 2021; Sacchi et al., 2022). Consistent with this interpretation, the renewable electricity share shows a significant negative association with CEI in both models reported in Table 1, confirming that cleaner electricity systems contribute independently to lower carbon intensity. Importantly, EV penetration remains negative and significant after controlling for the share of renewable electricity, indicating that the mitigation effect is not solely driven by differences in grid cleanliness across countries.

GDP per capita is significantly and negatively associated with CEI, supporting the EKC argument that later development stages are often associated with technological upgrading, efficiency improvement and cleaner production structures (Grossman and Krueger, 1995; Kaplowitz et al., 2013). By contrast, urbanisation is significantly and positively associated with CEI, consistent with evidence that urban expansion intensifies energy demand and carbon pressure (Sun et al., 2022; Wang et al., 2022b; Miao, 2017; Sufyanullah et al., 2022). These findings suggest that EV penetration alone may be insufficient to offset the broader emissions pressures associated with rapid urban development. PD remains statistically insignificant, suggesting that demographic effects may be nonlinear and development-dependent. This outcome is consistent with Liu et al. (2017) and Wang and Li (2021), who show that density effects vary across development contexts.

Finally, R&D showed a positive but statistically insignificant relationship with CEI. Although innovation can support emissions reduction and green technological progress (Wang and Zhang, 2020; Ma et al., 2022; Su et al., 2022), aggregate national R&D includes both green and non-green innovation and may not adequately capture directed technological change toward low-carbon outcomes. Additional lagged R&D specifications remain insignificant, suggesting that aggregate R&D investment does not directly translate into short- to medium-term reductions in carbon intensity. These findings imply that R&D effectiveness depends heavily on policy design and the direction of innovation spending (Chen et al., 2021).

To examine whether EV effects differ across development stages, the sample is divided into HIC and MIC based on World Bank income classifications (IEA, 2025; World Bank, 2023). The estimation results are reported in Table A4a and Table A4b ^[1].

For HICs, EV penetration remains significant and negative, indicating that EV diffusion is associated with lower CEI in developed economies. For MICs, EV stock becomes marginally significant at the 10% level (β = −0.021, p < 0.1), while EV sales remain statistically insignificant. This improvement relative to the original 8-country subsample suggests that the expanded MIC coverage and extended time period provide additional statistical power. However, the magnitude of the association remains substantially smaller than that observed in HIC. This difference provides direct support for the argument raised in the Introduction and Literature Review sections that EV decarbonisation outcomes depend on socio-economic conditions and regional heterogeneity (Doluweera et al., 2020; Shen et al., 2019). It is also consistent with lifecycle-oriented arguments that EV effectiveness varies across regions depending on the electricity mix and supporting conditions (Yu et al., 2022).

The heterogeneity results, therefore, imply that EV policy should not be treated as universally effective. Instead, its success depends on whether complementary conditions, such as cleaner electricity, stronger infrastructure and greater affordability, are in place. This interpretation is further supported by the finding that the renewable electricity share is significantly and negatively associated with CEI in HIC but not in MIC, suggesting that cleaner grid structures contribute to decarbonisation primarily in economies where renewable integration has reached sufficient scale. The MIC subsample contains 11 countries. Although the group count remains modest for System GMM estimation, it represents a meaningful improvement over the original eight-country subsample. The collapse option is applied to keep instrument counts below the number of groups, and diagnostic tests support model validity. Nevertheless, the MIC results should still be interpreted with caution given the limited cross-sectional dimension.

To further examine the mechanisms underlying these heterogeneity results, interaction term models are estimated. The interaction between EV stock and UR was positive and significant (β = 0.052, p < 0.05), indicating that the carbon mitigation effect of EV penetration weakens as urbanisation increases. Conversely, the interaction between EV stock and the share of renewable electricity is negative and significant (β = −0.022, p < 0.05), suggesting that EV penetration yields stronger reductions in carbon intensity in countries with cleaner electricity systems. These findings support the argument that the environmental effectiveness of EV adoption depends heavily on broader structural conditions, particularly urbanisation pressure and electricity generation structure. The full estimation results are reported in Table A5 ^[1].

EV adoption is frequently framed as a straightforward solution to transport decarbonisation. Yet the real climate value of EV penetration depends on whether electrification actually reduces CEI at the national level, rather than simply relocating emissions to upstream electricity and manufacturing systems. Addressing this issue, the present study examines 32 major EV-adopting countries over the period 2010 to 2023 using an extended STIRPAT framework and System GMM estimation. The findings indicate a robust negative association between EV penetration and carbon intensity, consistent with the argument that electric mobility can play a meaningful role in climate mitigation and support the UN's SDG 13 (Climate Action).

The results show that increasing EV penetration is associated with lower CEI, and this relationship remains robust when EV penetration is measured using either EV stock or EV sales. However, the estimated effect of EV stock is slightly larger than that of EV sales, suggesting that sustained fleet transformation may contribute somewhat more consistently to long-term decarbonisation than short-term market expansion. This outcome aligns with lifecycle concerns that rapid technological iteration and depreciation expectations may shorten EV service life and weaken total lifecycle mitigation benefits (Yang et al., 2020). Therefore, EV policies should not focus solely on accelerating new purchases but also on supporting stable utilisation and long-term fleet transition (Ledna et al., 2022).

At the same time, the strong role of the energy structure reveals why EVs cannot be treated as a standalone decarbonisation strategy. Primary EC significantly increases CEI, while a higher share of renewable electricity significantly reduces it, confirming that EV mitigation depends heavily on the electricity generation structure. EV deployment in countries with higher shares of renewable electricity is associated with stronger reductions in carbon intensity, reinforcing the argument that EV transitions must be coordinated with broader power-sector decarbonisation. Previous studies similarly show that EV mitigation can be negligible when electricity used for charging remains carbon-intensive but becomes more meaningful under cleaner electricity systems (Hofmann et al., 2016; Nuez et al., 2022; Tao et al., 2025; Singh et al., 2024). These findings are consistent with research emphasising that national grid conditions strongly shape EV lifecycle emissions outcomes (Franzò and Nasca, 2021; Sacchi et al., 2022). Without parallel progress in the clean electricity transition, EV diffusion may deliver weaker carbon-intensity reductions than expected, even as adoption grows rapidly (Tang et al., 2022; Guo et al., 2022).

The socio-economic drivers in the model further clarify why EV benefits vary across countries. GDP per capita is significantly and negatively associated with CEI, consistent with the EKC mechanism, in which later development stages often enable cleaner production structures, greater efficiency and stronger environmental governance (Grossman and Krueger, 1995). This finding reinforces the broader view that economic growth and environmental protection are not necessarily incompatible when growth quality improves and cleaner technologies diffuse (Kaplowitz et al., 2013). However, urbanisation significantly increases CEI, highlighting a major pressure that EV adoption alone does not automatically offset. This outcome is consistent with evidence that urban expansion increases carbon emissions through rising energy demand, consumption intensity and infrastructure growth (Sun et al., 2022; Wang et al., 2022b; Miao, 2017; Sufyanullah et al., 2022). EV promotion should therefore be embedded within broader low-carbon urban planning strategies, particularly in rapidly developing urban systems where emission pressures may rise faster than technology transitions can reduce them.

Aggregate R&D investment remains statistically insignificant after controlling for the electricity generation structure. This may reflect the inclusion of both green and non-green innovation activities and highlights the importance of directing innovation support toward environmentally beneficial outcomes (Chen et al., 2021; Su et al., 2022; Jiang et al., 2023; Fang, 2023). Green finance may further support renewable energy development and sustainable investment (Subramaniam and Loganathan, 2024; Goh et al., 2025).

Finally, the heterogeneity analysis confirms that EV-driven decarbonisation is not uniform across development levels. EV penetration is significantly and negatively associated with CEI in HIC and becomes only marginally significant in MIC, supporting the argument that regional heterogeneity shapes EV mitigation outcomes (Doluweera et al., 2020; Shen et al., 2019; Yu et al., 2022). This divergence likely reflects differences in electricity mix, infrastructure readiness, affordability and institutional capacity. The interaction term analysis further shows that the mitigation effect of EV penetration weakens with higher urbanisation and strengthens with higher shares of renewable electricity. Provincial-level evidence from China similarly demonstrates that EV lifecycle emissions are substantially higher in coal-dependent regions than in areas with cleaner electricity systems (Hu et al., 2024; Zhang and Hanaoka, 2021). Consequently, EV policy targets should be development-sensitive. High-income economies can pursue more aggressive EV scaling strategies, while middle-income economies may require stronger progress in electricity transition and infrastructure readiness before EV diffusion can generate comparable mitigation outcomes (IEA, 2025; World Bank, 2023).

Overall, this study provides cross-country evidence that EV penetration is associated with lower CEI. While recent evidence suggests that EV adoption may increase total carbon emissions in the absence of sufficient renewable energy (Tao et al., 2025), the present findings indicate that EV penetration is associated with lower CEI, suggesting that the environmental benefits of EVs may be more apparent when evaluated relative to economic output rather than aggregate emissions. These findings are based on 32 major EV-adopting countries and should therefore be interpreted as evidence from established and emerging EV markets rather than as a universal global pattern. Whether these findings extend to low-adoption economies or countries at earlier stages of EV transition remains an important avenue for future research. More broadly, the findings highlight that the climate value of EV diffusion depends on complementary structural conditions, including clean electricity supply, sustainable urban development and effective innovation governance. From a business and economic policy perspective, the results suggest that the environmental returns to transport electrification are closely linked to broader energy transition and sustainable development strategies. This conclusion is consistent with studies emphasising that the environmental benefits of transport electrification are closely linked to the electricity generation structure and broader energy transition progress (Franzò and Nasca, 2021; Sacchi et al., 2022). Without these supporting conditions, EV transitions may underperform their environmental promise and generate uneven mitigation outcomes across development stages.

This study does not involve human participants or animals and therefore does not require ethical approval.

The supplementary material for this article can be found online.