Equity implications of electric vehicles: A systematic review on the spatial distribution of emissions, air pollution and health impacts

We use the Context-Interventions-Mechanisms-Outcome (CIMO) framework to identify papers relevant to this review [19]. CIMO is commonly used to search policy-focused literature. We select this strategy because it aligns well with our goal to find papers that examined the mechanisms that drive the emission and health impact of EVs.

• For context ('C'), our goal is to identify papers that examined the emission and health impacts of EVs in major world regions and countries.
• For interventions ('I'), we include policy and investment decisions that encourage the development of EVs across its value chain.
• For mechanisms ('M'), we focus on identifying the mechanisms that determine the sub-national or cross-regional distribution of emission and health impacts of EVs.
• For outcomes ('O'), we are interested in three main outcomes: (i) emissions, including the emissions of CO₂, SO₂, NO_x , primary PM_2.5, and volatile organic compounds (VOC); (ii) pollution exposures, including ambient PM_2.5 and ozone concentration; and (iii) health impacts, especially premature deaths.

A detailed CIMO-based analytical framework is presented in supplementary note 1.

Based on the CIMO framework, we searched for peer-reviewed papers published in English between January 2010 and August 2022. We focused on papers published in 2010 and onwards because EV is a rapidly evolving technology, and its uptake has increased significantly only in the past decade.

Specifically, we took two steps to identify papers included in our review.

Step 1: we first searched for relevant papers on ScienceDirect and Google Scholar using multiple relevant search strings. As our focus is not only on the emission and health impacts of EVs, but also on the spatial distribution of those impacts, we added distribution-related keywords in our search strings. For example, to search for papers that studied the distribution of impacts by location, we used the following search string—'emission' AND 'electric vehicle' AND 'regional variation'. We observed that using search strings with multiple keywords narrowed down the search considerably and provided a small number of results.

Step 2: we then expanded the search in two dimensions. First, we added additional keywords to search for papers on distributional impacts. We included keywords such as 'equity', 'justice', and 'disparity' in our search strings. Second, to search papers for the world regions that were underrepresented in the initial search (e.g. Europe, India, and Africa), we added specific geographic regions in the search strings. We did this because our initial search mostly yielded papers focused on China and the USA. The complete list of search strings for both steps is provided in supplementary note 2.

Regarding our inclusion criteria, we only included papers that provided a quantitative assessment of the distribution of at least one outcome of interest (e.g. emissions, pollution, and/or health). Since studies that examined the distributions across populations often also provide a geographic angle related to where people live, we included papers that studied the impacts across different population groups in addition to different locations. Regarding our exclusion criteria, we did not include papers that did not perform a quantitative assessment of the distribution of emission and/or health impacts.

To ensure that we did not miss out on key papers, we reviewed the papers identified in step 1. We then applied the same inclusion and exclusion criteria to the new papers identified in step 2, and arrived at the final list of papers for review.

As shown in figure 1, we found 9838 papers from our initial two-step search. After removing duplicates, we were down to 2334 papers. We then screened the titles and abstracts of these papers. We excluded the papers that were not specifically focused on EVs (e.g. papers that studied the transportation sector in general) or did not provide a quantitative assessment of the distribution of emissions or health impacts of EVs (e.g. those life-cycle analyses that only focused on processes but not the location of those processes). Given our research interest, we only included papers that examined the impacts of interventions focused on EVs. These interventions include policy or investment decisions to increase the adoption of EVs. These could also include future scenarios designed to project the emission and/or health impacts under varying levels of EV penetration. Based on these criteria, we arrived at a list of 79 papers. This initial list of 79 papers was then carefully screened by a group of eight people. The full paper was read to further assess whether the paper indeed provided relevant insights that are essential for this review, particularly on geographic variations.

Zoom In Zoom Out Reset image size

After the initial screening, we narrowed down to 47 papers that offered insights to our core research questions [3–6, 10, 15–18, 20–57]. These 47 papers were again read and reviewed by the same group of eight people to sythesize the key insights. Then, a second round of review was conducted by three authors of this study. We divided these papers in such a manner that every single paper was read by at least two different people. In case there were discrepancies or disagreements about the coding for the paper, we discussed these papers in detail among all the four co-authors and arrived at a decision based on our consensus. The final list of papers included in this review is provided as a supplementary data file.

Specifically, for each paper, the reviewer was asked to summarize and synthesize the information for the following aspects (table 1):

Here we summarize the research scope of the 47 papers included in our review and highlight four general patterns (figure 2).

Zoom In Zoom Out Reset image size

First, the global south countries were understudied, despite their central importance in shaping the climate and environmental future. More than 50% of the papers included in our study were focused on developed countries, mainly the United States [4, 15–18, 20–23, 25, 28, 29, 32–34, 57] and European countries [6, 30, 36, 46–48, 51, 52]. About 25% of papers examined emerging economies like China [3, 5, 10, 26, 31, 39, 41, 45, 50] and India [42, 53], with more focused on China (about 20%). The remaining 25% of the paper had a multi-region [24, 27, 38, 40, 43, 49, 55, 56] or global focus [35, 37, 44, 45]. On the one hand, studying these major world regions is reasonable because the cross-regional inequality concerns are particularly important for large countries/regions that are geographically diverse. On the other hand, the lack of attention on some important world regions, such as Africa, Southeast Asia, and Latin America, requires future efforts. These regions are emerging new markets for EVs and are projected to be important contributors to future carbon emissions and air-pollution-related deaths [58, 59]. Understanding the equity implications of transport electrification in these global south regions will be critical for tackling both local and global sustainability challenges.

Second, although many studies considered the impacts beyond the transportation sector, the sectoral coverage was still incomprehensive in most studies. The adoption of EVs, first and foremost, influences the emissions and pollution impacts of the transportation sector. Yet, cross-sectoral linkages are important to understand the life-cycle impacts of the whole EV value chain. For the papers included in our review, 90% of the studies covered both transportation sector and power sectors, as electricity generation activities have strong ties to support EV operations [3–6, 10, 15–18, 20–23, 25–36, 38–43, 45–49, 51, 53, 55–57]. About half of the studies further examined gasoline production and vehicle manufacturing activities: 20% of the studies considered the changes in gasoline production and electricity generation activities to support the EV transition [21, 27, 31, 33, 35, 36, 38, 43, 53], while 10% of the studies considered the emissions and pollution impacts associated from the vehicle manufacturing activities [15, 17, 20, 26, 46], including the production of batteries for EVs and the manufacturing of materials required for vehicles (e.g. steel). However, only 20% of the studies included all four key sectors (i.e. transportation, power, gasoline production, and manufacturing) [4, 23, 25, 28–30, 42, 48, 49]. This pattern indicates that although the research community has started to recognize the importance of cross-sectoral linkages, a lot more needs to be done to account for the multi-sector, economy-wide impacts of EVs.

Third, studies at fine spatial resolution were insufficient. More than 60% of the papers compared emissions and pollution impacts across different countries or regions, without demonstrating the potential differential impacts across states, provinces, or counties within one region [5, 6, 15, 16, 18, 20–23, 25, 27, 30, 35–40, 43, 44, 46–49, 51, 52, 54–56]. About 15% of the papers examined the differential impacts across provinces [3, 26, 29, 31, 41, 42, 50]. This spatial scale is helpful in informing real-world decisions since many energy and climate decisions are made by provincial/state governments and relevant decision-makers. Only about 25% of the papers took a further step to study finer resolution impacts that vary across counties or spatial grids that are finer than states/provinces [4, 10, 17, 24, 28, 32–34, 45, 53, 57]. Most of these finer-scale studies used air quality modeling methods to simulate the geographic distributions of pollution concentrations. To estimate the gridded emissions as input to air quality models, these studies often treated power sector emissions as point sources (based on current locations of plants) and transportation sector emissions as non-point sources (based on information on road infrastructure and traffic volumes). Despite the efforts to assess fine-scale emission impacts, studies still often make simplified assumptions based on present-day spatial patterns, i.e. scaling down the baseline power/transportation sector emissions across all the spatial grids within the same province/state. These simplifications, though justifiable in most cases [60], do not properly account for the nuance of EV impacts depending on which power plants and vehicle choices are being impacted [61, 62].

Finally, the majority of the studies that examined the distribution of EV impacts were based on scenario analyses, e.g. designing counterfactual scenarios for the past or policy-relevant scenarios for the future. These scenario-based analyses allowed them to estimate and compare how sectoral activity could change with a higher share of EVs in the vehicle fleet, which in turn changes the emissions of both greenhouse gases and air pollutants. To account for the pollution and health impacts from these changes in emissions, some studies used air pollution models (e.g. WRF-CMAQ or WRF-Chem [3, 5, 33, 45]) and health impact assessment methods (e.g. concentration-response relationships from epidemiological studies [5, 32–34]) to estimate the corresponding exposure and health damages.

We focus on spatial distributions as one key aspect of equity consideration. For the outcomes examined in each study (i.e. emissions, air pollution, and/or health), we classified the findings into three broad categories: (1) all winning, i.e. as a result of EV transition, the paper found all the locations being covered in the study would always benefit from a reduction in emissions, pollution or health damages; (2) winners and losers, i.e. in all scenarios examined in the study, it is always the case that some locations would benefit while others may suffer from the EV transition; and (3) scenario dependent, i.e. the paper found that losers may occur only in a specific setting and the distributional consequences would depend on the scenario designs.

To illustrate how we classified the studies, here we use one paper as an example, i.e. Ji et al 2015, 'Environmental Justice Aspects of Exposure to PM_2.5 Emissions from EV Use in China' [10]. This paper found that the primary PM_2.5 emissions would decrease in regions where EVs are adopted, making these regions 'winners'. In contrast, increased EV adoption would result in an increase in primary PM_2.5 emissions in regions where electricity for running EVs is generated, making them 'losers'. These patterns of winners and losers were found in all the scenarios that they examined. Thus, in our synthesis, we categorize this study as 'winners and losers' for their key outcome on 'primary PM_2.5 emissions'.

Based on our synthesis of all 47 papers, we highlight two key insights (figure 3). First, while the majority of the studies did quantify the varying CO₂ emissions across regions or sub-national units, very few studies quantified the disparities in air pollution and health. As we were interested in examining both carbon and air pollution emissions, we included papers that only assessed the spatial distribution of CO₂ emissions and did not examine air quality or health impacts. We found that more than 90% of the papers provided evidence of spatial variations in CO₂ emissions as a result of EV transition [3, 4, 15–18, 20–49, 51–57]. In comparison, less than 35% of the papers provided information on the spatial variations in precursor air pollutant emissions (e.g. SO₂, NO_x , primary PM_2.5, or VOC) [3–6, 10, 22, 31–33, 40–42, 45, 50, 53, 55], exposure to ambient air pollution (e.g. ambient PM_2.5 or O₃) [3–5, 22, 32–34, 41, 45, 53], or the health impacts (e.g. premature mortality) [3, 5, 22, 32–34, 41, 45]. Since air pollution and health impacts are highly local, this finding underscores the importance of analyzing the pollution and health disparities in order to understand the local effects of transport electrification.

Zoom In Zoom Out Reset image size

Second, among the studies that examined the spatial variations in addition to the aggregate impacts, many identified potential unequal outcomes where the EV transition creates winners and losers across different locations. For some studies [4–6, 10, 15–18, 20–23, 26, 28, 30, 31, 36, 37, 40, 42–44, 49, 52, 55–57], such tradeoffs exist in all scenarios being examined, while other studies found tradeoffs only in a subset of the scenarios under certain conditions [3, 29, 33–35, 38, 41, 50, 54]. For instance, Peng et al 2018 found a net reduction in air pollutants and CO₂ emissions across all provinces in China when the penetration of low-carbon sources reached 40% of the electricity mix; otherwise, the deployment of EVs could exacerbate the emissions and pollution impacts in some provinces [3]. Our synthesis thus highlights the importance of going beyond the aggregate impacts that are often positive (e.g. a net reduction in emissions or health exposure/damages for the whole region). A detailed assessment of location-varying impacts is needed to understand the unequal local impacts of EV transition.

We further delve into the subset of the papers that examined the location-varying impacts (i.e. the non-grey segments in figure 3). The goal here is to synthesize key mechanisms that may contribute to inequitable distributions. We find that the drivers, processes and mechanisms that connect the EV transition with the resulting distributional impacts are quite complex (figure 4). In addition, the conclusions are often determined by which sectors and mechanisms are being considered; and only a small number of papers examined how these spatial patterns might change over time.

Zoom In Zoom Out Reset image size

First, the adoption rate of EVs is affected by many location-specific factors, where multiple energy-related sectors are further influenced by the operation and manufacturing of EVs. EV adoption is determined by several factors such as supporting policies like government subsidies and financial incentives, as well as the consumers' preference and acceptance to EVs, their socio-economic conditions (e.g., income level), and their driving patterns and habits. The increase in EV adoption directly affects the transportation sector (e.g. lowering gasoline and diesel consumption). In the meantime, the increase in EV adoption also brings changes to the power sector (e.g. increasing electricity generation) and the industrial sector (e.g. lowering gasoline and diesel production, and increasing battery and material production). In this step, the key determinants for the different geographic impacts are the locations of the relevant energy activities, shaped by the energy markets and EV supply chains [3–6, 10, 15–18, 20–57].

Second, the amount of CO₂ and air pollutant emissions vary across sectors and relevant energy activities, which in turn results in differential spatial distributions. The emission impacts also depend on whether gasoline or diesel ICEVs are being displaced by EVs. In terms of tank-to-wheel emissions (i.e. the emissions from driving vehicles), gasoline ICEVs have lower levels of NO_x and SO₂ emissions, but higher levels of non-methane volatile organic compounds (NMVOC) and CO₂ emissions than diesel ICEVs. In terms of well-to-tank emissions (i.e. the emissions from production, processing, and delivery of transport fuels), gasoline ICEVs have higher levels of emissions for CO₂ and air pollutants (including NO_x , SO₂, PM_2.5 and NMVOC) than diesel ICEVs. Regarding upstream sectors, coal-based power generation emits substantial amounts of SO₂ and NO_x , and the mining activities as part of the battery production process also emits SO₂. The emissions from these upstream activities may lead to unintended impacts in locations different from where EVs are being deployed. In addition, we find the spatial differences in the emission factors (i.e. emissions per unit energy activities) also play an important role, especially for the air pollutant emissions. Such differences are driven not only by fuel type and efficiency of locally-used technologies, but also by the stringency of air pollution control policies for the transportation, power, and industrial sectors [3–6, 10, 15–18, 20–57].

Third, the precursor air pollutant emissions influence the concentrations of air pollution, such as fine particulate matter (PM_2.5) and ozone (O₃), affecting the pollution exposure level of residents. In this step, the spatially-varying impacts can be further influenced by the non-linear atmospheric processes, such as local meteorological conditions that would affect the wind transport of pollution, as well as the chemical formation of secondary pollutants, including secondary aerosols and O₃ [3–5, 32–34, 41, 45, 53].

Finally, the human health impacts from the exposure to ambient air pollution are also affected by the characteristics of the exposed population. Here, the additional factors that could widen the spatial variations include the size and vulnerability of the populations [3–5, 32–34, 41, 45]. In addition to the long-term exposure to PM_2.5 and ozone, NO₂ from transportation sector emissions has been found to have short-term health damages, such as increasing risks of asthma attacks, allergies, and other respiratory diseases [63–65]. However, none of our reviewed studies on EVs estimated these short-term health damages from NO₂ exposure, highlighting an important area for future research

More broadly, most of the studies included in our review do not clearly state what definition of equity they adopted and what metrics were used to measure the disparities. To this end, we select three papers as examples to demonstrate what geographic distributions and relevant mechanisms they studied, and what could be the relevant equity definitions/metrics the authors adopted implicitly (figure 5):

Zoom In Zoom Out Reset image size

• Example 1: Wu and Zhang [55] compared the emissions of CO₂ and air pollutants between developed countries (Germany, France, the US, and Japan) and developing countries (China, Russia, India, and Brazil). Compared to conventional gasoline-based vehicles, they found that utilizing EVs reduces transport-related CO₂ emissions in all regions. However, the associated air pollutant emissions may increase in developing countries due to their dependence on thermal power generation combined with the high line loss rates in electricity transmission.Here equity could be defined as countries emit their fair share of CO2 emissions and have similar levels of air pollutant emissions on per capita basis. The equity assessment could be operationalized by comparing the CO₂ emissions and air pollutant emissions across countries at different developmental stages and income levels.
• Example 2: Ji et al [10] examined the PM_2.5-related health impacts due to the urban use of EVs in China. Although deploying EVs reduces emissions from the transportation sector in urban areas, it increases the air pollutant emissions from the power sector in rural areas where most power plants are located. This further leads to differential health impacts.Here equity could be defined as rural and urban populations face similar levels of pollution exposure and health impacts. The equity assessment could be operationalized by comparing the air pollutant emissions and ambient PM_2.5 concentrations in rural versus urban areas.
• Example 3: Hung et al [30] studied the impacts of electricity trade on the carbon footprints across different regions in Europe. They found that the trade of electricity could relocate the electricity generation activities to other regions, thus influencing the distribution of CO₂ emissions across regions.Here equity could be defined as countries are responsible for similar amount of consumption-based CO2 emissions. The equity assessment could be operationalized by comparing the production versus consumption-based emissions across regions.