Journal of Cleaner Production 459 (2024) 142574

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

Electrifying the road: Navigating the transition to electric vehicles in

Connecticut through hybrid insights and fleet evolution
Adam Gallaher a, *, 1, Marcello Graziano b, Carol Atkinson-Palombo a, Lyle Scruggs c
a
Department of Geography, University of Connecticut, Austin Building, 215 Glenbrook Rd #422, Storrs, CT, 06269, USA
b
Ruralis Institute for Rural and Regional Research, University Center Dragvoll N-7491 Trondheim, Norway
c
Department of Political Science, University of Connecticut, Oak Hall, 365 Fairfield Way, Unit 1024, Storrs, CT, 06269, USA

A R T I C L E I N F O A B S T R A C T

Handling Editor: Cecilia Maria Villas Bôas de The adoption of light duty electric vehicles (EVs) in the past 5 years has accelerated dramatically. As EVs
Almeida continue penetrating the U.S. market, it is critical to understand what a sustainable transition within the
transportation sector will look like in relation to mitigating impacts of climate change. This work aims to fill two
Keywords: existing gaps in the current literature: 1) investigate the positive and negative potential influences of the
Transportation
emerging EV transition within the U.S. market and 2) investigate the types of vehicles EVs are replacing. We
Electric vehicles
fulfill these objectives by utilizing a unique attribute dataset of all vehicles in Connecticut (USA) on the road
Spatial statistics
Adoption between 2013 and 2020. Our findings suggest that the penetration rate of traditional hybrid vehicles (e.g.,
Climate change Toyota Prius) is an important predictor of EV adoption. Furthermore, our results show that in Connecticut most
Energy transitions of that transition is occurring among hatchbacks (e.g., Chevrolet Bolt), a body style where internal combustion
engine (ICE) versions have relatively high fuel efficiencies compared with other major vehicle types (e.g., pickup
trucks or SUVs). Our results offer important implications for policymakers in terms of maximizing the deploy­
ment of EVs and EV infrastructure and the path in which the automotive sector is progressing towards
decarbonization.

1. Introduction decarbonizing personal transportation will be vital for addressing the

climate crisis.
The finite supply of fossil fuels, national energy security, and the To address these concerns, nations around the world have outlined
challenge of climate change have all led to a growing interest in electric policies designed to transition the transportation sector from fossil fuel
vehicles (EVs) (Banister, 2008; Muratori et al., 2021). Overall, EVs energy sources to cleaner technologies (Rockström et al., 2017). The
contain the potential to provide multiple benefits to society such as United States for example has recently outlined a goal of shifting 50% of
increased energy efficiency of passenger vehicles, diversification of en­ new vehicle sales towards zero-emission vehicles—battery electric,
ergy resources (i.e., electricity vs fossil fuels), improved local and plug-in hybrid, or hydrogen fuel cell—by 2030, and to further decar­
regional air quality, and potential to reduce global greenhouse gas bonize the entire transportation sector by 2050 (White House, 2021;
(GHG) emissions (Wu et al., 2015; Teixeira and Sodré, 2018; Choma DOE et al., 2022). Furthermore, passage of the US Inflation Reduction
et al., 2020). From 1990 to 2021, transportation was the second largest Act (IRA) of 2022 will further bolster EV adoption, infrastructure
contributor to global CO2 emissions and in 2021 contributed around development, and production supply chains in the coming decade
23% of total emissions (IEA, 2023a). Furthermore, the sector is expected (White House, 2023). As of March 2023, cars and vans within the Eu­
to experience increasing demand and subsequent emissions (Creutzig ropean Union are required to reduce emissions by 55% and 50%
et al., 2015; Davis et al., 2018). With public transport use in the United respectively by 2030 compared to 2021 levels and 100% across the
States declining (Erhardt et al., 2022), and Covid-19 further pushing board by 2035 (IEA, 2023b). In yet another example, China has
users to personal transportation (Liu et al., 2020; Loa et al., 2022), launched a target to electrify 80% of all public sector vehicles while
identifying and deploying a range of technologies and policies aimed at simultaneously build out charging stations by 2025 (IEA, 2023a).

* Corresponding author.
E-mail address: [email protected] (A. Gallaher).
1
Department of Natural Resources and the Environment, Cornell University, 111 Fernow Hall, 226 Mann Dr, Ithaca, New York 14,853.

https://doi.org/10.1016/j.jclepro.2024.142574
Received 13 September 2023; Received in revised form 9 April 2024; Accepted 13 May 2024
Available online 14 May 2024
0959-6526/© 2024 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Although ambitious, the dynamics between innovative EV technologies, reductions. Unfortunately, options for passenger van EVs are limited
and consumer adoption behavior remain complex and vary geographi­ thereby posing a barrier to realizing emission reduction in this ripe
cally (Tamayao et al., 2015; Yuksel et al., 2016; Ryghaug and Skjølsvold, segment.
2022). The remainder of this paper is as follows: section 2 provides an
To date, EVs have emerged as the most likely alternative to fossil overview of relevant EV adoption literature ranging from purely tem­
fuels within the personal passenger transportation sector. Throughout poral studies to those investigating the influence of geography, section 3
this paper we use the term EV to represent a combination of battery is a description of the study area and dataset used in the analysis, section
electric vehicles (BEVs) and plug-in hybrid vehicles (PHEVs) (see Sup­ 4 provides an overview of the methodological approach, followed by
plemental Table S2 for detailed description). Although there are distinct section 5 which presents the results, and lastly section 6 contains our
differences in terms of technology, climate mitigation potential, and conclusions and highlights the studies limitations and suggestions for
incentive structures, both vehicles use battery technology and place future research.
added demand on the electricity grid. For this reason, we have combined
them into one category. The widespread replacement of internal com­ 2. Theoretical framework & literature
bustion engine (ICE) vehicles with EVs offers a promising route to
decarbonize the transportation sector (Casals et al., 2016; Teixeira and The diffusion of innovative technologies rarely takes a direct path.
Sodré, 2018). Between 2010 and 2022 the global share of EV sales grew Rather, multiple paths work together or against one another leading to
from 0.011% to 14%; despite this rapid growth, the overall share of EVs the success or failure of integration within society. Geels (2002)
remains minimal at only 2.1% as of 2022 (IEA, 2023c). Although the last described how the success or failure of innovative technologies, under
decade brought about favorable conditions (i.e., diversity of vehicle certain conditions (e.g., exogenous shocks to the socio-technical regime)
options, government adoption incentives, and a maturing technology), shape and transform societal interactions from existing technologies to
successfully decarbonizing personal transportation faces multiple bar­ emergent ones (e.g., transition from fossil fuel vehicles to EVs). Given
riers (Berkeley et al., 2017). There currently exists a rich body of liter­ the relatively recent introduction of modern EVs, much remains to be
ature on factors influencing EV adoption (Kumar and Alok, 2020). Like learned about what factors will influence EV adoption. Currently, much
other renewable energy technologies (e.g., rooftop solar) (Graziano and of the research related to this question is often viewed through a tem­
Gillingham, 2015), EV adoption is associated with a mix of general and poral lens, employing models to predict adoption rates and uptake tra­
regional factors (see e.g., Lebrouhi et al., 2021 and the regional works jectories (Shepherd et al., 2012; Tran et al., 2013; Gnann and Plötz,
cited within). The potential negative impact of a rapid EV transition on 2015; Kumar and Alok, 2020). Less attention has been given to spatial
US carbon emissions could further complicate overall decarbonization elements of EV adoption, specifically how EVs propagate over time
efforts and the ongoing energy transition. A recent assessment suggests across different geographies (however see e.g., Delmas et al., 2017;
that an EV transition in the United States could increase electricity Morton et al., 2018; Lyu, 2023; Rode, 2024). Understanding EV adop­
consumption by 35% thereby placing additional stress on efforts to tion variation at local scales and considering the potential for spatial
decarbonizing the electricity grid to fully realize the benefits of EVs spillovers (LeSage and Pace, 2021) can provide public officials more
(Galvin, 2022). In addition, the types of vehicles that are replaced by effective policymaking tools, such as incentive programs, electricity grid
EVs will influence the perceived benefits of EVs as they relate to climate planning, and other supportive infrastructure that can facilitate a suc­
change goals. cessful transition to a fully electrified vehicle fleet (Broadbent et al.,
To fill these gaps, and to further understand the regional patterns of 2017). Coffman et al. (2017) review 50 studies on factors that influence
adoption of EVs in North America we aim to fulfill two main objectives EV adoption. They distinguish internal factors, like vehicle ownership
in this paper. Firstly, we estimate what factors (i.e., socioeconomic, costs, driving range, and charging time, from external factors like fuel
demographic, and technological) might influence EV adoption, how prices, consumer characteristics, availability of public charging stations,
those factors vary intraregionally, and what influence, if any, do spatial public visibility, and social norms. They conclude that high purchase
spillovers have on local adoption rates (Dharshing, 2017). Secondly, we prices are a major barrier to adoption, and that certain consumer
evaluate vehicle fleet replacement dynamics and CO2 emission impli­ characteristics, such as income, education, and age, significantly influ­
cations by body style for personal passenger vehicles. To answer our ence interest in EV ownership. In a separate review, of 40 articles pub­
main research questions, we rely on two datasets, a spatial panel dataset lished after 2011, Li et al. (2017) largely corroborates those results, but
of vehicle counts by fuel type, sociodemographic, and charging infra­ also note that income is not a key factor when focusing specifically on
structure data and another dataset of vehicle makes, models, fuel type, battery electric vehicles. More recent studies also confirm that more
body style, and year. Both datasets span the years 2013–2020. We education, being male, and being closer to charging stations increase EV
leverage panel and spatial-panel econometric approaches to identify adoption (Sierzchula et al., 2014; She et al., 2017; Melliger et al., 2018;
factors associated with EV adoption and lifecycle analysis approaches to Sovacool et al., 2018; Gillingham et al., 2023). Consumer profile studies
evaluate fleet wide replacement dynamics by body style. suggest that younger people are less likely to adopt an EV, possibly
In brief, our findings suggests that work modality, charging infra­ owing to the mix of capital and marginal costs for ICE vehicles and EVs
structure, and the share of traditional hybrid vehicles (e.g., Toyota (Sierzchula et al., 2014; Mukherjee and Ryan, 2020; Fevang et al.,
Prius) are associated with EV adoption. Contrary to previous research, 2021).
our results show that areas with increased shares of multi-car single Exploring both the spatial and temporal aspects that charging
family homes are not positively associated with EV adoption (Gärling infrastructure, situational contexts, and the influence neighboring
and Thøgersen, 2001; Campbell et al., 2012; Morton et al., 2018), raising geographic units have on the spatial diffusion of EVs can enhance our
questions about the idea that EV adoption is part of a “two car” model (i. overall understanding of an emerging EV transition (Morton et al.,
e., EVs are adopted for short distances only when an ICE vehicles is on 2018). For example, Campbell et al. (2012) used a spatial cluster model
tap for longer trips). With respect to fleet replacement dynamics, we to identify locational hot spots of consumers most likely to adopt an EV
found between 2013 and 2020 vehicles on the road in Connecticut within Birmingham, UK. Using UK census data, the authors found that
increased by 7.85% with the passenger van (aka minivan or sociodemographic data associated with that of an “ideal” EV adopter
multi-purpose vehicle) segment representing the largest increase. was clustered to surrounding Birmingham suburbs (primarily to the
Focusing specifically on EVs, we found that the hatchback body style (e. north of the city center), while the city center region of Birmingham was
g., Volkswagen Golf) is undergoing the largest shift towards electrifi­ part of the cluster least likely to adopt EVs. Overall, Brimingham was an
cation. Considering current EV availability, efforts towards electrifying unlikely location for EV adoption, the authors found variability in EV
the passenger van fleet would likely result in the greatest emission adoption when disaggregating to smaller spatial units of analysis.

2
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Morton et al. (2018) expands the study EV adoption in the UK by Table 1

analyzing neighborhood effects. By examining the whole of the UK, Summary of research on the consumer adoption of electric vehicles (EVs).
compared to a single city, the authors found higher concentrations of EV Source Region Country/ Method Covariates
ownership in urban settings which contained higher incomes, education, State
and shares of traditional hybrid vehicles. In the United States, Liu et al. Mukherjee Europe Ireland Non-linear Distance to
(2017) also found units with higher education had higher shares of and Ryan regression; charging point;
traditional hybrid vehicles, possibly linking the two technologies (i.e., (2020) general nesting dealerships;
traditional hybrid vehicles and EVs) among potential adopters. In yet spatial models Large homes;
Long commutes;
another study, Plötz et al. (2014) found German EV adopters tended to Renters; Higher
live in non-urban locations (presumably due to increased likelihood of ed.; age;
owning a garage), own multiple vehicles (i.e., two car model), and who population
economically benefited from EV ownership compared to urban owners Fevang Europe Norway Regression Income; region
et al. (urban);
whose timeline for return on investment via vehicle miles traveled was
(2021) nationality;
far greater. Similar geographic patters were observed in Norway and education;
Sweden, finding higher rural EV ownership rates (Kester et al., 2020). selected
Schulz and Rode (2022) found that the presence of public charging education
infrastructure in Norway increased the rate of EV adoption by 200% degrees; age;
distance to work;
over five years. Similarly, Namdeo et al. (2014) found higher urban EV travel
adoption provided there was adequate public charging infrastructure a characteristics;
feature of an EV transition that could act as a barrier if not properly employment
planned (Biresselioglu et al., 2018). sector
Campbell Europe United Cluster analysis Age; home
Among the literature reviewed, some clear trends emerge. Socio­
et al. Kingdom ownership; home
demographic variables such as income, age, education level, employ­ (2012) structure;
ment, gender, and commuting patterns are consistently evaluated as commute mode;
possible factors associated with EV adoption. Less studied, although multi-car
equally important, is the association that public charging and multi-car ownership;
occupation;
ownership has on EV adoption. A summary of literature covering the
education
adoption of EVs can be found in Table 1. Plötz et al. Europe Germany Logistic Gender; age;
(2014) regression location of
residence;
3. Study area and data
household size;
number of
We focus our research on Connecticut for several reasons. First, children;
Connecticut is a relatively early mover in the EV adoption race, ranking employment;
15th in share of EVs in the US (Fig. 1). Second, the state faces a strained commute
distance; income;
electrical grid, and the existing research on EV adoption in the US, has education
not paid much attention to regions prone to unreliable grid conditions Kester et al. Europe Norway & Survey; expert Gender;
(Graziano et al., 2020; Gallaher et al., 2021). Third, when assessing (2020) Sweden interview; focus household size;
sustainable transitions, regionalism and spatial considerations provide a group country; income;
education;
valuable perspective in that each location is distinct and might require a
employment;
different path to decarbonization (Truffer and Coenen, 2012). Thus, a political
focus on Connecticut provides the opportunity to explore sustainable affiliation;
transitions within the New England transportation sector while simul­ environmental
taneously engaging with possibly implications to electricity grid reli­ values
Liu et al. North United States Spatial Population;
ability. Beyond the abovementioned motivations for studying EV (2017) America (Ohio) autoregressive; number of
adoption using Connecticut as a case study, the state has faced a unique spatial error; homes; commute
barrier in how popular EVs (e.g., Tesla) are purchased. geographically mode; vehicle
Connecticut, along with 13 other states are subject to a direct-to- weighted ownership;
regression income
consumer sale ban2 on EVs sold outside of traditional dealership orga­
Morton Europe United Ordinary least age; education;
nizations. As the only state in New England to enact such a restriction on et al. Kingdom squares; spatial employment;
purchasing an EV, it highlights the unique nature of this study within the (2018) Durbin model median income;
regional context. Currently, purchasing an EV, specifically a BEV (e.g., population
Nissan Leaf or Tesla Model 3) requires individuals to travel beyond density; hybrid
vehicles per 1000
Connecticut’s borders to states like New York or Rhode Island so that
cars; EV charging
they may take delivery of their vehicle, ordered directly from the auto points; housing
manufacture. In fear of losing business, dealerships have leaned on structure;
policymakers to craft restrictions against taking delivery of vehicles commute mode;
single vehicle
ordered online directly from the manufacture (i.e., Tesla and Rivian).
house
Realizing the barrier placed on Connecticut residents, policy makers Schulz and Europe Norway Difference in Population
have introduced state bill 214 An Act Concerning the Sale of Electric Ve­ Rode Difference density; income;
hicles in The State to spur adoption of EVs. (2022) median age;
However, the bill was ultimately stifled, potentially further delaying education;
environmental
political
affiliation
2
New Hampshire, Massachusetts, and Rhode Island who had similar laws (continued on next page)
have all since updated those in some capacity to allow for direct-to-consumer
sales.

3
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Table 1 (continued ) manufacturers (e.g., General Motors or Ford), there remains a large gap
Source Region Country/ Method Covariates in achieving EV policy goals.
State In 2020, the Connecticut Department of Energy and Environmental
Namdeo Europe United Spatial hotspot Gender; age;
Protection (CT DEEP) developed a policy framework designed to
et al. Kingdom analysis occupation; accelerate EV adoption in tandem with climate and clean air goals.
(2014) household Connecticut, in coordination with 10 other states,3 have agreed to a goal
income; multi- of deploying 125,000 to 150,000 EVs by 2025 (CT, 2020). However, as
car household;
of 2020 there are 12,380 EVs in the state of Connecticut, representing
housing
structure; between 8.25% and 9.90% of 2025 goals. Moreover, the Governor’s
commuting mode Council on Climate Change (GC3) determined that in order to meet goals
Sovacool Europe Denmark; Survey; vehicle outlined in the 2008 Global Warming Solutions Act 20% of the light duty
et al. Finland; frequency background; vehicle (LDV) fleet, or 500,000 vehicles, would need to be replaced by
(2018) Iceland; analysis; single mobility
Norway; level statistical patterns;
EVs by 2030 (CT, 2020). Given the unique connection between elec­
Sweden analysis willingness to tricity, buying, and the large gap in EV policy goals, elucidating the
pay; preference various drivers and barriers of EV adoption in Connecticut could help
for vehicle bring to light challenges similar regions might face when attempting to
attributes;
electrify personal transportation.
gender; age;
occupation; To evaluate factors associated with EV adoption and the spatial
political patterns in Connecticut, we rely on a dataset constructed from
affiliation; combining data on vehicular, infrastructural, and sociodemographic
education; information. The summary statistics of the variables used in this analysis
household
income
are listed in Table 2 (see Supplemental Table S3 for list of sources). The
Melliger Europe Switzerland; Focus groups Gender; age; geographic unit of analysis is the US Census Zip Code Tabulation Areas
et al. Finland education; multi- (ZCTAs). ZCTAs are a generalized representation of US Postal Service zip
(2018) car household code service areas, which are not geographic units, but rather service
Axsen et al. North British Stated choice Gender; age;
routes used for mail delivery. The use of ZCTAs provides a stable
(2016) America Columbia survey household
income; geographical unit over time in locations that, in Connecticut, experi­
education; enced low population change: 0.9% growth between 2010 and 2020
housing (Proto, 2022). Variables used to evaluate factors associated with EV
structure; vehicle adoption are based on commonly used indicators from the literature.
ownership;
The dataset covers, annual vehicle registrations by energy/fuel type and
charging access
Nayum Europe Norway Survey Gender; age; vehicle body style, American Community Survey (ACS) 5-year averages
et al. household size; of sociodemographic data, and information about the EV charging
(2016) material status; infrastructure recorded by the Department of Energy’s Alternative Fuels
education;
Data Center station locator (AFDC, 2023) for years 2013–2020. We
occupation;
household dropped 19 ZCTAs that contained missing data for total population, total
income; multi- households, or were identified as PO boxes or special designation areas
car household (e.g., Universities or major companies). This resulted in a 6.7% reduc­
She et al. Asia China Survey Gender; age; tion in the original data going from 282 to 263 ZCTAs across all time
(2017) household size;
periods.
driving
experience; Data on vehicle registrations at ZCTA level is provided by IHS Markit
vehicle (now S&P Global). Contained within our vehicle registrations dataset is
ownership; information related to vehicle make, model, sub-model, engine size,
education;
body style (e.g., pickup truck, wagon, sedan, etc.), fuel type (e.g., gas­
income
Sierzchula Global 30 countries Regression Incentives; EV
oline, electric vehicle, etc.), and vehicle year. Vehicle year is obtained
et al. charging points; from the Vehicle Identification Number (VIN). Data was aggregated to
(2014) environmental the ZCTA level for our initial EV adoption analysis by fuel type and data
regulation; fuel year (other details from the IHS Markit dataset are used in the later
prices; EV
analysis of fleet replacement dynamics). To obtain sociodemographic
manufacturing;
income; data, we used the end year of each ACS 5-year average from 2009 to
education; 2020. For example, we use the 2009–2013 ACS 5-year average to
vehicles per represent sociodemographic information for the year 2013. Lastly, we
capita; electricity
summarize the cumulative number of charging stations (e.g., level 1, 2,
prices; EV market
price; urban
or DC fast charger) per ZCTA in each year, with all chargers installed
density prior to 2013 included in the 2013 data year.
Gillingham North United States Regression Income,
et al. America household size, 4. Methodology
(2023) education,
population
density This study takes a two-stage approach, whereby we begin with a
regression analysis of EV adoption in Connecticut from 2013 to 2020.

meeting clean air and climate targets set out by the state. Although, each
year auto manufacturers are offering an increasing number of EVs in 3
California, Maine, Maryland, Massachusetts, New Jersey, New York, Ore­
their lineup such as the Ford Mustang Mach-E, Lightning F-150, Kia EV6,
gon, Rhode Island, and Vermont. See: State Zero-Emission Vehicle Programs
etc. which are sold through traditional dealership networks. Despite an https://www.nescaum.org/documents/zev-mou-10-governors-signed-201911
increasing number of EV models available through traditional auto 20.pdf.

4
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Fig. 1. Market share of EVs as of 2020.25th percentile is equal to 0.16% and 75th percentile is equal to 0.57% with a median value of 0.30% (AFDC, 2022).

Table 2
Summary statistics of data used in EV adoption analysis. See Table S3 in appendix for list of data sources.
Variable Obs. Periods Mean Std. Dev. Min Max Rationale

Electric vehicles per 1000 LDV (%) 2104 7 210 269 0 3254 Dependent variable
Hybrid vehicles (%) 2104 7 1.78 0.77 0.12 8.16 Morton et al. (2018)
Median Household Income ($ 10,000) 2104 7 8.64 3.38 1.13 25.00 Morton et al. (2018)
Housing density (%) 2104 7 633.44 1016.54 3.29 8779.38 Campbell et al. (2012); Plötz et al. (2014); Kester et al. (2020)
Active age 25–44 (%) 2104 7 22.76 6.11 3.39 58.26 Liu et al. (2017); Coffman et al. (2017)
Long commute (>60min) (%) 2104 7 7.14 4.93 0 58.33 Plötz et al. (2014); Kester et al. (2020)
Female Pop (%) 2104 7 50.92 2.83 28.21 64.41
Service Sector (%) 2104 7 77.78 7.06 0 100
Multi-Car (%) 2104 7 63.99 15.89 0.84 100 Plötz et al. (2014); Kester et al. (2020)
Charging Stations (#) 2104 7 2.09 4.85 0 60 Namdeo et al. (2014); Schulz and Rode (2022)
Work from home (%) 2104 7 5.76 4.05 0 29.67

This is followed by an estimation of greenhouse gas implications of an 4.1. Temporal specification

emerging EV transition in Connecticut. For the regression analysis, our
methodological approach follows that of Gallaher et al. (2021), in that Our panel model can be parsimoniously stated as:
we begin by evaluating EV adoption from a temporal perspective.
EVsi,t = α + β1− 11 γ i,t + β12 δi,t + β13 φi,t + ηi + εi,t
However, geographic data tends to exhibit spatial dependence (i.e., the
presence of spatial autocorrelation or a clustering of similar values) and Whereby: EVsi,t is the share of EVs per light duty vehicles (LDV)
thus is best handled using spatial statistical methodologies. Unique to occurring in ZCTA (i) in each year (t); α is our intercept; β1− 11 γi,t rep­
this approach is the ability to understand the influence neighboring resents each of our sociodemographic variables (see Table 2); β12 δi,t
geographic units have on one another (i.e., spatial spillovers). Before we contain the total number of public charging stations ports within each
describe the methods used in our analysis, we briefly describe a directed ZCTA-year; β13 φi,t contains the share of traditional hybrid vehicles per
acyclic graphic (DAG) model to elucidate the underlying assumptions of LDV (replacement effect); ηi is a ZCTA level fixed effect; and εi,t is our
panel data models and the use of fixed effects (Fig. 2). Our variable of error term.
interest (share of EVs per 1000 light-duty vehicles (LDV) henceforth
referred to as share of EVs) has repeated measurements at the same unit
4.2. Spatial-panel models and specification
(Yit), and we have included a set of independent variables (Dit). Within
panel data, there is also a single-unit specific unobserved variable (Ui)
Although the temporal model can evaluate the association between
and some unit-specific time-invariant variable (Xi). We know from the
independent variables and EV adoption, geographic data often exhibits
DAG and the works of Imai and Kim (2019) that Di1 influences both Yi1
spatial dependence (i.e., spatial autocorrelation), which, if not accoun­
and Di2 and we note that the unobserved cofounder, Ui influences all Y
ted for, result in biased parameter estimates (Chun and Griffith, 2013).
and D variables. Therefore, we can conclude that D is endogenous
Like traditional correlation, spatial autocorrelation can be positive (i.e.,
because Ui is unobserved and absorbed into the structural error term of
similar values are clustered in space) or negative (i.e., dissimilar values
the regression model. Lastly, we know that past outcomes do not directly
are clustered in space). Spatial autocorrelation within geographic
affect current outcomes or current treatments and past treatments do not
datasets results from one of the foundational laws within the field of
directly affect current outcomes. It is under these assumptions that we
Geography, everything is related but near things are more related than
can use the fixed effects panel method to understand the relationship
distant things (Tobler, 1970). In essence, geographic data violates the
between D and Y.
basic assumptions associated with ordinary least squared (OLS) models
(e.g., independence and identically distribution of data). Therefore,
leveraging methods designed to account for spatial autocorrelation in

5
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

(see LeSage (2008) for more details). Most importantly, the SAR model
has been used in many studies (see e.g., Balta-Ozkan et al., 2015;
Dharshing, 2017; Graziano et al., 2019; Gallaher et al., 2021 for addi­
tional examples) and can accommodate the analysis of spatial re­
lationships within panel datasets (see e.g., Elhorst, 2014). The unique
characteristics of spatial models is the construction of a weighting ma­
trix, we use a row standardized queen’s weighting matrix, because such
a configuration allows for the greatest number of spatial interactions
(see e.g., Cabral et al., 2017; Dharshing, 2017; Lan et al., 2020).
Our spatial panel model can be parsimoniously stated as the
following:
EVsi,t = α + ρWi,t EVsi,t + β1− 11 γ i,t + β12 δi,t + β13 φi,t + ηi + εi,t (4)

Equation (4) mirrors equation 1, with the addition of the row stan­
dardized queen’s spatial weighting matrix appended to the dependent
variable. Moreover, SAR models account for spatial dependency among
neighboring geographic units (e.g., ZCTAs) through the spatial weight­
ing matrix. Parameter estimates in SAR models cannot be interpreted
directly from output coefficients as with an OLS model. Therefore, ap­
proaches have been developed to interpret output coefficients from SAR
models either directly (i.e., within our unit of analysis), or indirectly (i.
e., units of analysis neighboring any one unit), or total (i.e., the sum of
direct and indirect impacts). Additional details can be found in LeSage
(2008) and LeSage and Pace (2009).
Fig. 2. Directed acyclic graphic model of EV adoption in Connecticut. Yi rep­
resents share of EVs per ZCTA per year from 2013 to 2020, Di represents a
matrix of independent variables (Table 2) which vary over time. Ui is a single 4.3. Estimation of fleet-wide greenhouse gas emissions
unit-specific unobserved variable that varies across units but not across time
and a unit-specific time-invariant variable Xi that is observed unlike Ui. For EVs have long been promoted as an ideal solution for mitigating
additional information on panel-data DAGs see Cunningham (2021) chapter 8 GHGs within the transportation sector, specifically among LDVs. How­
and Imai and Kim (2019). ever, research conducted on the replacement dynamics of EVs is limited.
We estimate vehicle fleet emissions from the ten most common vehicles
geographic data results in less biased estimates of regression by body style on the road in Connecticut between 2013 and 2020. To
coefficients. assess fleet-wide replacement (i.e., the composition of vehicle fleet by
There are several methods used in testing for spatial autocorrelation; body style) we implemented a change analysis whereby we evaluated
however, the most widely used is Moran’s I. One of the key benefits of the vehicle stock in 2013 and compared to the stock in 2020. Our
Moran’s I is that it can index data across all four measurement scales: objective is to understand the GHG implications of an electrified vehicle
normal, ordinal, interval, and ratio (Griffith, 2010). Moran’s I is a fleet in Connecticut and determine which body style has been driving
normalized index whereby values fall between − 1 (perfect negative vehicle electrification.
spatial autocorrelation) and 1 (perfect positive spatial autocorrelation). Our estimated values for emissions (grams of CO2 per mile) originate
We can define Moran’s I by the following equation. from the US Department of Energy fuel economy database (DOE, 2023).
∑∑ ( ) For ICEVs we documented the tailpipe and upstream emissions. Tailpipe
N i j wi,j (xi − x) xj − x emissions include those that are emitted from the vehicle during normal
I= ∑ (2)
W i (xi − x)
2
operations and are represented in grams per mile of CO2. Upstream
emissions include CO2, methane, and nitrous oxide emitted from all
Where N is the number of spatial units indexed by i and j, x is our var­ steps in the use of fuels from production and refining to the distribution
iable of interest, in this case the share of EVs, x bar is the mean and w is and final use of fuel. Method and nitrous oxide are converted to a CO2
our weighting matrix. Global measures of spatial autocorrelation pro­ equivalent value (see EPA, 2022 for additional details). To calculate
vide a nice overview of how clustered the data are; however, we are emissions for EVs we used a similar approach, the main difference is we
assuming that any spatial autocorrelation is uniform across space. In this do not consider emissions during vehicle operation (e.g., tire or break
case, a more detailed measure of local spatial autocorrelation can pro­ emissions). To calculate upstream emissions from EV operations (i.e.,
vide valuable insights. One such method is the Local Indicators of Spatial charging) we used the US EPA’s emissions calculator which uses
Association (LISA) first introduced by Luc Anselin (1995). Based on this regional electricity emissions based on pounds per megawatt hour of
approach, we can know the location of where spatial autocorrelation is electricity. Because we focus on Connecticut, the calculator pulled data
occurring in addition to establishing a proportional relationship be­ from the Independent System Operator New England (ISO-NE). This
tween the sum of each local statistic and a corresponding global statistic data is based on values from 2021. We used an average of 15,000 miles
(e.g., Moran’s I) (Anselin, 1995). Building on the global spatial auto­ driven each year by those ages 20 to 54 as reported by the US Depart­
correlation equation, we can define LISA by the following equation: ment of Transportation for our unit of analysis (US DOT, 2022).
∑ ( )
N j wi,j (xi − x) xj − x
Ii = ∑ (3) 5. Results
W i (xi − x)
2

To investigate EV adoption in Connecticut, and to correct for spatial We first provide the temporal and spatial-temporal results followed
autoregressive correlation we deploy a Spatial Autoregressive model by results on the fleet wide replacement dynamics from 2013 to 2020
(SAR; Cliff and Ord, 1970). The SAR model is motivated based on and the estimated implications it has on greenhouse gas emissions.
time-dependency; in other words, we model the space-time lagged Overall, our results show of the factors considered eight are positively
values of the dependent variable via the spatial autoregressive process associated with and five are negatively associated with EV adoption

6
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

under our preferred specification. We observe different results from our unsurprising as they serve as a necessary component of an EV transition.
preferred specification, under a two-way fixed effects model with four While charging stations exhibit a limited influence on adoption of EVs β
factors positively associated with and seven factors negatively associ­ of 0.010, they still serve as a statistically significant indicator. Diverging
ated with EV adoption. Across all four specifications we found charging from previous literature our results show β of − 0.007 for multi-car
infrastructure, working from home, median household income, long households. Interestingly, longer commutes β of 0.007 and working
commutes, and housing density to be positively associated with EV from home β of 0.019 are positive indicators of EV adoption. Addi­
adoption (Table 3). tionally, we found that median household income β of 0.107 positively
influences EV adoption. ZCTAs where the share of incomes is below the
5.1. Panel results median, large precent of the population work in the service sector, and
are of an active age (i.e., 22 to 44), and are female are less likely to
When assessing the relationship between the share of traditional influence EV adoption.
hybrid vehicles and EVs, we found consistent results for models 1 and 3
(2 and 4 were not significant). Model 1 has a β of 0.127 which is sta­ 5.2. Spatial-panel results
tistically significant at the 99% confidence level and model 3 has a β of
0.049 and is statistically significant at the 90% confidence level. This Geography represents an interesting opportunity to explore re­
result suggests that on average if the share of traditional hybrid vehicles lationships between variables of interest while simultaneously incor­
increases by 1% then the share of EVs is predicted to increase by 0.127% porating inherent spatial interactions. The motivation to use spatial
or roughly by 1572 new EVs on the road. This implies that when the methodologies results from identification of spatial autocorrelation.
share of one technology, in this case traditional hybrid vehicles in­ Having tested for spatial autocorrelation via Moran’s I (Fig. 3) and LISA
creases, the share of EVs will also increase; either through shear pres­ (Fig. 4) we found that the share of EVs in Connecticut exhibited positive
ence on the road or by owners trading in “older” technologies for newer spatial autocorrelation with a statistically significant Moran’s I value for
ones (i.e., trading in a Toyota Prius for a Plug-in Hybrid Toyota Prius or years 2015 through 2020 (Fig. 3). Zooming in, we found clustering of
Chevy Bolt). Furthermore, we found the number of charging stations to high-high and low-low share of EVs relative to the mean distributed
also be associated with the adoption of EVs, although this result is across the state (Fig. 4). Much of the spatial clustering of EV adoption
occurred in the Fairfield County region of Connecticut, a region char­
acterized by high median household income (22.14% above the state
Table 3
average) and high median values of owner-occupied homes (54.29%
Results of panel regression of the share of EVs per 1000 LDVs in Connecticut
from 2013 to 2020. Model 1 represents a one-way fixed effects using unit fixed above the state average).
effects while model 2 represents two-way fixed effects. Standard errors are re­ Based on the panel results, our main variables of interest are the
ported in parenthesis, model 3 standard errors are clustered by town id. See share of traditional hybrid vehicles and the number of charging stations
equation 1 in section 4.1 for detailed description of OLS model and Table 2 in (Table 4). Our results carry over from the panel analysis to the spatial-
section 3 for motivation of variables used in the model. panel analysis in that we found positive correlation between EV adop­
Variables (1) (2) (3) (4) tion and independent variables of interest (i.e., share of traditional
hybrid vehicles and charging stations). Our results are consistent across
Unit Fixed Two-way Clustered First Diff.
Effects Fixed Effects Errors all three spatial-panel models although have diminishing significance.
We found that the share of traditional hybrid vehicles is positively
Hybrids (%) 0.127c − 0.0214 0.0498a − 0.00452
associated with the share of EVs by 0.071% (β 0.071) within a ZCTA and
(0.0146) (0.0148) (0.0302) (0.0121) by 0.118% (β 0.118) on neighboring ZCTAs (i.e., spatial spillover).
Active Age 22–44 − 0.685c − 0.270a − 0.436b 0.188 Simply put, if the share of traditional hybrid vehicles increases by 1%
(%) (0.181) (0.158) (0.173) (0.133) within a ZCTA we would expect the share of EVs to almost double in the
Housing Density 0.000451c 0.000215b 8.50e-06 − 1.03e-05
surrounding ZCTAs. Furthermore, when considering the association
(%) (9.86e-05) (8.59e-05) (1.02e-05) (6.86e-05)
Female Pop. (%) − 0.00440a − 0.00180 0.000368 0.000660 charging stations have on EV adoption, we found within a ZCTA an in­
(0.00225) (0.00195) (0.00178) (0.00184) crease of 0.004% (β of 0.004) and within neighboring ZCTAs (or spill­
Service Sector − 0.00107 − 0.000585 0.000414 − 0.000196 over) an increase of 0.008% (β of 0.008). From this we can conclude that
Employment (0.00112) (0.000972) (0.00138) (0.000912) presence of charging stations within a ZCTA has less of an association
(%)
Multi-Car Home − 0.00766c − 0.00423c − 0.00419c 0.000375
with EV adoption compared to neighboring ZCTAs; suggesting charging
(%) (0.00116) (0.00101) (0.000961) (0.000716) stations might only need to be near potential adopters and not directly
Charging Stations 0.0105c 0.00299b 0.00270 0.00253b within their geographic unit. Practically, our results show that on
(#) (0.00151) (0.00134) (0.00195) (0.00124) average if the share of traditional hybrid vehicles increases by 1% the
Long Commute 0.00768c 0.00187 0.00135 0.00216a
total number of EVs within a ZCTA will increase by 1864 and indirectly
(>60 min) (%) (0.00144) (0.00127) (0.00127) (0.00118)
Work-from-home 0.0192c 0.00883c 0.00781c 0.00433c by 3098. Although small, we found that hybrid vehicles as a “green”
(%) (0.00168) (0.00155) (0.00245) (0.00128) technology are positively associated with the adoption of EVs within any
Median Household 0.107c 0.0557c 0.0385c 0.00553 given ZCTA in addition to indirectly represented by spatial spillovers.
Income (0.00481) (0.00468) (0.00705) (0.00446)
($10,000)
Constant − 0.464c − 0.114 − 0.0913 0.0633c
5.3. Change in vehicle fleet from 2013 to 2020
(0.161) (0.140) (0.156) (0.00162)
As a way of visualizing changes in the statewide CT vehicle fleet over
Observations 2104 2104 2104 1841 time, we mapped out the distribution of fuel type and body style in 2013
R-squared 0.451 0.592 0.5814 0.023
and 2020 (Fig. 5; Table S1). From Fig. 5, two things are apparent, the
Number of ZCTA 263 263 263 263
Unit FE YES YES NO YES first is that the share of fossil fuel vehicles declines from 2013 to 2020
Year FE NO YES YES NO across all body styles. Secondly, the share of alternative fuel vehicles
Clustered Errors NO NO YES YES increases quite a lot for passenger vans. In the context of Connecticut,
Standard errors in parentheses. such a result is interesting, as there are a limited number of ethanol (e.g.,
a
p < 0.1. E− 85) fueling stations in the state. Currently, there are 13 ethanol
b
p < 0.05. fueling stations that support alternative fuel vehicles (i.e., Flex-Fuel)
c
p < 0.01. (AFDC, 2023). Although the share of alternative fuel vehicles

7
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Fig. 3. Annual results of Moran’s I test for spatial autocorrelation, positive values indicate the share of EVs are spatial correlated. Results for 2015 to 2020 are
statistically significant.

Fig. 4. LISA cluster map showing the share of EVs in 2020. High-High clusters indicate a high share of EVs relative to the mean, conversely Low-Low indicates a low
share of EVs relative to the mean. Low-High and High-Low outliers indicating high or low values relative to neighbors. All values are relative to the mean and not
absolute values. Note that ZCTAs are not drawn from county boundaries and therefore do not perfectly align. The inclusion of counties is used to contextualize the
comparison of median household income and share of EVs.

8
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Table 4 (Yuksel et al., 2016).

Results from spatial-panel regression of share of EVs per 1000 LDVs in Con­ We identified the 10 most common vehicles on the road in Con­
necticut from 2013 to 2020. Our main variables of interest are the share of necticut by body style for 2013 and 2020 and compared them with the
traditional hybrid vehicles and number of charging stations. For list of variables 10 most common EVs on the road in 2020. Table 5 provides the list of 10
see equation (4) in section 4.2 and Table 1 in section 3 for motivation of each most common body styles for 2013 and 2020, their average emissions
variable.
reported in grams per mile, annual emissions, and the emissions asso­
(1) (2) (3) ciated with the 10 most common EV models. Values for emissions
Direct Model Type SAR SEM SDM (grams/mile) are a result of comparing side by side tailpipe and up­
Effects Dependent Electric Electric Electric stream emissions as reported by US Department of Energy fuel economy
Variable Vehicles (%) Vehicles (%) Vehicles (%) database. Our findings for EV emissions reflect a relatively “clean”
Hybrid 0.0712c 0.0294b 0.0235
electricity grid mix in ISO-NE with natural gas and nuclear providing
Vehicles (%)
(0.0143) (0.0126) (0.0145) most of the electricity in the region. We found that on average if any one
Charging 0.0048b 0.0057c 0.0056c of the 160 most common vehicles (10 per body style per year) on the
Stations road in Connecticut were replaced with an EV, from an annual basis,
(0.0014) (0.0014) (0.0014)
emissions would reduce by around 82.22% with the largest reductions
Indirect Effects
Hybrid 0.1183c 0.1160b 0.1602b resulting from replacing passenger vans with alternative electrified vans
Vehicles (%) (e.g., Tesla model Y or Ford Mustang Mach-E) (see Table 4). Between
(0.0235) (0.0268) (0.0694) 2013 and 2020, passenger vans had very little change in overall emis­
Charging 0.0080b 0.0227c 0.0216c sions suggesting a possible gap in increased efficiency of that vehicle
Stations
body style. It is important to note that this does not supplement a full life
(0.0024) (0.0056) (0.0056)
Total Effects cycle analysis and a more detailed statewide analysis should be con­
Hybrid 0.1895c 0.1454b 0.01837b ducted to fully capture changes in emissions because of widespread fleet
Vehicles (%) transition from ICE vehicles to EVs. Furthermore, because of data limi­
(0.0372) (0.0599) (0.0736)
tation on registration, this analysis looks at the whole fleet replacement
Charging 0.0128b 0.0284c 0.0273c
Stations
pattern, rather than individual choices. Therefore, we are unable to
(0.0038) (0.0069) (0.0069) capture movement assoicated with de-registered vehicles entering or
leaving Connecticut. However, this analysis serves as a starting point in
N 2104 2104 2104 grasping a holistic understanding of the impacts a clean energy transi­
Standard errors are in parentheses. tion will have on regional greenhouse gas emissions from within the
p < 0.1. transportation sector.
b
p < 0.05.
c
p < 0.01.
6. Discussion and conclusion

increased from 77,414 to 155,108 between 2013 and 2020 (100% in­ In 2012, there were only 11 EV models available for sale in the US
crease), most of these vehicles are not maximizing their fuel benefits, and at the start of 2013 availability of EV models remained low, only
with the exception of compressed natural gas and propane vehicles adding six models (AFDC, 2022). However, by 2019 the landscape began
which cannot run on gasoline. From 2013 to 2020 the total vehicles in to change with 45 EV models available for sale in the US (AFDC, 2022).
Connecticut increased from roughly 2.4 million to 2.8 million with the As an increasing number of auto manufactures began to realize the de­
largest value change occurring within the fossil fuel segment. However, mand and benefit of EVs, model availability will only continue to grow
when considering the percent change, EVs increased by over 2000% (Muratori et al., 2021). With a limited variety of vehicles available to
from 588 vehicles in 2013 to 12,399 vehicles in 2020. Although, in 2020 consumers between 2013 and 2020, we found that Connecticut is still in
EVs in Connecticut only represent a fraction of the vehicles on the road, the early stages of EV adoption and that the levels to which EVs are being
at less than 0.5%. adopted do not align with what policy makers are aiming for. Our results
Given the large increase in EVs overall (Gillingham et al., 2023), we indicate that of the factors considered in this analysis, the share of
find it important to evaluate that increase within vehicle body segments. hybrid vehicles had the strongest positive association with EV adoption
Doing so could allow for a better understanding of emission implications in Connecticut over the study period and across space (i.e., spatial
of an EV transition. When comparing body styles by fuel type we found spillovers). Similar results were presented in Liu et al. (2017) whereby
between 2013 and 2020 fossil fuel vehicles experienced a decline of they assessed neighbor effects of hybrid vehicles on subsequent hybrid
28.2%. Put another way, for each body style and fuel type class fossil vehicle adoption in Ohio. This finding points towards the potential that
fuel vehicles are declining albeit slowly. Interestingly, when examining when the share of one technology, in this case traditional hybrid vehi­
which fuel type is replacing fossil fuels for each body style, we found that cles, increases, the shares of EVs will also increase: either through shear
EV hatchbacks contribute to the largest fuel shift. Across the board presence on the road or by owners trading in “older” technologies for
however, alternative fuel vehicles (i.e., E− 85, CNG, and propane) are newer ones (i.e., trading in a Toyota Prius for a Plug-in Hybrid Toyota
the fuel types most commonly replacing fossil fuels, limiting the po­ Prius or Chevy Bolt). Such spillover effects have been observed across
tential emission savings of fleet wide transitions (Fig. 5; Table S1). other renewable energy technologies as well, for example rooftop solar
EVs have a goal of serving as an alternative transportation technol­ (Graziano and Gillingham, 2015). However, we found that the spatial
ogy aimed at reducing greenhouse gas emissions over the long run. spillovers of EV adoption in Connecticut are not uniform, as visualized in
However, from the perspective of a sustainable energy transition such Fig. 6. Connecticut is just behind New York as one of the most unequal
claims must be tested. Several studies from a well-to-wheel analysis states, in terms of income, with a Gini index4 of 0.5 as of 2020 (US
show that EVs reduce greenhouse gas emissions anywhere from 10% to Census Bureau, 2021). For example, some neighboring ZCTAs with very
41% on average compared to similar ICE vehicles (Muratori et al., different median household income values could contribute to the lack
2021). One of the more important factors when considering the levels of
greenhouse gas emission reductions from EVs compared to ICE vehicles
is the electricity mix of the region, this is particularly true in the US 4
The Gini Index is a measure from 0 to 1 summarizing income inequality
where electricity mixes vary by region (Reichmuth, 2020). Furthermore, across the entire income distribution for a given region or spatial unit. See US
these values can change based on differing climate and weather patterns Census Bureau (2020) for more information.

9
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Fig. 5. Bookend vehicle composition in Connecticut (2013)(A) and 2020(B) by body style and fuel type, percent of vehicle body style fleet. Also see Table S1
in appendix.

Table 5
Assessment of the 10 most common vehicles on the road in Connecticut by body style for data years 2013 and 2020. EV emissions are the average emissions from the 10
most common EVs for 2023.
Body Style Year Average emissions Average annual emissions Relative change in average emissions (g/mi) EV emissions (g/ EV transition emissions
(g/mi) (g/mi/yr.) (%Δ 2013–2020) mi) (g/mi) (%)

Convertible 2013 509.5 7,642,500

Convertible 2020 510.8 7,662,000 0.26% 76.25 − 85.07%
Coupe 2013 377.5 5,662,500
Coupe 2020 373.4 5,601,000 − 1.09% 76.25 − 79.58%
Hatchback 2013 289.7 4,345,500
Hatchback 2020 242.1 3,631,500 − 16.43% 76.25 − 68.50%
Passenger 2013 574.4 8,616,000
Van
Passenger 2020 565.2 8,478,000 − 1.60% 76.25 − 86.51%
Van
Pickup Truck 2013 698.9 10,500,000
Pickup Truck 2020 605 9,075,000 − 13.44% 145 − 76.03%
SUV 2013 627.8 9,417,000
SUV 2020 474.4 7,116,000 − 24.43% 76.25 − 83.93%
Sedan 2013 459.4 6,891,000
Sedan 2020 388.1 5,821,500 − 15.52% 76.25 − 80.35%
Wagon 2013 503.3 7,549,500
Wagon 2020 437.6 6,564,000 − 13.05% 76.25 − 82.58%
Average 477.3 7,160,813 84.84 − 82.22%

of EV adoption across the state, writ large (Fig. 6). and work modality (e.g., working from home) to all be positively asso­
Furthermore, we found housing density, number of charging sta­ ciated with EV adoption. Working from home substantially increased as
tions, long duration commuters, median household income ($10,000), a result of covid-19 lockdowns and, for some industries, has continued

10
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Fig. 6. Bivariate map showing the share of EVs and median household income in Connecticut as of 2020. Median household income data is from the American
Community Survey 5-year average 2020 table. Note that ZCTAs are not drawn from county boundaries and therefore do not perfectly align. The inclusion of counties
is used to contextualize the comparison of median household income and share of EVs.

post-covid. Given this new relationship with employment and replacement dynamics within the EV transition (i.e., identifying which
commuting, there could be a larger proportion of consumers who vehicles, by body style and or fuel type, are replaced with EVs) (Xing
consider switching to an EV without much consideration for any et al., 2021). By understanding the replacement dynamics of a vehicle
perceived impacts to their commuting times, potentially limiting the fleet researchers can provide a baseline for evaluating the sustainability
influence of “range anxiety”5 on EV ownership. Despite only a subset of of transitioning the vehicle fleet from fossil fuels to electricity and the
the economy having an ability to work from home, our results offer subsequent impact on global emissions. Our results suggest under a
potential pathways for identifying locations of future EV adoption based rapidly electrifying vehicle fleet, EVs have the potential to substantially
off work modalities. Conversely, we found active age groups, females, curb emissions from the transportation sector. However, this depends on
service sector workers, and households with more than one vehicle to be the energy profile of the electricity grid (Li et al., 2016). New England,
negatively associated with EV adoption in Connecticut over the study like most of the US, has high shares of natural gas thereby limiting the
period. Previous research identifying the potential for a gender gap in resultant release of greenhouse gas emissions compared to regions that
EV ownership similarly found that women are less likely to own an EV are highly reliant on more polluting energy sources like coal. Evaluating
compared to men, although this is not indicative of their attitudes to­ vehicle emissions by body style reveals some interesting avenues for
wards environmental issues (Sovacool et al., 2019). Although identi­ future decarbonization. However, Connecticut, much like other regions
fying factors associated with EV adoption provides insights for (Stokes and Breetz, 2018), continues to advance and work towards
stakeholders, holistically evaluating the sustainability of an EV transi­ increasing shares of renewable energy thereby minimizing the envi­
tion by assessing impacts on greenhouse gas emissions provides yet ronmental impact of EVs from a “refueling” perspective. Passenger vans
another layer of understanding in a rapidly changing mobility (e.g., Chrysler Pacifica or Honda Odyssey) with 1.6%, coupes (e.g.,
landscape. two-door cars) with 1%, and convertibles with 0.26% experienced the
A sector-wide transition to EVs has the potential to result in various smallest change in emissions from 2013 to 2020. Vehicles within these
impacts to the climate, the local environment, and transportation energy body style segments pose the greatest potential for electrification owing
burden (i.e., percent of income spent on vehicle fuels see Vega-Perkins to there minimal advancements in emission reductions over time.
et al., 2023), any impacts resulting from a transition in the short run will Currently, the Chrysler Pacifica ($52,495) is the only EVs available on
be dependent on the dynamics between vehicle charging and electricity the market for the passenger van segment and with the current average
production along with incentive programs (Gan et al., 2021; Vega-­ price of passenger vans at $47,958, the Chrysler Pacifica has a nearly
Perkins et al., 2023). Recent studies have begun investigating fleet 10% mark up compared to the average (Cox Auto, 2024). While there
are more available EVs for coupe body style vehicles, those available
tend to be high end vehicles except for Hyundai which offers two models
5
Range anxiety is the fear of not knowing if there is enough “fuel” in the of reasonable value. Therefore, the average consumer, particular fam­
battery to get someone from point A to point B given a lack of suitable charging ilies, have little choice when deciding if their next vehicle will be an EV.
infrastructure. Importantly, we outline some limitations of this work. Decisions and

11
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

factors that influence the purchasing of a vehicle are extremely complex the work reported in this paper.
and sometimes may go against conventional thinking (e.g., the high
penetration of American made vehicle ownership in Michigan despite Data availability
poor reliability (Barber and Darrough, 1996). Thus, while we provide
context for the spatial diffusion of EVs over time, we cannot be certain of The authors do not have permission to share data.
the characteristics of the individual consumers. Additionally, when
assessing the replacement dynamics of the vehicle fleet in Connecticut Acknowledgements
from 2013 to 2020, we are unable to determine if a vehicle was sold or
moved between ZCTAs. The ability to obtain transactional data over the The authors would like to thank Dr. Fred Carstensen and Dr.
study period would greatly aid in assessing environmental implications Chuanrong Zhang for their support in securing the necessary datasets for
of an EV transition. Furthermore, although we leverage lifecycle analysis this analysis. This work is based upon support through the Team-TERRA
approaches our evaluation did not follow conventional lifecycle meth­ program at the University of Connecticut, which is supported by the
odologies and would benefit from further, separate analysis. Given these National Science Foundation under Grant DGE-2022036. This work is
limitations, we leave the door open for future investigations of EVs from also based upon support through the Institute of the Environment Center
a regional perspective. for Environmental Sciences and Engineering at the University of Con­
A possible future area of research could include the construction of a necticut and the Energy and Environment Specialty Group of the
social network, thereby providing insights into understanding how EV American Association of Geographers.
adoption occurs over time and space, at the individual level. Previous
work has studied the effects of such social networks and found they do Appendix A. Supplementary data
affect individual perceptions of EV adoption (Axsen and Kurani, 2012).
The increasing use of social media either directly or via interest groups Supplementary data to this article can be found online at https://doi.
(e.g., the EV Club of Connecticut) has opened the door for large scale org/10.1016/j.jclepro.2024.142574.
evaluations of the public’s perceptions of EVs. Along a similar vein,
future research could involve a survey of individuals on their percep­ References
tions of EVs, willingness to purchase, and experience with EV ownership
either via social networks or geographic proximity to neighbor owner­ AFDC, 2022. U.S. Plug-In electric vehicle sales by model. https://afdc.energy.gov/d
ata/10567.
ship. As of now we were able to extend the previous literature on the AFDC, 2023. Alternative fueling station locator. https://afdc.energy.gov/stations#/fin
spatial diffusion of EVs from a regional perspective in addition to d/nearest.
providing ground for holistically analyzing the sustainability of an EV Anselin, L., 1995. Local indicators of spatial association-LISA. Geogr. Anal. 27, 93–115.
https://doi.org/10.1111/j.1538-4632.1995.tb00338.x.
transition and what it means for climate and energy goals. Axsen, J., Goldberg, S., Bailey, J., 2016. How might potential future plug-in electric
Our results have relevance to policy makers and transportation vehicle buyers differ from current “Pioneer” owners? Transportation Research Part D
planners. The ability to identify regions experiencing hybrid vehicle 47, 357–370. https://doi.org/10.1016/j.trd.2016.05.015.
Axsen, J., Kurani, K.S., 2012. Interpersonal influence within car buyers’ social networks:
adoption will likely follow with EV adoption and thus possibly require applying five perspectives to plug-in hybrid vehicle drivers. Environ. Plann. 44,
additional charging infrastructure. We can extend these insights to the 1047–1065. https://doi.org/10.1068/a43221x.
power sector for capacity and distribution planning so as not to fall Balta-Ozkan, N., Yildirim, J., Connor, P.M., 2015. Regional distribution of photovoltaic
deployment in the UK and its determinants: a spatial econometric approach. Energy
behind on what will become increasingly critical infrastructure in an age
Econ. 51, 417–429. https://doi.org/10.1016/j.eneco.2015.08.003.
of transportation electrification, this is particularly important given the Banister, D., 2008. The sustainable mobility paradigm. Transport Pol. 15, 73–80. https://
variability of charging patterns of EVs (Kapustin and Grushevenko, doi.org/10.1016/j.tranpol.2007.10.005.
2020). Moreover, policy makers can draw from these results to better Barber, B.M., Darrough, M.N., 1996. Product reliability and firm value: the experience of
American and Japanese automakers, 1973-1992. J. Polit. Econ. 104, 1084–1099.
tailor incentive programs designed to leverage spatial spillovers on Berkeley, N., Bailey, D., Jones, A., Jarvis, D., 2017. Assessing the transition towards
subsequent vehicle technology adoption. Regions suffering from poor air battery electric vehicles: a multi-level perspective on drivers of, and barriers to, take
quality have the potential to greatly reduce emissions from the trans­ up. Transport. Res. Pol. Pract. 106, 320–332. https://doi.org/10.1016/j.
tra.2017.10.004.
portation sector provided a large-scale transition to EVs. Of course, this Biresselioglu, M.E., Kaplan, M.D., Yilmaz, B.K., 2018. Electric mobility in Europe: a
is largely dependent on the fuel profile of the electricity sector. An comprehensive review of motivators and barriers in decision making. Transport. Res.
increasing number of countries, regions, and states have outlined Part A 109, 1–13. https://doi.org/10.1016/j.tra.2018.01.017.
Broadbent, G.H., Drozdzewski, D., Metternicht, G., 2017. Electric vehicle adoption: an
renewable energy goals and targets aimed at reducing shares of fossil analysis of best practice and pitfalls for policy making from experiences of Europe
fuels within energy profiles. As the US continues to experience a wave of and the US. Geographic Compass 12, 12358. https://doi.org/10.1111/gec3.12358.
incentives designed to decarbonize personal transportation and the Cabral, J. de A., Legey, L.F.L., Freitas, CabralM.V.de, 2017. Electricity consumption
forecasting in Brazil: a spatial econometrics approach. Energy 126, 124–131.
electricity grid, EVs and the impact they could have on emissions and the https://doi.org/10.1016/j.energy.2017.03.005.
environment continues to be of interest for global sustainability efforts. Campbell, A.R., Ryley, T., Thring, R., 2012. Identifying the early adopters of alternative
fuel vehicles: a case study of Birmingham, United Kingdom. Transport. Res. Pol.
Pract. 46, 1318–1327. https://doi.org/10.1016/j.tra.2012.05.004.
CRediT authorship contribution statement
Casals, L.C., Martinez-Laserna, E., García, B.A., Nieto, N., 2016. Sustainability analysis of
the electric vehicle use in Europe for CO2 emissions reduction. J. Clean. Prod. 127,
Adam Gallaher: Writing – review & editing, Writing – original draft, 425–437. https://doi.org/10.1016/j.jclepro.2016.03.120.
Software, Methodology, Investigation, Funding acquisition, Conceptu­ Choma, E.F., Evans, J.S., Hammitt, J.K., Gómez-Ibáñez, J.A., Spengler, J.D., 2020.
Assessing the health impacts of electric vehicles through air pollution in the United
alization. Marcello Graziano: Writing – review & editing, Writing – States. Environ. Int. 144, 106015 https://doi.org/10.1016/j.envint.2020.106015.
original draft, Resources, Investigation, Conceptualization. Carol Chun, Y., Griffith, D.A., 2013. Spatial Statistics & Geostatistics: Theory and Applications
Atkinson-Palombo: Writing – review & editing, Writing – original for Geographic Information Science & Technology, SAGE Advances in Geographic
Information Science and Technology. SAGE, Los Angeles.
draft, Investigation, Funding acquisition. Lyle Scruggs: Writing – re­ Cliff, A.D., Ord, K., 1970. Spatial autocorrelation: a review of existing and new measures
view & editing, Writing – original draft, Investigation, Funding with applications. Econ. Geogr. 46, 269. https://doi.org/10.2307/143144.
acquisition. Coffman, M., Bernstein, P., Wee, S., 2017. Electric vehicles revisited: a review of factors
that affect adoption. Transport Rev. 37, 79–93. https://doi.org/10.1080/
01441647.2016.1217282.
Declaration of competing interest Cox Auto, 2024. Data tables for january 2024 kelly blue book average transaction prices
report. https://www.coxautoinc.com/wp-content/uploads/2024/02/January
-2024-Kelley-Blue-Book-Average-Transaction-Price-tables.pdf.
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influence

12
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Creutzig, F., Jochem, P., Edelenbosch, O.Y., Mattauch, L., van Vuuren, D.P., Lan, H., Cheng, B., Gou, Z., Yu, R., 2020. An evaluation of feed-in tariffs for promoting
McCollum, D., Minx, J., 2015. Transport: a roadblock to climate change mitigation. household solar energy adoption in Southeast Queensland, Australia. Sustain. Cities
Science 350 (6263), 911–912. https://doi.org/10.1126/science.aac8033. Soc. 53, 101942 https://doi.org/10.1016/j.scs.2019.101942.
Cunningham, S., 2021. Panel data. In: Cunningham, S. (Ed.), Causal Inference: the CT DEEP, 2020. Electric Vehicle Roadmap for Connecticut: A policy framework to
Mixtape. Yale University Press, New Haven, CT, pp. 386–405. accelerate electric vehicle adoption. https://www.dpuc.state.ct.us/DEEPEnergy.
Davis, S.J., et al., 2018. Net-zero emissions energy systems. Science 360 (1419). https:// nsf/c6c6d525f7cdd1168525797d0047c5bf/f7ed4932eec438d0852585520001c8
doi.org/10.1126/science.aas9793. 1b/$FILE/EV%20Roadmap%20for%20Connecticut.pdf.
Delmas, M.A., Kahn, M.E., Locke, S.L., 2017. The private and social consequences of Lebrouhi, B.E., Khattari, Y., Lamrani, B., Maaroufi, M., Zeraouli, Y., Kousksou, T., 2021.
purchasing an electric vehicle and solar panels: evidence from California. Res. Econ. Key Challenges for a Large-Scale Development of Battery Electric Vehicles: A
71 (2), 225–235. https://doi.org/10.1016/j.rie.2016.12.002. Comprehensive Review. Journal of Energy Storage 44 (B), 130273. doi: 10.1016/j.
Dharshing, S., 2017. Household dynamics of technology adoption: a spatial econometric est.2021.103273.
analysis of residential solar photovoltaic (PV) systems in Germany. Energy Res. LeSage, J.P., Pace, R.K., 2021. Interpreting spatial econometric models. In: Fischer, M.
Social Sci. 23, 113–124. https://doi.org/10.1016/j.erss.2016.10.012. M., Nijkamp, P. (Eds.), Handbook of Regional Science. Springer, Berlin, Heidelberg.
DOE, 2023. Fuel Economy Data. https://www.fueleconomy.gov/feg/download.shtml. https://doi.org/10.1007/978-3-662-60723-7_91.
DOE, DOT, EPA, HUD, 2022. The U.S. National blueprint for transportation LeSage, J., Pace, R.K., 2009. Introduction to Spatial Econometrics, 0 ed. Chapman and
decarbonization: a joint strategy to transform transportation. https://www.energy. Hall/CRC. https://doi.org/10.1201/9781420064254.
gov/sites/default/files/2023-01/the-us-national-blueprint-for-transportation-decar LeSage, J.P., 2008. An introduction to spatial econometrics. rei 19–44. https://doi.org/
bonization.pdf. 10.4000/rei.3887.
Elhorst, J.P., 2014. Spatial panel data models. In: Spatial Econometrics, SpringerBriefs in Li, W., Long, R., Chen, H., Geng, J., 2017. A review of factors influencing consumer
Regional Science. Springer, Berlin Heidelberg, Berlin, Heidelberg, pp. 37–93. intentions to adopt battery electric vehicles. Renew. Sustain. Energy Rev. 78,
https://doi.org/10.1007/978-3-642-40340-8_3. 318–328. https://doi.org/10.1016/j.rser.2017.04.076.
EPA, 2022. The emissions & generation resource integrated database: eGRID technical Li, N., Chen, J.-P., Tsai, I.-C., He, Q., Chi, S.-Y., Lin, Y.-C., Fu, T.-M., 2016. Potential
guide with year 2022 data. https://www.epa.gov/system/files/documents/ impacts of electric vehicles on air quality in Taiwan. Sci. Total Environ. 566–567,
2024-01/egrid2022_technical_guide.pdf. 919–928. https://doi.org/10.1016/j.scitotenv.2016.05.105.
Erhardt, G.D., Hoque, J.M., Goyal, V., Berrebi, S., Brakewood, C., Watkins, K.E., 2022. Liu, L., Miller, H.J., Scheff, J., 2020. The impacts of COVID-19 pandemic on public
Why has public transit ridership declined in the United States. Transport. Res. Part A transit demand in the United States. PLoS One 15 (11). https://doi.org/10.1371/
161, 68–87. https://doi.org/10.1016/j.tra.2022.04.006. journal.pone.0242476.
Fevang, E., Figenbaum, E., Fridstrøm, L., Halse, A.H., Hauge, K.E., Johansen, B.G., Liu, X., Roberts, M.C., Sioshansi, R., 2017. Spatial effects on hybrid electric vehicle
Raaum, O., 2021. Who goes electric? The anatomy of electric car ownership in adoption. Transport. Res. Transport Environ. 52, 85–97. https://doi.org/10.1016/j.
Norway. Transport. Res. Transport Environ. 92, 102727 https://doi.org/10.1016/j. trd.2017.02.014.
trd.2021.102727. Loa, P., Hossain, S., Liu, Y., Habib, K.N., 2022. How has the COVID-19 pandemic affected
Gallaher, A., Graziano, M., Fiaschetti, M., 2021. Legacy and shockwaves: a spatial the use of ride-sourcing services? An empirical evidence-based investigation for the
analysis of strengthening resilience of the power grid in Connecticut. Energy Pol. Greater Toronto Area. Transport. Res. Part A 155, 46–62. https://doi.org/10.1016/j.
159, 112582 https://doi.org/10.1016/j.enpol.2021.112582. tra.2021.11.013.
Galvin, R., 2022. Are electric vehicles getting too big and heavy? Modeling future vehicle Lyu, X., 2023. Are electric cars and solar panels complements? Journal of the Association
journeying demand on a decarbonized US electricity grid. Energy Pol. 161, 112746 of Environmental and Resource Economists 10 (4), 1019–1057. https://doi.org/
https://doi.org/10.1016/j.enpol.2022.112746. 10.1086/723494.
Gan, Y., Wang, M., Lu, Z., Kelly, J., 2021. Taking into account greenhouse gas emissions Melliger, M.A., van Vliet, O.P.R., Liimatainen, H., 2018. Anxiety vs reality – sufficiency
of electric vehicles for transportation de-carbonization. Energy Pol. 155, 112353 of battery electric vehicle range in Switzerland and Finland. Transport. Res. Part D
https://doi.org/10.1016/j.enpol.2021.112353. 65, 101–115. https://doi.org/10.1016/j.trd.2018.08.011.
Gärling, A., Thøgersen, J., 2001. Marketing of electric vehicles. Bus. Strat. Environ. 10, Morton, C., Anable, J., Yeboah, G., Cottrill, C., 2018. The spatial pattern of demand in
53–65. https://doi.org/10.1002/1099-0836(200101/02)10:1<53::AID- the early market for electric vehicles: evidence from the United Kingdom.
BSE270>3.0.CO;2-E. J. Transport Geogr. 72, 119–130. https://doi.org/10.1016/j.jtrangeo.2018.08.020.
Geels, F.W., 2002. Technological transitions as evolutionary reconfiguration processes: a Mukherjee, S.C., Ryan, L., 2020. Factors influencing early battery electric vehicle
multi-level perspective and a case-study. Res. Pol. 31, 1257–1274. https://doi.org/ adoption in Ireland. Renew. Sustain. Energy Rev. 118, 109504 https://doi.org/
10.1016/S0048-7333(02)00062-8. 10.1016/j.rser.2019.109504.
Gillingham, K.T., van Benthem, A.A., Weber, S., 2023. Has consumer acceptance of Muratori, M., Alexander, M., Arent, D., Bazilian, M., Cazzola, P., Dede, E.M., Farrell, J.,
electric vehicles been increasing? Evidence from microdata on every new vehicle Gearhart, C., Greene, D., Jenn, A., Keyser, M., Lipman, T., Narumanchi, S.,
sale in the United States. AEA Papers and Proceedings 113, 329–335. https://doi. Pesaran, A., Sioshansi, R., Suomalainen, E., Tal, G., Walkowicz, K., Ward, J., 2021.
org/10.1257/pandp.20231065. The rise of electric vehicles—2020 status and future expectations. Prog. Energy 3,
Gnann, T., Plötz, P., 2015. A review of combined models for market diffusion of 022002. https://doi.org/10.1088/2516-1083/abe0ad.
alternative fuel vehicles and their refueling infrastructure. Renew. Sustain. Energy Namdeo, A., Tiwary, A., Dziurla, R., 2014. Spatial planning of public charging points
Rev. 47, 783–793. https://doi.org/10.1016/j.rser.2015.03.022. using multi-dimensional analysis of early adopters of electric vehicles for a city
Graziano, M., Gillingham, K., 2015. Spatial patterns of solar photovoltaic system region. Technol. Forecast. Soc. Change 89, 188–200. https://doi.org/10.1016/j.
adoption: the influence of neighbors and the built environment. J. Econ. Geogr. 15, techfore.2013.08.032.
815–839. https://doi.org/10.1093/jeg/lbu036. Nayum, A., Klöckner, C.A., Mehmetoglu, M., 2016. Comparison of socio-psychological
Graziano, M., Fiaschetti, M., Atkinson-Palombo, C., 2019. Peer effects in the adoption of characteristics of conventional and battery electric car buyers. Travel Behaviour and
solar energy technologies in the United States: an urban case study. Energy Res. Society 3, 8–20. https://doi.org/10.1016/j.tbs.2015.03.005.
Social Sci. 48, 75–84. https://doi.org/10.1016/j.erss.2018.09.002. Plötz, P., Schneider, U., Globisch, J., Dütschke, E., 2014. Who will buy electric vehicles?
Graziano, M., Gunther, P., Gallaher, A., Carstensen, F.V., Becker, B., 2020. The wider Identifying early adopters in Germany. Transport. Res. Part A 67, 96–109. https://
regional benefits of power grids improved resilience through tree-trimming doi.org/10.1016/j.tra.2014.06.006.
operations evidences from Connecticut, USA. Energy Pol. 138, 111293 https://doi. Proto, J., 2022. Connecticut population change: 2010 to 2020 (No. 2022- R- 0009).
org/10.1016/j.enpol.2020.111293. Office of Legislative Research.
Griffith, D.A., 2010. The Moran coefficient for non-normal data. J. Stat. Plann. Inference Reichmuth, D., 2020. Are electric vehicles really better for the climate? Yes. Here’s why.
140, 2980–2990. https://doi.org/10.1016/j.jspi.2010.03.045. Union of Concerned Scientists The Equation. https://blog.ucsusa.org/dave-reichmu
IEA, 2023a. CO2 emissions in 2022. https://www.iea.org/reports/co2-emissions-in th/are-electric-vehicles-really-better-for-the-climate-yes-heres-why/.
-2022. Rockström, J., Graffney, O., Rogelj, J., Meinshausen, M., Nakicenovic, N.,
IEA, 2023b. Electric vehicles. https://www.iea. Schellnhuber, H.J., 2017.A. Roadmap for Rapid Decarbonization: Emissions
org/energy-system/transport/electric-vehicles#tracking. inevitably approach zero with a "carbon law. Science 355 (6331), 1269–1271.
IEA, 2023c. Global EV data explorer. https://www.iea.org/data-and-statistics/data-tools https://doi.org/10.1126/science.aah3443.
/global-ev-data-explorer. Rode, J., 2024. Solar photovoltaics and battery electric vehicles. available at: SSRN:
Imai, K., Kim, I.S., 2019. When should we use unit fixed effects regression models for Work. Pap. https://ssrn.com/abstract=4725095.
causal inference with longitudinal data? Am. J. Polit. Sci. 63 (2), 467–490. https:// Ryghaug, M., Skjølsvold, T.M., 2022. How policies and actor strategies affect electric
doi.org/10.1111/ajps.12417. vehicle diffusion and wider sustainability transitions. Proc. Natl. Acad. Sci. USA 120
Kapustin, N.O., Grushevenko, D.A., 2020. Long-term electric vehicles outlook and their (47). https://doi.org/10.1073/pnas.2207888119.
potential impact on electric grid. Energy Pol. 137, 111103 https://doi.org/10.1016/ Schulz, F., Rode, J., 2022. Public charging infrastructure and electric vehicles in Norway.
j.enpol.2019.111103. Energy Pol. 160, 112660 https://doi.org/10.1016/j.enpol.2021.112660.
Kester, J., Sovacool, B.K., Noel, L., Zarazua de Rubens, G., 2020. Rethinking the spatiality She, Z.Y., Sun, Q., Ma, J.J., Xie, B.C., 2017. What are the barriers to widespread adoption
of Nordic electric vehicles and their popularity in urban environments: moving of battery electric vehicles? A survey of public perception in Tianjin, China.
beyond the city. J. Transport Geogr. 82, 102557 https://doi.org/10.1016/j. Transport Pol. 56, 29–40. https://doi.org/10.1016/j.tranpol.2017.03.001.
jtrangeo.2019.102557. Shepherd, S., Bonsall, P., Harrison, G., 2012. Factors affecting future demand for electric
Kumar, R.R., Alok, K., 2020. Adoption of electric vehicle: a literature review and vehicles: a model based study. Transport Pol. 20, 62–74. https://doi.org/10.1016/j.
prospects for sustainability. J. Clean. Prod. 253, 119911 https://doi.org/10.1016/j. tranpol.2011.12.006.
jclepro.2019.119911. Sierzchula, W., Bakker, S., Maat, K., van Wee, B., 2014. The influence of financial
incentives and other socio-economic factors on electric vehicle adoption. Energy Pol.
68, 183–194. https://doi.org/10.1016/j.enpol.2014.01.043.

13
A. Gallaher et al. Journal of Cleaner Production 459 (2024) 142574

Sovacool, B.K., Kester, J., Noel, L., Zarazua de Rubens, G., 2018. The demographics of Truffer, B., Coenen, L., 2012. Environmental innovation and sustainability transitions in
decarbonizing transport: the influence of gender, education, occupation, age, and regional studies. Reg. Stud. 46 (1), 1–21. https://doi.org/10.1080/
household size on electric mobility preferences in the Nordic region. Global Environ. 00343404.2012.646164.
Change 52, 86–100. https://doi.org/10.1016/j.gloenvcha.2018.06.008. US Census Bureau, 2020. Gini index. https://www.census.gov/topics/income-poverty/i
Sovacool, B.K., Kester, J., Noel, L., Zarazua de Rubens, G., 2019. Are electric vehicles ncome-inequality/about/metrics/gini-index.html.
masculinized? Gender, identity, and environmental values in Nordic transport US Census Bureau, 2021. 2016-2020 American Community Survey 5-year Average.
practices and vehicle-to-grid (V2G) preferences. Transport. Res. Part D 72, 187–202. US DOT, 2022. Average Annual Miles Per Driver by Age Group.
https://doi.org/10.1016/j.trd.2019.04.013. Vega-Perkins, J., Newell, J.P., Keoleian, G., 2023. Mapping electric vehicle impacts:
Stokes, L.C., Breetz, H.L., 2018. Politics in the U.S. energy transition: case studies of greenhouse gas emissions, fuel costs, and energy justice in the United States.
solar, wind, biofuels and electric vehicles policy. Energy Pol. 113, 76–86. https:// Environ. Res. Lett. 18, 014027 https://doi.org/10.1088/1748-9326/aca4e6.
doi.org/10.1016/j.enpol.2017.10.057. White, House, 2023. Building a Clean Energy Economy: A Guidebook to the Inflation
Tamayao, M.A.M., Michalek, J.J., Hendrickson, C., Azevedo, I.M.L., 2015. Regional Reduction Act’s Investments in Clean Energy and Climate Action (No. 2).
variability and uncertainty of electric vehicle life cycle CO2 emissions across the White, House, 2021. Fact Sheet: President Biden Announces Steps to Drive American
United States. Environ. Sci. Technol. 49, 8844–8855. https://doi.org/10.1021/acs. Leadership Forward on Clean Cars and Trucks.
est.5b00815. Wu, X., Freese, D., Cabrera, A., Kitch, W.A., 2015. Electric vehicles’ energy consumption
Teixeira, A.C.R., Sodré, J.R., 2018. Impacts of replacement of engine powered vehicles measurement and estimation. Transport. Res. Transport Environ. 34, 52–67. https://
by electric vehicles on energy consumption and CO 2 emissions. Transport. Res. doi.org/10.1016/j.trd.2014.10.007.
Transport Environ. 59, 375–384. https://doi.org/10.1016/j.trd.2018.01.004. Xing, J., Leard, B., Li, S., 2021. What does an electric vehicle replace? J. Environ. Econ.
Tobler, W.R., 1970. A computer movie simulating urban growth in the Detroit region. Manag. 107, 102432 https://doi.org/10.1016/j/jeem.2021.102432.
Econ. Geogr. 46, 234. https://doi.org/10.2307/143141. Yuksel, T., Tamayao, M.A.M., Hendrickson, C., Azevedo, I.M.L., Michalek, J.J., 2016.
Tran, M., Banister, D., Bishop, J.D.K., McCulloch, M.D., 2013. Simulating early adoption Effect of regional grid mix, driving patterns and climate on the comparative carbon
of alternative fuel vehicles for sustainability. Technol. Forecast. Soc. Change 80, footprint of gasoline and plug-in electric vehicles in the United States. Environ. Res.
865–875. https://doi.org/10.1016/j.techfore.2012.09.009. Lett. 11, 044007 https://doi.org/10.1088/1748-9326/11/4/044007.

14