Content uploaded by Sergey Naumov
Author content
All content in this area was uploaded by Sergey Naumov on Nov 03, 2022
Content may be subject to copyright.
RESEARCH ARTICLE
Accelerating vehicle fleet turnover to achieve sustainable
mobility goals
Sergey Naumov
1
| David R. Keith
2
| John D. Sterman
2
1
Smeal College of Business,
The Pennsylvania State University,
State College, Pennsylvania, USA
2
MIT Sloan School of Management,
Cambridge, Massachusetts, USA
Correspondence
Sergey Naumov, Smeal College of
Business, The Pennsylvania State
University, State College, PA, USA.
Email: snaumov@psu.edu
Handling Editors: Merieke Stevens,
David R. Keith, Jose Holguin-Veras
Abstract
Achieving societal climate goals requires rapid reductions in greenhouse gas
(GHG) emissions from transportation. Recent efforts by policymakers have
focused on increasing consumer adoption of electric vehicles (EVs). Neverthe-
less, EV sales remain low. Worse, even if EV market share jumped dramati-
cally, it would take decades to replace the existing vehicle fleet, during which
time vehicle GHG emissions would continue, worsening climate change. Con-
sequently, some argue for policies to accelerate the retirement of inefficient
fossil-powered vehicles through “cash-for-clunkers”(C4C) programs. We
examine C4C policies through a behavioral model of vehicle fleet turnover and
EV market development in the United States. We find C4C policies can sub-
stantially reduce vehicle fleet emissions at reasonable cost per tonne of CO
2
.
To meet emissions reductions goals, C4C policies should apply only when con-
sumers replace their fossil-powered vehicles with EVs. C4C policies incentiviz-
ing EVs accelerate cost reductions through scale economies, charging
infrastructure deployment, model variety, and consumer awareness, boosting
EV adoption beyond the direct effect of vehicle replacement. The result is a
substantial synergy amplifying the impact of C4C and lowering unit cost of
emissions reductions. C4C is further amplified when deployed together with
complementary policies promoting renewable electricity production and a gas
tax or carbon price.
KEYWORDS
accelerated vehicle retirement, automobile manufacturing, public policy, sustainable
mobility
Highlights
•Promoting sales of electric vehicles will not be sufficient to meet 2050 cli-
mate goals due to slow turnover of the existing vehicle fleet.
•Incentivizing the early retirement of gasoline vehicles (“cash-for-clunkers”)
and replacing them with new electric vehicles can achieve greater emissions
reduction at reasonable cost.
Received: 3 October 2020 Revised: 25 September 2021 Accepted: 20 January 2022
DOI: 10.1002/joom.1173
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2022 The Authors. Journal of Operations Management published by Wiley Periodicals LLC on behalf of Association for Supply Chain Management, Inc.
J Oper Manag. 2022;1–31. wileyonlinelibrary.com/journal/joom 1
•The environmental benefits of cash-for-clunkers can be enhanced further
through complementary policies such a carbon price and faster deca-
rbonization of the electric grid.
1|INTRODUCTION
Reducing greenhouse gas (GHG) emissions to achieve cli-
mate goals and limit global warming to no more than
2C requires effective policies addressing the most pollut-
ing sectors of the world economy. However, global GHG
emissions are not on track to achieve these goals. The
United Nations Environmental Programme finds that the
world is failing to meet the goals of the Paris Agreement
(UNEP, 2020). Significantly, faster and deeper emissions
cuts are needed (Farmer et al., 2019) throughout the
economy. In April 2021, the United States announced its
new Nationally Determined Contribution (NDC) under
the Paris Agreement, pledging to reduce “its net green-
house gas emissions by 50–52 percent below 2005 levels
in 2030”(UNFCCC, 2021). Rapid reductions in carbon
emissions from transportation are crucial to achieving
these goals: the transportation sector is responsible for
29% of United States GHG emissions as of 2019, and
58% of these emissions arise from the roughly 250 mil-
lion light-duty vehicles (cars, SUVs, and pickup trucks)
on the roads today (U.S. EPA, 2019a).
Most efforts by policymakers and automakers have so
far focused on increasing the share of new vehicle sales
going to low- or zero-tailpipe emissions vehicles, for
example, by mandating the minimum fuel economy of
new vehicles through the corporate average fuel economy
(CAFE) standards (NHTSA, 2021a). The transportation
section of the new U.S. NDC calls for
tailpipe emissions and efficiency standards;
incentives for zero emission personal vehi-
cles; funding for charging infrastructure to
support multi-unit dwellings, public charg-
ing, and long-distance travel; and research,
development, demonstration, and deploy-
ment efforts to support advances in very low
carbon new-generation renewable fuels
(UNFCCC, 2021).
In addition to direct consumer subsidies, cost-
effective policies might involve government investment
in charging infrastructure or providing owner benefits,
for example, special license plates with parking or driving
privileges (Li et al., 2021).
Many other nations have adopted similar policies
designed to increase the market share of electric vehicles
(EVs). The most successful, Norway, achieved 64.5% EV
market share in 2021 (Elbil, 2021a), the result of aggres-
sive subsidies including zero VAT and purchase taxes on
EVs, charge point deployment, and abundant, inexpen-
sive hydropower (averaging approximately US$0.07–
$0.10/kWh, inclusive of taxes (Statistics Norway, 2021).
In August 2021, President Biden signed executive order
14037 establishing “a goal that 50 percent of all new pas-
senger cars and light trucks sold in 2030 be zero-[tailpipe]
emission vehicles….”(Federal Register, 2021).
However, the impacts of policies designed to acceler-
ate EV adoption are inherently constrained by the slow
turnover of the vehicle fleet. The average light-duty vehi-
cle (LDV) in the United States has a useful life of about
17 years (Keith et al., 2019), and many remain in use for
30 years or more (especially more-polluting light trucks).
Even if policies such as CAFE succeed in improving the
fuel economy of new gasoline vehicles, and sales of zero-
tailpipe emissions vehicles grow, decades will be required
to replace the existing vehicle fleet, during which time
GHG emissions will continue (Keith et al., 2019). Accel-
erating retirement and replacement of the vehicle fleet
will be required to achieve the GHG emission reduction
goals needed to limit climate change.
Policies seeking to accelerate the turnover of the vehi-
cle fleet have most commonly focused on boosting the
retirement of the oldest and most polluting vehicles from
the fleet, known colloquially as “cash-for-clunkers”
(C4C). And by stimulating new vehicle sales, C4C poli-
cies can have significant operational co-benefits, provid-
ing incentives that can accelerate the formation of the EV
market (e.g., increasing EV sales that builds consumer
awareness and creates demand for the roll-out of charg-
ing stations), and stimulating vehicle sales that sustain
valuable blue-collar manufacturing jobs (Adamson &
Roper, 2019; Kalkanci et al., 2019; Rothstein, 2016;
Ton, 2014). Thus, properly designed C4C policies could
yield “triple-bottom-line”benefits for profit, people, and
the planet (Kleindorfer et al., 2009). Increasing interest in
C4C in the mainstream media (Plumer et al., 2021) and
C4C proposals in nations such as India (Shah &
Monnappa, 2021) suggest C4C is gaining momentum as
governments seek to accelerate the decarbonization of
automotive transportation. However, policies to date
have not been optimized from a climate perspective as
yet, with real-world implementations pursuing con-
flicting objectives (such as reducing pollution and
2NAUMOV ET AL.
increasing vehicle production) and limited by the set of
fuel-efficient and alternative-fuel vehicles available to
consumers (Morrison et al., 2010). It therefore remains
an open question as to how C4C policies should be opti-
mally designed to maximize GHG emissions reduction
with manufacturing and employment co-benefits. In par-
ticular, should we incentivize the adoption of fuel-
efficient gasoline vehicles that are more attractive to con-
sumers in the short-run, but have more limited long-run
emissions benefit, or incentivize emerging zero-emissions
vehicle (ZEV) technologies such as EVs and hydrogen
fuel cell vehicles that are yet to achieve mainstream
adoption?
Here, we extend a fully calibrated extant model of the
US light-duty vehicle fleet turnover (Keith et al., 2019) to
analyze the effectiveness of C4C policies. We simulate
the turnover of the US light-duty vehicle fleet out to 2050
under different C4C policies. We provide guidance on the
practical steps needed to achieve meaningful GHG emis-
sions reduction from automotive transportation, contrib-
uting to the literature on sustainable and behavioral
operations and climate change mitigation. We outline the
opportunities and challenges in designing a C4C policy
that addresses environmental, financial, and manufactur-
ing objectives concurrently.
We find that C4C programs in which participants can
purchase either EVs or fuel-efficient gasoline vehicles
lead to small emissions reductions at a high cost. Much
larger and more cost-effective emissions reductions are
possible when C4C requires consumers replace their old,
fossil-fuel powered vehicle with an EV. C4C policies
incentivizing EVs yield benefits larger than the direct
effect of C4C on fleet replacement by stimulating a set of
powerful reinforcing feedbacks that speed the develop-
ment of the EV market. EV-focused C4C policies boot-
strap EV adoption by (i) accelerating EV cost reductions
through scale economies and learning throughout the EV
supply chain, (ii) expanding EV make and model variety,
(iii) speeding deployment of charging infrastructure, and
(iv) building consumer awareness and willingness to con-
sider EVs. These improvements foster additional EV pur-
chases beyond the direct impact of C4C on fleet turnover,
which then drive further improvements in the attractive-
ness of EVs, yielding a substantial synergy that amplifies
the direct benefits of C4C. These reinforcing feedbacks
also increase the benefits of complementary policies that
speed grid decarbonization. A gas tax or carbon price fur-
ther enhances the emissions reductions from C4C, and
the revenue can be used to help offset program costs or
be rebated to the public to address equity issues. We per-
form a wide range of sensitivity analyses over the key
uncertainties to examine the robustness of our results
and establish guidelines for the most effective and
efficient policy design to leverage synergies with comple-
mentary environmental policies.
2|CASH FOR CLUNKERS
2.1 |The road to a low emission vehicle
fleet
A variety of policies have been employed in the
United States to promote sales of alternative, low- or
zero-tailpipe emissions vehicles such as EVs. Tax credits
for buyers of EVs offer up to $7500 from the federal gov-
ernment, and additional incentives of up to $2500 in sev-
eral states (U.S. DOE, 2019). California and nine other
states follow the ZEV mandate, requiring that the market
share of ZEV increases from 4.5% in 2018 to 22% in 2025
(CARB, 2018, 2019). Despite this substantial policy sup-
port over many years, alternative fuel vehicles have only
achieved low single-digit market shares in the
United States to date (Figure 1).
Along with technology-specific policies, the CAFE
standards aim to improve the average fuel economy of
new light-duty vehicles (cars and light trucks). The CAFE
standards were weakened in 2020 under the safer afford-
able fuel-efficient (SAFE) vehicles rule. Adjusting for
credits and accounting for the difference between test
and real driving conditions, those standards required the
average fuel economy of new light-duty vehicle to
increase by 1.5% per year with a goal of 37 miles per gal-
lon (MPG) by 2026, with the fuel economy of new cars
and light trucks to be 43.7 and 31.3 MPG by 2026, respec-
tively (NHTSA, 2021a). On January 20, 2021, President
Biden issued Executive Order EO 13990 directing the
FIGURE 1 Sales of low- and zero-emission vehicles in the
United States. Source: Auto Alliance, 2018; HybridCars.com, 2018;
Automotive News, 2019
NAUMOV ET AL.3
government to establish “Ambitious, Job-Creating Fuel
Economy Standards”by “considering whether to propose
suspending, revising, or rescinding”the SAFE standards,
which apply through the model year 2026, and on August
5, 2021, the U.S. Department of Transportation's National
Highway Traffic Safety Administration (NHTSA) pro-
posed new CAFE standards for 2024–2026 that would
boost the yearly increase in CAFE stringency to 8% per
year from 1.5% per year set under the SAFE standards
(NHTSA, 2021b). On August 5, 2021, President Biden
issued EO 14037, “Strengthening American Leadership
in Clean Cars and Trucks,”directing the government to
to establish new multi-pollutant emissions
standards, including for greenhouse gas
emissions, for light- and medium-duty vehi-
cles beginning with model year 2027 and
extending through and including at least
model year 2030.
EO 14037 also established the goal that half of new vehi-
cle sales be electric by 2030 (Federal Register, 2021).
However, even if new vehicles sold in 2050 became 75%
more fuel-efficient than today (implying average fuel
economy of 130 MPG), the expected reduction in fleet
emissions will not exceed 40%, owing to the slow rate of
fleet turnover (Keith et al., 2019).
2.2 |Lessons from the US CARS “Cash
for Clunkers”program
Programs that accelerate the replacement of old and inef-
ficient vehicles currently in the fleet provide an opportu-
nity to reduce GHG emissions from the vehicle fleet. C4C
programs began in the 1990s, including Sweden,
Denmark, France, the United Kingdom, Canada (“Retire
Your Ride”), China (“old swap new”), and others
(ARC, 2012; BBC, 2009; Morrison et al., 2010;
Zhang, 2009), with stated objectives of either reducing
criteria pollutant emissions including GHGs, or stimulat-
ing the auto industry. All these accelerated vehicle retire-
ment programs offered customers incentives to scrap
their old cars, but differences in implementation time,
program duration, incentives, budget, and mixed objec-
tives have led to varying opinions about their success
(Alberini et al., 1995; Dill, 2004; Lachapelle, 2015;
Marin & Zoboli, 2020; Miller et al., 2020; Morrison
et al., 2010; Zhou et al., 2019). To provide further context
for this work, we summarize below the results of the
most widely studied program: the US government's Cars
Allowance Rebate System (CARS), commonly known as
cash for clunkers.
CARS was introduced in 2009 to boost sales of new
and domestically manufactured vehicles after the precipi-
tous market decline during the economic crisis of 2008,
when new vehicle sales plummeted 18%, to the lowest
rate in a decade. US manufacturers were hit hardest, with
sales declining between 28% and 36% (Vlasic, 2008). The
CARS program provided incentives between $3500 and
$4500 per vehicle to consumers who traded in a low fuel-
efficiency vehicle and purchased a new and higher fuel-
efficiency vehicle. Trade-ins were decommissioned (and
the materials recycled) so that they could not re-enter the
used vehicle market or be exported. Lasting only 8 weeks,
from July 1, 2009, to August 24, 2009, CARS stimulated
the replacement of 680,000 vehicles at a cost of $3 billion
(Li et al., 2013).
While the government declared the program a success
(U.S. DOT, 2009), subsequent research was less positive.
The program did nominally achieve its stated objective of
inducing additional vehicle sales. However, researchers
estimated that only 370,000 of the 680,000 vehicles sold
under the program would not have happened otherwise
(54%), and sales of new vehicles dropped after the pro-
gram expired (Mian & Sufi, 2012). Moreover, more than
half the incentives are estimated to have gone to house-
holds that would likely have purchased a new vehicle in
the next 2 months anyway (Hoekstra et al., 2017).
Because the program required the purchase of a new
fuel-efficient vehicle, many consumers chose to buy
cheaper and more fuel-efficient vehicles such as the
Toyota Corolla, boosting the sales of Japanese cars, and
reducing the average selling price of new vehicles sold
(Hoekstra et al., 2017; Simon, 2009), undermining the
goal of boosting domestic car sales.
The CARS program required replacement vehicles
purchased to be more fuel-efficient than the average vehi-
cle available on the market, increasing the average fuel
economy of all new vehicles by 0.6–0.7 MPG (Sivak &
Schoettle, 2009) and preventing emissions of 4.4 million
tonnes of CO
2
(tCO
2
) (Lenski et al., 2010). However, esti-
mates suggest program costs exceeded the value of the
environmental benefits by about $2000 per vehicle, calcu-
lated as the difference between the subsidy and environ-
mental benefits based on the then-current social cost of
carbon (SCC) (Abrams & Parsons, 2009). Estimates of the
cost per tonne of avoided CO
2
emissions vary widely,
from about $90/tCO2 to over $500/tCO
2
(Knittel, 2010; Li
et al., 2013). Further, 50% of the pollution reduction ben-
efits went to just 2% of US counties—mostly urban cen-
ters (Lenski et al., 2013).
Although CARS and similar policies (see Morrison
et al., 2010) provide important evidence regarding the
impact of accelerated vehicle retirement programs, their
impacts do not generalize to the current US market. New
4NAUMOV ET AL.
studies are needed to inform current (in California) or
proposed (in Massachusetts) government programs
(BAR, 2021; Rogers & Schmid, 2021).
For example, the short duration of CARS, only
8 weeks, allowed strategic consumers to accelerate
replacement of an older car slightly so as to qualify for
the program, undermining its net sales impact. The C4C
program we examine would last far longer, on the order
of a decade, limiting opportunities for strategic purchase
timing. Further, the higher average fuel economy of vehi-
cles in the fleet today, expanded makes and models of
low emissions vehicles (including more low- and zero-
tailpipe emission vehicles—103 in 2018 versus 20 in 2009
[see SAFE Vehicles Rule NHTSA, 2021a, p. 24237]),
higher fuel economy standards, and more stringent cli-
mate objectives, create a different playing field.
Every older, inefficient fossil-powered vehicle rep-
laced with an efficient electric vehicle generates direct
environmental benefits (Kim et al., 2003; Morrison
et al., 2010; Spitzley et al., 2005; Van Wee et al., 2000).
However, C4C and other policies to promote EV adoption
can also generate indirect benefits by accelerating market
formation. To date, EV adoption in the US and most
nations has been limited because EVs are unfamiliar to
many consumers, who do not include them in their con-
sideration sets when choosing a new vehicle (Hauser &
Wernerfelt, 1990, explore the concept of consideration
sets; Struben & Sterman, 2008, show how unfamiliarity
limits adoption of alternative vehicles. See also Shafiei
et al., 2012; Silvia & Krause, 2016; Hardman et al., 2020;
Khurana et al., 2020; Shetty et al., 2020; Wu et al., 2020).
Policies stimulating EV adoption will increase consumer
familiarity through social exposure and network effects
(Struben & Sterman, 2008). Further, increasing EV sales
will stimulate R&D, learning-by-doing, and scale econo-
mies that lower EV costs and improve range, stimulate
the deployment of charging infrastructure, and induce
automakers to offer a wider variety of makes and models.
These effects create self-reinforcing feedbacks that boot-
strap the nascent EV market towards maturity and poten-
tially amplify the impact of policies including C4C (Keith
et al., 2020). To the best of our knowledge, no existing
studies consider these indirect feedback effects of acceler-
ated vehicle retirement programs.
3|MODELING THE DYNAMICS
OF ACCELERATED FLEET
TURNOVER
To explore the dynamics of US vehicle fleet turnover with
C4C policies, we extend and modify a fully calibrated
model of the US light-duty vehicle fleet (Keith
et al., 2019) to include the effect of C4C policies (see
Figure 2). The fully documented model and data files
(Rahmandad & Sterman, 2012) are available in an elec-
tronic supplement (Naumov et al., 2022) (Data S1). The
model tracks cars and light trucks in the US fleet from
initial sale until discard. The model incorporates
established formulations from fleet diffusion and urban
mobility models (Keith et al., 2017; Keith et al., 2020;
Naumov et al., 2020; Struben & Sterman, 2008), discrete
consumer choice (e.g., Archer & Wesolowsky, 1996; Ben-
Akiva & Lerman, 1985; Brownstone et al., 2000;
Coltman & Devinney, 2013; McFadden, 1981; Pullman
et al., 2001; Verma et al., 2006), and behavioral decision-
making (Bendoly et al., 2006; Morecroft, 1985; Sterman
et al., 2015; Sterman & Dogan, 2015).
The model is calibrated to represent the light duty
vehicle fleet of the US, which comprises two vehicle
classes as defined by the US Department of Transpor-
tation (NHTSA, 2021a), cars and light trucks:
Z¼ Cars,Light trucksfg. We further consider two vehicle
powertrain platforms, internal combustion engine (ICE)
and electric (EV): P¼ ICE,EV
fg
available in each vehi-
cle class. The model therefore tracks four fleets: two ICE
(cars and light trucks), and two EV (cars and light tru-
cks). The age structure of each of the four fleets is repre-
sented with one-year cohorts. The ICE fleets are
calibrated to actual cohort-specific rates of vehicle retire-
ment and vehicle-miles traveled (VMT) per year in the
US light-duty vehicle fleet (Williams et al., 2017). We
assume the same age-specific hazard rates of retirement
and VMT per year for both the ICE and EV platforms.
The EV platform includes both battery electric vehi-
cles (BEV) and plug-in hybrid electric vehicles (PHEVs).
PHEVs have a smaller battery than BEVs, with typical
all-electric range of about 40 miles, but include a gasoline
engine as a range extender. Approximately 55% of all
miles driven by PHEVs are electric (AFDC, 2020),
increasing GHG emissions relative to BEVs, but they also
reduce range anxiety and require less time to charge. The
effect of PHEVs is thus twofold—on the one hand, they
increase GHG emissions from the combined EV fleet, but
they also pose fewer barriers to adoption, accelerating EV
market formation by building consumer awareness, cre-
ating demand for EV charging infrastructure, and foster-
ing scale economies in EV components and supply
chains, including motors, batteries, controls, and related
subsystems (Keith et al., 2020; Struben, 2006).
3.1 |Vehicle fleet turnover
The total installed base of vehicles of class iZwith
powertrain pPsums over Nindividual age cohorts a:
NAUMOV ET AL.5
Vip ¼X
N
a¼1
Vipa ð1Þ
In the absence of the C4C program, the total installed base
is changed through vehicle retirements ripa from each
cohort aof class iwith powertrain p, and the addition of
vehicle purchases nip entering the cohort of new vehicles:
dVip
dt ¼nip X
N
a¼1
ripa ð2Þ
where ripa is a function of the rate of vehicle retirements
through natural turnover αipa, increasing with the age of
the cohort a, estimated based on the existing data of the
US fleet (Williams et al., 2017):
ripa ¼Vipaαipa ð3Þ
C4C policies add outflows of vehicles retired under the
C4C program from qualifying age cohorts dipa and the
inflow of mandatory replacement purchases mip :
dVip
dt ¼nip X
N
a¼1
ripa þmip X
N
a¼Q
dipa ð4Þ
where Qis the minimum age cohort qualifying for the
C4C program.
Total new vehicle purchases for each vehicle and fuel
type, nip, are given by the replacement of all discarded
vehicles of each type, the share of replacements going to
each platform (EV or ICE), σip, and the increase in mar-
ket size due to fleet growth, assumed to grow at fractional
rate λ:
nip ¼σip X
N
a¼1
ripa þVip λ
!
ð5Þ
3.2 |Changing vehicle mix
Decommissioned vehicles are replaced with new vehicles.
People can replace their old ICE vehicle with either a
new ICE or a new EV. We do not allow for the possibility
of switching back from EVs to ICE vehicles in light of
recent mandates by many governments banning sales of
gasoline vehicles in the next years. We model consumer
choice based on the utility of EVs, ui,EV , versus ICE vehi-
cles, ui,ICE within vehicle class iZ. Following the litera-
ture on new vehicle platform diffusion (Struben &
Sterman, 2008; Keith et al., 2017, 2020), we use discrete
consumer choice formulations (e.g., McFadden, 1981;
FIGURE 2 The C4C policy and fleet turnover. Signs at arrowheads indicate the polarity of the causal relationship: “+”indicates an increase in
the independent variable causes the dependent variable to increase, ceteris paribus (and a decrease causes a decrease): X!þY≝∂Y=∂X>0.
Similarly, “–”indicates that an increase in the independent variable causes the dependent variable to decrease (and a decrease causes an
increase): X!Y≝∂Y=∂X< 0. Boxes represent stocks (accumulations); double arrows “⟹”with valves “⧖”represent flows, for example,
Vehicle Fleet tðÞ¼
Rt
t0NewðVehicle Sales sðÞVehicle Retirements sðÞÞds þVehicle Fleet t0
ðÞ
6NAUMOV ET AL.
Ben-Akiva & Lerman, 1985; Brownstone et al., 2000;
Verma et al., 2006) using a binomial logit model for the
market share of each vehicle platform ICE,EV
fg
for
each class iZ. Defining the utility of EVs, ui,EV ≝ui
and, without loss of generality, the utility of an ICE plat-
form ui,ICE ¼0, 8iZ, the market share of each plat-
form is:
σi,EV ≝σi¼#1
1þeui
σi,ICE ¼#1σi
8
<
:ð6Þ
Following standard logit models, EV utility is a linear
function of vehicle attributes, s, valued by consumers, xis ,
weighted by coefficients βs, with homoscedastic indepen-
dent and identically distributed (i.i.d.) extreme value
errors ϵi:
ui¼X
s
βsxs
iþϵið7Þ
We assume no change in the mix of light trucks and cars.
Some people might decide to purchase a light truck when
they trade in an old car, as seen in the recent shift from
sedans to SUVs (EIA, 2018), or potentially opt for a more
efficient car when they trade in an old light truck. We
leave consideration of this mechanism, and disaggrega-
tion to additional market segments (e.g., pickup trucks,
full size SUVs, small SUVs, crossovers, etc.) for future
studies, which may be feasible as more data become
available.
We include covariates for the inconvenience of the
EV platform relative to the ICE platform (reflecting lower
driving range, longer refueling time, lack of recharging
infrastructure, low consumer awareness, etc.), the price
of EVs relative to ICE vehicles, and the relative fuel sav-
ings from driving an EV versus ICE vehicle, such that
consumer utility becomes:
ui¼βmarketxmarket
iþβpricexprice
iþβfuelxfuel
iþϵið8Þ
The EV market is not yet mature, and multiple feed-
backs exist that may cause the utility of EVs to increase
as EV sales grow (Struben & Sterman, 2008; Keith
et al., 2020). Growing EV sales generate more funds for
R&D, speed learning-by-doing, and drive scale econo-
mies that reduce EV costs and improve range, inducing
automakers to offer a wider range of EV makes and
models. More EVs on the road increase demand for
charging, stimulating the deployment of charging infra-
structure, making EVs even more convenient. In addi-
tion, more EVs on the road increase social exposure to
them, building familiarity with and consumer accep-
tance. All these effects increase EV sales further,
forming reinforcing feedbacks potentially bootstrapping
the EV market. For parsimony we do not disaggregate
these multiple feedbacks, instead of aggregating them
into the self-reinforcing Market Formation feedback, R1,
shown in Figure 3 (Sterman (2000) provides an explana-
tion of causal diagramming notation and the concepts of
self-reinforcing (positive) and self-correcting (negative)
feedbacks).
We capture these effects as reductions in the price of
EVs and the inconvenience of the EV platform. We use a
standard power-law learning curve in cumulative produc-
tion experience, which serves as a proxy for the aggregate
effect of all sources of learning (Argote & Epple, 1990).
For the price attribute, we include both local market pro-
duction of EVs, L, (here, US production), and rest-of-
world EV production, W:
xprice
i¼Pi,0 W
W0
L
L0
γprice
ð9Þ
where W0and L0are reference levels of rest-of-world and
local production respectively, Pi,0 is the price premium of
EVs at the reference level of experience, and γprice is the
strength of the learning curve for price. We further
assume that world production Wgrows at rate λw:
dW
dt ¼Wλwð10Þ
We capture the effect of local market EV production on
market formation dynamics, reflected in consumer accep-
tance, recharging infrastructure availability, quality of
vehicle service, and other factors, jointly represented by
the inconvenience of the EV platform. Because the EV
platforms comprise both BEVs and PHEVs, the covariates
of the utility of the EV platforms are modeled as a linear
combination of the corresponding attributes of PHEVs
and BEVs with the behavioral parameter PHEV Consid-
eration, π, (Equations (11) and (16)) reflecting the avail-
ability, performance, and popularity of PHEVs as
perceived by consumers. This formulation allows us to
analytically determine the share of PHEVs versus BEVs
in the EV platform (Equation (19)) and accurately cap-
ture EV market formation and the GHG impact of C4C
policies. We assume that the inconvenience of PHEVs
relative to ICE vehicles is 0, since PHEV drivers can also
use the existing gasoline refueling infrastructure. The
perceived inconvenience of an EV is therefore lower
when PHEVs are present than in the absence of PHEVs
through lower range anxiety, shorter charge times, and
so forth:
NAUMOV ET AL.7
xmarket
i¼Ii,0 1πðÞ
L
L0
γmarket
ð11Þ
where Ii,0 is the reference inconvenience of BEVs at the
reference level of experience, and γmarket is the strength of
the learning curve for market formation.
Annual fuel costs for a new ICE vehicle depend on
miles driven per year, l, the price of gasoline, pg, and ICE
fuel economy, determined by the CAFE standard,
FECAFE
i. Annual fuel costs for a new EV depend on EV
miles driven per year, l’, which may differ from ICE miles
per year (Equation (12)), the price of electricity, pe, and
EV fuel use (including electric power and, for PHEVs,
gasoline for the non-electric miles driven), FU0
i,EV :Rela-
tive fuel cost savings for new EVs, xfuel
i, are then given by:
xfuel
i¼1FU0
i,EV l0pe
.1
FECAFE
i
lp
g
ð12Þ
It is well documented that people adjust their driving
behavior when the fuel price, pchanges relative to the
reference fuel price, p0, with the elasticity of miles driven
to fuel price εl(Wang & Chen, 2014):
l¼l0p
p0
εl
ð13Þ
We model the fuel use of the aggregate EV platform,
FU0
i,EV , as an average of the fuel use of new BEVs and
PHEVs, weighted by the share of PHEVs in new EV sales,
ρ, and accounting for the electricity and gasoline use of
PHEVs. We assume the same fuel economy for BEVs and
PHEVs. However, the gasoline fuel economy of new
PHEVs is higher than that of new ICE vehicles by factor,
ωe, for two reasons. First, the internal combustion
engine in a PHEV is more efficient than that of a
conventional ICE vehicle because the engine is smaller
and optimized for the PHEV (Zhang et al., 2019). Second,
PHEV buyers often adjust their driving and charging
behavior to use more all-electric miles, a phenomenon
known as “gas anxiety”—the “desire of PHEV drivers
to avoid using gasoline”(Ge et al., 2018). The
parameter λeis the share of total PHEV miles driven on
electricity. Thus, the average fuel use of the aggregate
new EV is:
FU0
i,EV ¼ρþ1ρðÞλe
ðÞ
1
FEBEV
i
þ1ρðÞ1λe
ðÞ
1
FECAFE
iωe
ð14Þ
We calculate the share of BEVs in new EV sales, ρ,in
two steps: (1) compute what EV market share would be
in the counterfactual scenario where the EV platform
consists only of BEVs, and (2) compare it to the market
share of the composite EV platform, which includes both
BEVs and PHEVs. In the absence of PHEVs, the utility of
the EV platform, u0
i, depends on the attributes, b
xs,
of BEVs:
FIGURE 3 FIGURE EV market formation
8NAUMOV ET AL.
u0
i¼βmarketb
xmarket
iþβpriceb
xprice
iþβfuelb
xfuel
iþϵið15Þ
Specifically, pure BEVs increase the inconvenience of the
EV bundle relative to a mix of BEV and PHEVs because
BEVs induce greater range anxiety:
b
xmarket
i¼xmarket
i
1πðÞ ð16Þ
We assume BEV and PHEV prices are not materially
different:
b
xprice
i¼xprice
ið17Þ
However, pure BEVs increase fuel savings of the EV plat-
form relative to ICE because BEVs are more cost-efficient
per mile than PHEVs:
b
xfuel
i¼11
FEBEV l0pe
=1
FECAFE
i
lp
g
ð18Þ
The share of BEVs vs. PHEVs in aggregate EV sales for
each vehicle class is then calculated from the utility to
consumers of these two cases:
ρ¼σ0
i
σi
¼1þeui
1þeu0
i
ð19Þ
where uiis the utility of an average EV of class iwhen
PHEVs are present, given by Equation (8), and u0
iis the
utility of an average EV in class iin the absence of
PHEVs, given by Equation (15).
3.3 |Vehicle discards in the C4C
program
Only ICE vehicles are retired through the C4C program,
by vehicle age a,dipa (Equation (4):
dipa ¼Viaβia ,#p¼ICE
0,#p¼EV
ð20Þ
where βia are the age-dependent hazard rates of
vehicle retirement due to the C4C program. The hazard
rate for ICE vehicle retirement through C4C is an age-
dependent rate, β0
ia, modified by the impact of the C4C
policy, ηia:
βia ¼β0
ia ηia ð21Þ
The base hazard of retirement via C4C, β0
ia, increases
with age, a:
β0
ia ¼β0
iaεAð22Þ
where β0
iis the hazard of early retirement via C4C for
1 year old vehicles, and rises with vehicle age, with sensi-
tivity of the hazard rate to vehicle age, εA.
The impact of C4C on early retirement, ηia, is the
product of three factors: the effect of the C4C incentive,
θI, the effect of C4C stringency, θS
i, and the effect of addi-
tional fuel savings from the C4C program, θFE
ia :
ηia ¼θI
iθS
iθFE
ia ð23Þ
The incentive effect θI
ireflects how much the C4C incen-
tive, Ii, affects the propensity of people to participate in
the program:
θI
i¼Ii
I0
εI
ð24Þ
where εI> 0 is the sensitivity of the incentive effect and
Ii=I0is the incentive relative to a reference value.
The C4C program mandates the minimum fuel econ-
omy for new vehicles purchased under the program,
FEmin
i. The stringency effect θS
icaptures the fact that
higher minimum fuel economy standards for replace-
ment vehicles reduce the number of eligible replacement
vehicles people can choose, therefore reducing the likeli-
hood of participating in C4C:
θS
i¼FEmin
i
FECAFE
i
εS
ð25Þ
where εS< 0 is the sensitivity of the stringency effect to
the ratio of the minimum mandated fuel economy to the
average fuel economy of new vehicles under the CAFE
standard.
However, people choosing to participate in C4C
might opt for an even more fuel-efficient vehicle than the
minimum, especially if the C4C standard is not much
higher than fuel economy under CAFE. We capture this
effect by calculating the average fuel economy of new
vehicle sales induced by C4C, FEC4C
i, as the minimum
mandated fuel economy, FEmin
i, adjusted upward by a
fraction, δFE, of the effect of C4C stringency θS
ion the
likelihood of participating in C4C:
NAUMOV ET AL.9
FEC4C
i¼FEmin
i1þδFEθS
i
ð26Þ
The less stringent the C4C program, that is, the higher
θS
i, the higher the actual fuel efficiency of new cars pur-
chased under C4C will be.
The effect of fuel savings on the hazard rate of C4C
participation, θFE
ia , is based on the average fuel cost per
mile of a new vehicle purchased under C4C, FCC4C
i, ver-
sus that of an average ICE vehicle in the existing fleet
FCi,ICE,a, by age cohort and platform:
θFE
ia ¼1FCC4C
i
FCi,ICE,a
εF
ð27Þ
where εF> 0 is the strength of the fuel savings effect.
New vehicles purchased under C4C can be either ICE
or EV (as determined by market share σip in
Equation (6)). Therefore the average fuel cost per mile for
new vehicles purchased under C4C is a linear combina-
tion of actual fuel cost per mile of an ICE vehicle pur-
chased under C4C, FEC4C
i, and perceived average fuel
cost per mile of an EV, calculated as a weighted sum of
BEVs electricity cost and PHEV electricity and gasoline
cost (FEi,EV (Equation (14)):
FU0
i,EV ¼ρþ1ρðÞλe
ðÞ
1
FEBEV
i
þ1ρðÞ1λe
ðÞ
1
FECAFE
iωe
FCC4C
i¼1σi
ðÞ
1
FEC4C
i
pgþσi
1
FEi,EV
peð28Þ
The average fuel cost per mile of an average vehicle in
the existing fleet depends on the average fuel use of an
ICE vehicle in the existing fleet, zi,ICE,a, by age cohort and
platform:
FCi,ICE,a¼zi,ICE ,apgð29Þ
The model tracks the fleet and fuel economy of each
cohort, and these are used to calculate total and average
fuel economy of each platform and for the entire fleet
(Appendix B: Co-flow Formulations).
3.4 |Vehicle replacement purchases
under C4C program
We assume all vehicles traded in through C4C, dipa, are
replaced with new vehicles, either ICE or EV, according
to the market share of each platform, similar to
Equation (5):
mip ¼σip X
N
a¼Q
dipa ð30Þ
Under different realizations of C4C people might qualify
and be paid to discard their vehicle without buying a new
one, and, having decided to go carless, fulfill their mobil-
ity needs through carpooling, ridesharing, bicycling, or
other means. Estimating the potential for such mode-
shifting and the resulting change in travel habits is
beyond the scope of this paper, so we do not consider this
possibility, establishing a lower bound on the potential
reduction in GHGs achieved through the program.
3.5 |Greenhouse gas accounting
We capture full lifecycle emissions of vehicles in the
model, including vehicle manufacturing, assembly, and
disposal, fuel production (well-to-pump or plug), and tail-
pipe emissions from fuel consumption. We use the
GREET 2020 model (ANL, 2020) to account for vehicle
emissions from manufacturing and assembly, μmfg
ip , and
disposal, μdis
ip , of vehicles, with the EV platform including
the GHG footprint of the Li-Ion battery.
ELCA
ip ¼Eip þnip þmip
μmfg
ip þX
N
a¼1
ripa þX
N
a¼Q
dipa
!
μdis
ip
ð31Þ
We compute the emissions by vehicle class and
powertrain platform, Eip, by tracking fleet and emissions
of each platform by each age cohort (Appendix B: Co-
flow Formulations).
The emissions of new EVs are a weighted sum of the
emissions attributable to BEVs and PHEVs. Emissions
per mile from the electric drive of PHEVs are assumed to
equal those of a BEV, but the emissions from the PHEV
internal combustion engine are lower than those of a
conventional ICE vehicle by a factor ωeas discussed in
Section 3.2. On average, tailpipe emissions of new
EVs are:
μi,EV ¼ρþ1ρðÞλe
ðÞνeηe
þ1ρðÞ1λe
ðÞ
νgas
FECAFE
iωe
þν0
gas
ð32Þ
with the emission factor of gasoline vehicles, νgas (U.S.
EPA, 2019b), in grams CO
2
per gallon, the fuel economy
10 NAUMOV ET AL.
required by CAFE, FECAFE
i, well-to-pump emissions of
gasoline, ν0
gas (ANL, 2020), the emissions factor of electric-
ity, νe, the share of total miles PHEVs drive on electricity,
λe, and the energy efficiency of the electric drivetrain, ηe,
in kWh per mile.
Thus, tailpipe emissions from the composite EV plat-
form (Equation (32)) is an implicit function of the mix of
BEVs and PHEVs in EV sales, which depends on the
PHEV consideration parameter, π. PHEVs comprise about
26% of all EV sales in the United States (Gohlke &
Zhou, 2020; U.S. DOT, 2020), but are more popular
in other markets (e.g., some European countries
[EEA, 2021]). However, PHEVs are likely to be phased
out to meet climate goals. For example, the European
Commission proposes “all new cars registered as of
2035 will be zero-[tailpipe] emission”(European
Comission, 2021). Phase-out is likely to be slower in the
United States. We therefore set πsuch that PHEVs repre-
sent about one-quarter of all EV sales at the beginning of
the simulation and decrease it linearly to 0 by 2050.
We assume that the electric grid used to charge EVs
becomes “greener”over time, so that the GHG emissions
factor of electricity, νe, in grams CO
2
per kWh
(ANL, 2020; G
omez Vilchez & Jochem, 2020), falls from
the initial emissions intensity of the grid, ν0, to the emis-
sions factor of renewable power, νr, as the share of
renewable, low GHG electricity, σr
t, increases over time:
νe¼ν01σr
t
þνrσr
tð33Þ
Fuel-related emissions of new vehicle sales, μip , are:
μip ¼
νgas
FECAFE
i
þν0
gas,#p¼ICE
μi,EV ,#p¼EV
8
<
:ð34Þ
Average fuel-related emissions of vehicle sales induced
by the C4C program, μC4C
ip , is calculated similarly using
the average fuel economy in Equation (26):
μC4C
ip ¼
νgas
FEC4C
i
þν0
gas,#p¼ICE
μi,EV ,#p¼EV
8
<
:ð35Þ
4|ANALYSIS
After describing model parameters and initialization
(Section 4.1), we simulate fleet evolution and emissions
from 2021 through 2050 through natural fleet turnover,
given the CAFE standards and EV adoption in the
absence of any additional policies (Section 4.2). Next, we
consider C4C policies with different incentive levels
(Section 4.3) and contrast C4C policies allowing replace-
ment vehicles to be ICE or EVs versus requiring replace-
ment purchases to be EVs (Section 4.4). In Section 4.5,
we consider the impact of complementary policies
including accelerating the transition to renewable elec-
tricity and introducing a gasoline tax or price on carbon
(Section 4.6). In Section 4.7, we explore the sensitivity of
results to the magnitude of the C4C incentive, minimum
qualifying age, and required fuel efficiency of replace-
ment vehicles, and to major uncertainties: consumer
responsiveness to C4C incentives and the strength of the
EV Market Formation feedbacks.
4.1 |Parameterization
To parameterize the model we use the best available data
and previous work on vehicle fleet turnover (Keith
et al., 2017, 2019), data on scrap rates, and vehicle miles
traveled from NHTSA and Oak Ridge National Laboratory
(NHTSA, 2006; Williams et al., 2017; Davis &
Boundy, 2021), vehicle life-cycle emissions (ANL, 2020),
and multiple studies of the 2009 US CARS program
(Abrams & Parsons, 2009; Sivak & Schoettle, 2009;
Morrison et al., 2010; Lenski et al., 2013; Li et al., 2013;
Hoekstra et al., 2017). However, the EV market is still
nascent. Data related to consumer EV adoption, the impact
of access to charging, and market development are sparse.
To estimate the parameters governing market formation
and consumer choice we use prior analyses and data from
similar settings (e.g., Keith et al., 2020) and recent research
estimating the impact of charging station availability on
electric vehicle utility (Wei et al., 2021). Where empirical
estimates are not possible, we make plausible evidence-
based assumptions, placing a premium on assumptions
that are qualitatively and directionally robust, recognizing
that we are studying the future behavior of the vehicle mar-
ket where prior relationships might not hold (Appendix A:
Parameterization of the Model), and explore the sensitivity
of the results to key assumptions (Section 4.7).
4.2 |Baseline (CAFE only)
First, we consider the baseline emissions reduction
resulting from rising fuel efficiency for new ICE vehicles
under the CAFE standards. We assume that CAFE stan-
dards follow the rule proposed by the EPA under Presi-
dent Biden's executive order 13990, which would lower
fleet-wide CO
2
emissions standards from 220 gCO
2
/mile
for model year 2022 to 171 gCO
2
/mile for model year
2023–2026 (U.S. EPA, 2021). Standards for later years
NAUMOV ET AL.11
have not yet been proposed. We therefore assume a con-
servative scenario in which emissions per mile fall an
additional 25% by 2050. Real-world driving is about 20%
less fuel-efficient than laboratory EPA tests (Lattanzio
et al., 2020), so we adjust fuel economy for new ICE vehi-
cles accordingly. Finally, we apply CAFE to gasoline
vehicles only, excluding EVs so that growing EV sales do
not have the unintended consequence of allowing auto-
makers to sell less fuel-efficient vehicles, as current regu-
lations permit (Jenn et al., 2019). In the baseline
simulation (Figure 4), total new vehicle sales grow to
about 22 million/year by 2050. EV market share rises to
44% by 2050, but the EV share of the fleet rises only to
27% of the installed base by 2050 due to the slow rate of
vehicle turnover. The emissions of an average new vehi-
cle fall 59% due to the effect of CAFE and the increasing
share of EVs, but total annual fleet emissions drop only
21%, the result of the long life of vehicles and the
assumed growth in total fleet size.
4.3 |C4C for the purchase of efficient
gasoline or electric vehicles
We now introduce an accelerated vehicle retirement pro-
gram (C4C). We assume the program operates for
10 years, starting in 2022 (Figure 5). To qualify we
assume that the trade-in vehicle must be at least 5 years
old, and that the replacement vehicle be at least 50%
more efficient than the CAFE standard at that time
(which qualifies efficient gasoline vehicles and EVs). In
the results below, the cumulative reduction in CO
2
emis-
sions through 2050 in each scenario is defined as the
accumulation of the difference between emissions in the
baseline case and emissions in the policy scenario (see,
e.g., the right panel in Figure 5). The unit cost of C4C per
tonne of avoided CO
2
emissions ($/tCO
2
) of each policy
scenario is given by the cumulative cost of C4C vehicle
incentives divided by the cumulative emissions reduction
in each scenario, both assessed through 2050.
Setting the incentive at $4000 per vehicle, roughly
equal to the average incentive in the 2009 CARS program
(Li et al., 2013), the C4C policy, denoted C4C ICE/EV,
replaces 12.3 M vehicles by 2032 and reduces cumulative
CO
2
emissions in 2050 by 0.08 GtCO
2
, 0.3% of simulated
cumulative emissions in the baseline case, at a unit cost
of $613/tCO
2
. Annual fleet emissions jump immediately
after the C4C policy is introduced relative to the baseline
as a result of additional emissions incurred with acceler-
ated scrappage and new vehicle manufacturing
(Figure 5). When we increase the C4C incentive to $8000
per vehicle, similar to the federal incentives available
FIGURE 4 Baseline scenario (CAFE only)
12 NAUMOV ET AL.
today for purchasing alternative fuel vehicles (U.S.
DOE, 2019), the policy replaces 20.4 M vehicles by 2032
and reduces cumulative emissions by 2050 by 0.13 GtCO
2
(Figure 5), at a unit cost of $1233/tCO
2
. A larger incen-
tive encourages more people to participate and increases
total avoided emissions, but the unit cost of those emis-
sions reductions increases substantially.
As expected, the introduction of the C4C policy in
2022 immediately increases new vehicle sales, a “shot in
the arm”for automotive production and manufacturing
jobs. However, new vehicle sales drop below the Baseline
in 2032 when C4C ends. Because C4C increases new
vehicle sales, the average age of the fleet falls relative to
the Baseline, beginning in 2022 (Figure 5). Consequently,
when the C4C policy ends in 2032, the fleet is younger,
reducing vehicle retirements. Sales eventually return to
the baseline level as vehicles age (Figure 5). Cumulative
sales under the C4C policy in 2032 when C4C ends are
about 13 million higher than in the baseline and stabilize
at 9.2 million vehicles above baseline by 2050. C4C,
therefore, induces a permanent increase in cumulative
auto production.
Despite the overall boost for the auto industry, C4C
could create transient challenges. The demand surge
when the policy begins would require production, and
possibly manufacturing capacity, to ramp up, and when
the program ends, the temporary drop in demand would
require production to fall for some period. These impacts
would ripple throughout the automotive supply chain
(Vickery et al., 2003; Holweg & Pil, 2008). While the pull-
forward effect on vehicle sales has been identified else-
where, it's impact has been downplayed as insignificant,
or affecting only “distant future”sales (Böckers
et al., 2012; Li et al., 2013). In contrast, our results suggest
policymakers should consider how to design C4C to miti-
gate these disequilibrium shocks. The start-up sales surge
could be moderated by gradual phase-in, for example, by
starting the program with eligibility restricted to older
vehicles, then expanding it over a few years. Similarly,
gradual phase-out of the incentive over a few years could
moderate the demand reduction on program termination.
Such policies would reduce the likelihood of transient
production and employment problems on program start
and sunset and provide time for effective planning and
coordination across the supply chain.
Although outside the boundary of our model, strate-
gic buyers already considering a new vehicle might wait
until the C4C program starts, reducing sales just before-
hand and increasing them further in the first few months
of the program. Similarly, strategic buyers might acceler-
ate a purchase they intended to make in the months after
the program ends so as to qualify for C4C. As discussed
FIGURE 5 C4C policy mandating replacement with efficient ICE vehicles
NAUMOV ET AL.13
above, such behavior was prevalent in the 2009 CARS
program. However, in contrast to the eight-week CARS
program of 2009, the long duration of the program mini-
mizes the opportunity for such behavior—while some
can defer the replacement of an old vehicle for a few
months, fewer can delay replacement for years. To the
extent strategic purchase timing occurs, it would slightly
increase the short-run volatility in automobile purchases
compared with the results in Figure 5. Coordination on
program timing between policymakers and automakers
would provide the time for auto OEMs and their supply
chain partners to prepare for C4C, reducing the need for
rapid, unplanned changes in production schedules.
4.4 |Targeting electric vehicle adoption
with C4C
How effective would C4C be if it were specifically
targeted to encourage the purchase of BEVs only? In the
next scenario, denoted C4C EV, we assume people
receive the incentive only if they purchase a BEV to
replace their old vehicle. We retain the minimum qualify-
ing age of 5 years, and keep the C4C incentive at $8000,
on the order of the $7500 federal tax credit available
today for many EVs.
As expected, mandating C4C trade-ins be replaced
with BEVs leads to a smaller increase in new vehicle sales
compared with the policy allowing replacement vehicles
to be ICE or EV (Figure 6), because fewer people are will-
ing to purchase a BEV due to their initially lower utility
(e.g., fewer makes and models, shorter range, limited
availability of recharging stations). Similarly, the drop in
sales when the program ends is smaller under C4C
EV. Note that the C4C EV $8 K policy replaces 9.9 M vehi-
cles by 2032, more than 50% fewer than when consumers
can choose ICE or EVs under C4C, but reduces cumula-
tive emissions through 2050 by 0.64 GtCO
2
, about 385%
more. Further, the smaller initial jump in total sales mod-
erates the overall production and supply chain challenges
automakers would face on program startup, but could cre-
ate bottlenecks for batteries and other EV-specific compo-
nents. Coordination between the government,
automakers, and supply chain partners will be critical to
mitigate the possibility of transient shortages, spot price
increases, and other startup problems.
The larger drop in cumulative emissions arises for
two reasons. First, BEV emissions are lower than emis-
sions from ICE vehicles meeting the CAFE standard. Sec-
ond, and more importantly, C4C EV increases EV sales
more than C4C ICE/EV, bootstrapping the overall EV
market by driving costs down faster and speeding growth
in make and model variety, charging infrastructure
deployment, and consumer familiarity with EVs. The
result is far more EV sales than C4C induces alone
(Figure 7). The synergy created by the reinforcing market
formation feedbacks is large compared with the direct
impact of C4C on vehicle replacement, contributing 84%
of the total impact of C4C EV $8 K on the cumulative
emissions reduction. The additional EV sales induced by
the self-reinforcing market formation feedbacks drive the
unit cost of emissions reductions down to $124/tCO
2
.
4.5 |C4C with complementary policies
We next consider a policy that combines the C4C pro-
gram with a gasoline tax and policies accelerating the
decarbonization of the electric grid. A gas tax or price on
CO
2
emissions would correct (part of) the unpriced car-
bon emissions externality and generate tax revenue to off-
set the cost of C4C implementation. Although gas tax
increases have historically been politically unpopular in
the United States (Hammar et al., 2004), even the Ameri-
can Petroleum Institute now supports carbon pricing to
help reduce GHG emissions (API, 2021). Combining C4C
FIGURE 6 C4C policy mandating replacement with EVs only
14 NAUMOV ET AL.
and a price on carbon or gas tax creates a “carrot and
stick”approach further increasing the attractiveness of
EVs. Further, the revenue from a carbon price can be
used to pay for the cost of C4C and/or rebated to con-
sumers as a carbon dividend to address concerns that gas
taxes are regressive. A large literature in operations man-
agement examines tax policy to promote environmentally
responsible technology (e.g., Zhu & Sarkis, 2004; Krass
et al., 2013; Drake et al., 2016). The results are largely
driven by assumptions about emissions prices and the
availability of additional incentives, such as production
and cost subsidies, consumer rebates, and so forth. For
example, Krass et al. write:
using a combination of the environmental
tax, consumer rebate, and fixed cost subsidy,
the regulator can always induce [the clean]
technology …and achieve the highest possi-
ble welfare value of social welfare achievable
with this technology.
Here, we consider the impact of a gas tax on the perfor-
mance of C4C. The broader welfare and equity implica-
tions of these policies, however, are beyond the scope of
our model.
Figure 8 shows the C4C EV $8 K policy together with
a gasoline tax. We assume the new tax rises linearly from
$0.00 to $0.30/gallon by 2050. Increasing fuel prices trig-
ger a behavioral response as consumers reduce miles
driven per Equation (13). However, the gas tax tested
here is small, amounting to only 10% of the price of gaso-
line in 2050, so with an assumed price elasticity of
demand of 0.3 (e.g., Gillingham & Munk-Nielsen, 2016),
the maximum impact is a 3% reduction in ICE vehicle
miles traveled (VMT) in 2050. Although the VMT reduc-
tion is small, the policy increases the attractiveness of
EVs relative to ICE vehicles, and leads to large emissions
reductions, particularly in later years (Figure 8).
Combining C4C EV and the gas tax raises the EV
share of new vehicle sales to 52% by 2050, with the EV
fleet share reaching 39%. Cumulative avoided emissions
rise to 0.97 GtCO
2
by 2050, and the cost falls to $81/tCO
2
.
That cost is the cumulative cost of the C4C incentives per
tonne of avoided emissions and does not include the rev-
enue from the gas tax. In that scenario, cumulative gas
tax revenues are approximately $318 billion, and could
pay for the C4C program with a large surplus remaining.
Because EVs have zero tailpipe emissions, their cli-
mate benefits are governed by the emissions intensity of
the electricity for recharging. We now consider the
impact of a faster transition to renewable, low-carbon
power production. The proposed Clean Electricity Pay-
ment Program seeks 80% low/zero emissions power by
2030 (Clean Air Task Force, 2021). We assume a more
modest transition, achieving about 50% by 2030 and a
maximum of 90% by 2040, compared with about 28% and
41% in the base case, respectively. Adding the renewable
electricity (RE) policy to C4C EV and the gas tax reduces
cumulative emissions by 1.31 GtCO
2
by 2050 (Figure 8),
and further reduces the average cost to $61/tCO
2
.
FIGURE 7 Comparison of C4C policies
NAUMOV ET AL.15
Figure 9 shows the impact of the C4C policies with the
complementary gas tax and renewable electricity policies.
The impact of faster electric system decarbonization is
much larger with the C4C EV policies: With C4C ICE/EV
$8 K, RE reduces cumulative emissions by 2050 by ≈0.03
GtCO
2
, but C4C EV $8 K increases the impact of renew-
able electricity to 0.32 GtCO
2
,nearly11timeslarger
(Figure 9). The synergy from the gas tax is negligible
because the tax is small, reaching only $0.09/gallon (<
$10/tCO
2
) by 2030 and $0.30/gallon (≈$34/tCO
2
) by 2050.
4.6 |Factors affecting C4C effectiveness
The impact and cost-effectiveness of C4C is contingent
on three main parameters: (1) the C4C incentive; (2) C4C
stringency, that is, the minimum fuel economy required
for replacement vehicles under C4C; and (3) C4C qualify-
ing age, that is, the minimum age of vehicles eligible to
participate in the C4C program.
Figure 10 shows the impact of C4C ICE/EV and C4C
EV as a function of the incentive. Larger incentives
induce large increases in C4C participation (the cumula-
tive number of vehicles traded in under the program).
Participation is much lower under C4C EV, but shows
similar responsiveness to the incentive. The magnitude
and sensitivity of emissions reductions to incentive size
are low when replacement vehicles can be either ICE or
EVs, while the average cost per tonne of avoided emis-
sions is high and rises steeply with larger incentives, even
when C4C ICE/EV is combined with the carbon price
and faster grid decarbonization. In contrast, the
FIGURE 8 C4C policy with EVs, a gas tax, and renewable electricity
FIGURE 9 C4C policy synergies
16 NAUMOV ET AL.
emissions reductions from C4C EV are far larger and rise
steeply as the incentive grows, and the average cost is far
lower and rises less rapidly. As expected, emissions
reductions show modest diminishing returns as the
incentive rises. Adding the complementary gas tax and
renewable electricity policies boosts emissions reductions
substantially for any incentive level and creates large
reductions in the average cost per tonne of avoided
emissions.
Program stringency is the required reduction in
replacement vehicle emissions per mile relative to the
CAFE standard at the time of trade-in and applies only to
the case where participants can choose either an EV or a
qualifying ICE vehicle. Perhaps a sufficiently stringent
C4C program, together with the reduction in emissions
per mile under the CAFE standards, would enable emis-
sions reductions under C4C ICE/EV to surpass those of
the EV only policy due to C4C higher participation when
participants can choose EVs or ICE vehicles. However,
while greater stringency lowers the emissions of replace-
ment ICE vehicles, it reduces people's propensity to par-
ticipate in the program because it will be harder to find
makes and models meeting people's needs. Figure 11
shows that greater stringency has weak effects on emis-
sions reductions because C4C participation falls as strin-
gency rises. As a result, although cumulative emission
savings rise for small increases in stringency, they remain
far below the emissions reductions from C4C EV even
FIGURE 10 Impact of the C4C incentive. Note the different scales for the two graphs of average costs (right panels)
NAUMOV ET AL.17
when the new ICE vehicles must be 50% better, or more,
than the average new vehicle under CAFE (compare
against Figure 9). The unit cost of emissions reductions
under C4C ICE/EV also remain far higher than the EV
only case and exhibit strong diminishing returns as strin-
gency increases. These results hold even when C4C
ICE/EV is combined with the complementary gas tax and
renewable electricity policies.
The effect of the minimum qualifying age is similarly
nonlinear (Figure 12). Older vehicles emit more CO
2
per
mile, improving emissions reductions and cost effective-
ness per vehicle, but limiting eligibility to very old vehi-
cles reduces participation because there are fewer older
vehicles, and reduces cost effectiveness because those
vehicles are already near end-of-life and replacement
without C4C. On the other hand, including younger vehi-
cles might be inefficient because they are less emission-
intensive than older ones, reducing the emissions bene-
fits of early retirement. As expected, participation in C4C
falls sharply as qualifying vehicle age increases from 1 to
15 years and approaches zero as qualifying age is
increased further. Consequently, emission reductions are
lowest for higher qualifying ages. Under C4C ICE/EV,
extending eligibility to the youngest vehicles yields a neg-
ligible improvement in cumulative emissions reductions,
and high cost per tonne of avoided emissions, because
most replacement vehicles chosen under that policy are
ICE, and the emissions intensity of young ICE vehicles is
closer to that of the replacement ICE vehicle. In contrast,
under C4C EV, extending eligibility to the youngest vehi-
cles yields a large increase in cumulative emissions
reductions, with only a small increase in the unit cost of
avoided emissions, because even young ICE vehicles gen-
erate far more emissions than an EV.
4.7 |Sensitivity analysis
Uncertainty is inherent in all models, but particularly rel-
evant for emerging technologies and markets where data
are not yet available. Key areas of uncertainty here
include parameters affecting EV market formation and
consumer responsiveness to C4C programs.
4.7.1 | EV market evolution
The growth and maturation of the EV market is con-
trolled by the strength of the learning curves governing
(i) EV cost reductions and (ii) market formation, the
FIGURE 11 C4C impact as it depends on program stringency. Efficiency of replacement vehicle under C4C relative to CAFE standard
at time of replacement
18 NAUMOV ET AL.
latter comprising the impact of growing vehicle range
and performance, make and model variety, charging
infrastructure deployment, consumer awareness and will-
ingness to consider EVs, and related factors. Figure 13
shows the effect of large variations on the assumed rate
at which EVs achieve cost parity with ICE vehicles.
Today, EVs are more expensive than comparable conven-
tional vehicles, largely due to the cost of batteries. EV
cost reductions in the model depend on the assumed
growth rate in rest-of-world EV sales, λw, Equation (10),
and strength of the cost reduction learning curve
strength, γprice, Equation (9). The baseline scenario (thick
solid line) takes the default values for the worldwide mar-
ket growth (10%/year) and strength of the cost reduction
learning curve (25% cost reduction per doubling of cumu-
lative sales). The Sluggish scenario assumes 50% slower
rest-of-world EV market growth rate and a 50% weaker
cost reduction learning curve. The optimistic case
assumes 50% faster market growth and a 50% stronger
learning curve, and the breakthrough case assumes 100%
faster market growth and a learning curve twice as strong
as the base case. The effect of these large changes is rela-
tively modest. In the base case, the EV share of the fleet
in 2050 is 27%. It varies from 21% in the sluggish scenario
FIGURE 12 C4C impact as it depends on program qualifying age. Note the different scales for the two graphs of average costs (right
panels)
NAUMOV ET AL.19
to 34% in the breakthrough scenario, with annual emis-
sions in 2050 varying by approximately ±3% of the base
value.
We next vary the strength of the reinforcing EV mar-
ket formation feedbacks by adjusting the strength of the
learning curve, γmarket, Equation (11). Figure 14 contrasts
the baseline scenario (γmarket ¼0:35Þagainst Sluggish,
Optimistic, and Breakthrough scenarios with
γmarket ¼0:175, 0:525, and 0:7, respectively. The impact is
large. The EV share of the fleet by 2050 is 27% in the base
case, but only 6% in the Sluggish case and 57% and 67%
in the optimistic and breakthrough scenarios, respec-
tively, reducing annual emissions, particularly after 2030.
Cumulative emissions by 2050 are 1.3% higher in the
Sluggish case than the base case and reduced by 4.5% and
8.3% for the Optimistic and Breakthrough cases
respectively.
How does the strength of the market formation feed-
backs affect the impact of C4C? Intuitively, C4C should
be less effective when these reinforcing feedbacks are
FIGURE 13 Effect of EV market on price
FIGURE 14 Effect of EV market formation
20 NAUMOV ET AL.
weak because the synergy of additional EV sales stimu-
lated by C4C will be smaller. However, C4C should be
less beneficial when the market formation feedbacks are
so strong that the EV market grows rapidly even without
the stimulating effect of C4C. Figure 15 shows how C4C
impact varies with the strength of the market formation
feedbacks (γmarket) for C4C policy with $8000 incentive
and additional complementary policies as in Figure 8. As
expected, cumulative emissions reductions rise, peak,
and then decline, and the average cost of these reductions
per tonne fall to a minimum, then rise.
Note, however, that the benefits of C4C fall only
slowly as the market formation feedbacks become very
strong. For example, when the market formation feed-
backs are twice as strong as the base case (as in the
Breakthrough scenario in Figure 14), cumulative emis-
sions reductions with the C4C EV $8000 policy in 2050
are still 0.49 GtCO2. Stronger EV market formation feed-
backs cause EV market share to rise faster even without
C4C, but has its greatest impact after 2030 and does not
affect the existing inefficient fleet today. By removing
older, more polluting vehicles in the fleet today, C4C
continues to have a large impact on emissions even under
highly optimistic assumptions about the speed of EV
market maturation. These results hold for C4C EV alone
and when the gas tax and renewable electricity policies
are added. With these complementary policies, emissions
reductions are much larger, and unit costs are far lower.
4.7.2 | Consumer responsiveness to C4C
Consumer response to C4C programs is governed by
three main and uncertain parameters: how responsive
people are to the incentive offered (εI), how much people
value the fuel savings from replacing their vehicle (εF),
and how much the likelihood of participating in C4C
increases with vehicle age (εA). The smaller the magni-
tude of each parameter, the smaller the increase in the
hazard rate of participating in C4C as incentives, fuel sav-
ings, or vehicle age increase. To explore the impact of
these uncertainties, we vary each parameter between
0 (not responsive, ignoring increasing benefits) and
1 (highly responsive, proportional to benefits), in
FIGURE 15 Effect of EV market formation feedbacks on C4C program
FIGURE 16 Full factorial sensitivity analysis of C4C program parameters. Darker to lighter shades encompass 50%, 75%, and 95% of all
runs, respectively
NAUMOV ET AL.21
increments of 0.05 in a full factorial sensitivity analysis gen-
erating 21
3
=9261 simulations. Figure 16 shows the results
for C4C EV $8 K with the gas tax and accelerated renewable
electricity policies. Importantly, C4C increases EV adoption
and cuts emissions under all combinations of parameters,
even those in which their values are least favorable.
While the full factorial design varies the parameters
governing the hazard rate of C4C participation indepen-
dently, they are likely to be highly correlated: all three
(incentive, fuel savings, and vehicle age) affect the finan-
cial impact of participating in C4C. Individuals more
aware of and responsive to incentive are also more likely
to be more aware of and respond to the fuel savings and
to the maintenance costs, breakdown risk, and lack of
modern safety features and other technologies in their
older vehicle. Figure 17 shows the results of assuming
the parameters governing the hazard rate of C4C partici-
pation are perfectly correlated, varying from 50% less
responsive to 50% more responsive than the base values.
With the base values of these parameters, the share of
EVs in the fleet in 2050 is 39% under the combined C4C
$8 K, gas tax, and renewable electricity policy, compared
with 27% in the baseline scenario without C4C. When all
three parameters vary from 50% to +50% of their base
values the EV share of the fleet in 2050 under the com-
bined C4C policies ranges from 36% to 43%, and annual
emissions range from 0.81 to 0.77 GtCO
2
/year, compared
with 27% and 0.93 GtCO
2
/year without C4C. The com-
bined C4C policies speed EV adoption and reduce annual
emissions even when the parameters governing C4C par-
ticipation are most pessimistic.
5|DISCUSSION
The long life of automobiles means policies promoting
purchase of more efficient and zero-tailpipe emissions
vehicles will not reduce fleet emissions fast enough to
meet current US goals. C4C programs can accelerate fleet
turnover and speed the transition to low-emissions trans-
portation. Designing a C4C policy that simultaneously
addresses environmental, financial, and manufacturing
objectives create opportunities but also challenges. Sev-
eral tensions must be resolved. Increasing the incentive
paid to each participant boosts participation, but raises
program costs and may create a surge of sales at the start
of the program that could strain automotive supply
chains. The more stringent the eligibility requirements
for C4C, the larger the emissions reductions from each
vehicle traded in under the program, but fewer people
will qualify, reducing participation. Allowing people to
choose a qualifying EV or ICE replacement vehicle
increases the variety of makes and models available,
increasing participation, but reduces the emissions sav-
ings per vehicle and slows the maturation of the EV
market.
Here we address these issues by extending an existing
simulation model of the US light-duty vehicle fleet to
include behavioral feedback governing consumer choice
among BEV, plug-in hybrid electrics (PHEV), and inter-
nal combustion engine (ICE) vehicles as they depend on
attributes such as price, fuel economy, and convenience,
the latter encompassing make and model variety, charge
time and availability, range anxiety, and other factors.
We also model the dynamics of EV market formation.
Higher EV sales lead to cost reductions and performance
improvements through scale economies, learning, and
R&D; boost growth in the variety of EV makes and
models and increase EV attractiveness to a heterogeneous
driving public; create demand for deployment of ubiqui-
tous charging infrastructure, reducing range anxiety; and
increase social exposure to EVs, increasing consumer
familiarity with them and people's willingness to consider
an EV in their next purchase. These effects create
FIGURE 17 Sensitivity analysis assuming parameters affecting C4C participation are perfectly correlated
22 NAUMOV ET AL.
multiple reinforcing feedback processes driving the for-
mation and maturation of the EV market. As EV sales
grow, these feedbacks improve the attractiveness of EVs,
leading to still more sales in a virtuous cycle.
Before summarizing the results, we consider the limi-
tations of this study and opportunities for additional
research. First, our model assumes existing patterns of
car ownership and use continue in the coming decades,
even though a wide range of future worlds are possible
with the emergence of technologies such as on-demand
mobility platforms, and self-driving cars. The impact of
these technologies is highly uncertain—the emergence of
safe and shared robotaxis could accelerate the removal of
vehicles off our roads, but they could also make driving
more attractive and more accessible, leading to an
increase in VMT (Naumov et al., 2020). The COVID-19
pandemic triggered a shift to work-from-home business
models and changes in settlement patterns, which, if they
persist, could have lasting impacts on patterns of vehicle
ownership and use. Second, our model uses an aggre-
gated representation of EV market formation, abstracting
away from dynamics such as the chicken-and-egg prob-
lem around charging infrastructure rollout that could be
a further barrier to EV adoption (see e.g., Struben &
Sterman, 2008, for a spatially explicit model of the coevo-
lution of alternative fuel vehicles and fueling infrastruc-
ture). Third, our analysis does not capture variations in
policies across the US caused by different state- and
municipality-level incentives, EV mandates, and other
preferences for EVs that could influence the effectiveness
of C4C. Fourth, other policies complementary to C4C,
gas or carbon taxes, and policies to promote zero- and
low-carbon power generation could be considered,
including subsidies for automakers to retool production
lines, retrain workers, and develop the new supply chains
and vehicle servicing capacity needed for EVs. Fifth,
other policies can increase the relative attractiveness of
EVs, including tolls and congestion pricing, registration
fees and taxes, in which EVs would be charged less or be
exempt. EVs can also be given access to HOV lanes with-
out passengers. Such policies are more common in
Europe than in the United States (e.g., “the polluter pays
principle”in Norway (Elbil, 2021b) and London's “con-
gestion charge”(Errity, 2021)), though some (e.g., HOV
lane access for low-emission vehicles) have been success-
fully deployed in various US states (AFDC, 2021). We
welcome additional research that addresses these
topics and brings climate-oriented C4C closer to
implementation.
Notwithstanding these opportunities for model exten-
sions, we find that C4C policies can contribute signifi-
cantly to US automotive emissions reductions. However,
the impact depends strongly on program characteristics.
The most important by far: we find C4C policies should
require trade-ins under the program be replaced with
electric vehicles, even with the stronger CAFE standards
implemented in 2021 and the assumption of continued
drops in ICE vehicle emissions per mile through 2050.
Requiring replacement vehicles purchased under C4C be
electric (“C4C EV”) reduces overall program participa-
tion, but has two critical benefits. First, C4C EV boosts
EV sales more than when participants can purchase an
EV or ICE vehicle (“C4C ICE/EV”), and EVs reduce
emissions more than ICE vehicles, even with the current
emissions intensity of the electric power system. Second,
and far more important, by increasing EV sales in the
near term, C4C EV speeds and strengthens the multiple
reinforcing feedback processes that accelerate the growth
of the EV market and supply chain. Consequently, C4C
EV increases EV sales far more than the direct impact it
has on vehicle replacement by bootstrapping the matura-
tion of the EV market, cutting emissions significantly
even after the C4C program ends and lowering the cost
per tonne of avoided emissions.
The impact of C4C EV is further enhanced when
deployed together with complementary policies including
a price on carbon (or gas tax), and policies that accelerate
the decarbonization of the electric power system. The
joint impact of C4C and these policies is larger than the
sum of their individual impacts because of the additional
synergy they generate by further promoting the multiple
market formation feedbacks described above.
How cost-effective could C4C programs be? The cost
per tonne of avoided emissions ranges from $124/tCO
2
for C4C EV with an $8000 per vehicle inventive to $61/
tCO
2
when deployed with the carbon price and acceler-
ated transition to renewable electricity (and with a $4000
incentive, $56/tCO
2
alone and just $24/tCO
2
, with the
complementary policies). These costs are within the
range of values for the SCC estimated by the EPA during
the Obama administration, $14–$138/tCO
2
in 2007$ (U.S.
EPA, 2016), $18–$175/tCO
2
in 2020$ using the CPI).
However, several considerations suggest those estimates
of the SCC are too low. First, SCC estimates are highly
sensitive to the social discount rate (SDR); EPA used
SDRs from 2.5 to 5%/year, while Drupp et al. (2018)
found “more than three-quarters”of experts surveyed
found “the median risk-free SDR of 2 percent accept-
able.”Lower social discount rates increase the SCC. Sec-
ond, SCC estimates are highly sensitive to the
assumed climate damage function. Later work
(e.g., Weitzman, 2010; Burke et al., 2015; Dietz &
Stern, 2015) supports damage functions and SCC esti-
mates much higher than those used by EPA a decade
ago. Third, models accounting for uncertainty in key cli-
mate and economic parameters, the possibility of
NAUMOV ET AL.23
crossing climate tipping points, and risks that emissions
reductions may be further delayed by technical, political,
or other factors yield much higher estimates of SCC:
Dietz et al. (2021) found that the risk of climate tipping
points, such as thawing permafrost, ice sheet disintegra-
tion, and changes in atmospheric circulation, increase
the expected value of SCC by about 25%. Daniel
et al. (2019), accounting for uncertainty in key climate
and economic parameters, and risks of delays in emis-
sions reductions estimate SCC through 2050 to be
approximately $100–$150/tCO
2
, even for low-end damage
functions and social discount rates >2%/year.
Bressler (2021) finds that including the “mortality cost of
carbon”—the economic value of the additional deaths
caused by the warming induced by a tonne of CO
2
emissions—yields an SCC up to $295/tCO
2
, while Kikstra
et al. (2021), accounting for climate-economy feedbacks
and temperature variability, estimate the SCC to be $307/
tCO
2
(interquartile range $147–$349/tCO
2
). We conclude
that properly designed C4C policies offer emissions
reductions at costs lower than or comparable to the value
of the avoided climate damage.
Additional complementary policies not examined
here could further reduce the costs of emissions reduc-
tions. Plautz (2021) reports that electrifying federal, state,
city, and postal service fleets could yield about $4 billion
in savings, savings that could be used to offset the costs
of C4C policies aimed at the general population. Even
more important, electrifying government fleets would
further speed EV market development, enhancing the
synergy arising from the reinforcing market formation
feedbacks and leading to larger synergies from C4C.
Even if C4C policies are cost effective, the distribu-
tional and equity impacts of these, and any, climate poli-
cies must be considered. Gas taxes and carbon prices are
often seen as regressive, disproportionately harming
lower-income individuals. The distributional impact of a
gas tax or carbon price, however, depends on the fate of
the revenue. Many proposals, across the political spec-
trum, call for 100% of the revenue be rebated to the
public as a “carbon dividend.”These include the Republi-
can Baker-Schultz plan (e.g., Climate Leadership
Council, 2019), the non-partisan Citizens Climate Lobby
plan (CCL, 2021), and the Energy Innovation and Carbon
Dividend Act (H.R. 2307, U.S. Congress, 2021), co-
sponsored by 40 Democratic representatives as of 5 May
2021, among others (Reynolds, 2021). The gasoline con-
sumption and carbon footprints of low-income individ-
uals are lower than those of the affluent. Consequently,
the net cost of a gas or carbon tax with dividend is low or
negative for lower-income individuals and positive for
the affluent. C4C programs will also primarily benefit
more affluent individuals who buy the majority of new
cars, while low-income individuals tend to purchase used
vehicles or forgo car ownership altogether, instead of
relying on public transportation. However, by accelerat-
ing fleet turnover, C4C policies speed reductions in
harmful tailpipe emissions. The adverse health impacts
of these emissions are disproportionately borne by the
poor and especially by people of color (Tessum
et al., 2021). Reductions in fossil fuel use induced by C4C
help these groups by reducing morbidity, mortality, days
of lost income, and health care costs, among other co-
benefits.
Intuition suggests that C4C programs should focus on
the oldest, most polluting vehicles, with relatively new
vehicles ineligible for the program. We find, however, that
C4C programs requiring the replacement vehicle be elec-
tric yield larger emissions reductions at reasonable cost
per tonne even when all existing ICE vehicles are eligible:
the emissions reductions of EVs are large even compared
with the newest ICE vehicles, and, by increasing participa-
tion, universal eligibility further speeds the maturation of
the EV market, boosting the synergies created by the rein-
forcing market formation feedbacks around EV costs,
range, performance, make and model variety, charging
infrastructure, and consumer awareness.
The potential to implement an environmental policy
that stimulates manufacturing employment is a particu-
larly appealing aspect of C4C. Our simulations show that
C4C will lead to an immediate increase in new vehicle
sales, and a permanent increase in cumulative produc-
tion even after C4C ends. However, the physics of fleet
turnover cannot be avoided. C4C accelerates trade-ins,
causing a surge in new vehicle sales when the program
starts, potentially stressing manufacturing capacity and
creating supply chain bottlenecks. Such bottlenecks
would be most severe for the EV supply chain, particu-
larly under C4C EV. Policymakers could mitigate the
potential for transient bottlenecks when the program
begins through gradual phase in, for example, by starting
the program with eligibility restricted to older vehicles
and expanding it over a few years. Likewise, a temporary
drop in sales is inevitable when the program ends. How-
ever, running C4C over a decade, as in our simulations,
both increases the cumulative emissions reductions from
the program and reduces the pull-forward effect and pos-
sibilities for strategic purchase timing by consumers,
reducing the sales decline when the program ends. The
temporary sales drop at program end and its transient
impacts on automakers and the supply chain could be
mitigated by, for example, gradually reducing the incen-
tive or eligibility. Coordination among government, auto-
makers, and suppliers will be important in any scenario.
The results also have methodological implications.
The large synergy created by the market formation
24 NAUMOV ET AL.
feedbacks means models that assume exogenous rates of
improvement and adoption of new technologies will
underestimate the impacts of policies such as C4C. Effec-
tive models should capture fleet turnover and other con-
straints imposed by the “physics”of the system, but
should also have a broad boundary that includes the
many endogenous feedback processes that condition con-
sumer choice and manufacturer behavior. Although we
examined the automobile industry here, the need for a
broad boundary including such behavioral feedback
would apply to models designed for any new technologies
and novel markets.
Limiting global warming and avoiding the worst
impacts of climate change require global GHG emis-
sions fall as quickly and deeply as possible. Transporta-
tion constitutes a major source of emissions, but the
long lifetime of vehicles means emissions reduction tar-
gets cannot be met even assuming optimistic reductions
in new vehicle emissions and a rapid increase in the
market share of zero-tailpipe emissions vehicles. C4C
programs designed to speed the transition to electric
vehicles can speed emissions reductions and bootstrap
the maturation of a robust zero-tailpipe emissions vehi-
cle fleet, automobile industry, and supply chain, helping
to make the transition to sustainable transportation a
reality.
OPEN RESEARCH BADGES
This article has been awarded Open Data Badge for
making publicly available the digitally-shareable data
necessary to reproduce the reported results. Data is avail-
able at Open Science Framework
This article has been awarded Open Materials Badge
for making the components of the research methodology
needed to reproduce the reported procedure and analysis
publicly available at Open Science Framework
ORCID
Sergey Naumov https://orcid.org/0000-0003-1882-1854
John D. Sterman https://orcid.org/0000-0001-7476-6760
REFERENCES
Abrams, B. A., & Parsons, G. R. (2009). Is CARS a clunker? The
Economists' Voice,6(8), 1–4. https://doi.org/10.2202/1553-3832.
1638
Adamson, M., & Roper, I. (2019). ‘Good’jobs and ‘bad’jobs: Con-
templating job quality in different contexts. Work, Employment
and Society,33(4), 551–559. https://doi.org/10.1177/
0950017019855510
AFDC. 2020. Data Sources and Assumptions for the Electricity
Sources and Emissions Tool. Retrieved from https://afdc.
energy.gov/vehicles/electric_emissions_sources.html
AFDC. 2021. Alternative Fuel Vehicles and High Occupancy Vehicle
Lanes. Retrieved September 12, 2021, from https://afdc.energy.
gov/laws/HOV
Alberini, A., Harrington, W., & McConnell, V. (1995). Determinants
of participation in accelerated vehicle-retirement programs.
The Rand Journal of Economics,26(1), 93. https://doi.org/10.
2307/2556037
ANL. (2020). The greenhouse gases, regulated emissions, and energy
use in technologies model. The Argonne National Laboratory's
Systems Assessment Center. Retrieved from. https://greet.es.
anl.gov/
API. 2021. API Outlines Path for Low-Carbon Future in New Climate
Action Framework. Retrieved from https://www.api.org/news-
policy-and-issues/news/2021/03/24/climate-action-framework
ARC. 2012. Retire Your Ride. Retrieved May 3, 2021, from https://
retireyourride.ca/
Archer, N. P., & Wesolowsky, G. O. (1996). Consumer response to
service and product quality: A study of motor vehicle owners.
Journal of Operations Management,14(2), 103–118. https://doi.
org/10.1016/0272-6963(95)00045-3
Argote, L., & Epple, D. (1990). Learning curves in manufacturing.
Science,247(4945), 920–924. https://doi.org/10.1126/science.
247.4945.920
Auto Alliance. 2018. Auto Alliance Advanced Technology Vehicle
Sales Dashboard. Retrieved from https://autoalliance.org/
energy-environment/advanced-technology-vehicle-sales-
dashboard/
Automotive News. 2019. Automotive News Data Center. Retrieved
from https://www.autonews.com/section/data-center
BAR. 2021. Consumer Assistance Program. Retrieved September
6, 2021, from https://www.bar.ca.gov/consumer/consumer_
assistance_program/
BBC. 2009. Scrappage scheme to be extended. Retrieved May 3, 2021,
from http://news.bbc.co.uk/2/hi/business/8278679.stm
Ben-Akiva, M., & Lerman, S. R. (1985). Discrete choice analysis: The-
ory and application to travel demand. MIT Press.
Bendoly, E., Donohue, K., & Schultz, K. L. (2006). Behavior in oper-
ations management: Assessing recent findings and revisiting
old assumptions. Journal of Operations Management,24(6),
737–752. https://doi.org/10.1016/j.jom.2005.10.001
Böckers, V., U. Heimeshoff, A. Müller. 2012. Pull-forward effects in the
German car scrappage scheme: A time series approach PDF logo.
Retrieved from https://www.econstor.eu/handle/10419/59001
Bressler, R. D. (2021). The mortality cost of carbon. Nature Commu-
nications,12(1), 4467. https://doi.org/10.1038/s41467-021-
24487-w
Brownstone, D., Bunch, D. S., & Train, K. (2000). Joint mixed logit
models of stated and revealed preferences for alternative-fuel
vehicles. Transportation Research Part B: Methodological,34(5),
315–338. https://doi.org/10.1016/S0191-2615(99)00031-4
Burke, M., Hsiang, S. M., & Miguel, E. (2015). Global non-linear
effect of temperature on economic production. Nature,
527(7577), 235–239. https://doi.org/10.1038/nature15725
CARB. 2018. California applauds multi-state coalition's releases new
Zero-Emission Vehicle Action Plan. Retrieved from https://ww2.
arb.ca.gov/news/california-applauds-multi-state-coalitions-
releases-new-zero-emission-vehicle-action-plan
CARB. 2019. Zero-Emission Vehicle Program. Retrieved April
18, 2019, from https://ww2.arb.ca.gov/our-work/programs/
zero-emission-vehicle-program/about
NAUMOV ET AL.25
CCL. 2021. The Basics of Carbon Fee and Dividend. Retrieved May
5, 2021, from https://citizensclimatelobby.org/basics-carbon-
fee-dividend/
Clean Air Task Force. 2021. Clean electricity payment program.
Climate Leadership Council. 2019. The Four Pillars of Our Carbon
Dividends Plan. Retrieved May 5, 2021, from https://clcouncil.
org/our-plan/
Coltman, T., & Devinney, T. M. (2013). Modeling the operational
capabilities for customized and commoditized services. Journal
of Operations Management,31(7–8), 555–566. https://doi.org/
10.1016/j.jom.2013.09.002
Daniel, K. D., Litterman, R. B., & Wagner, G. (2019). Declining CO
2 price paths. Proceedings of the National Academy of Sciences,
116(42), 20886–20891. https://doi.org/10.1073/pnas.1817444116
Davis, S., & Boundy, R. (2021). Transportation energy data book
(39th ed.). Oak Ridge, . Retrieved from. https://www.osti.gov/
servlets/purl/1767864/
Dietz, S., Rising, J., Stoerk, T., & Wagner, G. (2021). Economic
impacts of tipping points in the climate system. Proceedings of
the National Academy of Sciences,118(34), e2103081118.
https://doi.org/10.1073/pnas.2103081118
Dietz, S., & Stern, N. (2015). Endogenous growth, convexity of dam-
age and climate risk: How Nordhaus' framework supports deep
cuts in carbon emissions. The Economic Journal,125(583), 574–
620. https://doi.org/10.1111/ecoj.12188
Dill, J. (2004). Estimating emissions reductions from accelerated
vehicle retirement programs. Transportation Research Part D:
Transport and Environment,9(2), 87–106. https://doi.org/10.
1016/S1361-9209(03)00072-5
Drake, D. F., Kleindorfer, P. R., & Van Wassenhove, L. N. (2016).
Technology choice and capacity portfolios under emissions reg-
ulation. Production and Operations Management,25(6), 1006–
1025. https://doi.org/10.1111/poms.12523
Drupp, M. A., Freeman, M. C., Groom, B., & Nesje, F. (2018). Dis-
counting disentangled. American Economic Journal: Economic
Policy,10(4), 109–134. https://doi.org/10.1257/pol.20160240
EEA. 2021. New registrations of electric vehicles in Europe. Retrieved
May 6, 2021, from https://www.eea.europa.eu/data-and-maps/
indicators/proportion-of-vehicle-fleet-meeting-5/assessment
Elbil. 2021a. Norwegian EV market. Retrieved February 8, 2022,
from https://elbil.no/english/norwegian-ev-market/
Elbil. 2021b. Norwegian EV policy. https://elbil.no/english/
norwegian-ev-policy/
Errity, S. 2021. Congestion Charge exempt cars: do electric cars have
to pay? Retrieved September 12, 2021, from https://www.
drivingelectric.com/your-questions-answered/90/congestion-
charge-exempt-cars-do-electric-cars-have-pay
European Comission. 2021. European Green Deal: Commission pro-
poses transformation of EU economy and society to meet climate
ambitions. Retrieved September 19, 2021, from https://ec.
europa.eu/commission/presscorner/detail/en/ip_21_3541
Farmer, J. D., Hepburn, C., Ives, M. C., Hale, T., Wetzer, T.,
Mealy, P., …Way, R. (2019). Sensitive intervention points in
the post-carbon transition. Science,364(6436), 132–134. https://
doi.org/10.1126/science.aaw7287
Federal Register. 2021. 2021 Joseph R. Biden Jr. Executive Orders.
Retrieved September 19, 2021, from https://www.
federalregister.gov/presidential-documents/executive-orders/
joe-biden/2021
Ge, Y., MacKenzie, D., & Keith, D. R. (2018). Gas anxiety and the
charging choices of plug-in hybrid electric vehicle drivers.
Transportation Research Part D: Transport and Environment,
64, 111–121. https://doi.org/10.1016/j.trd.2017.08.021
Gillingham, K., A. Munk-Nielsen. 2016. A tale of two tails: Com-
muting and the fuel Price response in driving. NBER Working
Paper (December). Retrieved from http://www.nber.org/
papers/w22937
Gohlke, D., Y. Zhou. 2020. Assessment of light-duty plug-in electric
vehicles in the United States, 2010–2019. Argonne, IL
(United States). Retrieved from http://www.osti.gov/servlets/
purl/1642114/
G
omez Vilchez, J. J., & Jochem, P. (2020). Powertrain technologies and
their impact on greenhouse gas emissions in key car markets.
Transportation Research Part D: Transport and Environment,80-
(July 2019), 102214. https://doi.org/10.1016/j.trd.2019.102214
Hammar, H., Löfgren, Å., & Sterner, T. (2004). Political economy
obstacles to fuel taxation. The Energy Journal,25(3), 1–17.
https://doi.org/10.5547/ISSN0195-6574-EJ-Vol25-No3-1
Hardman, S., Kurani, K. S., & Chakraborty, D. (2020). The usual
policy levers are not engaging consumers in the transition to
electric vehicles: A case of Sacramento. California. Environmen-
tal Research Communications,2(8), 085001. https://doi.org/10.
1088/2515-7620/aba943
Hauser, J. R., & Wernerfelt, B. (1990). An evaluation cost model of
consideration sets. Journal of Consumer Research,16(4), 393.
Hoekstra, M., Puller, S. L., & West, J. (2017). Cash for corollas: When
Stimulus reduces spending. American Economic Journal: Applied
Economics,9(3), 1–35. https://doi.org/10.1257/app.20150172
Holweg, M., & Pil, F. K. (2008). Theoretical perspectives on the
coordination of supply chains. Journal of Operations Manage-
ment,26(3), 389–406. https://doi.org/10.1016/j.jom.2007.08.003
HybridCars.com. 2018. Hybrid Market Dashboard. Retrieved from
http://www.hybridcars.com/market-dashboard/
Jenn, A., Azevedo, I. L., & Michalek, J. J. (2019). Alternative-fuel-
vehicle policy interactions increase U.S. greenhouse gas emis-
sions. Transportation Research Part A: Policy and Practice,124-
(April), 396–407. https://doi.org/10.1016/j.tra.2019.04.003
Kalkanci, B., Rahmani, M., & Toktay, L. B. (2019). The role of
inclusive innovation in promoting social sustainability. Produc-
tion and Operations Management,28(12), 2960–2982. https://
doi.org/10.1111/poms.13112
Keith, D. R., Houston, S., & Naumov, S. (2019). Vehicle fleet turn-
over and the future of fuel economy. Environmental Research
Letters,14(2), 021001. https://doi.org/10.1088/1748-9326/aaf4d2
Keith, D. R., Naumov, S., & Sterman, J. D. (2017). Driving the
future: A management flight simulator of the US automobile
market. Simulation & Gaming,48(6), 735–769. https://doi.org/
10.1177/1046878117737807
Keith, D. R., Struben, J. J. R., & Naumov, S. (2020). The diffusion of
alternative fuel vehicles: A generalized model and future
research agenda. Journal of Simulation,1–18, 260–277. https://
doi.org/10.1080/17477778.2019.1708219
Khurana, A., Kumar, V. V. R., & Sidhpuria, M. (2020). A study on
the adoption of electric vehicles in India: The mediating role of
attitude. Vision: The Journal of Business Perspective,24(1), 23–
34. https://doi.org/10.1177/0972262919875548
Kikstra, J. S., Waidelich, P., Rising, J., Yumashev, D., Hope, C., &
Brierley, C. M. (2021). The social cost of carbon dioxide under
26 NAUMOV ET AL.
climate-economy feedbacks and temperature variability. Envi-
ronmental Research Letters,16(9), 094037. https://doi.org/10.
1088/1748-9326/ac1d0b
Kim, H. C., Keoleian, G. A., Grande, D. E., & Bean, J. C. (2003). Life
cycle optimization of automobile replacement: Model and
application. Environmental Science & Technology,37(23), 5407–
5413. https://doi.org/10.1021/es0345221
Kleindorfer, P. R., Singhal, K., & Wassenhove, L. N. (2009). Sustain-
able Operations Management. Production and Operations Man-
agement,14(4), 482–492. https://doi.org/10.1111/j.1937-5956.
2005.tb00235.x
Knittel, C. R. (2010). The implied cost of carbon dioxide under the
cash for clunkers program. SSRN Electronic Journal. https://
doi.org/10.2139/ssrn.1630647
Krass, D., Nedorezov, T., & Ovchinnikov, A. (2013). Environmental
taxes and the choice of green technology. Production and Oper-
ations Management,22(5), 1035–1055. https://doi.org/10.1111/
poms.12023
Lachapelle, U. (2015). Using an accelerated vehicle retirement pro-
gram (AVRP) to support a mode shift: Car purchase and modal
intentions following program participation. Journal of Trans-
port and Land Use,8(2), 107–123. https://doi.org/10.5198/jtlu.
2015.659
Lattanzio, R. K., L. Tsang, B. Canis. 2020. Vehicle fuel economy
and greenhouse gas standards: Frequently asked questions. Key
Congressional Reports for August 2019: Part IV.
Lenski, S. M., Keoleian, G. A., & Bolon, K. M. (2010). The impact of
‘cash for clunkers’on greenhouse gas emissions: A life cycle
perspective. Environmental Research Letters,5(4), 044003.
https://doi.org/10.1088/1748-9326/5/4/044003
Lenski, S. M., Keoleian, G. A., & Moore, M. R. (2013). An assess-
ment of two environmental and economic benefits of ‘Cash for
Clunkers’.Ecological Economics,96, 173–180. https://doi.org/
10.1016/j.ecolecon.2013.10.011
Li, S., Linn, J., & Spiller, E. (2013). Evaluating “cash-for-clunkers”:
Program effects on auto sales and the environment. Journal of
Environmental Economics and Management,65(2), 175–193.
https://doi.org/10.1016/j.jeem.2012.07.004
Li, S., Zhu, X., Ma, Y., Zhang, F., & Zhou, H. (2021). The role of gov-
ernment in the market for electric vehicles: Evidence from China.
SSRN Electronic Journal. https://doi.org/10.2139/ssrn.3908011
Marin, G., & Zoboli, R. (2020). Effectiveness of car scrappage
schemes: Counterfactual-based evidence on the Italian experi-
ence. Economics of Transportation,21, 100150. https://doi.org/
10.1016/j.ecotra.2019.100150
McFadden, D. (1981). Econometric models of probabilistic choice. In
C. F. Manski & D. L. McFadden (Eds.), Structural analysis of dis-
crete data and econometric applications (pp. 198–272). MIT Press.
Mian, A., & Sufi, A. (2012). The effects of fiscal Stimulus: Evidence
from the 2009 cash for clunkers program. The Quarterly Journal
of Economics,127(3), 1107–1142. https://doi.org/10.1093/qje/
qjs024
Miller, K. S., Wilson, W. W., & Wood, N. G. (2020). Environmental-
ism, stimulus, and inequality reduction through industrial pol-
icy: Did cash for clunkers achieve the trifecta? Economic
Inquiry,58(3), 1109–1128. https://doi.org/10.1111/ecin.12889
Morecroft, J. D. W. (1985). Rationality in the analysis of behavioral
simulation models. Management Science,31(7), 900–916.
https://doi.org/10.1287/mnsc.31.7.900
Morrison, G. M., Allan, A., & Carpenter, R. (2010). Abating green-
house gas emissions through cash-for-clunker programs. Trans-
portation Research Record: Journal of the Transportation
Research Board,2191(1), 111–118. https://doi.org/10.3141/
2191-14
Naumov, S., Keith, D. R., & Fine, C. H. (2020). Unintended conse-
quences of automated vehicles and pooling for urban transpor-
tation systems. Production and Operations Management,29(5),
1354–1371. https://doi.org/10.1111/poms.13166
Naumov, S., Keith, D. R., & Sterman, J. D. (2022). Accelerating vehi-
cle fleet turnover to achieve sustainable mobility goals: Model
and data. Ann Arbor, MI, Inter-university Consortium for
Political and Social Research [distributor]. https://doi.org/10.
3886/E161761V4
NHTSA. 2006. Vehicle survivability and travel mileage schedules.
National Highway Traffic Safety Administration 22. HS 809 952
NHTSA. 2021a. CAFE - Fuel Economy. Retrieved from http://www.
nhtsa.gov/fuel-economy
NHTSA. 2021b. Technical Support Document: Proposed Rulemaking
for Model Years 2024–2026 Light- Duty Vehicle Corporate Aver-
age Fuel Economy Standards. Retrieved from https://www.
nhtsa.gov/sites/nhtsa.gov/files/2021-08/CAFE-NHTSA-2127-
AM34-TSD-Complete-web-tag.pdf
Plautz, J. 2021. Electrifying 97% of the federal fleet by 2030 could
save billions: report. Retrieved September 6, 2021, from https://
www.smartcitiesdive.com/news/electrifying-97-of-the-federal-
fleet-by-2030-could-save-billions-report/605228/
Plumer, B., N. Popovich, B. Migliozzi. 2021, March 10. Electric cars
are coming. How long until they rule the road? The New York
Times. Retrieved from https://www.nytimes.com/interactive/
2021/03/10/climate/electric-vehicle-fleet-turnover.html?
action=click&module=TopStories&pgtype=Homepage
Pullman, M. E., Verma, R., & Goodale, J. C. (2001). Service design
and operations strategy formulation in multicultural markets.
Journal of Operations Management,19(2), 239–254. https://doi.
org/10.1016/S0272-6963(00)00059-0
Rahmandad, H., & Sterman, J. D. (2012). Reporting guidelines for
simulation-based research in social sciences. System Dynamics
Review,28(4), 396–411. https://doi.org/10.1002/sdr.1481
Reynolds, M. 2021. For a win on climate, let's put our best player in
the game. Retrieved May 5, 2021, from https://thehill.com/
blogs/congress-blog/energy-environment/548801-for-a-win-on-
climate-lets-put-our-best-player-in-the
Rogers, D. M., P. A. Schmid. 2021. Massachusetts H4082. An act rel-
ative to retiring high emissions vehicles. Retrieved September
6, 2021, from https://trackbill.com/bill/massachusetts-house-
bill-4082-an-act-relative-to-retiring-high-emissions-vehicles/
2140250/
Rothstein, J. S. (2016). When good jobs go bad: Globalization, de-
unionization, and declining job quality in the North American
auto industry. Rutgers University Press.
Shafiei, E., Thorkelsson, H.,
Asgeirsson, E. I., Davidsdottir, B.,
Raberto, M., & Stefansson, H. (2012). An agent-based modeling
approach to predict the evolution of market share of electric
vehicles: A case study from Iceland. Technological Forecasting
and Social Change,79(9), 1638–1653. https://doi.org/10.1016/j.
techfore.2012.05.011
Shah, A., C. Monnappa. 2021. India to proceed soon with “cash for
clunkers”scheme to boost car sales. Reuters. Retrieved from
NAUMOV ET AL.27
https://www.reuters.com/article/india-budget-autos/india-to-
proceed-soon-with-cash-for-clunkers-scheme-to-boost-car-sales-
idUSL4N2K73LV
Shetty, D. K., Shetty, S., Raj Rodrigues, L., Naik, N., Maddodi, C. B.,
Malarout, N., & Sooriyaperakasam, N. (2020). Barriers to wide-
spread adoption of plug-in electric vehicles in emerging Asian
markets: An analysis of consumer behavioral attitudes and per-
ceptions. Cogent Engineering,7(1), 1796198. https://doi.org/10.
1080/23311916.2020.1796198
Silvia, C., & Krause, R. M. (2016). Assessing the impact of policy
interventions on the adoption of plug-in electric vehicles: An
agent-based model. Energy Policy,96, 105–118. https://doi.org/
10.1016/j.enpol.2016.05.039
Simon, B. 2009, August 17. Cash-for-clunkers boost Japanese car
sales. Financial Times. Toronto. Retrieved from https://www.ft.
com/content/e4a67592-8b74-11de-9f50-00144feabdc0
Sivak, M., B. Schoettle. 2009. The effect of the “cash for clunkers”
program on the Overall fuel economy of purchased new vehi-
cles. Ann Arbor, MI.
Spitzley, D. V., Grande, D. E., Keoleian, G. A., & Kim, H. C. (2005).
Life cycle optimization of ownership costs and emissions reduc-
tion in US vehicle retirement decisions. Transportation
Research Part D: Transport and Environment,10(2), 161–175.
https://doi.org/10.1016/j.trd.2004.12.003
Statistics Norway. 2021. Electricity prices. Retrieved May 5, 2021,
from https://www.ssb.no/en/elkraftpris
Sterman, J. D. (2000). Business dynamics: Systems thinking and
modeling for a complex world. Irwin McGraw-Hill. https://doi.
org/10.1057/palgrave.jors.2601336
Sterman, J. D., & Dogan, G. (2015). I'm not hoarding, I'm just stocking
up before the hoarders get here. Journal of Operations Manage-
ment,39–40(1), 6–22. https://doi.org/10.1016/j.jom.2015.07.002
Sterman, J. D., Oliva, R., Linderman, K., & Bendoly, E. (2015). Sys-
tem dynamics perspectives and modeling opportunities for
research in operations management. Journal of Operations Man-
agement,39–40(1), 1–5. https://doi.org/10.1016/j.jom.2015.07.001
Struben, J. (2006). Essays on transition challenges for alternative pro-
pulsion vehicles and transportation systems. Massachusetts Insti-
tute of Technology. Retrieved from. https://dspace.mit.edu/
handle/1721.1/37159
Struben, J., & Sterman, J. D. (2008). Transition challenges for alter-
native fuel vehicle and transportation systems. Environment
and Planning. B, Planning & Design,35(6), 1070–1097. https://
doi.org/10.1068/b33022t
Tessum, C. W., Paolella, D. A., Chambliss, S. E., Apte, J. S.,
Hill, J. D., & Marshall, J. D. (2021). PM 2.5 polluters dispropor-
tionately and systemically affect people of color in the
United States. Science.Advances,7(18), eabf4491. https://doi.
org/10.1126/sciadv.abf4491
Ton, Z. (2014). The good jobs strategy: How the smartest companies
invest in employees to lower costs and boost profits. Houghton
Mifflin Harcourt.
U.S. Congress. 2021. Energy Innovation and Carbon Dividend Act.
Retrieved May 5, 2021, from https://www.congress.gov/bill/
117th-congress/house-bill/2307
U.S. DOE. 2019. Electric Vehicles: Tax Credits and Other Incen-
tives. Retrieved April 18, 2019, from https://www.energy.gov/
eere/electricvehicles/electric-vehicles-tax-credits-and-other-
incentives
U.S. DOT. (2009). Cash for clunkers wraps up with nearly 700,000
car sales and increased fuel efficiency, U.S. transportation secre-
tary LaHood declares program “wildly successful”.Office of
Public Affairs. Retrieved from. https://web.archive.org/web/
20091007021106/http:/www.cars.gov:80/files/08.
26PressRelease.pdf
U.S. DOT. 2020. Hybrid-Electric, Plug-in Hybrid-Electric and Electric
Vehicle Sales. Retrieved February 8, 2022, from https://www.
bts.gov/content/gasoline-hybrid-and-electric-vehicle-sales
U.S. EIA. 2018. Crossover utility vehicles overtake cars as the most
popular light-duty vehicle type. Retrieved April 15, 2019, from
https://www.eia.gov/todayinenergy/detail.php?id=36674
U.S. EIA. 2020. State Electricity Profiles. Retrieved February 8, 2022,
from https://www.eia.gov/electricity/state/
U.S. EPA. 2016. Social Cost of Carbon. Retrieved from https://www.
epa.gov/sites/production/files/2016-12/documents/social_cost_
of_carbon_fact_sheet.pdf
U.S. EPA. 2019a. Fast Facts on Transportation Greenhouse Gas Emis-
sions. Retrieved February 8, 2022, from https://www.epa.gov/
greenvehicles/fast-facts-transportation-greenhouse-gas-emissions
U.S. EPA. 2019b. Greenhouse Gas Emissions from a Typical Passenger
Vehicle. Retrieved April 15, 2019, from https://www.epa.gov/
greenvehicles/greenhouse-gas-emissions-typical-passenger-vehicle
U.S. EPA. 2021. Revised 2023 and Later Model Year Light-Duty Vehi-
cle Greenhouse Gas Emissions Standards. Retrieved September
19, 2021, from https://www.govinfo.gov/content/pkg/FR-2021-
08-10/pdf/2021-16582.pdf
UNEP. 2020. The Emissions Gap Report 2020. Nairobi. Retrieved
from https://www.unep.org/emissions-gap-report-2020
UNFCCC. 2021. The United States of America Nationally Deter-
mined Contribution. Retrieved from https://www4.unfccc.int/
sites/ndcstaging/PublishedDocuments/UnitedStatesofAmerica
First/UnitedStatesNDCApril212021Final.pdf
Van Wee, B., Moll, H. C., & Dirks, J. (2000). Environmental impact
of scrapping old cars. Transportation Research Part D: Trans-
port and Environment,5(2), 137–143. https://doi.org/10.1016/
S1361-9209(99)00030-9
Verma, R., Louviere, J. J., & Burke, P. (2006). Using a market-utility-
based approach to designing public services: A case illustration
from United States Forest Service. Journal of Operations Manage-
ment,24(4), 407–416. https://doi.org/10.1016/j.jom.2005.06.004
Vickery, S. K., Jayaram, J., Droge, C., & Calantone, R. (2003). The
effects of an integrative supply chain strategy on customer ser-
vice and financial performance: An analysis of direct versus
indirect relationships. Journal of Operations Management,
21(5), 523–539. https://doi.org/10.1016/j.jom.2003.02.002
Vlasic, B. 2008, July 2. Car sales at 10-year low. The New York
Times. Retrieved from https://www.nytimes.com/2008/07/02/
business/02auto.html
Volpe. 2015. CAFE Compliance and Effects Model. Retrieved from
http://www.nhtsa.gov/Laws+&+Regulations/CAFE+-+Fuel
+Economy/CAFE+Compliance+and+Effects+Modeling+
System:+The+Volpe+Model
Wang, T., & Chen, C. (2014). Impact of fuel price on vehicle miles
traveled (VMT): Do the poor respond in the same way as the
rich? Transportation,41(1), 91–105. https://doi.org/10.1007/
s11116-013-9478-1
Wei, W., Ramakrishnan, S., Needell, Z. A., & Trancik, J. E. (2021).
Personal vehicle electrification and charging solutions for high-
28 NAUMOV ET AL.
energy days. Nature Energy,6(1), 105–114. https://doi.org/10.
1038/s41560-020-00752-y
Weitzman,M.L.(2010).Whatisthe“damages function”for global
warming —And what difference might it make? Climate Change
Economics,01(01), 57–69. https://doi.org/10.1142/S2010007810000042
Williams, S. E., Davis, S. C., & Boundy, R. G. (2017). Transportation
energy data book: Edition 36. Oak Ridge, . Retrieved from.
http://www.osti.gov/servlets/purl/1410917/
Wu, J., Liao, H., & Wang, J.-W. (2020). Analysis of consumer atti-
tudes towards autonomous, connected, and electric vehicles: A
survey in China. Research in Transportation Economics,80,
100828. https://doi.org/10.1016/j.retrec.2020.100828
Zhang, F., Hu, X., Langari, R., & Cao, D. (2019). Energy manage-
ment strategies of connected HEVs and PHEVs: Recent pro-
gress and outlook. Progress in Energy and Combustion Science,
73, 235–256. https://doi.org/10.1016/j.pecs.2019.04.002
Zhang, L. 2009. China: “Cash for Clunkers”Programs. Retrieved
May 3, 2021, from https://loc.gov/law/foreign-news/article/
china-cash-for-clunkers-programs/
Zhou, J., Wang, J., Jiang, H., Cheng, X., Lu, Y., Zhang, W., Bi, J.,
Xue, W., & Liu, N. (2019). Cost-benefit analysis of yellow-
label vehicles scrappage subsidy policy: A case study of
Beijing-Tianjin-Hebei region of China. Journal of Cleaner Pro-
duction,232,94–103. https://doi.org/10.1016/j.jclepro.2019.
05.312
Zhu, Q., & Sarkis, J. (2004). Relationships between operational
practices and performance among early adopters of green sup-
ply chain management practices in Chinese manufacturing
enterprises. Journal of Operations Management,22(3), 265–289.
https://doi.org/10.1016/j.jom.2004.01.005
SUPPORTING INFORMATION
Additional supporting information may be found in the
online version of the article at the publisher's website.
How to cite this article: Naumov, S., Keith, D.
R., & Sterman, J. D. (2022). Accelerating vehicle
fleet turnover to achieve sustainable mobility goals.
Journal of Operations Management,1–31. https://
doi.org/10.1002/joom.1173
TABLE A1 Model parameters
Parameter Value
Initial price surcharge of EVs relative to ICE 0.25
Reference inconvenience of EVs (Car) 1
Reference inconvenience of EVs (light truck) 1.8
Weight of market maturity 2
Weight of price surcharge 1
Weight of fuel savings 1
Sensitivity of market share to utility 2
Grid electricity cost, $/kWh 0.1054 (U.S. EIA, 2020)
Current share renewable electricity 0.15
Max share renewable electricity 0.9
Learning curve strength price 0.25
Learning curve strength market formation 0.35
Aggregate market growth rate, 1/year 0.01
Elasticity of VMT to fuel price (Gillingham & Munk-Nielsen, 2016)
Hazard rate of vehicle discards, 1/year (Williams et al., 2017; Davis & Boundy, 2021)
Initial vehicle fleet, vehicles (Keith et al., 2019; Davis & Boundy, 2021)
Initial emissions, grams CO
2
/mile*vehicles (Keith et al., 2019)
Baseline fuel price, $/gallon 3.0
CAFE standards (Volpe, 2015; Keith et al., 2019; NHTSA, 2021a; U.S. EPA, 2021)
Reference C4C hazard rate 0.002
(Continues)
APPENDIX
A. Parameterization of the model
NAUMOV ET AL.29
B. Co-flow formulations
We calculate the emissions and fuel use of vehicles in the
existing fleet using standard co-flow formulations
(Sterman, 2000). The co-flow structure tracks fuel use
from each age cohort arising from age-specific fuel-effi-
ciency, improving fuel-efficiency of new vehicles, and
changing the mix of vehicle platforms. Total fleet fuel use
sums over Nindividual age cohorts a:
FUip ¼X
N
a¼1
FUipa ðB1Þ
Emissions from each cohort accumulate and deplete
following vehicle sales and retirements:
dFUip
dt ¼nipzip X
N
a¼1
ripazipa þmip zC4C
ip X
N
a¼Q
dipazipa ðB2Þ
where zip and zC4C
ip are fuel use of new vehicle sales from
natural fleet turnover and induced by the C4C program
respectively for vehicle class i, gallons per mile, and zipa
is the average fuel use of vehicles of class iwith
powertrain pin age cohort a,:
zipa ¼FUipa
Vipa
ðB3Þ
Fuel use of new EVs is a reciprocal of the average fuel
economy of a new EV which we calculate as a weighted
sum of the fuel economy of BEVs and PHEVs. The fuel
economy of the electric drive of a PHEV equal that of a
BEV, but the fuel economy of the PHEV internal com-
bustion engine is higher than that of a conventional ICE
vehicle by a factor ωeas discussed in Section 3.2. On aver-
age, fuel use of new EVs is:
zi,EV ¼1
ρþ1ρðÞλe
ðÞFEi,EV þ1ρðÞ1λe
ðÞFECAFE
iωe
ðB4Þ
with the share of total miles PHEVs drive on electricity,
λe, and the share of PHEVs in new EV sales, ρ, calculated
in Equation (19).
Fuel use of new vehicle sales, zip, is:
zip ¼
1
FECAFE
i
,#p¼ICE
zi,EV ,#p¼EV
8
<
:ðB5Þ
Average fuel use of vehicle sales induced by the C4C pro-
gram, zC4C
ip , is calculated similarly using the average fuel
economy in Equation (5):
zC4C
ip ¼
1
FEC4C
i
,#p¼ICE
zi,EV ,#p¼EV
8
<
:ðB6Þ
Fuel-related vehicle GHG emissions from the vehicle
fleet use the same co-flow structure. Total emissions sum
over Nindividual age cohorts a:
TABLE A1 (Continued)
Parameter Value
Sensitivity of C4C hazard rate to vehicle age 0.5
Reference C4C incentive value, $/vehicle $4000
Sensitivity of incentive effect 0.75
C4C policy start year 2022
C4C policy end year 2032
Minimum age to qualify for C4C, years 5
C4C stringency multiplier 1.5
Gas tax ramp start year 2022
Gas tax ramp end year 2050
Gas tax start value, $/gallon 0
Gas tax end value, $/gallon 0.3
Year of fully renewable electricity 2075
Maximum share renewables 0.9
Sensitivity of C4C stringency effect 0.75
Sensitivity of fuel savings effect 0.75
30 NAUMOV ET AL.
Eip ¼X
N
a¼1
Eipa ðB7Þ
The emissions accumulate and deplete following vehicle
sales and retirements:
dEip
dt ¼nipμip X
N
a¼1
ripaμipa þmip μC4C
ip X
N
a¼Q
dipaμipa ðB8Þ
where μip and μC4C
ip are emissions of new vehicle sales
from natural fleet turnover and induced by the C4C pro-
gram respectively for vehicle class i, in grams CO
2
per
mile, and μipa is the average emissions of vehicles of class
iwith powertrain pin age cohort a:
μipa ¼Eipa
Vipa
ðB9Þ
NAUMOV ET AL.31