The Abatement Cost of Methane Emissions from Natural Gas Production

55 Pages·2017·1.74 MB·English

The Abatement Cost of Methane Emissions from
Natural Gas Production
∗
Levi Marks
PRELIMINARY AND INCOMPLETE; PLEASE DO NOT CITE
Abstract
The fact that natural gas produces only about half as much carbon dioxide as coal
when burned to produce electricity has led many to believe gas will play a vital role
in the transition to sustainable energy. However, whether gas could serve as a “bridge
fuel” critically depends on the cost of controlling methane emissions from the gas sup-
ply chain. This paper examines firm behavior in response to changes in gas prices to
empiricallyestimatetherelationshipbetweenpricesandmethaneemissionratesforthe
extraction segment of the industry. Because optimally-behaving firms mitigate leakage
onlyuptothepointatwhichtheirmarginalcostofabatementequalsthewholesalegas
price, this estimated relationship can be mapped to an abatement cost curve. Results
indicate that methane emissions from natural gas production can be reduced at very
low cost relative to greenhouse gas emissions from other sectors. In particular, I esti-
mate that a relatively modest emissions tax on methane equivalent to a $5/ton price
on carbon would decrease methane emissions by 56 percent. Furthermore, extending
these results to make an out-of-sample prediction, I estimate that a tax designed to
fully internalize the social cost of methane would decrease emissions by 75 percent
while increasing the net cost of gas extraction by less than one percent. This finding
indicates natural gas is likely to remain highly competitive as an energy source under
methane regulation.
JEL: D22, H23, Q35, Q4, Q54
∗UniversityofCalifornia,SantaBarbara. Email: [email protected]. IamgratefultoOlivierDeschenes,
Matthew Zaragoza-Watkins, Kyle Meng, Ted Frech, Kristina Mohlin, David Lyon, Catherine Hausman,
Cl´ement de Chaisemartin, Sarah Bana, and Matthew Wibbenmeyer for invaluable feedback and suggestions
that helped shape this paper. I would also like to thank the Environmental Defense Fund for assistance in
acquiring dataon natural gasproduction andprices. All errors are mine. This version: September 19, 2018.
The most recent version can be downloaded here.
1
Introduction
Naturalgashasmanyfeaturesthatmakeitanattractiveenergysourceformitigatingclimate
change. It is abundantly available, the technology and infrastructure necessary to utilize it
on a large scale are already in place today, and it is cost-competitive with coal for electricity
generation while only producing about half the burn-point greenhouse gas (GHG) emissions.
Furthermore, quick ramp rates make gas-fired generation well-suited for balancing out the
intermittency of wind and solar power. These features have motivated the idea the transition
to low-carbon energy could be made more efficient by switching first from other fossil fuels
to gas and later from gas to renewables.1 However, this “bridge fuel” narrative has recently
been called into question by a growing body of scientific studies uncovering the problem of
methane leakage.
Methane (CH ), the principal component of natural gas, is itself a greenhouse gas that
4
is shorter-lived than carbon dioxide (CO ) but vastly more potent, with warming effects
2
about 30 times stronger than CO on a 100-year time horizon.2 Alvarez et al. (2012) cal-
2
culate that if more than 3.2 percent of total production escapes into the atmosphere before
being combusted for power production, natural gas is actually more harmful than coal on
a warming-per-MWh basis. While there is still considerable uncertainty as to how much
methane is presently being emitted, the true figure is likely to be in the neighborhood of
that threshold, as most estimates range from 2 to 4 percent of total production (Alvarez
et al., 2018; Sanchez & Mays, 2015; Moore et al., 2014).
A comprehensive climate change mitigation policy that fails to account for methane
leakage will therefore not only pass over low-cost abatement opportunities, but will also
result in over-substitution to gas from other fossil fuels. On the other hand, if climate
regulation does account for methane leakage, the continued cost-competitiveness of gas will
1See, for example, Jenner & Lamadrid (2013), Levi (2013), and Brown et al. (2009).
2This figure is drawn from the IPCC’s Fifth Assessment Report and is the factor used by the EPA’s Green-
house Gas Reporting Program, a primary data source for this paper. Estimates of the 100-year warming
potential of CH range from about 28-36 times that of CO (IPCC, 2013).
4 2
2
depend on the cost of reducing CH emissions from the gas supply chain. This paper informs
4
this question and others related to the development of methane regulation by estimating
a marginal abatement cost curve (MACC) for the natural gas extraction industry, which
currently accounts for nearly three-fourths of total methane emissions from the gas supply
chain in the United States (EPA, 2017).
The estimated MACC implies substantial low-cost GHG abatement opportunities com-
pared to other sectors. In particular, I predict that an emissions tax or permit price on
methane equivalent to a $5/ton carbon price would reduce emissions by roughly 56 percent.3
This represents a decrease of about 46 million tons of CO -equivalent emissions per year at
2
an annual net cost of $73 million, which is less than 0.1 percent of the wholesale value of all
gas produced in the United States.4 Furthermore, though it is an out-of-sample prediction,
I estimate that a methane price designed to fully internalize its social cost would reduce
emissions by about 75 percent at a net cost of only $280 million.5
My empirical strategy relies on spatially linking natural gas production facilities to
geographically-dispersedtradinghubstoexaminehowCH emissionratesrespondtochanges
4
in wholesale gas prices. I estimate this relationship using data on emissions from the EPA’s
Greenhouse Gas Reporting Program (GHGRP), a federally-mandated survey covering the
period from 2011-2016, and data on market prices and extraction activity from SNL and
DrillingInfo, respectively.6 Because in this setting the pollutant itself is also a priced com-
3Note that accurately monitoring CH emissions from the gas supply chain presents a significant challenge
4
to successfully implementing an emissions tax or trading program at this time. Unlike smokestack CO
2
emissions,fugitiveCH emissionsareinherentlydifficulttomeasure. However,technologicaladvancements
4
arerapidlyloweringmonitoringcostsandmarket-basedinstrumentsmaystillbeeffectivelydeployedunder
conditions of imperfect measurement (Montero, 2005; Cremer & Gahvari, 2002).
4This figure represents only physical costs and sets aside questions of how tax or permit revenue might be
distributed.
5Average annual gas prices range from about $2-$6/Mcf over the study period, whereas the social cost of
methane is about $27/Mcf. This figure is for emissions generated in 2020 assuming a 3 percent discount
rate and normalized to 2018 dollars (EPA, 2016).
6While the GHGRP is the most comprehensive dataset of methane emissions currently available, it is
importanttonotethatitdoesnotprovideadirectmeasurementofemissions,butratheranestimatebased
onequipmentcountsandemissionfactors,recordsoffirmactivity,andvariousotherinputs. Asisdiscussed
in detail in Section 3, in-depth scientific measurement studies have shown that it is relatively noisy fails to
capture some important sources of CH emission sources altogether. The empirical strategy used in this
4
paper is designed to address these challenges to make use of the signal that is available in the GHGRP.
3
modity, the estimated relationship between abatement and price can be directly mapped to a
relationship between abatement and cost. In other words, if firms are rational and there are
no market failures, they will choose a level of leakage abatement that equates the expected
value of a potentially captured unit of gas equal to the cost of the abatement activity re-
quired to capture it. A hypothetical tax or permit price applied to methane emissions would
therefore have the same effect on firms’ emitting behavior as would an increase in wholesale
gas prices. I use this condition to simulate how firms’ emissions would change following the
implementation of methane pricing, then aggregate these results to construct a sector-wide
marginal abatement cost curve.
Instead of calculating abatement costs using a bottom-up engineering approach, this em-
pirical strategy relies on actual firm behavior. While engineering estimates for the costs of
various abatement activities exist and are useful,7 they are limited in their ability to ex-
trapolate abatement costs for entire sectors, firms, or even individual facilities because they
cannot appropriately account for opportunity costs, learning costs, and other unobserved
factors. This is well-documented for GHG abatement through investments in energy effi-
ciency (Fowlie et al., 2018; Gillingham & Palmer, 2014; Allcott & Greenstone, 2012) and
carbon sequestration (Lubowski et al., 2006; Stavins, 1999). In contrast, by implicitly cap-
turing the firm’s decision-making process to employ the most efficient abatement measures
first, the approach used in this paper is able to account for all factors that are observed
by the firm but unobserved by the econometrician. This approach is particularly useful for
predicting the effect of regulating methane using an emissions tax or trading program, as
these instruments similarly incentivize firms to exploit the least costly abatement options
first.
This paper is broadly related to a sizable strand of literature in economics focused on
estimating abatement costs. Economy-wide GHG abatement cost curves are most commonly
estimated using computational general equilibrium models (Morris et al., 2012; Chen et al.,
7See, for example, ICF (2016) or EPA (2015).
4
2007; Klepper & Peterson, 2006), but have also been estimated empirically in at least one
instance (Meng, 2017). Sector-specific abatement costs have been empirically estimated by
Cullen & Mansur (2017), who recover an abatement cost curve for the US electricity sector,
and Anderson & Sallee (2011), who estimate the marginal abatement cost of tightening fuel
economy standards, among others. Another approach has been to estimate abatement costs
by using firm-level financial data to analyze how abatement expenditures respond to enacted
regulatory policies (Isaksson, 2005; Dasgupta et al., 2001). This paper introduces a novel
identification strategy based on the fact that the pollutant is also a priced commodity, and
it is the first to empirically estimate an abatement cost curve for methane emissions from a
sector of the natural gas industry. While I have applied this strategy to production, it may
potentially be employed to estimate CH abatement costs for the gas processing and storage
4
sectors in the future.8
This paper also contributes to an emerging literature on methane leakage. Although a
substantial literature has developed in science and engineering over the last decade exploring
the nature and extent of the problem,9 so far methane leakage has received little attention
from economists. One notable exception is Hausman & Muehlenbachs (2016), who reveal
and quantify regulatory distortions that allow gas distribution utilities to pass the cost of
leaked gas through to ratepayers, which results in inefficient levels of emissions. This paper
differs along a number of key dimensions. First, instead of estimating abatement costs using
public utilities’ financial data (which is not available for the production sector), I exploit
firms’ first order condition for leakage abatement (which is not applicable to the distribution
sector due to regulatory distortions). Second, instead of using “lost and unaccounted for
gas” (the difference between a utility’s purchases and sales, a measure that has its own
advantages and disadvantages), I use emissions data from the GHGRP. Finally, I estimate
abatement costs for the production sector, which generates the majority of CH emissions
4
8This method is not applicable to the transmission or distribution sectors, which (at time of publication)
are regulated such that pipeline owners are able to pass cost of lost gas through to their customers.
9See Alvarez et al. (2018) for a synthesis of recent studies and Brandt et al. (2014) for a compact overview
of earlier research.
5
from the US natural gas industry (EPA, 2017).
I begin by providing further background on methane leakage in the next section. Section
2 presents a model of firms’ extraction and leakage decisions that provides intuition for
the empirics. Section 3 describes data sources for emissions, production, price, and other
variables used in the analysis. Section 4 presents the empirical framework and results and
compares them to other estimates of abatement costs, and Section 5 concludes.
1 Background
Methane accounts for about 16 percent of greenhouse gas emissions worldwide, making it
the second most important greenhouse gas following carbon dioxide (IPCC, 2014). Similarly
to CO , significant quantities of methane are always being released into the atmosphere by
2
natural sources (primarily wetlands and organic decomposition), but recent human activity
has accelerated this flow to levels that are disrupting global equilibrium climate. The three
primary anthropogenic sources are agriculture, landfills, and natural gas systems, as gas used
for heating and electricity generation is composed of around 90 percent methane. Although
there is still considerable uncertainty as to how much gas is being emitted by each sector, the
EPA’s Greenhouse Gas Inventory (GHGI) estimates that natural gas systems are presently
the largest source in the US and that production is the largest source within the industry
(see Figure 1).10 Higher rates of CH emissions in production compared to other sectors are
4
consistent with a lower commodity value prior to processing, transport, and storage.
Production’s share of emissions has also increased dramatically over the last few decades,
up from about 50 percent in 1990 to about 72 percent in 2015. The shift is partly due to
improvements in the processing, transportation, storage, and distribution segments of the
supply chain, but also due to the rise in unconventional extraction technologies such as hy-
draulic fracturing (Howarth et al., 2011). Natural gas consumption in the United States has
10The GHGI is an EPA emissions monitoring project that is related to, yet distinct from, the GHGRP.
While the GHGRP is focused on accurately tracking emissions for high-emitting facilities, the GHGI is
focused on creating a comprehensive picture of all US emissions at the industry level.
6
Figure 1: Fugitive methane emissions from the various components of the natural gas supply
chain (EPA Greenhouse Gas Inventory 2015 Estimates). (return)
AdaptedwithpemissionfromAEMONGFR(2014)
similarly increased following the shale gas revolution, as low prices have prompted exten-
sive investment in gas-fired electricity generation. Gas is expected to continue to increase
its share of the US energy mix in the short and medium-term future (EIA, 2017), making
it imperative to account for fugitive methane emissions in any broad-based climate change
mitigation strategy.
Broadly speaking, CH is unintentionally released into the atmosphere at gas produc-
4
tion facilities through leaks in extraction, initial processing, and transmission equipment. It
is also intentionally vented during certain procedures in well completions, workovers, and
maintenance.11 Sites that employ hydraulic fracturing of shale formations release substan-
tially more methane in the well completion process, as hydraulic fluids return some gas to
the surface that is either uneconomical to capture or arrives before the pipes are installed to
connect the well to the rest of the network. There is a high degree of heterogeneity in leak-
age rates across facilities, which is reflected both in scientific measurement studies (Sanchez
& Mays, 2015; Subramanian et al., 2015) and in the GHGRP, where production facilities’
11Well completion consists of all activities between actual drilling and extraction of gas for sale, which
includes installing equipment and testing, as well as hydraulic fracturing and retrieval of fluids for tight-
gas reservoirs. Workovers describe major operations to repair or stimulate gas flow at existing wells.
7
emission rates vary from less than .01 percent to over 10 percent. Finally, natural gas is
often found alongside petroleum, in which case it may be either vented, flared, or collected
and sold (if it is economical to do so) by wells that primarily extract oil.12
As of now, regulations on methane emissions from oil and gas production are not well-
established in the United States. In late 2016, the EPA introduced performance standards
for new wells, processing plants, and compression stations. In 2018, however, the EPA’s new
administration proposed amendments that would greatly weaken these requirements. Also
in 2016, the Bureau of Land Management (BLM) finalized a policy to require wells located
on federal and tribal lands to capture high percentages of gas in place of venting and flaring
on the basis of conserving federal resources. However, this policy was never implemented
and its future remains uncertain. In terms of local regulations, in 2014 Colorado introduced
relatively strong performance standards for new and existing wells, including equipment
mandates, waste-reducingproceduresduringwellcompletion, andsemi-annualleakdetection
and repair.13
2 Theory
This section develops a theoretical model of the production and emission decisions faced by
natural gas production firms in order to motivate the empirical analysis of firms’ abatement
costs. I begin by deriving firms’ first order conditions for leakage and abatement and proceed
to demonstrate how a relationship between price and abatement costs can be mapped to a
relationship between a potential emissions tax and abatement costs.
12Becauseoilandgasaresooftenco-located,petroleumandnaturalgasproductionfacilitiesarenotconcisely
differentiated in the datasets used in this paper.
13The EPA regulations came into effect in August 2016. Because this policy affects all production in the
UnitedStates,itsimpactshouldbepickedupbytimefixedeffects. However,resultsarerobusttoexcluding
2016. I control for Colorado regulations in the empirical analysis.
8
2.1 The Firm’s Problem
Consider the instantaneous profit function of a gas production firm’s operations within a
single basin:
π = P (Q −L )−C(Q ,L ) (1)
t t t t t t
P is the price of gas in period t, Q is the quantity of gas the firm extracts in t, L is the
t t t
quantity of gas it leaks, and C(·) is its total cost. Because the quantity of gas leaked depends
on the amount of gas flowing through the facility’s equipment, it is useful to decompose
leakage into the product of extraction and a leakage rate R :
t
π = P (1−R )Q −C(Q ,R ) (2)
t t t t t t
In this framework, the firm’s problem consists of choosing how much to extract alongside
choosing how careful to be to avoid leaks. This characterization makes it possible to separate
C(·) into costs of extraction that are unrelated to the facility’s leakage rate (i.e., costs of
obtaining leases, capital costs for equipment gas does not pass through) and costs that
determine the leakage rate (i.e., the additional up-front capital costs for equipment that
emits less, labor costs for leak detection and repair). If we assume leakage-related costs
are separable for each unit of extraction, we can re-write the firm’s profit function as the
following:
π = P (1−R )Q −C (Q )−Q c (R ) (3)
t t t t 1 t t 2 t
Here, C (·) is the total cost of extraction not related to the leakage rate and c (·) is
1 2
the per-unit cost of having leakage rate R . This decomposition allows costs not associated
t
with leakage to be nonlinear in production. For example, one might imagine that the cost
of acquiring new leases in a given basin increases as the firm increases production because
9
the total number of leases is finite. On the other hand, costs associated the leakage rate are
assumed to be the same regardless of the firm’s level of production. For example, paying a
worker to inspect one well site for leaks is assumed to cost the same amount whether the firm
operates 50 wells or 5,000 wells. However, c (·) is explicitly nonlinear in R —in particular,
2 t
it is decreasing and convex such that it approaches infinity as the leakage rate approaches
zero. This captures the intuition that, due to diminishing returns, bringing the leakage rate
down from 5 percent to 4.5 percent will be significantly cheaper than bringing it down from
1 percent to 0.5 percent.
The firm’s first-order condition for Q sets the marginal revenue generated by extracting
t
one unit of gas equal to the marginal cost of extracting it:
∂C (Q )
1 t
P (1−R ) = +c (R ) (4)
t t 2 t
∂Q
t
Note that the firm’s marginal revenue for one unit of extraction is lower than just the gas
price, as only the portion that is not leaked may be sold. In the firm’s first-order condition
for R , however, the firm’s marginal revenue of avoiding one unit of leakage is simply the gas
t
price, since the whole unit may be sold:
∂c (R )
2 t
P = − (5)
t
∂R
t
Equation (5) forms the basis for the empirics: When maximizing profits, the firm chooses
a leakage rate that sets the price equal to their marginal cost of leakage abatement.14,15
Intuitively, if one unit of gas can be sold for P , the firm will be willing to expend up to P
t t
to prevent it from being lost.
14Note that −∂c2(Rt) is actually positive because c is decreasing in R .
∂Rt 2 t
15The one-period framework presented here is useful for setting up a tractable empirical model, but it
oversimplifies some important temporal aspects of the firm’s true decision making process. In Appendix
A.1, I extend this framework to a dynamic model and discuss empirical applications that may become
possible with better emissions data in the future.
10