Energy Policy 95 (2016) 2131

Contents lists available at ScienceDirect

Energy Policy
journal homepage: www.elsevier.com/locate/enpol

Leveling the playing eld of transportation fuels: Accounting

for indirect emissions of natural gas
Steven Sexton a,n, Jonathan Eyer b
a
Duke University, 201 Science Drive, Durham, NC 27708, USA
b
University of Southern California, 3710 S. McClintock Ave, Los Angeles, CA 90089, USA

H I G H L I G H T S

Natural gas used in transport causes indirect emissions in the electricity sector.

These emissions result from increased coal use in electricity generation.

They rival in magnitude indirect land use change (ILUC) emissions of biofuels.

Natural gas fuels are estimated to be as carbon intensive as the petroleum fuels.

Policy ignores indirect emissions from natural gas.

art ic l e i nf o a b s t r a c t

Article history: Natural gas transportation fuels are credited in prior studies with greenhouse gas emissions savings
Received 23 July 2015 relative to petroleum-based fuels and relative to the total emissions of biofuels. These analyses, however,
Received in revised form overlook a source of potentially large indirect emissions from natural gas transportation fuels, namely
17 March 2016
the emissions from incremental coal-red generation caused by price-induced substitutions away from
Accepted 16 April 2016
Available online 30 April 2016
natural-gas-red electricity generation. Because coal-red generation emits substantially more green-
house gases and criteria air pollutants than natural-gas-red generation, this indirect coal-use change
Keywords: effect diminishes potential emissions savings from natural gas transportation fuels. Estimates from a
Compressed natural gas parameterized multi-market model suggest the indirect coal-use change effect rivals in magnitude the
Liquied natural gas
indirect land-use change effect of biofuels and renders natural gas fuels as carbon intensive as petroleum
Transportation fuels
fuels.
Low carbon fuel standard
Greenhouse gas emissions & 2016 Published by Elsevier Ltd.
Indirect emissions

1. Introduction
Recent advances in shale gas production and growing demand
for greenhouse gas emissions reductions have spurred interest in
natural gas as a competitor to petroleum-based transportation
fuels. Though natural gas presently powers a mere 394,000 light-
duty vehicles and 40,000 heavy-duty vehicles in the U.S., its share
of transportation fuel rivals electricity and its eet far exceeds the
electric vehicle eet (EIA, 2015a).
Short-run substitutions to natural gas are constrained by lim-
ited refueling infrastructure and incompatible engine technology,
yet natural gas vehicle fuel consumption is increasing, and, by one
estimate, will globally displace 1.5 million barrels of oil per day by
2030 (IHS, 2015). Natural gas vehicle fuel consumption is expected
n
Corresponding author.
E-mail address: steven.sexton@duke.edu (S. Sexton).
to grow faster than any other transportation fuel to 2040 (EIA,
2015d, 2015e). Motivated largely by concern about climate impacts
of gasoline and diesel combustion, policy in the U.S. and elsewhere
endeavors to overcome obstacles to expanded natural gas use in
transportation, a sector responsible for nearly one-third of an-
thropogenic greenhouse gas emissions in the U.S. (EPAct, 2005;
Knittel, 2012; EPA, 2013b; Obama, 2014; EPA, 2012). Federal tax
expenditures in the U.S., for instance, promote natural gas vehicle
sales and refueling infrastructure investments. In his 2014 State of
the Union address, U.S. President Barack Obama touted natural gas
as a bridge fuel that can power our economy with less of the
carbon pollution that causes climate change, and he advocated
policy to shift more cars and trucks from foreign oil to American
natural gas. Federal Corporate Average Fuel Economy standards
also favor natural gas vehicles by allowing manufacturers to count
each natural gas vehicle sale as more than one vehicle sale in its
compliance calculations, effectively subsidizing natural gas vehicle

http://dx.doi.org/10.1016/j.enpol.2016.04.023
0301-4215/& 2016 Published by Elsevier Ltd.
22 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131
production. States also grant natural gas vehicles access to high-
occupancy-vehicle lanes and provide other incentives to en-
courage adoption of a technology that is generally considered to be
less carbon-intensive than petroleum-based fuels combustion.
These and other policies are catalogued in CRS (2014). Still, some
analyses suggest policy is necessary to level the playing eld of
transportation fuels that disadvantages natural gas, e.g., Knittel
(2012).
Compressed natural gas (CNG) and liqueed natural gas (LNG)
fuels are typically ascribed lower life-cycle carbon emissions than
gasoline and diesel, for which they respectively substitute (Burn-
ham et al., 2011; Larive et al., 2011; Kyle and Kim, 2011; Ste-
phenson et al., 2012; Chan et al., 2013; Nigro and Jiang, 2013; Ou
and Zhang, 2013), though Tong et al. (2015a, 2015b) estimate
greater emissions from natural gas than conventional fuels for all
heavy-duty truck technologies and for some medium and light-
duty truck technologies depending upon modeling assumptions.
Natural gas emissions are also generally thought to be lower than
the sum of direct and indirect emissions of biofuels from corn and
soy (Fargione et al., 2008; Searchinger et al., 2008; ARB, 2009;
Hertel et al., 2009; Laborde, 2011; European Commission, 2012;
Khanna and Crago, 2012).
The conventional wisdom that natural gas is a clean, alternative
transportation fuel implicitly assumes, however, the absence of
considerable indirect emissions, like the indirect land-use change
(ILUC) emissions of biofuels that have attracted considerable at-
tention in scholarly research and policy making since 2008, e.g.,
Fargione et al. (2008), Searchinger et al. (2008), ARB (2009), Hertel
et al. (2009), Lapola et al. (2010), Laborde (2011), European Com-
mission (2012) and Khanna and Crago (2012). Despite the interest
in indirect biofuel emissions, the indirect emissions from other
fuels like natural gas have been virtually ignored. In fact, in im-
plementing a rst-of-its-kind low carbon fuel standard that relies
upon accurate measures of carbon intensities across fuels, Cali-
fornia regulators identied only land use change as a signicant
source of indirect emissions. This paper, however, presents a
parameterized analytical model that estimates signicant indirect
emissions from CNG and LNG that are ignored in previous life cycle
assessments and in existing policy. These indirect emissions are
estimated to rival and even exceed the ILUC emissions of biofuels
and to render natural gas fuels no less carbon intensive than the
petroleum fuels they are intended to displace.
Intuitively, an exogenous increase in natural gas consumption
in transportation due to policy, for instance, must be met by an
increase in the quantity of natural gas supplied, a decrease in
natural gas consumption elsewhere in the economy, or by some
combination of supply-side and demand-side responses. If the
increased demand for natural gas in transportation were fully met
by an equivalent increase in natural gas production, i.e., if natural
gas supply were perfectly elastic, then indirect emissions from
natural gas would likely be nominal. If, however, the marginal cost
of natural gas production increases in the quantity of production
(so that natural gas supply is upward sloping in price-quantity
space), as ample empirical evidence suggests, then increased de-
mand for natural gas raises the price of natural gas (Brown and
Krupnick, 2010; Arora, 2014; Ponce and Neumann, 2014; Hausman
and Kellogg, 2015). Higher natural gas prices, in turn, induce
substitutions to other, potentially dirtier energy feedstocks else-
where in the economy, including in electricity generation.
Electricity generation is responsible for more than forty percent
of natural gas consumption in the U.S. (EIA, 2013). As natural gas
prices rise in response to increased demand for its use in trans-
portation, power generators are expected to substitute away from
natural-gas-red electricity generation and toward generation
from other feedstocks like coal. A number of life-cycle analyses
have concluded that a typical coal plant emits considerably more
greenhouse gases than a typical natural gas plant (Spath and
Mann, 2000; Burnham et al., 2011; Fulton et al., 2011). A con-
sequence of natural gas use in transportation, therefore, may be
increased carbon emissions in power generation. While life-cycle
emissions of electricity generation are estimated with uncertainty,
much like the life-cycle emissions of transportation fuels, a 2011
Intergovernmental Panel on Climate Change (IPCC) review of more
than 90 credible life-cycle assessments concluded that natural gas
plants generate 53% fewer carbon dioxide-equivalent emissions
than coal plants 469 g per kilowatt-hour (kWh) versus 1001 g
per kWh (von Stechow et al., 2011; Caulton et al., 2014). Coal
plants also emit far greater quantities of criteria pollutants than
natural gas plants. These include nitrous oxides, sulfur dioxide,
and particulate matter that are deleterious to human health (Jar-
amillo et al., 2007).
Thus, an indirect effect of greater natural gas use in transpor-
tation is greater carbon and criteria pollutant emissions in the
power sector. This effect is analogous to the well-studied and
controversial ILUC effect of biofuels: greater demand for agri-
cultural output raises the rental rate of cropland, which induces
conversion of undeveloped land to farmland, releasing the carbon
stored in the ground and biomass. To our knowledge, however, the
magnitude of indirect coal-use change (ICUC) emissions from
natural gas has not been previously estimated.
Were a global and economy-wide carbon price imposed to in-
ternalize the climate change externality posed by carbon emis-
sions, e.g., via an optimal carbon tax or optimal tradable permit
standard, then an accounting of indirect effects of natural gas use
in transportation would be of little value to policy makers, as
would be the analogous accounting for biofuels. Interest in the
ILUC, however, derives from the implementation of second and
third-best policies to promote low-carbon technologies, including
low carbon fuel standards.
So long as carbon emissions are not priced appropriately across
sectors and jurisdictions, and so long as planners pursue alter-
native carbon emissions abatement strategies, the ICUC is pre-
sumably as relevant to policy as is the ILUC. This is true even
though some jurisdictions regulate carbon emissions in the elec-
tricity sector. Because natural gas ows relatively uninhibited
within the U.S., and at greater cost, around the world, sub-national
carbon policy is ineffective in avoiding indirect emissions from
natural gas transportation use. The emissions leak from the
regulated jurisdiction to unregulated jurisdictions. While direct
emissions leakage in the electricity sector may be constrained by
restrictions on electricity imports to the regulated jurisdiction, a
sector-specic policy such as those in place today, cannot contain
leakage in other sectors. Intuitively, if a unit of natural gas is used
to power cars in a jurisdiction that prices carbon emissions in the
electricity sector, e.g. California, it raises the price of natural gas to
electricity generators in the regulated and unregulated jurisdic-
tions. Substitutions toward coal will come from the unregulated
jurisdictions where such substitutions are unencumbered by car-
bon policy imposed in the regulated jurisdictions.
This paper proceeds in turn to develop an analytical model of
ICUC emissions, to calibrate and simulate the model, to consider
key uncertainties in estimating indirect emissions, to discuss im-
plications of the analysis, and to address limitations of the model
before concluding.
2. A model of indirect emissions from natural gas
In order to estimate the magnitude of ICUC emissions, we de-
velop and parameterize a multi-market model in which natural
gas is produced according to an upward sloping supply function. It
is demanded in the residential, commercial, industrial,
 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131 23

transportation and electricity sectors of the economy. Quantity equilibrium. Hence,3

demanded of natural gas is assumed to fall as prices rise. Because
P C P NG
our interest centers on the electricity market response to in- dQC = NG dQ ENG + dQ TNG .
creased demand for natural gas in the transportation sector, we Q C Q (1)
specify the residual supply of natural gas to electricity and trans- Given full pass-through of costs in rate setting, the change in
portation as total supply net of demand by the industrial, re- electricity price is equal to the marginal cost increase indicated in
sidential, and commercial sectors.1 Coal is also produced according (1), and the change in electricity demand is determined by its
to an upward sloping supply function. Electricity demand is as- price responsiveness as follows:
sumed to be inelastic in the short run, but declining in price in the
long run, as a sufcient time horizon allows end-users to adjust ED P NG
dED = dQ ENG + dQ TNG .
capital holdings in order to avoid higher costs, e.g., by investing in P E Q NG
energy efcient capital. Within this framework, ICUC emissions are
The equilibrium condition that changes in supply equal chan-
determined as the difference in equilibrium electricity sector
ges in demand, thus, yields:
emissions before and after an incremental increase in natural gas
use for transportation. 1 C 1 ED P NG
dQ + dQ ENG = E NG dQ ENG + dQ TNG .
In equilibrium, the quantity of electricity supplied, ES, is equal P Q (2)
to the quantity of electricity demanded, ED, in each period, and so
the change in electricity supplied is equal to the change in elec- As interest centers on changes in the electricity market equili-
tricity demanded: brium per unit change in natural gas demand in the transportation
dQC
sector, (1) and (2) are divided by dQNG
T and solved for , the
dE S = dED. dQ TNG
change in coal use in electricity generation per unit change in
The change in electricity supplied is equal to the change in dQ ENG
natural gas use in transportation, and , the change in natural
electricity generation from coal plus the change in electricity dQ TNG
generation from natural gas, i.e., gas use in electricity generation per unit change in natural gas use
1 C 1 in transportation. The former, for instance, equals:
dE S = dQ + dQ ENG,
QC P NG
C
where QC and QNG are, respectively, the quantities of coal and dQC P Q NG
E = .
natural gas consumed in electricity generation and s and are dQ TNG P NG E 2 QC P NG
1 2 NG E + 2 C
parameters reecting, respectively, the quantities of coal and Q P P Q NG
natural gas required to produce a unit of electricity.
The magnitude of the ICUC effect is the sum of the changes in
The change in electricity demanded is determined by the
coal and natural gas-red electricity generation multiplied by the
change in electricity price, PE, and by the responsiveness of elec-
emission intensities of the respective generation pathways:
tricity demand to electricity price, i.e.,
ED E dQC dQ ENG
dED = dP . NG
+ ,
P E dQ T dQ TNG

The change in electricity price is determined by the change in the where and are coefcients reecting the emissions per unit
cost of electricity generation and the rate at which the cost change generation from coal and natural gas, respectively. Given an up-
dQ ENG
is passed onto consumers. Cognizant of the regulated nature of ward sloping natural gas supply, then < 0, i.e., incremental
dQ TNG
electricity rates, it is assumed cost changes are fully passed onto
demand for natural gas in other sectors reduces gas-red elec-
consumers.2 We assume no other costs change as feedstock costs
tricity generation. In the short run, the decline in natural-gas-red
change; marginal costs other than feedstock costs are similar
generation is met by an equal increase in coal-red generation as
across coal and natural gas generation, covary in production levels,
the quantity of electricity demanded is unchanged, i.e.,
and constitute a small fraction of feedstock costs. Indeed, fuel costs NG
1 dQC 1 dQ E
are responsible for 77% and 86% of total variable costs of electricity dQ NG
+ dQ NG
= 0. But for downward sloping electricity demand
T T
generation from coal and natural gas, respectively (Brown and in the long run, the reduction in natural-gas-red generation ex-
Andres, 2014; Nuclear Energy Institute, 2013). ceeds the increase in coal-red-generation because the quantity of
Electricity generators are assumed to minimize energy feed- 1 dQC
NG
1 dQ E
electricity demanded declines in price, i.e., + < 0.
stock costs, taking prices as given, so that the marginal costs of dQ NG
T
dQ NG
T
electricity generation from natural gas and coal are equated in The parsimony of this model is afforded by the physical and
economic characteristics of electricity generation and natural gas
1
supply.4 In particular, and in order to avoid dependence upon
The residual supply elasticity facing the electricity and transportation sectors
Qtotal Qi Qr Qc
is dened as Sresid = S Di Dr Dc where Di , Dr , and cD are
Qresid Qresid Qresid Qresid 3
With thousands of rms producing natural gas, the wholesale natural gas
the elasticities of demand in the industrial, residential, and commercial sectors and market is highly competitive (Wells, 2006; EIA, 2015b; FERC, 2015). Hence, it is
Qtotal, Qresid, Qi, Qr , and Qc are total natural gas consumption, consumption in the assumed changes in wholesale prices are fully transmitted downstream to all
residual sector, i.e., electricity and transportation, and industrial, residential, and consumers regardless of the constant mark-up natural gas consumers face in dif-
commercial consumption, respectively. ferent sectors. This is similar to Linn et al. (2014) and Hausman and Kellogg (2015),
2
This is a conservative assumption. It assumes electricity demand declines for instance.
4
more in response to an increase in natural gas prices than if market power or rate This analysis abstracts from an assessment of regional impacts that would
regulation induced less than complete pass-through of feedstock cost increases. account for differences in electricity generating eets across control areas or in-
Electricity demand reduction in response to electricity price increases mitigates the terconnections because natural gas, unlike electricity, typically ows readily across
ICUC effect because the demand response avoids coal-powered generation. If such boundaries so that a CNG vehicle refueling in California may affect power
generators possess market power, costs would not be fully passed on to consumers sector emissions in New York. This is distinct from the effect of electric vehicles. In
because the power generator would face a tradeoff between selling at a reduced that context, economists have calculated region-specic marginal emissions from
margin if prices were not increased and selling fewer units if prices were increased. electric vehicles because the marginal emissions cannot leak beyond, e.g., Nuclear
24 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131

Fig. 1. Natural gas and coal generation shares. Depicted are the shares of total
electricity generation produced by coal and natural gas plants, respectively, from
2000 to 2012. Source: EIA.

many additional and uncertain economic relationships, it is as-

sumed that marginal increases in natural gas prices induce sub-
stitution toward coal-red electricity generation alone.
While marginal gas and coal plants have long-competed for
generation share, gas and coal plants historically served distinct
roles in the electric grid. Until recently, the former performed
primarily a load-following function because of the relatively high
marginal costs and relatively low ramping costs of gas turbine
technologies. Coal served base-load generation because of higher
ramping costs and lower marginal costs. Shale gas production,
however, has yielded low feedstock prices that, in concert with
adoption of combined-cycle technologies, make natural gas plants
competitive with coal plants for base-load generation (EIA, 2012).
At the same time, increasing demand for ramping capacity due to
intermittent renewable generation has impelled some coal plants
to undertake more frequent cycling. Consequently, the two feed-
stocks are increasingly substitutable over a broad range of elec-
tricity generation.
The substitutability of these feedstocks is exhibited in Fig. 1,
which shows the shares of coal and natural-gas-red electricity
generation in the U.S. from 2001 to 2013. The generation share of
natural gas is highly negatively correlated (r  89) with the
generation share of coal. Moreover, Fig. 2 depicts the order in
which generating units are dispatched in the Southeastern U.S. in
order to meet electricity demand at lowest cost, according to a U.S.
Energy Information Administration (EIA) model. The top panel
shows the dispatch curve for the year 2010 and demonstrates that
coal and natural gas plants substituted over a non-trivial range of
generation. The bottom panel shows the dispatch curve in 2012,
when natural gas prices were lower than in 2010. Substitution
between coal and natural gas generation occurred over the entire
range of coal and gas plant dispatch. Mason et al. (2015) sum-
marizes the literature on coal and natural gas substitutions in
electricity generation.
This model assumes a continuous and monotonic natural gas
supply function that relates higher levels of production to higher
marginal costs. Thus, even small changes in the quantity de-
manded of natural gas induce changes in natural gas prices.
Moreover, motivated by the high-degree of substitutability be-
tween coal and natural gas, this analysis also assumes that even
marginal increases in natural gas prices are sufcient to induce
marginal electricity generating units to respond, i.e., for a marginal
coal plant to ramp up production as a marginal gas plant ramps
down production due to a marginal increase in natural gas price.
Over the short-to-medium term, at least, it is unlikely genera-
tion sources other than coal plants substitute for gas-red gen-
eration. Generation from petroleum feedstocks is costly and
(footnote continued)
Energy Regulatory Commission regions or interconnections. See for instance Graff-
Zivin et al. (2014).
Fig. 2. EIA Southeastern U.S. dispatch curve. Depicted are modeled dispatch curves
which depict the order in which the electricity generating in eet is dispatched to
meet electricity load at lowest cost. The range of load over which coal and natural
gas substitute is considerable in 2010 (panel A) and greater in 2012 (panel B) when
natural gas prices were lower. Source: EIA (2012).
thereby called up to meet only a few percent of total generation
during peak demand. As natural gas and coal prices rise, these
peaking plants may be called up to serve base-load generation,
though this is unlikely given differences in plant efciency and fuel
costs (EIA, 2015c). Nuclear and renewable wind and solar gen-
eration are also unlikely to substitute for gas-red generation
because those resources effectively operate at full capacity (EIA,
2014b). These plants are characterized by high xed capital costs
but very low marginal costs of generation. Nuclear plants also have
little ability to ramp production, while renewables cannot be
dispatched, making them poor substitutes for the dispatchable
and rampable generation of coal and gas plants.
Nevertheless, greater wind, solar, or nuclear capacity that dis-
places marginal coal or natural gas base-load generation in the
long run would reduce the indirect emissions of natural gas
transportation fuels. The EIA, however, forecasts no considerable
change in the share of generation from these feedstocks by 2040.
Wind capacity is projected to increase from 6% of total capacity to
7%, solar from 0.3% to 1.5% and nuclear is projected to decline from
10.5% of capacity to 8.6% (EIA, 2014a).
In spite of the power plant emissions regulations promulgated
by the U.S. Environmental Protection Agency, i.e., the Clean Power
Plan, coal and natural gas are likely to continue to dominate
electricity generation (EIA, 2015a). The nuclear renaissance an-
ticipated in the 1990s and early 21st century has not materialized,
partly due to the nearly $10 billion cost of a typical reactor and to
renewed safety concerns following radioactive emissions from
Japan's Fukishima Daiichi nuclear plant in 2011. Four U.S. nuclear
generators were permanently closed in 2013, and the prospects of
capacity growth are dim (Lucas, 2012). Solar and wind generation
remain the most costly forms of power generation, even
 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131 25

accounting for the cost of pollution emissions. They are competi- Table 1
tive only with policy support and with back-up generation from Calibration parameters.
dispatchable fossil fuels (Sovacool, 2009; Frank, 2014).
Parameter description Parameter

Coal consumption (million short tons) 824,758

3. Results and discussion Natural gas price (2012 $US/MMBtu) 2.66
Natural gas consumption - electricity (million cubic feet) 9,144,417
Natural gas consumption - residential (million cubic feet) 4,914,327
ICUC emissions are estimated by parameterizing the foregoing Natural gas consumption - commercial (million cubic feet) 3,278,856
model. For simplicity, all supply and demand relationships are Natural gas consumption - industrial (million cubic feet) 7,413,918
assumed to be linear in price. Thus, elasticities and equilibrium
quantities are substituted into the expressions for changes in coal
and gas use in the electricity sector to determine the magnitudes Table 2
of changes for a unit change in natural gas use in transportation. Elasticity assumptions.

We calibrate the model using price and quantity data for 2013. We Parameter Estimate range Baseline estimate
solve for the price of coal using the price of natural gas and the
b
equilibrium condition equating the marginal costs of coal and Coal supply 1.86 to 4.9 3.0
natural gas generation. The coal price implied by the model ap- Natural gas supply 0.2 to 0.81c 0.81
Residential natural gas demand  0.2 to  0.53d  0.2
proximates but is lower than the price reported by the EIA$40.05
Commercial natural gas demand  0.21 to  0.53e  0.23
per ton compared to $45.77 per tonas expected given that the Industrial natural gas demand  0.57 to  1.24f  0.57
model ignores operating costs other than feedstock costs. These Electricity demand  0.32g  0.32
calibration parameters are reported in Table 1. Coal-gas CO2e differential (gCO2e/ 350 to 650h 532
kWh)a
Estimates from the economics literature and modeling as-
sumptions of regulatory agencies are employed to characterize the Reported are the ranges of parameter estimates identied in the economics and
price-responsiveness of commodity supplies and demands.5,6 engineering literatures and implemented in Monte Carlo simulations. Also reported
are baseline parameter values used to estimate ICUC emissions.
a
The difference in carbon dioxide emissions (CO2e) between coal plants and
5
For at least several decades, economists have considered long-run coal supply natural gas plants.
to be elastic, though empirical estimates are sparse (Zimmerman, 1981; Joskow, b
Beck et al. (1991), Brown (1998) and EIA (2013), and authors' own estimate.
1988). In a study of the Appalachia region of the U.S., Harvey (1986) estimated a See footnote 5.
long-run supply elasticity of 3.1. Brown (1998) rely upon analysis in Brown et al. c
Arora (2014), Ponce and Neumann (2014) and Hausman and Kellogg
(1995) to calibrate a macroeconomic model of climate change policy impacts; They (2015).
assume a coal supply elasticity of 1.86. Burniaux and Martins (2012), however, d
Davis and Muehlegger (2010), CEC (2012) and Hausman and Kellogg
assume a coal supply elasticity of 7 in their general equilibrium analysis of carbon (2015).
leakage from climate change policy. They rely on estimates from Mellish (1998), e
Davis and Muehlegger (2010), CEC (2012) and Hausman and Kellogg
who employs two-stage least squares estimates of price responsiveness to over- (2015).
come the simultaneity of supply and demand. Beck et al. (1991) use mine-level cost f
Davis and Muehlegger (2010), CEC (2012) and Hausman and Kellogg
data to estimate smaller long-run supply elasticities for Australian coal of 1.9, 2.7, (2015).
and 3.5. Their estimates do not account for part of the supply response to price, e.g., g
Dahl and Roman (2004).
ignoring the role of price expectations on supply decisions, and they also fail to h
Spath and Mann (2000), Jaramillo et al. (2007), Jiang et al. (2011), and von
control for simultaneous supply and demand determination, as Mellish (1998) Stechow et al. (2011).
does. More recently, the EIA has estimated the coal supply elasticity to be equal to
4.9 based on model-produced coal supply curves. Staff judgement and historical
estimates, however, impelled the agency to adopt a more modest elasticity esti- Similarly, estimates of the emission rates of coal and gas plants are
mate of 2.5 in its implementation of the National Energy Modeling System. obtained from the engineering literature and government reports.
We generate our own elasticity estimate using the EPA's Integrated Planning Because the price relationships and emission rates are estimated
Modeling (IMP) upon which EPA relies for regulatory impact assessments. The
with error and because the available record presents a range of
model includes mine-level supply curves generated for the agency by Wood
Mackenzie, a consultancy with extensive experience in preparing mine-by-mine estimates for key parameters, as shown in Table 2, the magnitude
estimates of cash operating costs, as well as access to public and proprietary data of ICUC emissions is estimated for various scenarios using a range
sources. A detailed discussion of how these supply curves are generated is provided of plausible values for key model parameters. Monte Carlo simu-
in EPA (2013a). Using the 2016 cost and production forecasts obtained from the lations are also used to estimate ICUC emissions across the range
IMP, we estimate a quadratic supply function over the relevant range of coal pro-
duction and compute a point elasticity at current production levels. Specically, we
of parametric uncertainty and to characterize the magnitude of
regress cumulative annual production on price and the square of price. The esti- uncertainty.
mated supply function includes an intercept of  1665.919, and coefcients on the
price variables of 178.4253 and  1.961519, respectively. These coefcients are 3.1. Model estimates of indirect coal use change emissions
signicant at the 10% level or higher. This regression yields an R-squared of 0.9695.
It suggests a point elasticity estimate at current production levels of 3.0, which lies
approximately in the middle of the range of estimates generated by the prior Based upon our baseline model parameterizations, it is esti-
studies. This coal elasticity estimate is an input into our baseline ICUC estimate. mated that ICUC causes 28.1 g of CO2 equivalent emissions per
6
Arora (2014) provides the most extensive, recent estimates of natural gas megajoule of natural gas (g CO2e/MJ) in the long run. In the short
supply elasticities in the U.S. Relying on weekly, monthly, and quarterly observa-
run, absent electricity demand response, the ICUC is estimated to
tions on production and producer price index from 1993 to 2013, Arora (2014)
estimates vector-autoregressive models that exploit plausibly exogenous changes cause 37.2 g CO2e/MJ. These estimates assume power plant emis-
in demand. For each market shock, a unique long-run elasticity is estimated. These sion intensities reported by the IPCC, natural gas supply and sec-
range from 0.1 to 0.5, depending upon the market shock, and the frequency and toral demand elasticities estimated by Hausman and Kellogg
range of data employed in the estimation. More recent data yield larger elasticity
(2015), and our original analysis of coal supply curves employed by
estimates, suggesting that shale gas production is more responsive to price than
conventional gas production. Ponce and Neumann (2014) estimate a long-run the EPA in its Integrated Planning Model for carbon policy impact
supply elasticity of 0.6, while the EIA's NEMS model employs an elasticity of 0.5 assessment (see footnote 5). These results are reported in Table 3,
(Arora, 2014). For our baseline ICUC emissions estimate, we rely on the gas supply
elasticity estimated by Hausman and Kellogg (2015), which uses lagged weather as
an exogenous shock to contemporaneous demand that operates via competitive (footnote continued)
storage markets. Hausman and Kellogg (2015) estimate a long run supply elasticity Given that the Hausman and Kellogg estimate is the highest among these recent
of 0.81 using monthly marketed production and spot prices from 2001 to 2014. estimates, its use is a source of conservatism in our estimates of the ICUC.
26 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131
Table 3
Indirect coal use change emissions (g/MJ) in short and long run (in bold).
Coal supply elasticity Natural gas supply elasticity
0.2 0.5 0.81
2.5 46.8 39.7 34.4
32.3 28.0 24.7
3.0 48.9 42.3 37.2
35.8 31.5 28.1
4.9 53.6 48.6 44.3
44.0 40.2 36.9
Reported are short-run and long-run (bold typeface) indirect coal use change
emissions for preferred coal and natural gas supply elasticity assumptions (3.0 and
0.81, respectively) and alternative elasticity estimates from the literature (see
footnotes 2 and 3).
along with model ICUC emission estimates for alternative coal and
gas supply elasticities as indicated in the table.
ICUC emissions are reported for a low-to-mid-range gas supply
elasticity estimate of 0.2 obtained from Arora (2014), the gas
supply elasticity of 0.5 employed by the EIA in its National Energy
Modeling System (NEMS), and the Hausman and Kellogg (2015)
estimate of 0.81. ICUC estimates are also reported for an adjusted
coal supply elasticity of 2.5 employed by the EIA in its modeling in
2014, our original point elasticity estimate of 3.0 derived from EPA
mining cost and production data, and for a pre-adjusted EIA esti-
mate of 4.9 derived from the coal market module of NEMS. Long-
run elasticity estimates are reported in bold typeface. The short-
run ICUC estimates differ from the long-run ICUC estimates only in
that electricity demand is assumed to be inelastic in the short run;
the supply elasticities for coal and natural gas are the long-run
elasticities employed in long-run ICUC estimates.
For these elasticity assumptions, long-run ICUC emissions vary
from 24.7 g CO2e/MJ with a relatively elastic natural gas supply
and relatively inelastic coal supply to 44.0 g CO2e/MJ for relatively
inelastic natural gas supply and relatively elastic coal supply.
Short-run ICUC emissions are estimated to vary from 34.4 g CO2e/
MJ to 53.6 g CO2e/MJ. If instead of IPCC coal and gas plant emission
intensities, we assume the smaller emissions differential esti-
mated by Jiang et al. (2011), long-run modeled ICUC emissions
range from 16.7 g CO2e/MJ to 31.5 g CO2e/MJ. If the mid-range
emissions differential of Jaramillo et al., (2007) were assumed, the
long-run ICUC estimate would be 19.836.5 g CO2e/MJ for the
elasticity assumptions made in Table 3. If the point estimate of
natural gas supply elasticity from Hausman and Kellogg (2015) is
increased by one standard error to 0.97 and other baseline as-
sumptions are maintained, then the long-run ICUC emissions es-
timate declines to 26.6 g CO2e/MJ; if the point estimate is in-
creased by two standard errors, ICUC emissions are 25.2 g CO2e/
MJ.7
As evidenced in Table 3, ICUC emissions estimates are sensitive
to parametric assumptions. To explore these sensitivities and the
magnitude of uncertainty in ICUC estimation, a uniform distribu-
tion is specied for each parameter across the range of estimates
available in the literature and employed by regulators in the U.S.
Then ICUC emissions are computed for each of 10,000 sets of in-
dependent draws from these distributions. The uniform parameter
distributions imply that a random draw of each parameter value
7
We are grateful to a reviewer for suggesting that we report ICUC estimates
accounting for uncertainty in the Hausman and Kellogg (2015) parameter estimate.
.05
.04
.03
Density
.02
.01
0
10 20 30 40 50 60
ICUC (gCO2e/MJ)
Fig. 3. Distribution of simulated ICUC emissions. Distribution of modeled ICUC
emissions from 10,000 independent draws of parameter values from uniform
parameter distributions.
within the specied ranges is equally likely, so that these simu-
lation estimates entail no researcher judgement. Similarly, the
Monte Carlo framework assumes independence among the alter-
native parameter distributions. Given that the different parameter
estimates which underpin the uniform distributions were gener-
ated from independent studies, this seems reasonable. This Monte
Carlo simulation yields an estimated average long-run ICUC effect
of 37.2 g CO2e/MJ and a range of 13.364.4 g CO2e/MJ. The fth,
50th, and 95th percentiles of the distribution are 24.4, 36.8, and
51.2 g CO2e/MJ, respectively. Fig. 3 reports the histogram of long-
run ICUC estimates from the 10,000 independent, random draws.8
The ICUC estimate based upon our baseline parameterization is
lower than the mean and median estimates from the simulation
because, in the baseline, we conservatively assumed the largest
natural gas supply elasticity available in the literature. ICUC
emissions are decreasing in the price responsiveness of natural gas
supply: the greater is the supply response to an increase in price,
the greater is the share of diverted natural gas that is replaced by
new natural gas supply and the smaller is the share replaced by
coal. The negative relationship between the natural gas supply
elasticity and ICUC emissions is depicted in Fig. 4 (a), which plots
simulated ICUC emissions against model inputs for the natural gas
supply elasticity. The strong linear relationship evidences the
sensitivity of modeled ICUC emissions to the gas supply elasticity.
Modeled ICUC emissions are also sensitive to, though increas-
ing in, the coal supply elasticity. The more elastic is coal supply,
the greater is the coal replacement of any natural gas diverted to
the transportation sector and the smaller is any price-induced
electricity demand reduction. This positive relationship between
coal elasticity and ICUC emissions is depicted in Fig. 4 (b), which
plots the simulated ICUC emissions against the coal elasticity in-
puts. The ICUC estimate produced by our preferred model para-
meters depends upon a mid-range coal supply elasticity estimate
(see footnote 5).
Differences in assumed carbon intensities across coal and nat-
ural gas generating units are also important determinants of
modeled ICUC emissions. Coal and gas emission rates are more
8
A reviewer correctly cautions against interpreting this histogram as a prob-
ability distribution for the magnitude of the ICUC. In order for it to have a prob-
ability-distribution interpretation, we would need complete information on the
distributions of the underlying parameters, including their covariance structure.
Instead, the histogram represents the range of ICUC estimates that can be gener-
ated from alternative combinations of estimates for the underlying parameters that
have been set forth in the literature.
 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131 27

ICUC Sensitivity to Input Parameters reected in California's initial low carbon fuel standard are similar
in magnitude, though somewhat larger than other estimates of
biofuel land-use change emissions (Hertel et al., 2009; Edwards
et al., 2010; Tyner et al., 2010; Marelli et al., 2011; Dunn et al.,
60
2013). Regulators recently revised ILUC emissions from corn
ethanol to 19.8 g CO2e/MJ. Indirect emissions from soy biodiesel
50 decline to 29.1 g CO2e/MJ (ARB, 2015).
If even the smallest estimate of the long-run ICUC effect re-
ported in Table 3 (24.7 g CO2e/MJ) is added to the direct emissions
40 of natural gas fuels, then CNG rivals gasoline and diesel in carbon
ICUC

intensity; LNG is more carbon intensive. The high-range estimate

30 of ICUC emissions suggests that natural gas is more carbon in-
tensive than all fuel pathways identied by California regulators
except some hydrogen fuels and electricity. Fig. 5 compares the
20 direct and indirect carbon emissions per MJ of select transporta-
tion fuels.
Because of differences in the efciency with which alternative
10
vehicle powertrains convert fuel into vehicle miles traveled, a
.2 .4 .6 .8
comparison of carbon emissions per vehicle mile traveled may be
Natural Gas Supply Elasticity
more appropriate for regulatory purposes like the implementation
(a) Sensitivity to Natural Gas Supply Elasticity of a low carbon fuel standard, though vehicle efciencies are es-
timated with some error. Fig. 6 compares the carbon emissions of
representative fuel pathways per vehicle mile for light-duty and
heavy-duty vehicles. Because natural gas powered vehicles are less
60
energy efcient than gasoline or diesel vehicles, natural gas fuels
compare even less favorably than they do on an emissions per MJ
50 basis (Burnham et al., 2011). CNG is a more carbon intensive fuel
than ethanol and gasoline, fuels for which it is intended to sub-
stitute to mitigate greenhouse gas emissions. Likewise, CNG and
40 LNG are more carbon intensive heavy-duty vehicle fuels than
ICUC

diesel and biodiesel. Yet, pursuant to California's low carbon fuel

30
standard, the natural gas fuels generate carbon creditsrather than
carbon debitsbecause ICUC emissions are ignored.
While the impetus for transportation fuels policy is pre-
20 dominantly carbon emissions mitigation, impacts on other pollu-
tant emissions should also enter into policy decisions. Substitution
toward coal-red electricity generation increases criteria pollutant
10
emissions because of the substantially higher criteria pollution
2 2.5 3 3.5 4 intensities of coal plants relative to natural gas plants. Natural gas
Coal Supply Elasticity
plants emit approximately 1/5th as much nitrous oxide, 1/25th as
(b) Sensitivity to Coal Supply Elasticity much sulfur dioxide, and 1/100th as much particulate matter per
unit generation as coal plants (Spath and Mann, 2000). Our own
Fig. 4. ICUC sensitivity to input parameters. Modeled ICUC emission estimates are
calculation of smokestack emissions using 2012 Continuous
plotted against natural gas supply elasticity inputs (panel a) and coal supply elas-
ticity inputs (panel b). The strong linear relationship among ICUC estimates and
input parameters indicates the sensitivity of model estimates to parameter choice.
ICUC estimates are decreasing in natural gas supply elasticity. They are increasing
in coal supply elasticity.

highly correlated with simulated ICUC emissions than other model

inputs (r 0.23 and 0.36, respectively, versus r 0.17 and 0.18, for
coal and natural gas supply elasticities, respectively). The model is
insensitive to uncertainty in natural gas sectoral demand elasti-
cities; only industrial demand registers a Pearson correlation dif-
ferent from zero to two decimal places (r 0.03).

3.2. Indirect coal use change emissions and transportation fuels

policy

These estimates of indirect emissions from natural gas fuels are

comparable in size to the estimated ILUC emissions of biofuels
(Fargione et al., 2008; Searchinger et al., 2008; ARB, 2009; Hertel
et al., 2009; Lapola et al., 2010; Laborde, 2011; European Com-
mission, 2012; Khanna and Crago, 2012). California regulators in-
itially imposed an indirect emissions penalty of 30 g CO2e/MJ on
corn ethanol, 46 g CO2e/MJ on Brazilian sugarcane ethanol, and
62 g CO2e/MJ on soy biodiesel (ARB, 2009). The indirect effects
Fig. 5. CO2 emissions per megajoule. Depicted above are CO2 emissions per
Megajoule for several transportation energy sources for light-duty vehicles, buses,
and trucks. For each fuel, emissions are separated between direct emissions and
any applicable indirect emissions. Direct emissions as reported by Burnham et al.
(2011), Rosenfeld and Jackson (2008), and the California Air Resource Board (ARB,
2014). Indirect land-use emissions are reported by the California Air Resource
Board in the 2015 Low Carbon Fuel Standard reauthorization (ARB, 2015). Indirect
coal-use emissions use lifecycle emissions estimates from Jiang et al. (2011), Jar-
amillo et al. (2007), and Edenhofer (2011). Emissions are presented per megajoule.
28 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131

Fig. 6. Depicted above are CO2 emissions per vehicle mile for several transportation energy sources for light-duty vehicles, buses, and trucks. For each fuel, emissions are
separated between direct emissions and any applicable indirect emissions. Direct emissions as reported by Burnham et al. (2011), Rosenfeld and Jackson (2008), and the
California Air Resource Board (ARB, 2014). Indirect land-use emissions are reported by the California Air Resource Board in the 2015 Low Carbon Fuel Standard reauthor-
ization (ARB, 2015). Indirect coal-use emissions use lifecycle emissions estimates from Jiang et al. (2011), Jaramillo et al. (2007), and von Stechow et al. (2011). Emissions are
presented per per vehicle mile.

Emission Monitoring data from the EPA nds that natural gas
plants emit only 14% of the nitrous oxide emitted from coal plants
and less than 1% of the sulfur dioxide. Assuming external damages
of $300/ton of nitrous oxide emissions and $1200/ton of sulfur
dioxide emissions (Muller and Mendelsohn, 2007), our preferred
long-run estimate of the ICUC implies that each gasoline gallon
equivalent unit of natural gas used in transportation causes $0.08
in indirect health damages from the power sector. This exceeds the
estimated $0.065 in direct health benets from replacing a gallon
of gasoline with an equivalent unit of CNG (NRC, 2010).
Though coal and natural gas are expected to dominate electricity
generation at least through mid-century, were high natural gas pri-
ces or policy to induce substantial investment in low-carbon gen-
eration capacity, the eet of generating units would change slowly.
Natural gas transportation policy, would, therefore, create an initial
carbon debt as coal replaces natural gas generation in the short-to-
medium run. Indeed, the carbon debt would not be repaid for ap-
proximately 20 years if fossil generation were replaced by new nu-
clear capacity, approximately the same time-horizon over which the
carbon debt from Brazilian sugarcane ethanol would be repaid
(Fargione et al., 2008). Even short-term increases in carbon emissions
could have negative effects if the atmospheric carbon concentration
surpasses climate tipping points. For example, at 2013 emissions
rates the carbon concentration will exceed 450 parts per million
between 2030 and 2038, causing detrimental and irreversible ocean
acidication (McNeil and Matear, 2008; Johnson et al., 2013).
3.3. Model limitations
The foregoing model of ICUC emissions has some limitations.
First, the analysis is of marginal changes in natural gas use in
transportation. Conceivably excess coal plant capacity may be ex-
ceeded in the short-to-medium run, inducing substitution toward
peak-load-plant generation, as well as necessarily signicant
price-induced reductions in electricity demand. However, because
the average capacity factor for coal plants declined from 73.4% to
59.7% over the past half-decade, capacity factors would remain
below 2008 levels even if the coal eet absorbed the natural gas
generation displaced by 14 million CNG vehicles (8% of U.S. pas-
senger vehicle eet) (EIA, 2014a).
Second, if policy effectively regulates carbon emissions from
smokestacks while leaving vehicle tailpipes unregulated, the ex-
ternal cost of indirect emissions is essentially nil as the carbon cost
is paid in the electricity sector (Fowlie et al., 2012). Thus, lower
emissions from tailpipes may be preferred to lower emissions at
power plants if the latter are effectively regulated, e.g., by an op-
timal sector and economy-wide carbon tax or tradable permit
standard. However, to the extent emissions can leak out of jur-
isdictions with regulated smokestacks, e.g., via a national market
for natural gas, then there is little reason to prefer smokestack
emissions to tailpipe emissions (Cuddington and Wang, 2006;
FERC, 2012; Joskow, 2013).
Third, this analysis estimates ICUC emissions under existing
coal and gas plant emission intensities, and, therefore, are in-
formative about the indirect effects of transportation fuels policy
given the existing generating eet. The generating eet, however,
is likely to evolve in ways that impact the average emission in-
tensities of generating sources. The EPA, for instance, promulgated
regulations in 2015 to lower carbon emissions from existing and
new power plants. The Carbon Pollution Standards for new power
plants would limit newly constructed coal plants to 500 g of CO2
per kilowatt hour when they are enacted. Newly constructed
natural gas plants would be limited to 454 g of CO2 per kilowatt
hour. Such a small relative emission differential would yield a
negative ICUC as price-induced electricity conservation would
dominate emissions from incremental coal generation. The carbon
pollution standards are sufciently onerous, however, that they
are likely to delay the retirement of existing coal plants while
discouraging new coal capacity. If new natural gas plants are built
under the policy, as seems likely given their relatively lax stan-
dards, then the emission intensities of the coal and gas eets are
likely to diverge, increasing the magnitude of ICUC emissions.
EPA also proposes to reduce the carbon intensity of existing
electricity generation capacity by 30%, though states are granted
exibility to achieve unique emission abatement requirements via
a variety of channels, including improved power plant efciency,
greater building efciency, better transmission infrastructure, and
increased biomass and renewable generation. Conceivably, State
Implementation Plans could achieve abatement obligations absent
changes in power plant efciencies. If coal and gas plant emission
rates were each reduced sufciently to comply with the regula-
tions, the emissions differential between a typical coal and gas
plant would decline from the IPCC's estimated 532 gCO2e/kWh to
372 gCO2e/kWh, comparable to the estimate from Jiang et al.
(2011). The long-run ICUC would still be estimated to be on the
order of 16.731.5 g CO2e/MJ. Even if all reductions were achieved
 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131
by reducing carbon intensity at existing coal plants, the ICUC
would be positive.
Fourth, as this model addresses policy interventions that pro-
pose to increase demand for coal for electricity generation, we do
not consider what alternative coal uses, i.e., coal exports, are
crowded out by the policy intervention. In much the same way,
Searchinger et al. (2008) did not consider the alternative uses for
converted land that would have occurred in the absence of bio-
fuel-induced demand for cropland expansion. Such considerations
would be speculative. Moreover, the emissions impacts of coal
exports are typically excluded from carbon policy impact assess-
ments because of unreliable data from which to infer the magni-
tude of potential leakage and because such leakage would un-
dermine the rationale for the policies themselves. For instance, a
perfectly elastic export demand for U.S. coal would imply that all
carbon emissions reductions from displaced coal demand in the
regulated jurisdiction would be fully offset by emissions from in-
creased coal use in the rest of the world. Emissions anywhere pose
the same rst-order threat of climate change, thus limiting the
benets of state-level (national) carbon policy.
Burniaux and Martins (2012) consider the role of coal exports
on carbon leakage from any state-level climate policy. They con-
clude that high transportation costs, differentiated coal products
and other factors limit the degree of integration of world coal
markets and, thereby, the potential for carbon leakage, particularly
for elastic coal supply. Using a simple general equilibrium model,
they estimate leakage of 20% for a world coal supply elasticity of
2.0. This leakage rate declines rapidly as coal supply becomes more
elastic. The world supply is expected to be more elastic than do-
mestic supply, for which we estimate an elasticity of 3.0.
Light et al. (1999) estimate virtually no leakage for a coal supply
elasticity of 3.0 and conventional estimates for trade elasticities.
Further, it is argued that foreign demand for coal may be fairly
inelastic, in which case reductions in coal demand at home do not
imply increases in coal exports abroad (Golden, 2013; Wolak,
2014). Nevertheless, were this analysis to account for an elastic
foreign demand for coal, the carbon leakage would lower the cost
of displacing natural gas in electricity generation, i.e. lower ICUC
emissions, while also lowering the potential gain from carbon
policy like a low carbon fuel standard.
Fifth, this model accounts for indirect emissions of CNG and
LNG in the electric power sector alone. However, natural gas and
coal compete in other sectors as well, including in the industrial
sector where both may be used to generate heat and power.
Likewise, in the residential sector, natural gas heat and electric
heat are substitutable. To the extent natural gas prices rise, in-
dustrial consumption may shift to coal-generated heat and power.
In the residential sector, homes may substitute away from natural
gas heat and toward electric heat which may rely on coal gen-
eration. Thus, indirect emissions from coal-use change across the
economy are likely to be greater than the indirect emissions esti-
mated here in only the electric power sector.
Finally, this partial equilibrium analysis can inform sector-
specic climate change mitigation and pollution abatement po-
licies, like low carbon fuel standards, but it also highlights the
inefciency of such policies that ignore their impacts in related
sectors and fail to equate the marginal cost of abatement across
sectors. The ICUC is among a growing number of indirect effects of
fuel policy identied in the literature, e.g., Hochman et al. (2011),
Rajagopal et al. (2011), Zilberman et al. (2013) and Smeets et al.
(2014), that demonstrate the complexity of partial equilibrium
modeling of sectoral carbon policy impacts and point toward ef-
ciency gains from economy-wide carbon abatement policy that
would, for example, impose a uniform tax on carbon emissions
regardless of the sectors in which they are emitted.
29
4. Conclusion and policy implications
In order to diminish pollution emissions from petroleum fuel
combustion, policy makers and researchers, including some
economists, advocate for policy that increases natural gas con-
sumption in the transportation sector. This analysis of the ICUC
effect, however, indicates that attendant emissions in the electric
power sector may reduce carbon abatement from low carbon fuel
policies like that of California, the EU, and British Columbia. In-
attention to these indirect emissions may yield policies that induce
a suboptimal balance of fuel consumption, e.g., too much natural
gas consumption and too little consumption of biofuel, gasoline,
and diesel, which are estimated to be no dirtier than CNG and LNG
when accounting for ICUC. The ICUC is estimated to rival in
magnitude the ILUC emissions of biofuels that have garnered
considerable scholarly and policy interest. The results of this
analysis suggest policy that accounts for ILUC emissions from
biofuels but not ICUC emissions from natural gas are biased
against biofuels, contrary to claims that policy is biased against
CNG and LNG. Moreover, as this analysis suggests alternative fuels
other than biofuels may be associated with substantial indirect
emissions, it augers for a comprehensive approach, e.g., a general
equilibrium approach, to evaluating transportation fuels policies
as a complement to the multi-market analysis undertaken here
and the partial equilibrium approach used in much of the ILUC
literature. Further exploration of other indirect emissions path-
ways is also warranted.
Because of uncertainty in values for model inputs, the magni-
tude of the ICUC is uncertain. This analysis has attempted to ac-
count for this uncertainty by estimating ICUC emissions for various
plausible model parameters and by conducting Monte Carlo si-
mulations across the range of estimated parameter values. How-
ever, further research is warranted to lessen the uncertainty in the
physical and economic relationships to which ICUC is sensitive.
Moreover, this uncertainty, and the potential of other substantial
indirect emissions from transportation fuels, demonstrates the
inefciency of second and third-best policies for carbon abate-
ment, including subsidies to specic fuels and technologies and
low carbon fuel standards that fail to account for general equili-
brium effects. An economy-wide carbon tax could avoid these
inefciencies and the implicit biases these policies may introduce.
Acknowledgments
The authors declare no conicts of interest. No outside funding
sources were used for this paper.
References
ARB, 2009. California Air Resources Board. Low Carbon Fuel Standard: Sub-
chapter 10. Climate Change Article 4. Regulations to Achieve Greenhouse Gas
Emissions Reductions Subarticle 7. California Air Resources Board.
ARB, 2014. California Air Resources Board. iLUC Analysis for the Low Carbon Fuel
Standard (Update), Technical Report, California Air Resources Board (accessed
June 2015) http://www.arb.ca.gov/fuels/lcfs/lcfs_meetings/iluc_presentation_
031014.pdf.
ARB, 2015. California Air Resources Board. Low Carbon Fuel Standard: Sub-
chapter 10. Climate Change Article 4. Regulations to Achieve Greenhouse Gas
Emissions Reductions Subarticle 7, Final Order, California Air Resources Board,
Sacramento, CA.
Arora, Vipin, 2014. Estimates of the price elasticities of natural gas supply and
demand in the United States (MPRA Paper no. 54232).
Beck, Tony, Jolly, Lindsay, Tomislav, Loncar, 1991. Supply Response in the Australian
Black Coal Industry. Australian Government Publishing Service, Australia.
Brown, Jason P., Andres, Kodaka, 2014. U.S. electricity prices in the wake of growing
natural gas production. Main Str. Econ. 2, 18.
Brown, Stephen P.A., 1998. Global warming policy: some economic implications.
Econ. Rev. Fed. Reserve Bank Dallas, 2635.
30 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131
Brown, Stephen P.A., Ycel, Mine K., et al., 1995. Energy prices and state economic
performance. Fed. Reserve Bank Dallas Econ. Financ. Policy Rev. Second Quart.,
1323.
Brown, Stephen P.A., Krupnick, Alan J., 2010. Abundant shale gas resources: long-
term implications for US natural gas markets, resources for the future. Dis-
cussion Paper 10-41. Washington, DC.
Burnham, Andrew, Han, Jeongwoo, Clark, Corrie E., Wang, Michael, Dunn, Jennifer
B., Palou-Rivera, Ignasi, 2011. Life-cycle greenhouse gas emissions of shale gas,
natural gas, coal, and petroleum. Environ. Sci. Technol. Technol. 46 (2), 619627.
Burniaux, Jean-Marc, Martins, Joaquim Oliveira, 2012. Carbon leakages: a general
equilibriumview. Econ. Theory 49 (2), 473495.
Caulton, Dana R., Shepson, Paul B., Santoro, Renee L., Sparks, Jed P., Howarth, Robert
W., Ingraffea, Anthony R., Cambaliza, Maria O.L., Sweeney, Colm, Karion, Anna,
Davis, Kenneth J., 2014. Toward a better understanding and quantication of
methane emissions from shale gas development. Proc. Natl. Acad. Sci. 23,
62376242.
CEC, California Energy Commission, 2012. 2011 Natural gas market assessment:
outlook. California Energy Commission, Sacramento, CA (accessed June 2015):
http://www.energy.ca.gov/2011publications/CEC-200-2011-012/CEC-200-
2011-012-SD.pdf.
Chan, Sabrina, Miranda-Moreno, Luis F., Ahsan, Alam, Marianne, Hatzopoulou,
2013. Assessing the impact of bus technology on greenhouse gas emissions
along a major corridor: a lifecycle analysis. Transp. Res. D: Transp. Environ. 20,
711.
CRS, Congressional Research Service, 2014. Natural Gas for Cars and Trucks: Options
and Challenges. Technical Report, Congressional Research Service, Washington,
DC.
Cuddington, John T., Wang, Zhongmin, 2006. Assessing the degree of spot market
integration for US natural gas: evidence from daily price data. J. Regul. Econ. 29
(2), 195210.
Dahl, C., Roman, C., 2004. Energy Demand Elasticities - Fact or Fiction: A Survey
Update. Unpublished manuscript. Colorado School of Mines Working Paper,
Golden, CO.
Davis, Lucas W., Muehlegger, Erich., 2010. Do Americans consume too little natural
gas? An empirical test of marginal cost pricing. RAND J. Econ. 41 (4), 791810.
Dunn, Jennifer B., Mueller, Steffen, Kwon, Ho Young, Wang, Michael Q., et al., 2013.
Land-use change and greenhouse gas emissions from corn and cellulosic
ethanol. Biotechnol. Biofuels 6 (1), 51.
Edwards, Robert, Mulligan, Declan, Marelli, Luisa, 2010. Indirect land use change
from increased biofuels demand. JRC Scientic and Technical Reports, 59771.
EIA, Energy Information Administration, 2013. Coal Market Module of the National
Energy Modeling System Documentation.
EIA, Energy Information Administration, 2013. Data: Natural Gas Consumption by
End Use. Energy Information Administration, Washington, DC (accessed June
2015) at url http://www.eia.gov/dnav/ng/ng_cons_sum_dcu_nus_a.htm.
EIA, Energy Information Administration, 2014a. Annual Energy Outlook 2014 with
Projections to 2040. Energy Information Administration, Washington, DC (ac-
cessed June 2015): http://www.eia.gov/todayinenergy/index.cfm?
tg AEO2014%20(Annual%20Energy%20Outlook%202014).
EIA, Energy Information Administration, 2014b. Capacity Factors for Utility Scale
Generators Not Primarily Using Fossil Fuels, January 2008March 2014. Energy
Information Administration, Washington, DC. Online: url http://www.eia.
gov/electricity/monthly/epm_table_grapher.cfm?t epmt_6_07_b.
EIA, Energy Information Administration, 2015a. Annual energy outlook 2015.
Technical Report, Energy Information Administration, Washington, DC.
EIA, Energy Information Administration, 2015b. Annual Report of Natural Gas and
Supplemental Gas Supply and Disposition.
EIA, Energy Information Administration, 2015c. Electric Power Monthly: Net Gen-
eration by Energy Source, Energy Information Administration. Washington, DC
(accessed June 2015) at url http://www.eia.gov/electricity/monthly/epm_ta
ble_grapher.cfm?t epmt_1_1.
EIA, Energy Information Administration, 2015d. Monthly energy review. Technical
Report. Energy Information Administration, Washington, DC.
EIA, Energy Information Administration, 2015e. U.S. Natural Gas Vehicle Fuel
Consumption.
EIA, Energy Information Administration, 2012. Cheaper Natural Gas Alters Gen-
eration Dispatch in Southeast.
EPA, U.S. Environmental Protection Agency, 2012. 2017 and later model year light-
duty vehicle greenhouse gas emissions and corporate average fuel economy
standards. In: Federal Register, Vol. 77 U.S. Government Printing Ofce Wa-
shington, DC. pp. 6262463200.
EPA, U.S. Environmental Protection Agency, 2013a. Documentation for EPA Base
Case v.5.13: Using the Integrated Planning Model, Technical Report, U.S. En-
vironmental Protection Agency (accessed June 2015): http://www.epa.gov/air
markets/documents/ipm/Documentation.pdf.
EPA, U.S. Environmental Protection Agency, 2013b. Inventory of U.S. Greenhouse
Gas Emissions and Sinks: 19902013. Environmental Protection Agency, Wa-
shington, DC (accessed June 2015): http://www.epa.gov/climatechange/pdfs/
usinventoryreport/US-GHG-Inventory-2015-Main-Text.pdf.
EPAct, 2005. H.R. 6109th Congress: Energy Policy Act of 2005 (accessed June
2015): https://www.govtrack.us/congress/bills/109/hr6.
European Commission, 2012. Proposal for a Directive of the European Parliament
and of the Council amending Directive 98/70/EC relating to the quality of petrol
and diesel fuels and amending Directive 2009/28/EC on the promotion of the
use of energy from renewable sources. European Commission, Brussels (ac-
cessed June 2015): http://ec.europa.eu/clima/policies/transport/fuel/docs/
com_2012_595_en.pdf.
Fargione, Joseph, Hill, Jason, Tilman, David, Polasky, Stephen, Hawthorne, Peter,
2008. Land clearing and the biofuel carbon debt. Science 319 (5867),
12351238.
FERC, Federal Energy Regulatory Commission, 2012. Energy Primer: A Handbook of
Energy Market Basics, Ofce of Enforcement, Federal Energy Regulatory Com-
mission, Washington, DC.
FERC, Federal Energy Regulatory Commission, 2015. Energy Primer: A Handbook of
Energy Market Basics, Washington, DC.
Fowlie, Meredith, Knittel, Christopher R., Wolfram, Catherine, 2012. Sacred cars?
Cost-effective regulation of stationary and nonstationary pollution sources. Am.
Econ. J.: Econ. Policy 4 (1), 98126.
Frank, Charles, 2014. The Net Benets of Low and No-Carbon Electricity Technol-
ogies. Brookings Institution, Washington, DC.
Fulton, Mark, Mellquist, Nils, Bluestein, Joel, 2011. Comparing Life-Cycle Green-
house Gas Emissions from Natural Gas and Coal. Deustsche Bank Group,
Frankfurty, Germany.
Golden, Mark, 2013. Reduce Greenhouse Gas by Exporting Coal? Palo Alto, CA
(accessed June 2015): http://news.stanford.edu/news/2013/january/coal-asia-
environment-011513.html.
Graff-Zivin, Joshua S., Kotchen, Matthew J., Mansur, Erin T., 2014. Spatial and
temporal heterogeneity of marginal emissions: implications for electric cars
and other electricityshifting policies. J. Econ. Behav. Organ. 107, 248268.
Harvey, Curtis E., 1986. Coal in Appalachia: An Economic Analysis. University Press
of Kentucky, Lexington, KY.
Hausman, Catherine, Kellogg, Ryan, 2015. Welfare and distributional implications of
shale gas, Working Paper Number 2115, National Bureau of Economic Research,
Cambridge, MA.
Hertel, Thomas, Golub, Alla, Jones, Andrew, O'Hare, Michael, Plevin, Richard,
Kammen, Daniel, 2009. Global Land Use and Greenhouse Gas Emissions Im-
pacts of US Maize Ethanol: The Role of Market-Mediated Responses. Center for
Global Trade Analysis, Department of Agricultural Economics, Purdue Uni-
versity, West Lafayette, IN.
Hochman, Gal, Rajagopal, Deepak, Zilberman, David, 2011. The effect of biofuels on
the international oil market. Appl. Econ. Perspect. Policy 33 (3), 402427.
IHS, 2015. LNG in transportation: challenging oil's grip. Technical Report. IHS,
Englewood, CO.
Jaramillo, Paulina, Grifn, W. Michael, Matthews, H. Scott, 2007. Comparative life-
cycle air emissions of coal, domestic natural gas, LNG, and SNG for electricity
generation. Environ. Sci. Technol. 41 (17), 62906296.
Jiang, Mohan, Grifn, W. Michael, Hendrickson, Chris, Jaramillo, Paulina, Van-
Briesen, Jeanne, Venkatesh, Aranya, 2011. Life cycle greenhouse gas emissions
of Marcellus shale gas. Environ. Res. Lett. 6 (3), 034014.
Johnson, Leo, Jonathan, Grant, Low, Lit Ping, 2013. Busting the carbon budget. Low
Carbon Economy Index 2013. Price Waterhouse Cooper (accessed June 2015):
http://www.pwc.co.uk/en_UK/uk/assets/pdf/low-carbon-economy-index-
2013.pdf.
Joskow, Paul, 1988. Price adjustment in long-term contracts: the case of coal. J. Law
Econ. 31 (1), 4783.
Joskow, Paul L., 2013. Natural gas: from shortages to abundance in the United
States. Am. Econ. Rev. 103 (3), 338343.
Khanna, Madhu, Crago, Christine L., 2012. Measuring indirect land use change with
biofuels:Implications for policy. Annu. Rev. Resour. Econ. 4 (1), 161184.
Knittel, Christopher R., 2012. Leveling the Playing Field for Natural Gas in Trans-
portation. Hamilton Project of the Brookings Institution, Washington, DC.
Kyle, Page, Kim, Son H., 2011. Long-term implications of alternative light-duty ve-
hicle technologies for global greenhouse gas emissions and primary energy
demands. Energy Policy 39 (5), 30123024.
Laborde, David, 2011. Assessing the land use change consequences of European
biofuel policies. International Food Policy Institute, Washington, DC (accessed
June 2015): http://ebrary.ifpri.org/cdm/ref/collection/p15738coll5/id/197.
Lapola, David M., Schaldach, Ruediger, Alcamo, Joseph, Bondeau, Alberte, Koch,
Jennifer, Koelking, Christina, Priess, Joerg A., 2010. Indirect land-use changes
can overcome carbon savings from biofuels in Brazil. Proc. Natl. Acad. Sci. 107
(8), 33883393.
Larive, R., Beziat, J.-F., Edwards, J.-C., 2011. Well-to-wheels analysis of future auto-
motive fuels and powertrains in the European context. Joint Research Centre,
European Commission, Brussels.
Light, Miles K., Kolstad, Charles D., Rutherford, Thomas F., 1999. Coal markets and
the Kyoto protocol. Discussion Paper in Economics. University of Colorado,
Boulder. Vol. 23.
Linn, Joshua, Muehlenbachs, Lucija, Wang, Yushuang, 2014. How do natural gas
prices affect electricity consumers and the environment? Discussion Paper RFF
DP 14-19. Resources for the Future, Washington, DC.
Lucas, Davis, 2012. Prospects for nuclear power. J. Econ. Perspect. 26 (1), 4966.
Marelli, Luisa, Ramos, Fabien, Hiederer, Roland, Koeble, Renate, 2011. Estimate of
GHG Emissions from Global Land Use Change Scenarios. Joint Research Centre,
European Commission, Brussels.
Mason, Charles F., Muehlenbachs, Lucija, Olmstead, Sheila M., 2015. The Economics
of Shale Gas Development. Resources for the Future Discussion Paper 1442.
Resources for the Future, Washington, DC.
McNeil, Ben I., Matear, Richard J., 2008. Southern Ocean acidication: a tipping
point at 450 ppm atmospheric CO2. Proc. Natl. Acad. Sci. 105 (48), 1886018864.
Mellish, Michael, 1998. Coal pricing methodology for the annual energy outlook.
Issues in Midterm Analysis and Forecasting Energy Information Administration,
Washington, DC.
 S. Sexton, J. Eyer / Energy Policy 95 (2016) 2131
Micaela, Ponce, Neumann, Anne, 2014. Elasticities of supply for the US natural gas
market. Discussion Paper no. 1372. German Institute for Economic Research,
Berlin.
Muller, Nicholas Z., Mendelsohn, Robert, 2007. Measuring the damages of air pol-
lution in the United States. J. Environ. Econ. Manag. 54 (1), 114.
Nigro, Nick, Jiang, Shelley, 2013. Lifecycle Greenhouse Gas Emissions from Different
Light-duty Vehicles and Fuel Pathways: A Synthesis of Recent Research. Center
for Climate and Energy Solutions, Arlington, CA (accessed June 2015): http://
www.c2es.org/publications/.
NRC, National Research Council, 2010. Hidden Costs of Energy: Unpriced Con-
sequences of Energy Production and Use. The National Academies Press.
Nuclear Energy Institute, 2013. Fuel as a percentage of electric power production
costs. Technical Report (accessed June 2015): http://www.nei.org/Knowledge-
Center/Nuclear-Statistics/Costs-Fuel,-Operation,-Waste-Disposal-Life-Cycle/
Fuel-as-a-Percent-of-Production-Costs.
Obama, B., 2014. State of the union address. In: 225th Annual State of the Union
Address The White House, Washington, DC.
Ou, Xunmin, Xiliang, Zhang, 2013. Life-cycle analyses of energy consumption and
GHG emissions of natural gas-based alternative vehicle fuels in China. J. Energy
2013, 18.
Rajagopal, Deepak, Gal, Hochman, David, Zilberman, 2011. Indirect fuel use change
(IFUC) and the lifecycle environmental impact of biofuel policies. Energy Policy
39 (1), 228233.
Rosenfeld, Jeffrey, Jackson, Michael, 2008. Life-Cycle Cost Model and Pollutant
Emissions Estimates Greenhouse Gas and Criteria Pollutant Emissions Esti-
mates. TIAX LLC, Cupertino, CA.
Smeets, Edward, Tabeau, Andrzej, van Berkum, Siemen, Moorad, Jamil, van Meijl,
Hans, Geert, Woltjer, 2014. The impact of the rebound effect of the use of rst
generation biofuels inthe EU on greenhouse gas emissions: a critical review.
Renew. Sustain. Energy Rev. 38, 393403.
Sovacool, Benjamin K., 2009. The intermittency of wind, solar and renewable
electricity generators: Technical barrier or rhetorical excuse? Util. Policy 17 (3),
288296.
Stephenson, Eleanor, Doukas, Alexander, Shaw, Karena, 2012. Greenwashing gas:
31
might a transition fuel label legitimize carbon-intensive natural gas develop-
ment? Energy Policy 46, 452459.
Spath, P.L., Mann, M.K., 2000. Life cycle assessment of a natural gas combined-cycle
power generation system. National Renewable Energy Laboratory, Golden, CO.
Timothy, Searchinger, Heimlich, Ralph, Houghton, Richard A., Dong, Fengxia, Elo-
beid, Amani, Fabiosa, Jacinto, Tokgoz, Simla, Hayes, Dermot, Yu, Tun-Hsiang,
2008. Use of US croplands for biofuels increases greenhouse gases through
emissions from land-use change. Science 319 (5867), 12381240.
von Stechow, C., Edenhofer, O., Pichs-Madruga, R.-Sokona, Y.-Seyboth, K., Mat-
schoss, P., Kadner, S., Zwickel, T., Eickemeier, P., Hansen, G., Schlomer, S., 2011.
Annex II: Methodology. In: IPCC Special Report on Renewable Energy Sources
and Climate Change Mitigation. Cambridge University Press, Cambridge, UK.
Tong, Fan, Paulina, Jaramillo, Azevedo, Ines L., 2015a. Comparison of life cycle
greenhouse gases from natural gas pathways for light-duty vehicles. Environ.
Sci. Technol. 29 (9), 60086018.
Tong, Fan, Paulina, Jaramillo, Azevedo, Ines L., 2015b. Prev. article next article table
of contents comparison of life cycle greenhouse gases from natural gas path-
ways for medium and heavy-duty vehicles. Environ. Sci. Technol. 49 (12),
71237133.
Tyner, Wallace E., Farzad, Taheripour, Qianlai, Zhuang, Dileep, Birur, Uris, Baldos,
2010. Land use changes and consequent CO2 emissions due to US corn ethanol
production: a comprehensive analysis. Report to Argonne National Laboratory
Purdue University, West Lafayette, IN.
Wells, Jim, 2006. Natural gas: factors affecting prices and potential impacts on
consumers. Technical Report, Government Accountability Ofce, Washington,
DC.
Wolak, Frank, 2014. Q&A with Frank Wolak on U.S. coal exports and climate
change. Property and Environment Research Center (accessed June 2015):
http://www.perc.org/blog/qa-frank-wolak-us-coal-exports-and-climate-
change.
Zilberman, David, Barrows, Geoff, Hochman, Gal, Rajagopal, Deepak, 2013. On the
indirect effect of biofuel. Am. J. Agric. Econ. 95 (5), 13321337.
Zimmerman, Martin B., 1981. The US Coal Industry: The Economics of Policy Choice.
MIT Press, Cambridge, MA.