eScholarship provides open access, scholarly publishing services to the University of California and delivers a dynamic research platform to scholars worldwide.
Title:
Lifecycle emission impacts of subsidies for energy efficiency: Evidence from Cash￿for￿Clunkers
Author:
Rajagopal, Deepak, University of California - Los Angeles
Publication Date:
03-22-2011
Series:
Recent Work
Publication Info:
Recent Work, Institute of the Environment and Sustainability, UC Los Angeles
Permalink:
http://escholarship.org/uc/item/7mq9j9bz
Keywords:
transportation, fuel, biofuel, energy efficiency, subsidy, pollution, lifecycle assessment
Abstract:
One popular policy option to address environmental and economic concerns arising from current patterns of energy use is to subsidize increase in energy efficiency or renewable energy. In this paper we evaluate the lifecycle environmental bene￿ts of the Car Allowance Rebate System
(CARS) (or commonly `Cash-for-clunkers') which provided a subsidy for voluntary early retirement and replacement of low fuel economy automobiles with new, higher fuel economy vehicles. We
￿nd that the estimates of bene￿ts hinge crucially on the assumption about what type of vehicle would have been purchased in the counterfactual scenario. Estimates are signi￿cantly less sensitive to assumptions about the remaining useful life of vehicles traded-in and the rebound e￿ect. Our prediction is that CARS program lead to a reduction of 9.1 to 17.8 million metric tonnes in greenhouse gas (GHG) emissions and 850 to 1600 million gallon reduction in gasoline use over a 13 year period. The average subsidy per tonne of avoided GHG lies between $142 and $278.
Disaggregation of bene￿ts based on the fuel economy of the clunker reveals opportunities for better aligning incentives and program bene￿ts in future
UNIVERSITY OF CALIFORNIA, LOS ANGELES
Working Paper Series
Lifecycle emission impacts of subsidies for energy efficiency: Evidence from Cash‐for‐Clunkers
D. Rajagopal
Mar 2011
WP#7
Lifecycle emission impacts of subsidies for energy eciency:
Evidence from Cash-for-Clunkers
D Rajagopal
Abstract
One popular policy option to address environmental and economic concerns arising from current patterns of energy use is to subsidize increase in energy eciency or renewable energy.
In this paper we evaluate the lifecycle environmental benets of the Car Allowance Rebate
System (CARS) (or commonly `Cash-for-clunkers') which provided a subsidy for voluntary early retirement and replacement of low fuel economy automobiles with new, higher fuel econ- omy vehicles. We nd that the estimates of benets hinge crucially on the assumption about what type of vehicle would have been purchased in the counterfactual scenario. Estimates are signi cantly less sensitive to assumptions about the remaining useful life of vehicles traded-in and the rebound eect. Our prediction is that CARS program lead to a reduction of 9.1 to
17.8 million metric tonnes in greenhouse gas (GHG) emissions and 850 to 1600 million gallon reduction in gasoline use over a 13 year period. The average subsidy per tonne of avoided
GHG lies between $142 and $278. Disaggregation of benets based on the fuel economy of the clunker reveals opportunities for better aligning incentives and program benets in future.
Keywords: transportation, fuel, biofuel, energy eciency, subsidy, pollution, lifecycle assess- ment. 1 Introduction
Current patterns of energy use are responsible for a range of societal concerns such as energy inse- curity, resource depletion, pollution and high cost of energy. While there exist a variety of dierent ways in which public policy can mitigate these concerns and with dierent implications for e- ciency and equity, such as establishing a fee for pollution, setting a cap on pollution and allocating tradable pollution rights, setting standards, mandating new technologies, providing information etc., a popular approach is to subsidize an increase in energy eciency or renewable/clean energy.
One such subsidy is to rebate the voluntary early replacement of energy intensive durable goods with more energy ecient substitutes. In this paper we evaluate the lifecycle environmental bene- ts of one such program, namely, Car Allowance Rebate System (CARS), commonly referred to as
`Cash-for-clunkers'. CARS was created with the passage of the Consumer Assistance to Recycle and Save Act in 20091. The program aimed to both provide a stimulus to the U.S. economy at a time of recession by boosting auto sales, and also reduce pollution from road transportation by causing the early retirement of fuel inecient vehicles or `clunker' with more ecient vehicles. To
Institute of Environment, University of California, Los Angeles, rdeepak@ioe.ucla.edu (corresponding author)
1www.cbo.gov/doc.cfm?index=10323
1 this end, it oered a rebate of either $3500 or $4500 depending the fuel-economy of the clunker and the increase in the fuel economy as a result of the trade-in. Government records indicate that as of
October 23, 2009 more than 677,081 clunkers were retired in exchange for a total rebate amount to
$2.85 billion. Judging by the consumer response, one may describe the CARS program as a success.
The economic and environmental merits of CARS are both however a topic of controversy with critics pointing out ineciencies resulting from either the premature abandonment of functioning goods from future use (Abrams and Parsons, 2009, Miron, 2011) or subsidizing purchases that would have regardless of the policy (Miron, 2011); incentives for getting rid of a rarely used vehicle
(Dill, 2001), higher driving than otherwise (Knittel, 2009); minimal to no eect of the economy and jobs (Mian and Su, 2010); and the high cost of greenhouse gas (GHG) abatement (Sachs, 2009).
Others however argue that the program was successful in simultaneously aiding economic recovery and job creation(Ching et al., 2010), achieving energy and environmental goals and demonstrating that rapid change toward sustainability is possible(Tyrrell and Dernbach, 2011).
The focus of this paper is on estimating the lifecycle impact of CARS on air emissions (both
GHG and non-GHG) and fuel consumption, and the subsidy provided per unit of avoided GHG emissions. There exists a large literature analyzing both the environmental impacts of CARS and previous experience with accelerated vehicle retirement programs. Lenski et al. (2010) calculate that the CARS avoided 4.4 million metric tonnes of carbon di-oxide equivalent (tCO2e) emissions.
Dierent from most of the earlier literature Lenski et al. compute the lifecycle eect by taking into account additional emissions attributable to vehicle manufacturing and disposal. Using the average fuel economy of the clunker, the average fuel economy of the new vehicle purchased, Abrams and
Parsons (2009) hypothesize that CARS avoided about 840 gallons per year per vehicle over the three remaining years of the average clunker while Sachs (2009) predicts 1390 gallons per year per vehicle over a ve period. The National Highway Trac Safety Administration (2009) estimates a total savings of 823 million gallons of gasoline, and avoided lifecycle GHG emissions of about 9.5 million metric tons over a 25 year period. Voluntary accelerated vehicle retirement programs have also previously been implemented by several national and sub-national agencies. A survey of several programs in the U.S. and abroad, suggests that unlike CARS which was designed as an economic stimulus, reduce fuel use to improve energy security and to reduce GHG emissions, the former were focussed on reducing air pollutants such as volatile organic compounds (VOC), nitrogen oxide
(NOx), carbon monoxide emissions (CO) and particulate matter (PM) (Dill, 2004). The survey reveals that while vehicle retirement programs may likely reduce emissions, emissions of NOx and
CO did not decline as expected. The survey also suggests that previous U.S. programs generally did not attempt to in uence the participants choice of replacement transportation although several
2
European programs, similar to CARS, required purchase of a new vehicle. Dill (2001) however points out that the estimates hinge crucially on the assumption about how the owner replaces the transportation provided by the scrapped vehicle.
Our paper diers from the previous literature in the following aspects. First, we estimate the lifecycle GHG and gasoline use reduction relative to dierent three dierent counter-factual scenarios and thus derive a range of estimates. Second, for each counterfactual scenario we analyze the sensitivity of impact to dierent assumptions about the remaining useful life of retired vehicles and to dierent magnitudes of the rebound eect. Third, we calculate the impact on emissions of criteria air pollutants. Fourth, we estimate the total benets by aggregating the estimates for every clunker traded-in instead of estimating simply based on total average or average by vehicle class.
We also disaggregate the impact based on the fuel-economy of the clunker and derive insights for improving program design. We begin by describing a simple model of the choice the owner of a clunker, or more generally a durable energy consuming good, faces.
2 Model
T=0 T=Δt 1 1+Δt 2 2+Δt c0 e0 cΔt c1+Δt c2+Δt e1 e2 Energy efficiency eΔt c1 e1+Δt e2+Δt c2 Time
Path B: c0 -> cΔt -> c1+Δt -> c1+Δt->…
Path D: c0 -> e0 -> c1 -> c2 ->…
Path P: c0 -> e0 -> e1 -> e2 ->…
Path C: c0 -> eΔt -> e1+Δt-> e1+Δt ->…
Figure 1: The choices the owner of an inecient durable good faces over time
Figure 1 describes a generic model of replacement choices over time for the owner of low- eciency durable good under a program such as CARS. Let c and e represent the energy use eciency of the low-eciency and the high-eciency good respectively. Let subscripts 0; 1; 2:: denote time.
t represent remaining life span of the inecient good at T = 0. It should be
3
pointed out that even if the current owner sells the inecient good prior to its end of useful life, the good would continue to be operated by a second-hand buyer and from a lifecycle emissions perspective, our concern is only with the utilization of the good rather than the its utilization by a given owner. This is signicant departure from the assumptions of previous literature. Without loss of generality, let one unit of time interval denote the expected life span of the durable good, which is assumed the same for both c and e, and therefore 0 <
t < 1. Let us also assume that goods that are not traded-in at T = 0 and all new newly purchased goods will not be retired before end of expected useful life and that the energy eciency for both goods increases over time i.e.,
@c
@t > 0; @e
@t > 0. Since the number of combinatorial choices increase exponentially with time, we depict the possibilities only through T = 2 +
t. Path P denoted by (c0 ! e0 ! e1 ! e2 !
:::) represents a combination of choices where in the owner of a clunker trades-in and switches permanently to a high-eciency vehicle i.e., his future purchases are also high eciency models.
Path B denoted by (c0 ! c
t ! c1+
t ! c2+
t ! :::) implies that owner of a clunker does not trade-in and continues to purchase low eciency vehicles into the future. Path C denoted by
(c0 ! e
t ! e1+
t ! e2+
t ! :::) implies that even in the absence of CARS, the owner whould have switched to a high eciency vehicle upon retirement of the clunker. Finally, path D denoted by (c0 ! e0 ! c1 ! c2 ! :::) implies that despite opting into the program, the owner reverts to a low-eciency vehicle for his next purchase.
Because the CARS program's primary objective was to stimulate the U.S. economy during a time of recession by boosting sales of automobiles and also because the future is uncertain, we restrict our analysis to time T 2 [0; 1]. However, we consider two types of counterfactual scenarios for owners of clunkers that were not to traded-in but in the near future i.e., T =
t, may need to be retired, namely, their next purchase is a clunker albeit more-ecient and that their next purchase is a high-eciency vehicle. Thus we conne our analysis to comparing emissions under paths P;B;C in the time interval [0; 1]2. To the extent that the program induces a permanent switch to a higher eciency category of vehicle, by conning our analysis to T 2 [0; 1] we under-estimate the program bene ts. Let  represent the emissions per gallon of gasoline and  represent the fuel eciency of a vehicle in miles per gallon. Let z = 
 , represent emissions per mile. Let L represent the expected mileage before vehicle retirement and L represent the odometer reading on a vehicle at the time of trade-in. Let  Z represent the emissions from vehicle production and disposal and z = Z
L represent the average emissions per mile attributable to production and disposal. We assume that each type of vehicles becomes more fuel-ecient with time i.e., k t2 > k t1 for any t2 > t1, where k 2 (c; e).
2Emission under paths P and D are equivalent
4
The emissions under the three paths are,
ZP = (ze
0 + z)L + z(L Lc)
ZB = (zc
0 + z)(L Lc) + (zc

t + zc

t)Lc (1)
ZC = (zc
0 + z)(L Lc) + (ze

t + ze

t)Lc
Accounting for rebound: Thus far we assumed that total VMT between T = 0 and T = 1 is xed and equal to L. However, it has been argued by several researchers that one unintended consequence of increase energy eciency is rebound in energy consumption (Greening et al., 2000,
Hertwich, 2005, Small and Van Dender, 2007, Sorrell and Dimitropoulos, 2008). Rebound arises from the reduction in the marginal cost of energy as a result of energy eciency. In the context of increase in automobile fuel economy, this implies a result of lower marginal cost of driving and increase in driving as a consequence. Furthermore, there may also be an additional rebound from switching to a newer, low-maintenance and more comfortable vehicle. We modify the system of equations (1) in the following manner to account for rebound. Let, pg denote price of gasoline; pk f = pg
k , k 2 fe; cg the cost per mile for a given type of vehicle;  < 0 the price elasticity of demand for gasoline. Then, the percentage increase in gasoline consumption as a result of increase in energy eciency and reduction in cost of driving (pe f < pc f since, e > c) assuming constant elasticity of demand is given by, dq q
=  dpf pf
=  pg e pg
c
pg
e
= 

c
e
1

(2)
Since q = L
 and holding  xed once a new vehicle has been purchased, the increase in driving with the new vehicle as a result of the increase in gasoline consumption is dL L
= dq q = 

c
e
1

(3)
Equations (2) and (3) show that the rebound eect on fuel use and driving is a function of the elasticity of demand and ratio of the fuel economy of the old and new vehicle. Since  < 0 and
e > c, the rebound eect increases with increase in elasticity of demand and increase in e
c . We
5
can now rewrite (1) as
ZP = (ze
0 + z)(L + dL(e
0)) + z(L + dL(c
0) Lc)
ZB = (zc
0 + z)(L + dL(c
0 ) Lc) + (zc

t + zc

t)(Lc + dL(c

t)) (4)
ZC = (zc
0 + z)(L + dL(c
0 ) Lc) + (ze

t + ze

t)(Lc + dL(e

t))
Gasoline use under each path is qP =
1
e
0
(L + dL(e
0))
qB =
1
c
0
(L + dL(c
0) Lc) +
1
c

t
(Lc + dL(c

t)) (5) qC =
1
c
0
(L + dL(c
0) Lc) +
1
e

t
(Lc + dL(e

t))
Equations (4) and (5) suggests that the benets of program increase with increase in fuel economy,e 0, of the newly purchased vehicle; decrease with increase in fuel economy, c
0 , of the clunker; decrease with increase in fuel economy of the vehicle purchased in the counterfactual c

t
or e

t
; decrease with increase in VMT of the clunker, Lc; and increase with increase in emission intensity of gasoline  (since z = 
 ). Our model under-estimates emission reduction benets of the vehicle retirement program in the counterfactual B in case the program induces a permanent switch to a higher eciency category of vehicle. Total avoided emissions and gasoline use over all vehicles traded-in, N, with respect to counterfactual j 2 B;C is given by ,
XN
i=1

Zi
P;j =
XN
i=1
Zi
P Zi j (6)
XN
i=1

qi
P;j =
XN
i=1 qi P qi j (7)
The average subsidy, per unit of emission reduction for clunker of given fuel economy,  = k, with respect to counterfactual j, is computed as
sk;j =
Pnk
i=1 si
Pnk
i=1
Zi
P;j
for each k 2 (; ) (8) where, nk is the number of clunkers with  = k and si is subsidy for vehicle i. It is worth pointing out that sk;j does not re ect the cost-eectiveness of GHG abatement which requires taking to consideration the dierence in purchase cost and lifecycle fuel cost. The average subsidy per unit
6
of emission avoided by the program is computed as,
sj =
P
k=
Pnk
i=1 si
P
k=
Pnk
i=1
Zi
P;j
(9)
3 Data, results and sensitivity analysis
! " # $
%
%&"
%&$
%&'
%&(
!
!&"
!&$
!&'
!&(
"
"&"
)*!%
+
,-./0123*14*5/6789/*.2-:/:!7;
/:*8-./0123*"
@A286->/:*8-./0123*#
@A286->/:*8-./0123*$
! " # $
%
"
$
&
'
!%
!"
!$
()*+,-./0-102+3456+0*.)7+7!48
(3)8,+048019+60+5-8-:/0480:46+;0