supply-side reforms to oil and gas production on federal landsnoncommercial-noderivatives 4.0...
TRANSCRIPT
Supply-Side Reforms to Oil and Gas Production on Federal LandsModeling the Implications for Climate Emissions, Revenues, and Production Shifts
Brian Prest
Working Paper 20-16 September 2020
Resources for the Future i
About the Author Brian Prest is a fellow at Resources for the Future specializing in climate change, electricity markets, and oil and gas economics. Prest uses economic theory and econometric models to improve energy and environmental policies by assessing their impacts on markets and pollution outcomes. His recent work includes evaluating the impacts of federal tax credits for coal use. He is also working to establish an empirical basis for determining discount rates used in the social cost of carbon. His past work includes econometric analysis of the US oil and gas industry, understanding the economic effects of rising temperatures, modeling the market dynamics of climate change policy under policy uncertainty, and assessing household responses to time-varying electricity pricing. His work has appeared in the Journal of the Association of Environmental and Resource Economists, Energy Economics, and The Energy Journal.
Prior to joining RFF, Prest earned his PhD at Duke University and previously worked in both the public and private sectors. At the Congressional Budget Office, he developed economic models of various energy sectors to analyze the effects of proposed legislation, including the 2009 Waxman-Markey cap-and-trade bill and related Clean Electricity Standards. At NERA Economic Consulting, he conducted electricity market modeling, project valuation, and discounted cash flow analysis of various infrastructure investments in the United States, Latin America, Europe, Africa, and Southeast Asia, with a focus on the power sector.
Insert title here on Master A ii
About RFFResources for the Future (RFF) is an independent, nonprofit research institution in Washington, DC. Its mission is to improve environmental, energy, and natural resource decisions through impartial economic research and policy engagement. RFF is committed to being the most widely trusted source of research insights and policy solutions leading to a healthy environment and a thriving economy.
The views expressed here are those of the individual authors and may differ from those of other RFF experts, its officers, or its directors.
Sharing Our WorkOur work is available for sharing and adaptation under an Attribution-NonCommercial-NoDerivatives 4.0 International (CC BY-NC-ND 4.0) license. You can copy and redistribute our material in any medium or format; you must give appropriate credit, provide a link to the license, and indicate if changes were made, and you may not apply additional restrictions. You may do so in any reasonable manner, but not in any way that suggests the licensor endorses you or your use. You may not use the material for commercial purposes. If you remix, transform, or build upon the material, you may not distribute the modified material. For more information, visit https://creativecommons.org/licenses/by-nc-nd/4.0/.
Supply-Side Reforms to Oil and Gas Production on
Federal Lands: Modeling the Implications for Climate
Emissions, Revenues, and Production Shifts
Brian C. Prest ∗
September 13, 2020
∗Resources for the Future, 1616 P St NW, Washington, DC 20036. [email protected]. I am grateful to
The Wilderness Society for financial support and Enverus for the data used in this study. I have no
relevant or material financial interests related to the research described in this paper.
Supply-Side Reforms to Oil and Gas Production on Federal
Lands: Modeling the Implications for Climate Emissions,
Revenues, and Production Shifts
Abstract
Over the last decade, 25 percent of US fossil fuel production came from lands
and waters directly managed by the federal government, and the resulting emis-
sions are equivalent to nearly a quarter of all US greenhouse gas (GHG) emissions.
Policy reforms targeting oil and gas production on federal lands have increasingly
attracted attention as an option to reduce emissions. Yet such policies are con-
troversial, in part due to concerns of “leakage,” in which reduced oil and gas pro-
duction on federal lands shifts to other producers. Accordingly, this paper models
the effects of three proposed policy reforms for federal oil and gas production:
raising royalty rates, carbon “adders” (fees) that internalize GHG externalities,
and a moratorium on new leasing. The model, which accounts for unprecedented
declines in oil prices associated with COVID-19, shows that raising royalty rates
has negligible effects on emissions but could raise an additional $1–$3 billion an-
nually. A moratorium reduces emissions from federal lands by an estimated 314
million metric tons of carbon dioxide equivalent (MMTCO2e) per year on aver-
age from 2020–2050 but also reduces royalty revenues by $5–$6 billion annually.
A carbon adder achieves two-thirds of the emissions reductions of a moratorium
(216 MMTCO2e annually from federal lands) and also raises $7 billion annually.
Although those emissions reductions are substantial, production shifts are also
large, implying smaller net emissions reductions of 85 to 147 MMTCO2e and 58
to 100 MMTCO2e annually for a moratorium and carbon adder, respectively. De-
spite sizable reductions, none of these policies would achieve the goal of net-zero
emissions from oil and gas on federal lands by 2040, as endorsed in the June 2020
report from the House Select Committee on the Climate Crisis. Achieving that
ambitious goal would therefore require modifying existing leases and/or additional
investments in carbon sequestration and renewable energy on federal lands.
Keywords: oil, gas, public lands, public finance, climate policy, emissions, leakage,
instrument choice
1 Introduction
Over the last decade, 25 percent of US fossil fuel production came from lands and waters
directly managed by the federal government, and the resulting emissions are equivalent
to nearly a quarter of all US greenhouse gas (GHG) emissions. Policy reforms target-
ing oil and gas production on federal lands have increasingly attracted attention as an
option to reduce emissions. However, disagreement remains, even among supporters
of climate policy, about the effectiveness of such policies. Critics argue that they are
undermined by emissions “leakage”—in which reduced fossil fuel production (and hence
emissions) in one region is offset by increased production and emissions in other regions.
Proponents argue that although leakage may reduce the efficacy of a policy, the net
effect is unlikely to be zero and that alternative, demand-side approaches can similarly
lead to leakage. Further, proponents argue, supply-side policies are simply more feasi-
ble to implement than demand-side policies for various political or institutional reasons
(Green and Denniss 2018). Indeed, federal coal and offshore oil and gas leasing was tem-
porarily suspended by the Obama administration, and a permanent suspension has been
endorsed by 2020 Democratic presidential candidate Joe Biden. Although such policies
are not first best from an economic perspective, economists are increasingly receptive
to incorporating political feasibility constraints into their assessments of second-best
policies (Goulder 2020).
Historically, the US government has leased federal lands1 to private firms that then
extract and sell federally owned resources. In exchange for the right to extract those
resources, firms pay the federal government royalties (a share of gross revenues, typically
12.5 to 18.75 percent), along with other payments, including bonus bids and rental pay-
ments. These revenues are shared between the states and the federal government. The
1Henceforth, I use the term “federal lands” as shorthand for both lands and waters where the mineralrights are owned by the federal government. This does not include Native American lands because therevenues from mineral extraction on those lands accrue to the relevant tribes and not to the federalgovernment.
1
federal share is used both as an unrestricted revenue stream and to fund land and water
conservation and water reclamation projects. Because the US federal government owns
large swaths of resource-rich land, fossil fuel production on federal land is a significant
contributor to greenhouse gas emissions. In particular, carbon emissions associated with
fossil fuels produced from federal lands represent 24 percent of US CO2 emissions (Mer-
rill et al. 2018), making it a large target for policymakers seeking to reduce emissions.
Further, the executive branch has broad authority under existing law to expand or re-
strict leasing for fossil fuel development on federal lands, without the need for legislative
action (Leshy 2019; Beaudreau, Schneider and Marnitz 2019).
Recent policy proposals that would reduce oil and gas development on federal lands
include increased royalty rates, carbon adders to internalize climate externalities, and a
leasing moratorium. Each policy was previously considered in context of the coal leasing
program during the Obama administration (CEA 2016; Gillingham et al. 2016; Krupnick
et al. 2016; Gillingham and Stock 2016). Ultimately, in 2016, the Obama administration
ordered a temporary moratorium on coal leasing while the program underwent a review.
The Trump administration revoked this moratorium and terminated the review.
In the years since, the US coal industry has been in decline, shifting the portfolio of
fossil fuel production on federal lands away from coal and toward oil and gas. According
to data from the Department of the Interior, although federal coal production has fallen
by nearly 30 percent from 2014 to 2019, federal oil production has actually risen by
about 40 percent.2 As a result, greenhouse gas emissions associated with oil and gas
produced on federal lands now exceed those associated with coal from those lands.3
This decline of coal and rise of oil and gas on federal lands has drawn attention to
reforming the federal oil and gas leasing program. For example, the majority staff of
the US House of Representatives’ Select Committee on the Climate Crisis (HSCCC)
2At the same time, federal natural gas production has declined by a relatively modest 10 percent.These data can be found at https://revenuedata.doi.gov/downloads/production-by-month/.
3This is based on the same data from the Department of the Interior, applying the emissions factorsdiscussed in section 2.2.6 for oil and gas and an emissions factor of 1.87 tCO2e per short ton for coal.
2
released a report (HSCCC 2020) that proposed a series of policies that aim to reach net-
zero greenhouse gas emissions on federal lands by 2040, including higher royalty rates and
a moratorium. Several pieces of legislation have been introduced that would implement
these proposals (H.R. 4364, S. 3330, H.R. 5186, S. 2906, and H.R. 5435). Many of
these changes could also be implemented through unilateral executive action by a future
administration. Indeed, every 2020 Democratic presidential candidate endorsed some
form of restrictions on federal oil and gas leasing, including a moratorium.
The HSCCC report also expresses more policy goals than simply reducing emissions.
On the one hand, the report frames these public lands policies as part of a broader climate
policy effort to reduce emissions, in this case by directly reducing oil and gas production.4
On the other hand, the report expresses a desire to raise additional royalty revenues for
the communities most affected by a reduction in fossil fuel extraction. That revenue
could also be used for other purposes, such as investing in research and development of
clean energy or reducing distortionary taxes. These dual goals of reducing emissions and
raising revenues create a tension that affects policy design. For example, a moratorium
may substantially reduce emissions, but it will also reduce revenues as production falls.
Although there is renewed interest in supply-side restrictions on federal oil and gas
production, there is a dearth of economic research that speaks to how effective these
policies would be. Gerarden, Reeder and Stock (2020) suggest that reducing federal coal
production by charging carbon adders—fees based on the marginal damages of carbon
emissions—could be effective at reducing emissions. But this result for the coal industry
4Although a key focus of these policies is greenhouse gas emissions, reducing oil and gas developmenton federal lands also has other important environmental and social benefits. This includes opening uppublic lands to alternative uses, such as conservation, preservation of biodiversity, renewable energydevelopment, and/or recreation. Although precluding oil and gas development also has economic costs,alternative land uses also yield economic benefits, such as for industries associated with recreation ortourism. For example, Walls, Lee and Ashenfarb (2020) find that designating public lands as nationalmonuments increased the growth of local business establishments. However, because reducing emissionsis the primary stated environmental goal for these policies, I focus on that as a measure of a policy’senvironmental effects. Because the emissions effects of these policies are strongly linked to the land useeffects, the size of the emissions impacts is also an approximate indicator for the size of these otherbenefits. However, I do not estimate the magnitude of these other benefits, nor do I estimate theeconomic costs of the proposed policies.
3
does not necessarily extend to oil and gas. The economics of oil and gas are simply very
different from that of coal. For one, oil and gas are less carbon intensive than coal. For
another, oil markets are much more globally linked than coal, largely due to relatively
low transportation costs. The market for US-produced gas is also increasingly global,
with the recent rapid construction of liquefied natural gas export facilities. Another
difference is that a much smaller share of US oil and gas production comes from federal
land (22 percent of oil and 12 percent of gas in 2019) compared to coal (about 40
percent). This distinction is only growing with the rise of oil and gas production from
shale formations, which are predominantly located on state and private land. Finally,
oil and gas production from shale is more price responsive than conventional production
(Newell, Prest and Vissing 2019; Newell and Prest 2019). All of these factors suggest
strong potential for leakage of production from federal lands to state, private, and tribal
lands, in addition to foreign producers.
One recent study (Erickson and Lazarus 2018) estimated the impacts of ending new
federal leasing of oil and coal (but not gas) in a static constant-elasticity model drawing
on supply elasticity estimates from the gray literature. That study estimated that a
moratorium on all new federal fossil fuel leasing could reduce global CO2 emissions by
280 million tons per year by 2030. However, most of this reduction was estimated to come
from reduced coal consumption, with only about 14 percent (39 million tons) estimated
to come from oil. This estimate may understate long-run effects however because it is
based on a static model for the year 2030. But the effects of changing federal leasing
policies generally occur more than a decade into the future, suggesting larger effects
beyond 2030. Federal oil and gas leases typically have a duration of 10 years, and oil
and gas firms typically do not develop these leases until the eighth, ninth, or tenth year,
that is, just prior to expiration (CBO 2016). Further, once a well is drilled, standard
leasing provisions extend the duration indefinitely so long as the well is producing oil
or gas. This means that wells drilled on federal leases continue to produce for decades
4
after the initial 10 year term. As a result, changes in federal leasing policy today (say
in 2020) primarily affect production more than a decade into the future (after 2030),
meaning their impacts are likely to be much larger beyond a 10 year period. For the
same reason, CBO estimated minor revenue effects from leasing reform but emphasized
that their small estimates primarily reflected their use of CBO’s standard 10 year budget
window and noted that the effects could be substantially larger after that point (CBO
2016).
These aforementioned two studies represent the two main efforts in the literature
to estimate the effects of reforming federal oil and gas leasing policies on greenhouse
gas emissions (Erickson and Lazarus 2018) or revenues (CBO 2016), highlighting the
extremely limited literature on the topic.5
This paper fills that gap in the literature by building on and extending the econo-
metric oil and gas supply methods developed in Newell, Prest and Vissing (2019) and
Newell and Prest (2019) to model the effects of several proposals that would reform leas-
ing policy regarding oil and gas production from US federal lands. I consider the three
key policy approaches that have recently attracted attention: (1) raising federal royalty
rates by 6.25 to 12.5 percentage points (from their current levels of 12.5 percent onshore
and typically 18.75 percent offshore6), (2) charging carbon adders equal to the social
cost of carbon of about $50 per ton of CO2 to internalize the externalities of greenhouse
gas emissions, or (3) establishing a complete moratorium on all new oil and gas leasing.
I focus on these three policies because they have each attracted attention for potential
reform. The Department of the Interior already charges royalty rates and has clear au-
thority to change them. Further, proponents of this approach argue that federal onshore
5U.S. Government Accountability Office (2017) also cites a draft paper (Enegis, LLC 2011) studyingthe effects of raising royalty rates on revenues, but it does not appear to be publicly available anywhere.
6Royalty rates on offshore wells depend on water depth. In recent years, offshore oil and gas de-velopment has increasingly focused on deepwater reservoirs, where the royalty rate is 18.75 percent.Although the statutory rates are typically 12.5 and 18.75 percent, these rates are often subject to al-lowances, deductions, and waivers that, under the current system, can reduce effective rates below thestatutory ones. See U.S. Government Accountability Office (2017).
5
royalty rates are very low (12.5 percent) relative to market rates typically charged on
state and private lands (often 18.75 to 25 percent). Raising royalty rates would ensure
that taxpayers receive returns on public resources commensurate with market rates. I
model a carbon adder for two reasons: first, it is an economically appealing approach
that approximates Pigouvian-style taxation for covered producers, and second, it dif-
ferentially disincentivizes oil versus gas production (in accordance with their different
carbon intensities), in contrast to royalty rates, which disincentivize oil and gas equally.
The carbon adder is set based on the social cost of carbon (SCC) as estimated by the
Interagency Working Group in 2016, which equals approximately $50 per ton in 2020
and rises at 2 percent annually in real terms.7 Finally, I model a moratorium because
this approach has been used repeatedly in recent decades on a temporary basis and
because policymakers are now considering a permanent one. For example, Joe Biden’s
2020 presidential campaign proposed some form of each of these policies, including “ban-
ning new oil and gas permitting on public lands and waters [and] modifying royalties
to account for climate costs.”8 I also model different variants and combinations of these
policies (e.g., increasing the royalty rate and charging carbon adders) to illustrate their
potential interactions.
First, I econometrically estimate how US drilling activity responds to oil and gas
prices, allowing for heterogeneous responses by type of well (e.g., oil-directed versus gas-
directed drilling, wells on federal versus nonfederal land). Then, I use these estimates to
simulate how drilling activity translates into oil and gas production over time, based on
the path of oil and gas prices (net of royalties and carbon charges). The model accounts
for key structural features of oil and gas markets, including both own-price and cross-
price responses (e.g., natural gas production depends on both gas prices and oil prices),
complementarities in production (such as so-called “associated gas” that is produced
7This reflects the estimate from the Interagency Working Group on the Social Cost of Carbon (IWG2016), after adjusting for inflation to 2020 dollars. The SCC values estimated by the IWG rise atapproximately 2 percent per year.
8See https://joebiden.com/climate-plan/, last accessed September 8, 2020.
6
alongside oil in oil-directed wells; see, e.g., Gilbert and Roberts 2020), and leakage (e.g.,
substitution from federal to nonfederal production).
Unlike work from the literature on the topic, the model extends beyond 2030. This
captures the long time lags between federal policy changes and realized production im-
pacts. This lag occurs because oil and gas wells drilled prior to a change in leasing policy
are unaffected by the policy but nonetheless may produce for many decades to come.
The model accurately reflects that, once drilled, a well may produce indefinitely so long
as it is capable of yielding oil or gas. The model also reflects that even after a change
in policy regarding new leases (including a moratorium), some wells may continue to be
drilled on leases that were issued before the policy change but had not yet been drilled.
After the primary terms of that stock of existing leases expire (say, after 10 years, as-
suming no extensions), policy changes to newly issued leases affect all new wells. Before
that point, only a fraction of newly drilled federal wells are covered by changes in lease
terms.9
Finally, the model is calibrated using data that include the recent shale boom that
has increased the price responsiveness of oil and gas supply. I estimate key model
parameters using a large well-level dataset on more than one million individual oil and
gas wells in the United States, representing nearly all operating wells in the country. The
estimated model is then used to simulate the effects of different sets of federal leasing
policies on oil and gas prices, production, emissions, and revenues from royalties and
carbon adders. The model accounts for endogenous production responses from non-US
foreign suppliers through a reduced form relationship based on modeling results from the
International Energy Agency. Importantly, and unlike other studies, the model accounts
9As described in more detail in section 2.2.6, I assume that the fraction of newly drilled federalwells that is covered by each policy phases in linearly over a 10-year period, where 10 years is thetypical statutory length of federal leases. This simplifying assumption is analogous to the assumptionin Gerarden, Reeder and Stock (2020). This implicitly assumes that no lease extensions are grantedand that existing lease sales are scheduled to expire in a uniform fashion over the next 10 years. Theresults are not very sensitive to this assumption, as discussed in section 2.2.6, affecting cumulativefederal emissions by about ±8 percent at the most.
7
for the dramatic decline in oil prices in 2020 associated with the COVID-19 pandemic
and the Russia-Saudi price war.
Figure 1: Policy Impacts on Emissions and Revenues, Annual Average 2020–2050, byDemand Elasticity Assumption
Note: “Nonfederal” includes both domestic and foreign producers.
The results demonstrate stark differences in the effects of the three policies on the
key outcomes of emissions and revenues. Figure 1 summarizes the results for three key
policies: a 25 percent royalty rate, a $50 carbon adder, and a moratorium. Raising
royalty rates at the levels commonly proposed would have little effect on federal oil
and gas production and hence relatively small effects on emissions associated with that
production, around 37 million tons of CO2e (MMTCO2e) annually on average from
federal lands,10 but could raise as much as $3 billion in additional royalty revenues per
10The emissions “from federal lands” refers to the CO2e emissions “embodied” in the oil and gasproduced—that is, these ultimately result from the combustion of the oil and gas produced on federallands. This is consistent with the terminology used in Merrill et al. (2018). These emissions technically
8
year.11 At the other extreme, a moratorium would lead to substantial reductions in
emissions from federal oil and gas production (an estimated 314 MMTCO2e annually on
average from federal lands) but at the loss of $5–$6 billion of royalty revenues per year.
A “middle ground” policy of carbon adders (which would internalize the externalities
of greenhouse gas emissions for federal production) would achieve about two-thirds of
the emissions reductions of a moratorium (216 versus 314 MMTCO2e annually) but also
raise, rather than lose, about $7 billion in additional royalty and carbon revenues per
year on average. Adding a royalty rate increase on top of a carbon adder (not shown
in Figure 1) is estimated to produce only slightly more emissions reductions (about 10
percent more) but also actually raises less revenue (about 10 percent less) than a carbon
adder alone because layering on this charge further reduces federal production.
Although those estimated emissions reductions from production on federal lands can
be large, leakage rates are also substantial, ranging between 53 and 74 percent, depending
on oil and gas demand elasticity assumptions, as indicated in Figure 1. This means that,
for example, the federal reductions of 314 MMTCO2e from a moratorium translate into
only 85 to 147 MMTCO2e worth of reductions in net global emissions. This leakage is in
part due to offsetting supply responses from production on state and private lands not
subject to federal restrictions and in part due to leakage to foreign producers. Leakage
to US producers on state and private lands constitutes about one-third of the total
leakage effect, despite the fact that those sources historically represented less than 15
percent of global oil and gas supply (in barrels of oil equivalent). This disproportionate
contribution to leakage is due to the projected rise in the nonfederal US share of global
occur at the point of combustion and not at the production site on the federal lands themselves.Nonetheless, for brevity, throughout this paper, the reported emissions reductions by supply source(such as “emissions reductions from federal lands”) correspond to this “embodied” CO2e measure.
11All revenue estimates represent the effects on gross royalty and carbon adder revenues collectedby the federal government (excluding revenues from tribal lands). Historically, federal royalty revenueshave been split approximately equally between the states and the federal government. I do not attemptto calculate the federal versus state shares of these incremental revenues raised under higher ratesbecause this would be a policy choice, but a 50/50 split would be a reasonable estimate.
9
oil and gas supply and the larger supply elasticities of nonfederal US supply (relative to
foreign producers), both of which can be attributed to the shale boom.
Despite the significant potential for emissions leakage, the results suggest that federal
oil and gas leasing policies can have larger effects on global emissions than previous
estimates indicate. However, even the most aggressive policy considered—a moratorium
on all new federal oil and gas leasing—would not drive oil and gas emissions from federal
lands to zero because production from wells on existing leases would remain unrestricted
(see Figure 2). Achieving the HSCCC report’s target of net-zero emissions on federal
lands by 2040 therefore would require modifying existing federal leases and/or a larger
role for carbon sequestration and renewable energy development on federal lands.
−10
0−
80−
60−
40−
200
Year
Per
cent
Red
uctio
n in
Ann
ual F
eder
al O
il an
d G
as E
mis
sion
s
2020 2025 2030 2035 2040 2045 2050
18.75% Onshore RR25% RR, Onshore Only25% RR, Onshore &Offshore
$50 Carbon Adder (2%)
$50 Carbon Adder (2%)& 25% RR
Moratorium
Figure 2: Federal Emissions Reductions by Policy and Year, as a Percent of Baseline
Notes: RR = royalty rate. Figure only shows emissions reductions from oil and gas produced onfederal lands. Values are presented as a percent of oil and gas emissions from federal lands in eachyear, not including emissions from other sources, such as coal.
10
2 Model and Results
The approach in this paper builds upon and extends the methods developed in Newell,
Prest and Vissing (2019) and Newell and Prest (2019). Each of those papers separately
models the three key stages of the oil and gas production process: (1) drilling wells,
(2) completing them (which may include hydraulically fracturing them) to bring them
online for production, and (3) production over time once the wells are online. Those
papers then combine the models of each of the three stages to simulate the change in
oil or gas production resulting from an exogenous change in prices. Those simulations
were somewhat stylized steady-state models designed to demonstrate estimated price
responses simply, whereas this paper extends them to incorporate key features relevant
to changing federal oil and gas leasing policies. This includes the potential for supply
substitutions across well types (including federal versus nonfederal wells) and a more
nuanced treatment of well-level production declines over time (which are important
to modeling the effects of a moratorium on new federal drilling and understanding the
feasibility of achieving the goal of net-zero emissions). Another extension is to endogenize
the price to policy changes, which is key to understanding the potential for policy leakage
and hence net emissions impacts.
2.1 Simulation Overview
The simulation model is depicted in a flowchart in Figure 3. I start with a given path
of projected oil and gas prices and assumptions about policies (royalty rates, carbon
adders, or a moratorium) over time (box 1). For example, oil and gas price paths in the
baseline scenario are based on observed futures prices, and royalty rates are assumed to
be unchanged from current levels (12.5 percent onshore, 18.75 percent offshore). These
price and policy paths are fed into the drilling module (box 2), which predicts future
drilling activity for each month into the future based on these price paths (adjusted for
11
royalty rates or carbon adders), separately for each of the eight well types. The drilling
module is based on the econometric model discussed in detail in the next section.
Figure 3: Simulation Model Overview
The resulting trajectory of newly drilled wells gradually translates into new wells
coming online for production (box 3) based on the empirical distribution of time from
the initiation of drilling to first production (again, separately by well type). Then, newly
operating wells produce oil and gas (box 4a) based on empirically estimated production
profiles over time (also known as “type curves”). Finally, existing wells that have already
been drilled will also continue to produce oil and gas for many years to come. These
production levels are estimated using the standard “Arps curve” approach (box 4b).
Production from new and existing wells is added together to arrive at total US oil and
gas production. Finally, the US results are combined with a rest-of-world (ROW) module
(box 5) to account for ROW supply responses to changes in US production, capturing
potential leakage effects. The methodology underlying each box in Figure 3 is explained
in detail in the next section.
12
The model begins in equilibrium under current projections of oil and gas supply,
demand, and prices.12 Then, to model the impacts of policy changes, I change the
relevant policy assumptions (federal royalty rates, carbon adders, or a moratorium) in
box 1 and simulate the remaining components of the model (boxes 2–5) under this
new policy assumption. Any of the three policy levers reduces total quantity supplied,
pushing the market out of equilibrium (quantity supplied less than quantity demanded)
under the baseline price paths for oil and gas. I then numerically solve for the rise
in oil and gas prices necessary to return the market to equilibrium, which yields the
equilibrium outcomes under the new policy scenario. The effects of the policy on various
outcomes, such as prices or emissions, are then calculated as the differences between the
two scenarios (baseline versus policy case).
2.2 Model Estimation and Calibration
In this section, I explain the estimation and calibration of the key components of the
simulation—that is, the estimation of the models in boxes 2, 3, 4a, and 4b in Figure
3. Research has demonstrated (Anderson, Kellogg and Salant 2018; Newell, Prest and
Vissing 2019; Newell and Prest 2019) that drilling is the key driver of long-run supply
responses to oil and gas prices, so this stage merits the most attention. The other
stages (time to production and production from existing wells) tend not to be very
responsive to price.13 Most drilling costs are up-front and fixed, whereas the marginal
cost of producing from an existing well is very low. This implies that it is almost always
rational for a firm to produce a well at its maximum flow capacity, suggesting little
adjustment of production from existing wells in response to price changes (Anderson,
Kellogg and Salant 2018). Rather, oil and gas producers respond to price increases and
12Oil and gas prices were based on futures prices as of June 25, 2020.13Further, even if these two stages were price responsive, this would primarily affect the timing of
production rather than the total amount of production realized in response to a price shock.
13
decreases by drilling more or fewer wells, respectively. For these reasons and more,14 I
focus on modeling the first stage as a function of oil and gas prices, while treating the
remaining stages as exogenous to prices.15
In the coming sections, I explain the estimation of the price response of drilling
activity, or “drilling elasticities” for short. Then I explain the estimations of the amount
of time it takes for a drilled well to begin production (the second stage) and how much
oil and gas each well produces in each month of its life (the third stage). Finally, I
explain how these three stages are combined to simulate the effects of changes in US
federal leasing policies on federal and nonfederal oil and gas production.
Because the estimation relies on several key features of the data, I provide an overview
of the data used in this study before moving on to discuss the estimation and simulation.
2.2.1 Data
The key data source is a well-level dataset from Enverus (formerly Drillinginfo) on more
than one million oil and gas wells in the United States. This data source has been widely
used in the economics literature (e.g., see Allcott and Keniston 2017; Feyrer, Mansur and
Sacerdote 2017; Bartik et al. 2019) because it is both highly detailed (e.g., well-level pro-
duction time series) and nationally comprehensive. The dataset I use includes all wells
in Enverus’s data that began production between January 1990 and February 2019.16
The dataset includes rich information on each well, including its latitude and longitude,
when it was drilled, completed, and began production, whether it is oil directed or gas
directed, and a monthly time series of its oil and gas production over time.
14In addition, most proposed changes to federal oil and gas leasing policies would only apply to newlydrilled wells (and only ones on new leases at that) and not to any existing wells, again suggestingfocusing on the development of newly drilled wells.
15Regression analyses of the second and third stages nonetheless confirm the findings of the previousliterature that these stages are not very price responsive.
16The Enverus data were downloaded in April 2020, but due to reporting lags, they are only generallycomplete with a one-year lag. Therefore, I end the sample period on February 2019, as the wells in thedata after this date likely represent a biased and incomplete sample.
14
This well-month panel dataset includes 121 million monthly observations on 1,044,817
wells, accounting for nearly all oil and gas production in the United States. For 2018,
the last full year of data, the total oil produced by the wells in the sample accounted for
93 and 97 percent of US oil and gas production, respectively, according to data reported
by the Energy Information Administration (EIA).17
In all econometric analyses and simulations, I calculate values separately by well
type. Specifically, I distinguish between wells along the following three dimensions:
1. Federal versus nonfederal
2. Oil-directed versus gas-directed
3. Onshore versus offshore
These three binary dimensions lead to 23 = 8 well types. The first dimension is
natural because the focus of this study is a set of policies that would directly affect
wells on federal land (and lead to indirect effects on nonfederal wells).18 I further dis-
tinguish between oil-directed and gas-directed wells because past literature has shown
that each type of well responds differently to oil versus gas prices. As would be ex-
pected, oil-directed drilling responds more to oil prices, whereas gas-directed drilling
responds more to gas prices (see Newell, Prest and Vissing 2019; Newell and Prest 2019;
Gilbert and Roberts 2020). Pooling well types in an econometric analysis would ignore
this heterogeneity and also reduce econometric precision. Finally, I also distinguish be-
tween onshore and offshore wells because the economics, engineering, and geology of
onshore and offshore drilling are quite distinct. This is also important because one of
the proposed policy changes would only affect the treatment of onshore federal wells.19
17These discrepancies owe to wells not in the sample, as Enverus is not a perfect census of every well.In my simulations, I account for these differences by scaling up oil and gas production projections byfactors of 1
0.93 and 10.97 respectively.
18I treat wells on Native American lands as nonfederal because the proposed policy changes wouldnot affect leases on those lands.
19This policy would raise the onshore federal royalty rate from 12.5 to 18.75 percent, which is therate already typically charged for offshore wells.
15
The Enverus data do not always indicate whether a well is on federal land or when
it is offshore, so I overlay GIS shapefiles representing federal land20 and the ocean21 to
geotag wells as federal versus nonfederal and onshore versus offshore. The identification
of a well as oil directed or gas directed is primarily based on the production type variable
in the Enverus data.22
Figure 4 shows the location of the wells in the data by type. Onshore federal wells
tend to be concentrated in the mountain west, where federal wells are predominantly
gas directed (in dark blue). However, the map illustrates the signs of the recent rise
in onshore oil drilling (in red) on federal lands in pockets of southeastern New Mexico
(overlaying the Permian basin), eastern Colorado and Wyoming (overlaying the Nio-
brara formation), and western North Dakota (overlaying the Bakken). Nevertheless,
most of the shale boom has occurred on private lands, which is particularly evident for
oil-directed drilling (in yellow) in west Texas and gas-directed drilling (light blue) in
Pennsylvania, Ohio, and West Virginia. Although federal onshore oil production is on
20I use a GIS shapefile representing federal surface ownership (available at https://www.arcgis.
com/home/item.html?id=26c2a38f94c54ad880ff877f884ff931), but this is not necessarily the sameas the owner of the subsurface mineral rights. Unfortunately, no comprehensive nationwide GIS shapefileexists on federal mineral rights ownership, so this is an approximation. However, I checked the accuracyof the geotagging approach by comparing the total production from wells geotagged as “federal” toofficial production statistics reported by the Department of the Interior’s Office of Natural ResourcesRevenue (ONRR). The aggregated production based on the geotagged Enverus data very closely matchesONRR data. For example, in 2018, based on the geotagged aggregation, oil production in 2018 (thelast full year of data) averaged 2.39 million barrels per day (mb/d), compared to 2.41 mb/d reportedby ONRR, a difference of less than one percent. The difference is larger for natural gas production,where the geotagged aggregation is about eight percent smaller than ONRR’s official statistics. Aninvestigation into this difference suggests it is likely due to some very old federal onshore gas wellsthat were drilled before 1990 that do not appear in the dataset. Such wells would not have a materialeffect on the paper’s results. First, they would not be useful in informing the identification of drillingresponses to prices during the sample period. Second, they would not affect the simulated effects ofleasing policy changes, which would apply to newly drilled wells.
21The ocean shapefile is available at https://www.naturalearthdata.com/downloads/
10m-physical-vectors/10m-ocean/. The Enverus data indicate whether wells are on federalwaters, meaning the ocean shapefile is only necessary to identify offshore nonfederal wells.
22Nearly all (92 percent) of wells are indicated as oil or gas wells in the raw data. The remaining wellshad more ambiguous values for production type, primarily labelled as “O&G.” These wells were labelledas gas directed if their gas-to-oil ratio is higher than the 90th percentile of the observed gas-to-oil ratioof oil wells; all other wells with ambiguous type were labelled as oil directed.
16
Figure 4: Location of Wells in Data by Well Type and Federal Lands
Sources: Well locations are from Enverus. The federal lands locations are based on the ArcGIS federallands shapefile available athttps://www.arcgis.com/home/item.html?id=26c2a38f94c54ad880ff877f884ff931. Datasetincludes Alaska, which is omitted from the map for space.
the rise, it remains small relative to oil production on state and private lands, and most
federal oil production still comes from offshore wells, primarily the Gulf of Mexico.
Not all well types are equally important in driving total US oil and gas production. To
illustrate the relative importance of each well type, Figure 5 shows historical production
of oil (top panel) and gas (bottom) by well type. Beginning around 2009, the sharp rise in
oil production from the shale boom is evident in the graph (see yellow line representing
nonfederal oil drilling, as the shale boom primarily took place on nonfederal lands).
The stall in production following the temporary crash in oil prices in 2014–2015 is also
17
evident. Onshore federal oil-directed drilling has also risen (solid red line in top panel),
which includes the New Mexico side of the Permian basin.
Nonetheless, offshore oil-directed production has historically been dominant on fed-
eral lands (dashed red line, top panel). The other categories of offshore wells have
contributed relatively little to US oil and gas production in recent years. Federal gas
production is dominated by onshore gas wells (bottom panel, solid dark blue line). In
general, US gas production is dominated by onshore nonfederal wells (light blue line)
and associated gas production from onshore nonfederal oil wells (yellow line in bottom
panel).
Although I model all eight well types to comprehensively account for all production
sources, Figure 5 illustrates that the key types driving long-run production are onshore
nonfederal oil wells (yellow solid line), onshore nonfederal gas wells (light blue solid),
offshore federal oil wells (red dashed), and onshore federal gas wells (dark blue solid).
Although historically relatively small, production from onshore federal oil wells (e.g.,
in the New Mexico portion of the Permian basin) is also expected to be important in
the future due to its recent rapid growth. Hence, in the subsequent analysis, the key
estimates meriting attention are those for these well types.
18
02
46
8
Oil
Pro
duct
ion
(mb/
d)
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
OnshoreOffshore
010
2030
4050
6070
Gas
Pro
duct
ion
(bcf
/d)
2000 2002 2004 2006 2008 2010 2012 2014 2016 2018
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
OnshoreOffshore
Figure 5: Historical Production of Oil (top) and Gas (bottom), by Well Type, 2000–2019
Notes: The sample includes wells drilled in 1990 or later.
In addition to the Enverus data, I also use data from the Federal Reserve Economic
Database (FRED) for historical oil (WTI) and gas (Henry Hub) prices and inflation
indexes.23 For the simulation, I also use futures prices for West Texas Intermediate
(WTI), Brent, and Henry Hub from CME Group as the market’s best-guess forecasts
23I also use copper prices as an instrument in the econometric analysis. The specific series used areDCOILWTICO, PNGASUSUSDM, PCOPPUSDM, and CPIAUCSL.
19
of future commodity prices24 and projections from the International Energy Agency’s
2019 World Energy Outlook (IEA 2019) for global oil and gas demand, ROW supply,
and international gas price spreads.
2.2.2 Econometric Estimation of Drilling Response (Box 2)
The drilling estimation uses standard time series methods as in Newell, Prest and Vissing
(2019) and Newell and Prest (2019). Namely, for each well type j, I estimate how the
number of wells drilled in month t responds to variation in contemporaneous and lagged
oil and gas prices. The estimating equation for each well type j is
∆ log(Wells Drilledj,t) =12∑`=0
ηoilj,`∆ log(WTIt−`) + ηgasj,` ∆ log(Henry Hubt−`) + λmoy + εj,t,
(1)
where WTI is the West Texas Intermediate crude oil price, Henry Hub is the natural gas
price, and λmoy represents month of year fixed effects to capture seasonality in drilling
activity. The time series of the variables in equation (1) are shown in Figure 6. The
graph suggests a slightly lagged response of drilling activity to prices. This is most likely
due to lag times, due to planning and logistics, between when drilling decisions are made
by firms and when drilling rigs are brought to the well site. Twelve months of lagged
prices are included, as in Newell, Prest and Vissing (2019) and Newell and Prest (2019),
but the results are robust to including more lags.
24These futures are available here: WTI, Brent, and Henry Hub. The prices used reflect the closingprice on June 25th, 2020.
20
1990 1995 2000 2005 2010 2015 2020
050
010
0015
0020
0025
0030
00
Month
Num
ber
of W
ells
Dril
led
050
100
150
Oil and G
as Price
(2020$ per barrel of oil equivalent)
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
WTI Oil PriceHenry Hub Gas Price
1990 1995 2000 2005 2010 2015 2020
020
4060
8010
012
0
Month
Num
ber
of W
ells
Dril
led
050
100
150
Oil and G
as Price
(2020$ per barrel of oil equivalent)
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
WTI Oil PriceHenry Hub Gas Price
Figure 6: Wells Drilled per Month, by Well Type, Onshore (top panel) and Offshore(bottom panel)
21
The ηoilj,` and ηoilj,` parameters in equation (1) are short-run drilling elasticities for a
well of type j (i.e., the effect of a change in log prices on log drilling activity with a
lag of ` months). The long-run, or cumulative, drilling elasticity with respect to the oil
price is simply the sum of the contemporaneous and lagged coefficients ηoilj =∑12
`=0 ηoilj,`.
The long-run drilling elasticity with respect to gas prices is defined analogously.
Endogeneity is not typically a major problem when estimating US drilling elasticities
(see e.g., Prest 2018) because the country has historically been considered a relatively
small producer, particularly for the oil market, and because surprise shocks to drilling
activity are only weakly related to contemporaneous changes in production, which arise
with a significant lag. As a result, drilling activity shocks tend to have little contempo-
raneous effect on oil and gas prices. This argument may be weaker in recent years with
the shale boom, which has arguably allowed US supply to have a larger influence on oil
prices in particular. Therefore, I instrument for the potential endogeneity of oil and gas
prices to drilling activity using approaches from the literature.25
Drilling Estimation Results
Each regression specification in equation (1) has 26 elasticity estimates (13 for oil and
13 for gas prices).26 With a separate regression for each of the eight types of wells, this
25The instrument for the oil price is the market price of copper, which acts as a proxy for globalcommodity demand, as used in Hamilton (2014); Prest (2018) and Newell, Prest and Vissing (2019).The instrument for natural gas prices is the twice-lagged level of log Henry Hub prices, based on one ofthe strongest instruments considered in Hausman and Kellogg (2015). Unfortunately, in this context,most standard instruments for natural gas prices, including those used in the literature, turn out tobe weak (e.g., see Hausman and Kellogg 2015, which faced a similar difficulty). After considering alengthy list of alternative natural gas instruments (average temperature, heating degree days, coolingdegree days, natural gas inventories, and their lags), the twice-lagged Henry Hub price turned out tobe the strongest and most conceptually defensible instrument. I use this price because it would beinappropriate to use the first lag, because the potentially endogenous variable (the first difference of logHenry Hub prices) is mechanically a function of the first lag.
26Each regression uses 325 monthly observations: Feb. 1992 to Feb. 2019. The Henry Hub priceseries begins in Jan. 1991, so with 12 lagged first differences in prices, the complete time series beginsin Feb. 1992. Each model is estimated by two-stage least squares (2SLS). The first stage is the same foreach regression; the first-stage F-test for the copper price instrument is 14.8, which is strong, whereasthe F-test for the lagged Henry Hub instrument is 5.1. The latter F-test for natural gas prices is similarin magnitude to the results from Hausman and Kellogg (2015). When using ordinary least squares(OLS) instead of 2SLS, the long-run elasticities are generally somewhat smaller in magnitude for allkey well types, but the differences in magnitudes are not large. The most important effect of the IVapproach is for the oil price elasticity for offshore oil wells, where the IV elasticity is positive (0.48)
22
leads to 26×8 = 208 coefficient estimates representing the time profile of drilling elastic-
ities. Because the long-run drilling response depends only on the cumulative elasticities
(i.e., the sum of the contemporaneous and lagged price coefficients), for the purposes of
exposition, I present the results more concisely by showing only these cumulative esti-
mates.27 This nonetheless leads to 16 cumulative elasticity estimates—one for oil prices
and one for gas prices, for each of the eight well types. These are shown in Figure 7 for
onshore wells and Figure 8 for offshore wells.
Before discussing the results, it is worth noting the expected signs of the elasticities.
Although “own-price” elasticities (e.g., oil prices on oil-directed drilling) should be pos-
itive, “cross-price” elasticities (e.g., gas prices on oil-directed drilling) could be positive,
negative, or even zero. For example, to the extent that associated gas coproduced by
oil-directed wells is valuable, higher gas prices could support oil-directed drilling, imply-
ing a positive elasticity. On the other hand, if rising gas prices lead to competition for
drilling rigs that increase (unobserved) costs for oil-directed drilling, rising gas prices
could lead to reduced oil drilling, implying a negative “cross-price” elasticity.
Within each figure, the own-price elasticities are found in the top left (oil prices on
oil wells) and bottom right (gas prices on gas wells) panels. The cross-price elasticities
are in the top right (gas prices on oil wells) and bottom left (oil prices on gas wells)
panels. The largest source of US oil production is onshore nonfederal oil wells, which are
estimated to have a long-run drilling elasticity of 1.04 with respect to oil prices (yellow
bar in the top left panel of Figure 7), which is also the most precisely estimated elasticity,
with a standard error of 0.30. The corresponding elasticity for federal onshore oil wells
is 0.93, slightly smaller but not statistically different from their nonfederal counterparts.
These own-price elasticity estimates for onshore gas wells (bottom right panel) are 0.7
for nonfederal and 1.2 for federal wells, although they are less precisely estimated. The
but not significant (standard error of 0.61) but the OLS elasticity is close to zero (-0.08, again notsignificant). All statistical inference uses Newey West covariance matrices.
27The full regression results are presented in appendix section A.
23
cross-price elasticities are all positive but small and statistically insignificant. Although
the literature has not typically estimated federal versus nonfederal drilling elasticities
specifically, these estimates are comparable in magnitude to the most appropriate ana-
logues in the literature (Hausman and Kellogg 2015; Anderson, Kellogg and Salant 2018;
Newell, Prest and Vissing 2019; Newell and Prest 2019; Gilbert and Roberts 2020).
Despite the substantial literature on onshore drilling elasticities, I am aware of no
recent literature estimating offshore drilling elasticities. This is perhaps due to the
small number of offshore wells drilled each year, leading to small sample sizes and hence
noisy estimates. Indeed, the standard errors on the offshore drilling elasticities are
wide. Although four offshore well types (federal versus nonfederal and oil versus gas)
are presented in Figure 8, the vast majority of offshore wells are federal. The own-price
elasticity estimate for this group is 0.5, with a standard error of 0.6, and the cross-price
elasticity is 0.2, also with a standard error of 0.6. Because the other types of offshore
wells are so few in number, their estimated elasticities matter very little to the simulation
results.
These large standard errors imply considerable uncertainty in the offshore drilling
elasticity; indeed, I cannot reject that it is zero. However, the implications for the
simulation modeling of leasing policies are smaller than may otherwise appear, because
simulating most policies of interest does not depend strongly on the offshore oil price
elasticity. The proposed increase in royalty rates to 18.75 percent would have no appre-
ciable effect for offshore wells (for which the rate is typically already 18.75 percent), and
the proposed moratorium is not a price-based instrument and therefore is not mediated
by an elasticity estimate.
24
Nonfederal Federal
Oil Wells
Cum
ulat
ive
Oil
Pric
e E
last
icity
−1
01
23
−1
01
23
Nonfederal Federal
Oil Wells
Cum
ulat
ive
Gas
Pric
e E
last
icity
−1
01
23
−1
01
23
Nonfederal Federal
Gas Wells
Cum
ulat
ive
Oil
Pric
e E
last
icity
−1
01
23
−1
01
23
Nonfederal Federal
Gas Wells
Cum
ulat
ive
Gas
Pric
e E
last
icity
−1
01
23
−1
01
23
Figure 7: Onshore Long-Run Drilling Elasticities with Respect to Oil Prices (left column)and Gas Prices (right)
Notes: Error bars represent 90 percent confidence intervals
25
Nonfederal Federal
Oil Wells
Cum
ulat
ive
Oil
Pric
e E
last
icity
−1
01
23
−1
01
23
Nonfederal Federal
Oil Wells
Cum
ulat
ive
Gas
Pric
e E
last
icity
−1
01
23
−1
01
23
Nonfederal Federal
Gas Wells
Cum
ulat
ive
Oil
Pric
e E
last
icity
−2
−1
01
23
−2
−1
01
23
Nonfederal Federal
Gas Wells
Cum
ulat
ive
Gas
Pric
e E
last
icity
−1
01
23
−1
01
23
Figure 8: Offshore Long-Run Drilling Elasticities with Respect to Oil Prices (left column)and Gas Prices (right)
Notes: Error bars represent 90 percent confidence intervals. Bottom left panel has a different scale.
26
2.2.3 Time from Drilling to First Production (Box 3)
Once a well is drilled, it must be completed (and potentially hydraulically fractured)
before it is ready to produce. Newell, Prest and Vissing (2019) and Newell and Prest
(2019) found that the completion time (or more specifically, the time in months between
the “spud date”—the date drilling began—to the first production date) did not strongly
depend on prices. This time lag does affect the timing of production responses, however
because it creates a lag between a rise in drilling activity and the realization of incre-
mental oil and gas production. The simulation accounts for this time lag by converting
changes in drilling activity to new wells coming online (box 3 in Figure 3) according
to the empirically estimated distribution of completion time. These distributions are
shown in Figure 9.28 Both onshore and offshore, the distributions of spud to first pro-
duction time is similar across oil, gas, federal, and nonfederal wells. Offshore wells tend
to take longer to come online (nearly two years, compared to four months on average
for onshore, although the offshore average is in part driven by the skewed distribution
with the long right tail).
28These estimated distributions do not include wells with completion times that appear to be dataerrors, such as wells that were reported to begin production before they were drilled or took longer toenter production than is reasonable (two years after drilling for onshore wells or 10 years for offshorewells), based on the length of a typical lease’s primary term. Wells included in this calculation repre-sent approximately 90 percent of oil and gas production, suggesting that the impact of excluding theremainder is minor.
27
0.00
0.02
0.04
0.06
0.08
0.10
Months from Spud to Production
Den
sity
0 6 12 18 24
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
0 20 40 60 80 100 120
0.00
0.02
0.04
0.06
0.08
0.10
Months from Spud to Production
Den
sity
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
Figure 9: Density Plot of Time from Drilling (“Spud”) to First Production, by WellType, Onshore (top panel) and Offshore (bottom panel)
28
2.2.4 Oil and Gas Production from New Wells, Over Time (Box 4a)
Once wells are online, they produce years or decades to come. Converting newly operat-
ing wells to oil and gas production requires modeling the profile of such production over
time (also called a “type curve”) for each well type. As research has shown, production
from existing wells is almost perfectly inelastic to oil and gas prices.29 A panel regression
of monthly well-level production on prices confirms this in this data as well.30 For this
reason, I model the average production profile per well as exogenous based on the av-
erage production profile, by well type, scaled to the average observed initial production
(IP) in 2019 (the first year of the simulation), also by well type.
More specifically, I calculate the average production profile by age of well in months
for wells beginning production in 2009 or later. This year coincides with the beginning
of the shale boom, is sufficiently recent that it captures the trends toward sharper
decline curves due to the growing focus on the development of shale formations, and is
sufficiently long ago to ensure an adequate number of wells in the data with a long enough
observed history to reliably estimate production profiles. These profiles are converted to
a percentage of IP and projected out to 30 years (the length of the simulation) using an
Arps curve fit on the first five years of the average production profile.31 The estimation
of Arps curves is discussed in more detail in the next section. These fitted curves (as a
29Although well shut-ins happen on rare occasions, this primarily affects the timing, rather than thelevel, of production.
30Results available on request.31I use the first five years to avoid noise that would be introduced by using data from a small number
of wells that contribute to the production profile in the final years of the sample. Namely, the only wellsfor which we observe the 120th month of production in December 2018 are those drilled in exactly themonth of January 2009. Using that small sample would lead to noisy estimates, composition bias in thefinal months of the production profile, and potentially divergent production projections. Nonetheless,the majority of a well’s cumulative oil and gas production is realized in the first five years (Figure 10).
29
percentage of IP) are then scaled to average IP values observed in the first year of the
simulation.32 The resulting production profiles, by well type, are shown in Figure 10.0
100
200
300
400
500
600
700
Months Since Initial Production
Oil
Pro
duce
d P
er D
ay (
barr
els
per
day)
0 60 120 180 240 300 360
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
020
0040
0060
0080
00
Months Since Initial Production
Gas
Pro
duce
d P
er D
ay (
mcf
per
day
)
0 60 120 180 240 300 360360
050
010
0015
0020
0025
0030
00
Months Since Initial Production
Oil
Pro
duce
d P
er D
ay (
barr
els
per
day)
0 60 120 180 240 300 360
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
010
0020
0030
0040
0050
00
Months Since Initial Production
Gas
Pro
duce
d P
er D
ay (
mcf
per
day
)
0 60 120 180 240 300 360360
Figure 10: Estimated Production Profiles, by Well Type, Oil Production (left panels)and Gas Production (right panels), Onshore (top) and Offshore (bottom)
32I assume that IP values by well type remain constant over time. Although IPs are unlikely toremain constant over time in reality, it is not clear whether they will rise or fall in the long run. On theone hand, technological innovation has driven large increases in IP in recent years, and it is possible thistrend could continue. On the other hand, IPs could decline as the most productive wells are exhausted.
30
2.2.5 Oil and Gas Production from Existing Wells, Over Time (Box 4b)
For a given price path and set of policies, the previous modules of the model (boxes 1-
4a) combined produce projections of oil and gas production over time from newly drilled
wells. Because the typical well produces oil and gas for decades after it is drilled, a
nontrivial share of total production at any given time is from existing (i.e., previously
drilled) wells. I calculate production from wells drilled before the beginning of the
simulation (2019) using well-specific Arps curve projections.
The Arps curve is the standard method used by petroleum engineers to forecast an
individual well’s future production. This approach involves estimating the following
nonlinear equation by nonlinear least squares:
qτ =q0
(1 + bd0τ)1b
+ ετ , (2)
where qτ is a well’s oil or gas production in month τ = 0, 1, 2, . . . of a well’s productive
life. The q0 term is the well’s IP rate, and the Arps parameters are d0 (which represents
the initial decline rate) and b (which represents how much the decline rate slows over
time).33 I estimate separate Arps curves for each well still producing as of the end of
2018 and use the fitted Arps curve to project production to 2050. I estimate two Arps
curves for each well—one each for oil and gas. For each well type, production is summed
across wells by calendar month to generate total projected oil and gas production from
existing wells. These projections are shown in Figure 11.
33The special case of b = 0 corresponds to constant exponential decline, qτ = q0e−d0τ , but it is
common for production to decline slower than exponentially (b > 0).
31
02
46
8
Oil
Pro
duct
ion
(mb/
d)
2019 2025 2030 2035 2040 2045 2050
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
OnshoreOffshore
010
2030
4050
6070
Gas
Pro
duct
ion
(bcf
/d)
2019 2025 2030 2035 2040 2045 2050
Oil Wells, NonfederalOil Wells, FederalGas Wells, NonfederalGas Wells, Federal
OnshoreOffshore
Figure 11: Projected Production of Oil (top) and Gas (bottom) from Existing Wells asof 2018
32
2.2.6 Simulating US Oil and Gas Supply in Each Scenario (Boxes 1–4)
With each stage of the US supply process modeled, I combine them as in Newell, Prest
and Vissing (2019) and Newell and Prest (2019) to simulate future US oil and gas
production for a given price path and set of federal leasing policies. I begin the simulation
at the start of 2019 (following the last full year of complete data). The oil (WTI) and
gas (Henry Hub) price paths are set equal to historical observed monthly average values
for 2019 and the first half of 2020, after which prices are set equal to the futures prices
for delivery in each future month (as of June 25, 2020).34 Those futures strips extend
through 2031–2032, and I extrapolate to 2050, accounting for seasonality.35 Baseline
federal royalty rates are set to 12.5 percent for onshore wells and 18.75 percent for
offshore wells.36
In addition to a baseline “business as usual” scenario, I simulate scenarios featuring
different combinations of federal leasing policy changes that take effect in 2020. Each
scenario reflects a different application of one or more of the three policy levers: raising
federal royalty rates, charging carbon adders, and a moratorium.
34It is common to use futures prices as proxies for expected future spot prices, based on the conceptthat arbitrage will ensure that futures reflect the market’s best guess of future spot prices. This ideaeffectively assumes that oil futures prices do not carry a risk premium. As documented by Baumeisterand Kilian (2016), such a risk premium can be positive or negative. Accounting for it would addconsiderable additional complexity, so I set it aside for simplicity. Given the broader uncertainty aboutthe outlook for oil prices in 2020, the potential bias from ignoring the risk premium is likely swampedby other uncertainties. Further, because this potential bias would be present in both the baseline andpolicy scenarios, it is unlikely to strongly affect the modeling results that reflect their difference.
35Futures prices were downloaded from the CME group reflecting closing prices on June 25, 2020,when futures prices were available for oil and gas for delivery through 2031 (WTI) and 2032 (HenryHub). I extrapolate the final point on each futures strip beyond 2030 based on a regression of the logdifference of futures prices on month of year indicators, which account for price seasonality for naturalgas. The procedure yields prices that rise at average annual growth rates after 2030 of of 2.0 and 2.5percent for WTI and Henry Hub, respectively. Figure A.2 depicts these baseline price projections.
36Nonfederal royalty rates are assumed to be unchanged over the simulation. Since only changeswould affect nonfederal drilling through equation (1), the precise assumption about nonfederal royaltyrates is immaterial. Nonetheless, I apply an 18 percent royalty rate, in the middle of the range ofnonfederal rates reported in U.S. Government Accountability Office (2017). For example, state ratesin Colorado, New Mexico, Utah, and Wyoming range from 12.5 to 20 percent, whereas rates in Texasare set at 25 percent. See also https://westernpriorities.org/wp-content/uploads/2015/06/
Royalties-Report_update.pdf.
33
I model the first two policies as affecting the net price of oil and gas received on
production from new federal wells. For example, raising the onshore royalty rate from
12.5 percent to 18.75 percent amounts to an (approximately) 6.25 percentage point drop
in the net price of oil and gas received, which is fed into the estimated model of drilling
elasticities in equation (1). Similarly, carbon adders are translated into oil- and gas-price
equivalent values. For example, a $50/tCO2 carbon adder translates into a $21.50 per
barrel charge, assuming an emissions rate of 0.43 tCO2 per barrel of oil combusted. The
corresponding charge for natural gas is $3.30 per mcf.37 Note that these carbon charges
are quite large relative to market value, suggesting that carbon adders are likely to lead
to large reductions in federal production, particularly for natural gas.38 The simulation
results demonstrate this effect.
For all well types, I use the applicable paths of net oil and gas prices (after ap-
propriately deducting royalties and carbon adders)39 to calculate the predicted values
of monthly log-changes in drilling activity from equation (1), in both the baseline and
policy cases, and convert this to predicted levels of wells drilled by month.
A moratorium on new federal oil and gas leasing is simpler to model because it is a
quantity instrument rather than a price instrument. After instituting a moratorium, new
leasing ends, meaning eventually new federal drilling activity must go to zero. Operators
may still have existing federal leases on which they have yet to drill. These leases last
up to 10 years, assuming no extensions. For this reason, I model the moratorium as a
gradual, linear 10-year decline in federal drilling activity that the model would otherwise
predict, and no new drilling is permitted after that time. That is, 0 percent of new federal
37The emissions rate for natural gas is 0.066 tCO2e per mcf. This is based on a 177 lbs of CO2 per mcffrom direct gas combustion, plus 28.55 lbs of CO2e from methane leakages, which is based on the 2.3percent methane leakage estimate from Alvarez et al. (2018) and a 100-year global warming potential.Together, this implies an emissions rate of (117 + 28.55 lbs CO2)× 1 metric ton
2204.62 lbs = 0.066 tCO2e per mcf.38To ensure simulated net prices never go negative, which would preclude the use of logged prices in
equation (1), I impose a floor on the net oil and gas prices equal to $1 per barrel of oil equivalent.39Naturally, for nonfederal wells, royalty rates are held constant and no carbon adders are charged.
Nonfederal wells are nonetheless affected by the endogenous oil and gas prices calculated in box 5, asexplained in the next section. This accounts for policy leakage.
34
drilling is assumed to be covered by the moratorium in year 1 of the policy change, 10
percent covered in year 2, and so on until 100 percent is covered in year 10.
Ten years is the standard statutory length of federal oil and gas leases.40 The linear
phase-in assumption effectively assumes that no existing-but-undrilled leases would be
renewed beyond their 10-year primary term. The royalty rate increase and carbon adder
policies are also modeled as being phased in linearly over 10 years for the same reason.
As in Gerarden, Reeder and Stock (2020), I assume a linear phase-in of royalty rates
and carbon adders that apply to all wells because explicitly modeling which subsets of
wells would face which royalty rates would add an unnecessary degree of complexity.
The results are not sensitive to this assumption because only about 15 percent of
business-as-usual federal emissions over the 30-year simulation horizon are from newly
drilled wells during this 10-year window. The linear phase-in effectively implies that
about half of this 15 percent of federal emissions (about 8 percent) are covered by the
policy. Alternative extreme assumptions that either none of these wells are covered or
all of them are would change the share of covered federal emissions by ∼ ±8 percent.
2.2.7 Rest of World Supply and Demand (Box 5)
The previous sections explain how US oil and gas supply is simulated for any given price
path and set of policies. However, because the United States is not a closed economy,
incorporating the potential for emissions leakage requires also modeling the responses of
foreign supply and demand. To account for these trade effects, I incorporate an ROW
model based on supply and demand projections from the International Energy Agency
(IEA)’s 2019 World Energy Outlook (WEO) (IEA 2019). I use the WEO’s central
“Stated Policies” scenario for global oil and gas demand and ROW supply, interpolated
to the monthly level to correspond with the time step in the US supply model.
40Ten years is the standard length of onshore leases and Alaskan and deepwater offshore leases (CBO2016). Although some offshore leases in shallow water have shorter lease terms (such as eight years),these account for relatively little of offshore oil and gas development.
35
Some technical adjustments to these projections must be made to render the US
supply model comparable to the values reported in the WEO. For example, the US
supply model simulates gross gas withdrawals, whereas the WEO projections are for
marketed gas production, which is a subset of gross withdrawals. In addition, I make
some adjustments to the WEO demand and ROW supply projections to be consistent
with the much lower oil and gas prices observed in 2020 relative to those assumed in the
WEO projections in 2019. These are discussed in appendix section C.
I model ROW supply as endogenous using the ROW supply elasticities implied by
the WEO projections. I infer these supply elasticities by comparing ROW oil and gas
production in the WEO’s base “Stated Policies” scenario to its “Current Policies” sce-
nario, which corresponds to a case with somewhat higher oil and gas demand. These
elasticities start at about +0.2 in the short run (2020) for both oil and gas supply and
rise gradually in the long run (2050) to +0.9 for oil and +1.2 for gas, owing to the rising
implied elasticities over time embedded in the WEO. On average, the ROW oil and gas
supply elasticities are +0.4 for oil and +0.5 for gas over the simulation horizon.41
For global oil and gas demand, I apply standard demand elasticities from the liter-
ature. In the main results, I use elasticities of -0.2 for both oil and gas demand, based
on the central case in Erickson and Lazarus (2018) for oil (which was in turn based
on literature reviews by Hamilton 2009 and Bordoff and Houser 2015) and empirical
estimates from Arora (2014) and Auffhammer and Rubin (2018) for gas. I also conduct
a “high-elasticity” sensitivity case where the elasticities are set to -0.51 for oil and -0.42
for gas. The gas demand elasticity is from Metcalf (2018), which in turn is based on
estimates in Hausman and Kellogg (2015). The oil demand elasticity is based on the
mean estimate from Balke and Brown (2018), but it also is very close to the value of
-0.50 used in Metcalf 2018 based on the findings of Allaire and Brown (2012).
41This is similar to the estimated long-run oil supply elasticity of 0.55 found in Balke and Brown(2018) from an empirically calibrated dynamic stochastic general equilibrium model. That paper didnot estimate a gas supply elasticity.
36
2.2.8 Solving for New Equilibrium Prices
When simulating a change in federal leasing policies that reduces federal oil and gas
production, I solve for the new oil and gas prices that clear the markets for both oil and
gas. The equilibrium concept I use is based on a standard no-arbitrage condition that
implies that changes in expected future commodity prices are immediately capitalized
into contemporaneous commodity prices.
This equilibrium concept is perhaps most easily understood as an application of the
result of the standard Hotelling model of nonrenewable resource extraction (Hotelling
1931), although it is not restricted to that model. In the Hotelling model, current and
(discounted) future oil prices in equilibrium are inseparably linked due to a no-arbitrage
condition, implying that the price in a future year is a fixed multiple of current prices.
This implies that an x% increase in the equilibrium price of oil due in the future to a
supply shock must also coincide with an equivalent x% increase today. More generally,
this inseparable, intertemporal link between prices over time is the result of no-arbitrage
condition for any storable asset (see, e.g., Fama and French 1987, 1988).42
Based on this result, I assume that the percentage change in the price of oil is
the same across all periods in the simulation horizon, and similarly so for the price of
gas. This theoretically inspired equilibrium mechanism also greatly simplifies solving for
new market-clearing prices because it only requires a two-dimensional optimization. A
key, desirable consequence of this assumption is that the expected effects of announced
policies are immediately capitalized into market prices, even before the policy has an
42The equilibrium price of a futures contract, Ft for a storable asset equals its spot price, S, grossedup by the discount factor, (1 +Rt), plus the marginal warehousing cost net of the marginal convenienceyield of holding the asset, Wt − Ct: Ft = S(1 + Rt) + Wt − Ct. With risk-neutral traders, the futuresprice should also reflect the expected future spot price. This demonstrates the link. The assumptionthat supply shocks have an equal percentage effect on current and future prices requires either thatthe marginal warehousing cost less the marginal convenience yield, Wt − Ct, is zero or that it scalesin proportion to the price. This is a reasonable approximation the purposes of this model. Finally,this equation is only valid in the presence of an interior solution for inventories, which ensures the no-arbitrage condition holds. Therefore, I track inventories in the model to ensure that they never becomenegative (which is impossible) or exceed physical limits. For the small policy changes (in the contextof global supply) that I consider, these constraints never become binding.
37
appreciable effect on realized production. As a result, the model includes a kind of
“green paradox” effect (Sinn 2008), whereby an anticipated tightening of policy in the
future leads to immediately higher prices and hence somewhat accelerated oil and gas
production.
3 Results
3.1 Simulation Results
I model the following six policy scenarios, including variants of policies previously con-
sidered for coal leasing reform in Krupnick et al. (2016) and Gerarden, Reeder and Stock
(2020) and those by HSCCC (2020).
1. A raise in onshore royalty rates to 18.75 percent (matching the current 18.75 rate
typically charged offshore);
2. A raise in onshore royalty rates to 25 percent (matching the high end of rates on
state and private lands);
3. A raise in onshore and offshore royalty rates to 25 percent;
4. A $50/tCO2e carbon adder, rising at 2 percent annually, both onshore and offshore;
5. A $50/tCO2e carbon adder and a 25 percent royalty rate, both onshore and off-
shore; and
6. A moratorium on new leasing, onshore and offshore.
Consistent with statutory lease terms, I assume the primary terms of existing, undrilled
leases expire on schedule (i.e., after 10 years). That is, I assume no extensions on
undrilled leases, but once drilled, those wells may continue to produce indefinitely, also
consistent with existing law.
38
Table 1 shows the impacts of each policy scenario on equilibrium oil and gas prices,
emissions by source (US federal, US nonfederal, ROW, and global), emissions leakage
rates, and changes in royalty and carbon adder revenues. All values shown are annual
averages over the full 2020–2050 window.43 Further, the table presents results using
both “base-case” demand elasticities and the “high-elasticity” sensitivity case.
Because US federal oil and gas production is a relatively small fraction of global oil
and gas supply (about 2 and 3 percent for oil and gas, respectively, in 2018), the price
impacts are small. Raising federal royalty rates would increase oil (WTI) and gas (Henry
Hub) prices by 0.1–0.3 percent, depending on the size of the rate increase. Federal oil
and gas production declines somewhat, albeit with a bit of lag due to the delay between
leasing policy changes and changes in realized production. On average, this leads to a
direct reduction in federal emissions of 16–37 MMTCO2e per year on average during
the 2020–2050 window (see column 3). However, this is offset by an increase of 3–9
MMTCO2e associated with production on nonfederal US lands (column 4) and another
6–18 MMTCO2e increase in emissions from foreign production (column 5). In other
words, about one-third of the leakage arises from US production from nonfederal lands
and the other two-thirds from foreign (ROW) supply.
43I focus on averages here and in Table 1 for simplicity and to reflect cumulative emissions effects,which are more relevant than emissions in any particular year from a climate perspective. The timeprofiles of US production are shown in Figures 12 and 13, and the full time paths of emissions impactsare shown in appendix Figures A.3 and A.4. By nature of the equilibrium concept, the percentagechanges in oil and gas prices are constant over the time horizon.
39
Tab
le1:
Pol
icy
Impac
tson
Oil
and
Gas
Pri
ces,
CO
2e
Em
issi
ons,
and
Roy
alty
&C
arb
onR
even
ues
(Annual
Ave
rage
s20
20–2
050)
Em
issi
ons
Ch
ange
(∆M
MT
CO
2e/
year
),P
rice
2020
–50
Ave
rage
Roy
alty
Ch
ange
(%)
US
Lea
kage
&C
arb
onB
ase
Dem
and
Ela
stic
itie
sO
ilG
asF
eder
alN
onfe
der
alR
OW
Glo
bal
rate
Rev
enu
e(∆
$b)
(εD oil
=−
0.2,εD g
as
=−
0.2)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
18.7
5%O
nsh
ore
Roy
alty
Rat
e(R
R)
0.1%
0.2%
-16
48
-473
.8%
$1.1
25%
RR
,O
nsh
ore
only
0.1%
0.3%
-31
815
-873
.8%
$2.0
25%
RR
,O
nsh
ore
and
Off
shor
e0.
2%0.
3%-3
79
18-1
073
.4%
$3.0
$50
Car
bon
Ad
der
(ris
ing
at2%
)1.
2%1.
6%-2
1652
106
-58
73.2
%$7
.1$5
0C
arb
onA
dd
er(2
%)
&25
%R
R1.
3%1.
7%-2
3857
117
-64
73.1
%$6
.5M
orat
oriu
m1.
9%1.
9%-3
1473
155
-85
72.9
%-$
5.5
Hig
hD
eman
dE
last
icit
ies
(εD oil
=−
0.51
,εD g
as
=−
0.42
)
18.7
5%O
nsh
ore
RR
<0.
05%
0.1%
-16
36
-755
.2%
$1.1
25%
On
shor
eR
R0.
1%0.
2%-3
16
11-1
455
.2%
$2.0
25%
RR
,O
nsh
ore
and
Off
shor
e0.
1%0.
3%-3
77
14-1
754
.4%
$3.0
$50
Car
bon
Ad
der
(ris
ing
at2%
)0.
8%1.
3%-2
1639
78-1
0053
.9%
$7.1
$50
Car
bon
Ad
der
(2%
)+
25%
RR
0.9%
1.4%
-239
4386
-111
53.7
%$6
.4M
orat
oriu
m1.
3%1.
6%-3
1454
113
-147
53.2
%-$
5.5
Not
es:
RO
W=
rest
ofw
orld
,R
R=
roya
lty
rate
.A
llva
lues
are
rela
tive
toth
eb
usi
nes
s-as
-usu
alsc
enar
io.
All
pol
icie
sex
cep
tfo
rth
efi
rst
two
app
lyb
oth
onsh
ore
and
offsh
ore
wel
ls.
The
carb
onad
der
grow
sat
2p
erce
nt
annu
ally
,in
lin
ew
ith
IWG
esti
mat
es.
Oil
and
gas
pri
cep
erce
nta
gech
ange
sar
ere
lati
veto
WT
Ian
dH
enry
Hu
b,
resp
ecti
vely
.C
olu
mn
(6)
equ
als
the
sum
ofco
lum
ns
(3),
(4),
and
(5).
Col
um
n(7
)eq
ual
s[(
4)+
(5)]
/|(3
)|.
40
The leakage rate for the royalty rate policies ranges from 54 to 74 percent of the
direct federal reductions. The magnitude of the leakage rate does not vary substantially
with the policy approach but rather depends on the demand elasticity.44 This is why I
consider the sensitivity of the results to alternative demand elasticity assumptions.45
The royalty rate policies lead to small effects in emissions because they have relatively
little effect on federal production; those higher royalty rates raise nontrivial amounts of
additional revenue, $1–$3 billion in incremental revenue annually on average.46 For
context, in 2016, when oil and gas prices were at low levels similar to those in 2020,
ONRR reported $3.9 billion in oil and gas royalty revenues ($1.5 billion onshore and $2.4
billion offshore), so the estimated $1–$3 billion in additional revenues can be substantial.
The carbon adder drives substantially larger reductions than royalty rates do. As
previously noted, carbon adders can be quite large as a fraction of the market value of oil
and (especially) gas. Yet because they are levied only on federal oil and gas producers
that represent a small fraction of the market, there is little ability to pass these costs
through to consumers in the form of higher prices. Therefore, production on federal
44This should not be surprising conceptually. First principles tells us that in the extreme case ofperfectly inelastic demand (εDoil = εDgas = 0), leakage would be 100 percent, whereas at the other extreme
of perfectly elastic demand εDoil = εDgas = −∞), leakage would be zero.45I also conduct a sensitivity analysis in which I reduce the ROW oil and gas supply elasticities by
half. The effects on federal emissions and revenues are almost exactly the same as in the main resultsin Table 1. This low ROW supply elasticity case leads to a leakage rate of around 64 percent, and themoratorium reduces global emissions by 113 MMTCO2e. Both estimates are about halfway in betweenthe base- and high-elasticity results in Table 1. With lower ROW supply elasticities, oil and gas pricesrise somewhat more (2.4–2.6 percent instead of 1.9 percent) and leakage is attributable in approximatelyequal parts to production from nonfederal US and ROW sources. Because the results are less sensitiveto the ROW supply elasticity than to the demand elasticity, Table 1 focuses on the demand elasticitysensitivity. Appendix Table A.9 also contains another sensitivity case in which oil and gas prices areassumed to remain at high prepandemic levels, based on IEA’s 2019 projections. The results are similarto the low ROW supply elasticity case–about halfway between the “base-” and “high-”demand elasticitycases.
46These revenue estimates do not include potential reductions to bonus bids driven by higher royaltyrates or carbon adders. This is likely to be a small effect, however. Further, fully accounting for thiseffect would require building an auction model embedded in a model of competitive versus noncom-petitive leasing and parcel choice, which is well beyond the scope of this paper. A substantial share offederal oil and gas leases are sold noncompetitively at the minimum allowable bonus bid of $2 per acre,which, on a typical 1,000 acre federal lease, amounts to only $2,000. Although leases sold competitivelyearn higher bids, bonus bids nonetheless represent a small fraction of federal oil and gas revenues (his-torically about 10 percent, less than $1 billion per year), so even if I were to assume that these bidswere driven to effectively zero, it would not change the order of magnitude of the revenue estimates.
41
lands falls sharply under a carbon adder, leading to a 216 MMTCO2e reduction in
direct emissions. Equilibrium oil and gas prices rise by 1–2 percent, spurring additional
production from nonfederal and foreign sources, leading to leakage rates of 53–73 percent.
The global reduction in emissions is much larger than under the royalty rate policies, now
58–100 MMTCO2e per year on average depending on the demand elasticity assumption.
Even though federal production would decline sharply under carbon adders, they
would nonetheless generate substantial amounts of incremental carbon revenues from
the federal wells that do get developed. This leads to an estimated $7 billion per year
more in combined royalty and carbon revenues than under business as usual.
Adding a royalty rate increase to a carbon adder leads to a slight increase in the
emissions reductions, but it also actually yields less revenue than a carbon adder alone
because of the depressing effect the combined charges have on production. This sug-
gests a trade-off between emissions reductions and revenue generation when considering
implementing overlapping policies.
Finally, compared to a carbon adder, a moratorium generates only modestly larger
emissions reductions—314 MMTCO2e from federal sources and 85–147 MMTCO2e glob-
ally. Put differently, the carbon adder alone generates about two-thirds as much emis-
sions reductions as a full moratorium because the costs of carbon adders are so large
that they would make much of federal drilling unprofitable anyway. This is particularly
true for gas-directed drilling, for which the carbon adder is large relative to the market
value of the gas produced. The incremental emissions reductions that a moratorium
achieves primarily comes from eliminating the remaining oil-directed drilling that would
continue to be profitable under a carbon adder.
Although the carbon adders can achieve much of the emissions reductions that a
moratorium could, the policies are not substitutes, because they have diverging implica-
tions for revenues. A carbon adder would raise about $7 billion annually in incremental
royalty and carbon revenue more than under business as usual, whereas a moratorium
42
would lose more than $5 billion by eliminating leases that would otherwise be paying roy-
alties. Although both policies reduce production and hence emissions, the carbon adder
affects behavior through a price instrument that raises funds, whereas the moratorium
has its effect through a quantity instrument that does not.
To the extent that policymakers have dual goals of reducing emissions and raising
revenues (for example, to be used for transition assistance to states and communities
dependent on extracting federal resources, investments in other approaches to reduce
emissions, such as R&D, or reducing other distortionary taxes), then a carbon adder
may be considered a superior alternative to a moratorium.
3.2 Discussion
3.2.1 Are These Effects Large?
A useful benchmark to which to compare these estimated reductions is the Clean Power
Plan (CPP). The CPP was projected to reduce emissions by approximately 400 MMTCO2e
annually once fully implemented in 2030.47 This is similar to the direct reduction in fed-
eral emissions that I estimate a moratorium could achieve—314 MMTCO2e. However,
policies targeting federal oil and gas production have a much greater potential for leak-
age than the CPP did. That policy pertained to the power sector, which is much more
regionally contained. Due to leakage, I estimate much smaller global emissions reduc-
tions of 85–147 MMTCO2e for a moratorium. In other words, my estimates suggest that
an oil and gas moratorium could produce global emissions reductions about one-quarter
of those projected for the CPP. Although this may appear relatively small, the CPP is
often referred to as the Obama administration’s “signature climate policy,”48 whereas a
47See estimates in EPA (2015) and Gerarden, Reeder and Stock (2020), which range from -357 to-413 MMTCO2e for the mass-based CPP.
48See, for example, ABC News, June 19, 2019, “EPA finalizes power plant rules to replace Obama’ssignature climate change policy.”
43
moratorium would likely be one of a basket of policies implemented for the oil and gas
sector (e.g., see the long list of policies proposed by HSCCC 2020).
Another way to assess the magnitude of these reductions is to consider their mon-
etized value using the social cost of carbon. These emissions reductions, valued at the
SCC of $50 in 2020 (rising at 2 percent annually), imply estimated climate benefits of
approximately $4–$7 billion annually for a carbon adder and $6–$10 billion annually for
a moratorium.
3.2.2 Caveat about Emissions Rates
The reported emissions impacts reflect the gross emissions associated with the produc-
tion and combustion of oil and gas. They do not account for the possibility that the fuel
produced may substitute for other fuels with higher carbon intensities, such as coal.49
This is primarily a concern for natural gas which, along with the growth of renewable
energy, has displaced coal-fired power generation in recent years in the United States
(Mohlin et al. 2018). However, the literature has consistently found that increased gas
supplies on their own have negligible effects on GHG emissions in the long run (Gilling-
ham and Huang 2019; Shearer et al. 2014; Newell and Raimi 2014; Huntington 2013;
Brown, Krupnick and Walls 2009). This negligible effect arises because, without climate
policy in place, gas displaces both zero-carbon energy and coal in similar measure. The
unexpectedly rapid closure of US coal plants in the recent years following that literature
suggests that future natural gas use may increasingly crowd out zero-carbon alternatives
like nuclear or renewable energy, at least in the United States. But this assumption may
still overstate the emission impacts of reducing natural gas consumption in other coun-
tries. Conducting a full substitution analysis for all possible global substitution patterns
is beyond the scope of this paper and a valuable avenue for future research.
49An interesting question for future research is how reforming federal oil and gas leasing policyalongside reforming coal leasing policy would interact in the power sector. The size and direction ofthis effect would depend on the degree of substitution between gas, coal, and zero carbon electricitysources, both federal and nonfederal.
44
For present purposes, however, I bound the effect of this assumption by calculating
what share of emissions reductions come from reducing oil (rather than gas) production,
where such a substitution effect is less of a concern and reduced demand is expected to
come from reduced energy use. For both the carbon adder and moratorium policies, the
majority of the estimated global emissions reductions come from reduced oil consumption
(63 and 70 percent of emissions reductions under the carbon adder and moratorium,
respectively). Therefore, when ignoring the emissions associated with reduced natural
gas consumption, the emission reductions from those policies would shrink by about
one-third. The reductions in oil and gas production are illustrated in the next section.
3.2.3 Effects on Federal Oil and Gas Production, by Well Type
Figures 12 and 13 show the effects of each policy on federal oil and gas production,
over time and disaggregated by onshore versus offshore production. Even absent any
policy change, the model projects a near-term decline in oil and gas production in the
baseline (dashed lines); this reflects the combined factors of the observed decline in
federal drilling activity in recent years (see Figure 6) and the large drop in oil prices in
2020. As prices are projected to recover over the coming decade (as indicated by futures
prices), production is projected to recover as well.
In each panel, production in the policy case is represented by solid lines. As seen in
Figure 12, the onshore-only policies naturally only reduce onshore production, although
the effect is modest, owing to the relatively modest change in royalty rates. By contrast,
the more aggressive policies shown in Figure 13 (carbon adders and a moratorium) lead
to substantial reductions in federal production. Most reductions do not appear until
after 10 years or more, however, because changing leasing policies today restricts new
drilling, leaving existing wells unaffected. The policies begin to have a large effect only
45
after about 10 years, at which point existing, undrilled leases have largely expired and
new federal wells are effectively all covered by a carbon adder or moratorium.50
The carbon adder strongly reduces oil and gas production because it represents a large
effective change in oil and gas prices. In the model, drilling activity responds to this
change in net prices based on the estimated elasticities in Figures 7 and 8. The reduction
in offshore oil drilling is very uncertain, however, due to the large standard errors on
this drilling elasticity (see Figure 8). Under the carbon adder, the policy under which
this elasticity is most important, the federal emissions reductions are roughly equally
attributable to onshore oil, onshore gas, and offshore oil (offshore gas does not change
much because it is small to begin with; see Figure 13). Therefore, if we were to model a
sensitivity case where offshore oil wells were completely unresponsive to price and hence
to carbon adders (an extreme case), this one-third of the emissions reductions from a
carbon adder would vanish and the total emissions impacts would be approximately two-
thirds as large. This elasticity uncertainty is primarily relevant to the royalty rate and
carbon adder policies because they are price instruments. By contrast, a moratorium is
more akin to a quantity instrument, which eliminates this source of uncertainty.
In the moratorium case, the production decline primarily reflects natural declines
from existing wells drilled prior to the policy change. Because wells typically continue
producing for decades after they were first drilled, a small but positive amount of oil
and gas production remains in both 2040 and 2050, well after the moratorium on new
leases has been fully phased in. This implies that a moratorium alone is not sufficient to
achieve the goal of net-zero emissions from public lands by 2040 in the HSCCC (2020)
report. That goal would require more aggressive policies than even a moratorium, such as
modifying existing leases (which is not considered in HSCCC 2020) and/or a substantial
role for carbon sequestration and renewable energy development on federal lands.
50As noted in section 2.2.6, this lag is modeled by linearly phasing in each policy over the first 10years of the simulation, reflecting the standard 10-year lease length.
46
4 Conclusion
Restricting oil and gas production on federal lands has increasingly attracted attention
from policymakers. The Obama administration imposed temporary moratoriums on
coal and offshore oil and gas leasing, and 2020 presidential candidate Joe Biden has
endorsed ending oil and gas development on federal lands. Because these policies can
be enacted without congressional action, they are more likely to implemented by an in-
terested administration. Yet such policies are controversial in part because of concerns
about leakage of oil and gas production to nonfederal sources. I estimate the impacts
of three key policies proposed to reform US federal oil and gas leasing: raising royalty
rates, charging carbon adders to internalize the externalities associated with greenhouse
gas emissions, and a moratorium on all new oil and gas leasing. Although raising royalty
rates is unlikely to have major effects on oil and gas production or emissions, a morato-
rium could have substantial effects, reducing direct federal emissions by 314 MMTCO2e.
However, production shifts to nonfederal sources lead to smaller net global emissions
of 85 to 147 MMTCO2e–that is, a leakage rate of 53–73 percent. A moratorium also leads
to significant losses in royalty revenues ($5–$6 billion annually). An alternative policy
approach of charging carbon adders to internalize climate externalities could achieve
about two-thirds of the emissions reductions that a moratorium would (216 MMTCO2e
from federal lands and 58–100 MMTCO2e globally on net) while also raising as much as
$7 billion annually in incremental royalty and carbon revenues. These revenues could be
used for other policy priorities, such as research and development, reductions in other
distortionary taxes, or transition assistance for affected states and communities that
are dependent on fossil fuel extraction. However, even the most aggressive policy, a
moratorium on new leasing, would not on its own achieve the HSCCC (2020) goal of
net-zero emissions from oil and gas on federal lands by 2040. Achieving that ambitious
goal would therefore require modifying existing leases and/or additional investments in
carbon sequestration and renewable energy development on federal lands.
47
18.7
5% O
nsho
re R
oyal
ty R
ate
(RR
)
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
25%
Ons
hore
RR
Fed
eral
Tot
alF
eder
al O
nsho
reF
eder
al O
ffsho
re
Bas
elin
eP
olic
y C
ase
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
25%
RR
, Ons
hore
and
Offs
hore
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
Fig
ure
12:
Pro
ject
edF
eder
alO
ilan
dG
asP
roduct
ion
by
Pol
icy–R
oyal
tyR
ate
Pol
icie
s(B
ase
Ela
stic
itie
s)
48
$50
Car
bon
Add
er (
2%)
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
$50
Car
bon
Add
er (
2%)
+ 2
5% R
R
Fed
eral
Tot
alF
eder
al O
nsho
reF
eder
al O
ffsho
re
Bas
elin
eP
olic
y C
ase
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
Mor
ator
ium
2020
2025
2030
2035
2040
2045
2050
01234
US
Fed
eral
Cru
de O
il P
rodu
ctio
n
mb/d
2020
2025
2030
2035
2040
2045
2050
051015
US
Fed
eral
Gas
Pro
duct
ion
bcf/d
Fig
ure
13:
Pro
ject
edF
eder
alO
ilan
dG
asP
roduct
ion
by
Pol
icy–C
arb
onA
dder
and
Mor
ator
ium
Pol
icie
s(B
ase
Ela
stic
itie
s)
49
References
Allaire, Maura, and Stephen P.A. Brown. 2012. “Eliminating Subsidies for FossilFuel Production: Implications for U.S. Oil and Natural Gas Markets.” Resources forthe Future Issue Brief.
Allcott, Hunt, and Daniel Keniston. 2017. “Dutch Disease or Agglomeration? TheLocal Economic Effects of Natural Resource Booms in Modern America.” The Reviewof Economic Studies, 85(2): 695–731.
Alvarez, Ramon A., Daniel Zavala-Araiza, David R. Lyon, David T. Allen,Zachary R. Barkley, Adam R. Brandt, Kenneth J. Davis, Scott C. Herndon,Daniel J. Jacob, Anna Karion, Eric A. Kort, Brian K. Lamb, ThomasLauvaux, Joannes D. Maasakkers, Anthony J. Marchese, Mark Omara,Stephen W. Pacala, Jeff Peischl, Allen L. Robinson, Paul B. Shepson, ColmSweeney, Amy Townsend-Small, Steven C. Wofsy, and Steven P. Hamburg.2018. “Assessment of Methane Emissions from the U.S. Oil and Gas Supply Chain.”Science, 361(6398): 186–188.
Anderson, Soren T., Ryan Kellogg, and Stephen W. Salant. 2018. “Hotellingunder Pressure.” Journal of Political Economy, 126(3): 984–1026.
Arora, Vipin. 2014. “Estimates of the Price Elasticities of Natural Gas Supply andDemand in the United States.” MPRA Paper No. 54232.
Auffhammer, Maximilian, and Edward Rubin. 2018. “Natural Gas Price Elastic-ities and Optimal Cost Recovery Under Consumer Heterogeneity: Evidence from 300Million Natural Gas Bills.” Energy Institute at Haas Working Paper 287.
Balke, Nathan S., and Stephen P.A. Brown. 2018. “Oil Supply Shocks and theU.S. Economy: An Estimated DSGE Model.” Energy Policy, 116: 357 – 372.
Bartik, Alexander W., Janet Currie, Michael Greenstone, and Christo-pher R. Knittel. 2019. “The Local Economic and Welfare Consequences of HydraulicFracturing.” American Economic Journal: Applied Economics, 11(4): 105–55.
Baumeister, Christiane, and Lutz Kilian. 2016. “A General Approach to Recov-ering Market Expectations from Futures Prices with an Application to Crude Oil.”manuscript.
Beaudreau, Tommy, Janice Schneider, and Joshua Marnitz. 2019. “The Pub-lic’s Interest and Durable Management of Energy Development on Public Lands.”Environmental Law Reporter News & Analysis, 49: 10735.
Bordoff, Jason, and Trevor Houser. 2015. “Navigating the U.S. Oil Export Debate.”
Brown, Stephen PA, Alan Krupnick, and Margaret A Walls. 2009. “NaturalGas: A Bridge to a Low-Carbon Future?” Issue brief, 09–11.
50
CEA. 2016. “The Economics of Coal Leasing on Federal Lands: Ensuring a Fair Returnto Taxpayers.” US Council of Economic Advisers.
Congressional Budget Office. 2016. “Options for Increasing Federal Income FromCrude Oil and Natural Gas on Federal Land.”
Enegis, LLC. 2011. “Benefit-Cost and Economic Impact Analysis of Raising the On-shore Royalty Rates Associated with New Federal Oil and Gas Leasing.”
EPA. 2015. “Regulatory Impact Analysis for the Clean Power Plan Final Rule.” U.S.Environmental Protection Agency.
Erickson, Peter, and Michael Lazarus. 2018. “Would constraining US fossil fuelproduction affect global CO2 emissions? A case study of US leasing policy.” ClimaticChange, 150(1–2): 29–42.
Fama, Eugene F, and Kenneth R French. 1987. “Commodity Futures Prices: SomeEvidence on Forecast Power, Premiums, and the Theory of Storage.” Journal of Busi-ness, 60(1): 55–73.
Fama, Eugene F, and Kenneth R French. 1988. “Business Cycles and the Behaviorof Metals Prices.” The Journal of Finance, 43(5): 1075–1093.
Feyrer, James, Erin T. Mansur, and Bruce Sacerdote. 2017. “Geographic Dis-persion of Economic Shocks: Evidence from the Fracking Revolution.” American Eco-nomic Review, 107(4): 1313–34.
Gerarden, Todd D, W Spencer Reeder, and James H Stock. 2020. “Federal CoalProgram Reform, the Clean Power Plan, and the Interaction of Upstream and Down-stream Climate Policies.” American Economic Journal: Economic Policy, 12(1): 167–99.
Gilbert, Ben, and Gavin Roberts. 2020. “Drill-Bit Parity: Supply-Side Links inOil and Gas Markets.” Journal of the Association of Environmental and ResourceEconomists, 7(4): 619–658.
Gillingham, Kenneth, and Pei Huang. 2019. “Is Abundant Natural Gas a Bridgeto a Low-Carbon Future or a Dead-End?” The Energy Journal, 40(2).
Gillingham, Kenneth, James Bushnell, Meredith Fowlie, Michael Green-stone, Charles Kolstad, Alan Krupnick, Adele Morris, RichardSchmalensee, and James Stock. 2016. “Reforming the US Coal Leasing program.”Science, 354(6316): 1096–1098.
Gillingham, Kenneth T, and James H Stock. 2016. “Federal Minerals LeasingReform and Climate Policy.” The Hamilton Project Policy Proposal.
Goulder, Lawrence H. 2020. “Timing Is Everything: How Economists Can BetterAddress the Urgency of Stronger Climate Policy.” Review of Environmental Economicsand Policy, 14(1): 143–156.
51
Green, Fergus, and Richard Denniss. 2018. “Cutting with Both Arms of the Scis-sors: The Economic and Political Case for Restrictive Supply-Side Climate Policies.”Climatic Change, 150(1–2): 73–87.
Hamilton, James D. 2009. “Understanding Crude Oil Prices.” Energy Journal,30(2): 179–206.
Hamilton, James D. 2014. “Oil Prices as an Indicator of Global Economic Condi-tions.” Econbrowser.
Hausman, Catherine, and Ryan Kellogg. 2015. “Welfare and Distributional Impli-cations of Shale Gas.” National Bureau of Economic Research Working Paper 21115.
Hotelling, Harold. 1931. “The Economics of Exhaustible Resources.” Journal of Po-litical Economy, 39(2): 137–175.
HSCCC. 2020. “Solving the Climate Crisis.” House Select Committee on the ClimateCrisis.
Huntington, Hillard G. 2013. “Changing the Game?: Emissions and Market Impli-cations of New Natural Gas Supplies.” Stanford Univ., CA (United States).
IEA. 2019. “World Energy Outlook 2019.” International Energy Agency.
Interagency Working Group on the Social Cost of Greenhouse Gases, UnitedStates Government. 2016. “Technical Support Document: Technical Update of theSocial Cost of Carbon for Regulatory Impact Analysis Under Executive Order 12866.”
Krupnick, Alan, Joel Darmstadter, Nathan Richardson, and KatrinaMcLaughlin. 2016. “Putting a Carbon Charge on Federal Coal: Legal and EconomicIssues.” Environmental Law Reporter News & Analysis, 46: 10572.
Leshy, John D. 2019. “Interior’s Authority to Curb Fossil Fuel Leasing.” Environmen-tal Law Reporter News & Analysis, 49: 10631.
Merrill, Matthew D., Benjamin M. Sleeter, Philip A. Freeman, Jinxun Liu,Peter D. Warwick, and Bradley C. Reed. 2018. “Federal lands greenhouse emis-sions and sequestration in the United States—Estimates for 2005–14.” US GeologicalSurvey Investigations Report 2018–5131.
Metcalf, Gilbert E. 2018. “The Impact of Removing Tax Preferences for US Oil andNatural Gas Production: Measuring Tax Subsidies by an Equivalent Price ImpactApproach.” Journal of the Association of Environmental and Resource Economists,5(1): 1–37.
Mohlin, Kristina, Jonathan R Camuzeaux, Adrian Muller, Marius Schneider,and Gernot Wagner. 2018. “Factoring in the Forgotten Role of Renewables in CO2
Emission Trends Using Decomposition Analysis.” Energy Policy, 116: 290–296.
52
Newell, Richard G., and Brian C. Prest. 2019. “The Unconventional Oil SupplyBoom: Aggregate Price Response from Microdata.” The Energy Journal, 40(3).
Newell, Richard G, and Daniel Raimi. 2014. “Implications of Shale Gas Develop-ment for Climate Change.” Environmental Science & Technology, 48(15): 8360–8368.
Newell, Richard G., Brian C. Prest, and Ashley B. Vissing. 2019. “Trophy Hunt-ing versus Manufacturing Energy: The Price Responsiveness of Shale Gas.” Journalof the Association of Environmental and Resource Economists, 6(2): 391–431.
Prest, Brian C. 2018. “Explanations for the 2014 oil price decline: Supply or demand?”Energy Economics, 74: 63–75.
Shearer, Christine, John Bistline, Mason Inman, and Steven J Davis. 2014.“The Effect of Natural Gas Supply on US RenewableEenergy and CO2 Emissions.”Environmental Research Letters, 9(9): 094008.
Sinn, Hans-Werner. 2008. “Public Policies Against Global Warming: A Supply SideApproach.” International Tax and Public Finance, 15(4): 360–394.
U.S. Government Accountability Office. 2017. “Oil, Gas, and Coal Royalties, Rais-ing Federal Rates Could Decrease Production on Federal Lands but Increase FederalRevenue.”
Walls, Margaret, Patrick Lee, and Matthew Ashenfarb. 2020. “National Monu-ments and Economic Growth in the American West.” Science Advances, 6(12).
53
Appendix
A Detailed Drilling Regression Results
Tables A.1 through A.8 show detailed regression results for each of the eight well types.
Recall from Figure 5 that the major sources of US oil and gas production are onshore
oil wells (Tables A.1–A.2), onshore gas wells (Tables A.3–A.4), and federal offshore oil
wells (Table A.6). These wells’ elasticities are all of the expected sign and magnitude,
but they are not always statistically significant. The elasticity estimates for the other
well classes (Tables A.5, A.7, and A.8) have little effect on the results because they
contribute very little to US production. For example, very few nonfederal offshore wells
exist, because the federal government owns the vast majority of offshore territories.
B Drilling Model Validation
To demonstrate the model’s ability to accurately predict drilling activity out of sample,
I compare the model’s simulated drilling activity following the unprecedented decline
in oil prices that occurred in 2020 to measures of drilling activity from Baker Hughes.
Recall that the Enverus data are only complete with an approximately one-year lag,
meaning I must start my simulations in 2019. Figure A.1 compares historical (pre-
2019) and simulated (2019–2020) drilling activity in my model, based on the estimated
elasticities in Tables A.1–A.8, to the Baker Hughes rig count, which is collected weekly
and represents the best real-time indicator of drilling activity.51
Figure A.1 shows that wells drilled (Enverus data) and rigs active (Baker Hughes)
are very strongly correlated historically, for both oil- and gas-directed wells. In my sim-
ulation, drilling declines sharply in response to the sharp decline on oil prices, based on
the estimated elasticities from equation (1). The simulated decline in drilling predicted
by my model strongly mirrors the real-time data on active rigs from Baker Hughes. This
implies that the model is doing a good job of predicting drilling rates out of sample.
51Although wells drilled and rigs active are very strongly correlated, they represent slightly differentconcepts because a single rig can drill more than one well per month. For this reason, Baker Hughesrig counts are shown on a secondary y axis.
54
2010 2012 2014 2016 2018 2020
050
010
0015
0020
0025
0030
00
Wel
ls d
rille
d pe
r m
onth
HistoricalPeriod
SimulationPeriod
Oil wells drilled (left)Oil rigs active (right)
Gas wells drilled (left)Gas rigs active (right)
Simulatedwellsdrilled
050
010
0015
0020
00R
igs
activ
e
Figure A.1: Model Validation: Wells Drilled (Actual and Simulated) versus BakerHughes Rig Counts (Actual)
55
Table A.1: Drilling Regression Results–Nonfederal Onshore Oil Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.001 0.153 0.996 0.208 0.206 0.3141 Lag 0.311 0.08 <0.001 *** 0.006 0.038 0.8682 Lags 0.129 0.085 0.129 0.025 0.047 0.5913 Lags 0.202 0.071 0.005 ** 0.046 0.043 0.2934 Lags 0.066 0.073 0.371 0.018 0.043 0.6775 Lags 0.254 0.080 0.002 ** -0.068 0.043 0.1176 Lags -0.047 0.079 0.550 0.004 0.039 0.9117 Lags 0.070 0.086 0.417 0.077 0.052 0.1398 Lags 0.063 0.074 0.397 -0.004 0.032 0.9029 Lags -0.126 0.090 0.161 0.033 0.042 0.44010 Lags 0.088 0.086 0.307 -0.004 0.050 0.93411 Lags -0.067 0.094 0.480 -0.063 0.040 0.11812 Lags 0.096 0.069 0.164 0.134 0.048 0.006 **Cumulative 1.038 0.303 0.001 *** 0.411 0.34 0.227
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.412Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 0.697 0.499
Table A.2: Drilling Regression Results–Federal Onshore Oil Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price -0.151 0.409 0.712 -0.283 0.265 0.2861 Lag 0.545 0.231 0.019 * 0.113 0.097 0.2472 Lags 0.256 0.268 0.340 0.115 0.085 0.1773 Lags 0.080 0.201 0.692 0.0004 0.112 0.9974 Lags -0.152 0.163 0.350 0.099 0.104 0.3415 Lags 0.114 0.188 0.546 0.118 0.140 0.4026 Lags 0.045 0.158 0.776 -0.121 0.130 0.3527 Lags 0.230 0.186 0.217 -0.125 0.138 0.3678 Lags -0.078 0.192 0.685 0.181 0.088 0.040 *9 Lags 0.103 0.223 0.645 -0.109 0.111 0.32510 Lags -0.020 0.162 0.902 -0.049 0.125 0.69311 Lags 0.047 0.193 0.808 0.242 0.083 0.004 **12 Lags -0.086 0.172 0.617 -0.174 0.153 0.257Cumulative 0.931 0.52 0.074 . 0.006 0.504 0.991
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.243Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 0.158 0.854
56
Table A.3: Drilling Regression Results–Nonfederal Onshore Gas Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.195 0.189 0.301 0.022 0.251 0.9321 Lag -0.004 0.075 0.955 0.168 0.045 <0.001 ***2 Lags 0.172 0.114 0.133 0.070 0.057 0.2213 Lags 0.058 0.097 0.551 0.067 0.051 0.1944 Lags 0.099 0.085 0.244 0.041 0.040 0.3095 Lags 0.015 0.086 0.865 0.048 0.050 0.3416 Lags 0.027 0.115 0.817 0.029 0.058 0.6217 Lags -0.078 0.114 0.493 0.004 0.066 0.9528 Lags 0.019 0.100 0.853 0.123 0.059 0.036 *9 Lags 0.062 0.112 0.580 -0.038 0.056 0.49210 Lags -0.053 0.116 0.648 0.087 0.076 0.25011 Lags -0.131 0.125 0.295 0.019 0.057 0.74412 Lags -0.019 0.069 0.78 0.050 0.076 0.515Cumulative 0.360 0.351 0.306 0.688 0.461 0.137
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.276Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 0.272 0.762
Table A.4: Drilling Regression Results–Federal Onshore Gas Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.194 0.551 0.726 0.504 0.445 0.2591 Lag -0.029 0.261 0.911 0.151 0.173 0.3822 Lags 0.225 0.272 0.409 -0.039 0.170 0.8183 Lags -0.486 0.211 0.022 * 0.245 0.160 0.1264 Lags 0.230 0.268 0.390 -0.321 0.145 0.028 *5 Lags 0.273 0.356 0.443 0.278 0.162 0.087 .6 Lags 0.286 0.369 0.440 -0.116 0.148 0.4337 Lags -0.641 0.358 0.074 . 0.212 0.131 0.1078 Lags 0.559 0.224 0.013 * 0.053 0.177 0.7669 Lags -0.206 0.336 0.541 0.206 0.204 0.31410 Lags -0.042 0.312 0.893 -0.159 0.189 0.40311 Lags -0.373 0.329 0.258 0.036 0.166 0.82712 Lags 0.371 0.276 0.180 0.180 0.165 0.276Cumulative 0.360 0.857 0.675 1.231 0.753 0.103
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.240Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 0.467 0.627
57
Table A.5: Drilling Regression Results–Nonfederal Offshore Oil Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.727 1.551 0.640 0.074 0.641 0.9081 Lag -0.530 0.639 0.407 0.184 0.329 0.5762 Lags 0.560 0.599 0.351 0.099 0.299 0.7413 Lags -0.007 0.523 0.990 0.127 0.314 0.6864 Lags 0.029 0.414 0.944 0.090 0.362 0.8035 Lags -0.523 0.581 0.369 0.112 0.334 0.7396 Lags 0.818 0.630 0.195 -0.029 0.269 0.9147 Lags 0.356 0.517 0.491 0.044 0.300 0.8838 Lags -0.629 0.53 0.236 -0.244 0.250 0.3299 Lags 0.561 0.554 0.312 0.358 0.309 0.24810 Lags -0.931 0.591 0.117 -0.007 0.389 0.98611 Lags -0.056 0.532 0.916 0.254 0.351 0.46812 Lags 0.219 0.477 0.646 0.005 0.401 0.990Cumulative 0.595 1.638 0.716 1.069 1.017 0.294
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.103Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 0.523 0.594
Table A.6: Drilling Regression Results–Federal Offshore Oil Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.382 0.661 0.564 -0.020 0.344 0.9541 Lag 0.072 0.447 0.872 0.142 0.153 0.3512 Lags 0.072 0.322 0.823 0.085 0.178 0.6343 Lags -0.253 0.246 0.305 0.309 0.194 0.1134 Lags 0.647 0.310 0.038 * -0.425 0.179 0.018 *5 Lags -0.255 0.365 0.485 0.008 0.126 0.9486 Lags 0.218 0.283 0.442 0.028 0.139 0.8447 Lags 0.125 0.269 0.643 -0.146 0.151 0.3368 Lags 0.287 0.318 0.366 0.223 0.189 0.2399 Lags -0.885 0.416 0.034 * 0.094 0.191 0.62510 Lags 0.368 0.304 0.227 -0.06 0.276 0.82711 Lags 0.184 0.330 0.578 0.029 0.229 0.89912 Lags -0.487 0.254 0.056 . -0.094 0.178 0.596Cumulative 0.476 0.611 0.437 0.172 0.565 0.762
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.109Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 1.33 0.266
58
Table A.7: Drilling Regression Results–Nonfederal Offshore Gas Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price -2.167 0.964 0.025 * 0.644 0.812 0.4281 Lag 1.664 0.582 0.005 ** 0.049 0.344 0.8862 Lags -1.175 0.575 0.042 * -0.174 0.340 0.6093 Lags 0.443 0.531 0.405 0.593 0.261 0.024 *4 Lags -0.336 0.617 0.587 0.154 0.291 0.5965 Lags 0.645 0.597 0.281 -0.348 0.364 0.3406 Lags -0.355 0.502 0.480 0.203 0.296 0.4947 Lags -0.423 0.613 0.490 -0.110 0.323 0.7358 Lags 0.150 0.640 0.815 0.354 0.282 0.2119 Lags -0.580 0.609 0.342 -0.055 0.295 0.85110 Lags -0.518 0.519 0.319 0.576 0.286 0.045 *11 Lags 0.863 0.527 0.103 -0.792 0.288 0.006 **12 Lags -0.505 0.537 0.348 0.777 0.331 0.020 *Cumulative -2.295 1.275 0.073 . 1.872 1.266 0.141
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.062Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 3.006 0.051
Table A.8: Drilling Regression Results–Federal Offshore Gas Wells
Dependent Variable: ∆log(Wells Drilled)Variable ∆Log(WTI) ∆Log(Henry Hub)
Estimate Std. Error Pr(>|t|) Estimate Std. Error Pr(>|t|)Current Price 0.084 0.997 0.933 0.669 0.524 0.2031 Lag 0.319 0.518 0.539 0.359 0.264 0.1752 Lags -0.815 0.482 0.092 . -0.171 0.234 0.4643 Lags 0.174 0.464 0.709 0.549 0.305 0.073 .4 Lags 0.611 0.563 0.279 -0.146 0.327 0.6565 Lags -0.418 0.512 0.414 -0.137 0.276 0.6206 Lags -0.098 0.435 0.823 -0.059 0.203 0.7707 Lags 0.283 0.481 0.557 0.205 0.259 0.4298 Lags 0.528 0.380 0.166 0.248 0.317 0.4349 Lags -0.775 0.396 0.051 . -0.079 0.290 0.78610 Lags -0.283 0.435 0.515 0.078 0.231 0.73411 Lags -0.188 0.456 0.681 0.041 0.249 0.86912 Lags 0.601 0.481 0.212 -0.104 0.239 0.664Cumulative 0.023 1.093 0.983 1.453 0.843 0.086 .
*** p < 0.001, ** p < 0.01, * p < 0.05, . p < 0.1.Notes: Regression includes month of year fixed effects. Instruments are copper prices and twice-lagged Henry Hub price level. Standard errors are Newey West.
Diagnostics Statistic p-valueR-Squared 0.100Number of Observations (Months) 325Weak Instruments (Log(WTI)) 14.774 <0.001 ***Weak Instruments (Log(Henry Hub)) 5.054 0.007 **Wu-Hausman 2.017 0.135
59
C Integrating US Supply Model with IEA WEO
Projections
This section explains the details of integrating the US supply model with the WEO pro-
jections and adjustments to the WEO projections to account for unprecedented events
related to COVID-19 and the Russia-Saudi oil price war that were not predicted by the
WEO.
First, some technical adjustments must be made to integrate and harmonize the
outputs of the US supply module with the values reported in the WEO projections. First,
I account for regional differences in oil and gas prices. I assume that ROW prices (Brent)
follow a fixed percentage premium on WTI equal to the observed percentage spread
between Brent WTI futures prices. This varies slightly over time, but it averages about
11 percent over the simulation horizon because the Brent/WTI spread has historically
been a stable percentage of the WTI price.
I set ROW natural gas prices equal to the Henry Hub price plus a spread that is
set equal to the simple average of the EU–US and China–US gas price spreads in the
WEO projections. This also varies slightly over time, but it averages about $5 per
million British thermal units (mmbtu) of the simulation horizon. I use a fixed, rather
than percentage, spread to reflect the reality that gas price spreads primarily owe to
liquefaction and transportation costs that do not scale with the value of the gas itself.
Lastly, the US model outputs gas supply in terms of gross withdrawals, whereas WEO’s
projections reflect marketed gas production.52 According to EIA data,53 88.4 percent of
gross gas withdrawals were marketed in 2018, so I convert US gross gas withdrawals to
marketed gas production by multiplying by a factor of 0.884.
Second, the 2019 WEO projections assumed much higher oil and gas prices than
have actually been realized in 2020, due to unprecedented recent events. As mentioned
in the text, I use baseline oil and gas prices based on observed futures strips in June
2020, when near-term WTI prices were around $40 per barrel and Henry Hub prices were
around $2 per mmbtu. This is far lower than price assumptions in the 2019 WEO model:
near-term crude oil prices around $70–80 per barrel and Henry Hub gas prices around
$3 per mmbtu. This difference owes to the WEO modelers understandably failing to
predict the demand reduction caused by the COVID-19 crisis and the surge in global
oil supply brought on by the Russia-Saudi oil price war. For these reasons, true oil and
52Marketed production is gross withdrawal minus gas that is vented, flared, used for repressurization,and nonhydrocarbon gases removed during processing.
53See https://www.eia.gov/dnav/ng/hist/n9050us2m.htm and https://www.eia.gov/dnav/ng/
hist/n9010us2m.htm.
60
gas demand has fallen well short of WEO’s 2019 projections, true US supply has fallen
short of WEO’s projections (which assumed higher oil and gas prices), and true ROW
supply has exceeded those projections (which did not predict the increase in Russian and
Saudi supply). Nonetheless, WEO 2019 is still the most up-to-date long-run projection
of global oil and gas demand and ROW supply.
To employ these projections and still reflect recent events, I make some adjustments
to the global demand and ROW supply projections via two steps.54 In the first step,
I run the simulation model with the raw, unadjusted WEO demand and ROW supply
projections, but I include the observed dramatic decline in oil and gas prices and permit
my US supply model to respond to this decline. In this naive simulation, quantity
supplied is no longer equal to quantity demanded, and the modeled “excess demand”
represents the combined effects of ignoring the recent shocks to global demand and ROW
supply. Because it is unclear what fraction of this net excess is attributable to declining
demand associated with COVID-19 and what fraction is associated with global supply
shocks, I simply apportion it equally. That is, in the second step, I shift global demand
in the WEO projections downward by half of the modeled excess oil and gas demand
and adjust WEO’s ROW oil and gas supply upward by the other half.
The result of this adjustment is a market that is in equilibrium, with global supply
equal to global demand for both oil and gas at the baseline oil and gas prices based
on observed futures markets. Although this 50/50 adjustment is admittedly an ad hoc
assumption, it is required to account for unprecedented recent events. This adjustment
also has very little effect on the responses of US production to US policy reforms, which
are primarily driven by the estimates of the US supply response.
54My US supply model explicitly estimates the response of US oil and gas supply to the new priceenvironment, so no additional adjustment is needed there.
61
D Supplemental Simulation Figures
D.1 Baseline Price Projections
2020 2030 2040 2050
020
4060
8010
0
Oil
Pric
e ($
/bar
rel)
HistoricalPrices
FuturesPrices
ProjectedFuturesPrices
02
46
810
Gas
Pric
e ($
/mm
btu)
WTI (left)Henry Hub (right)
Figure A.2: Baseline Oil (WTI) and Gas (Henry Hub) Prices (2020$)
D.2 Emissions Impacts over Time
Figures A.3 and A.4 show the change in emissions over time by source: US federal, US
nonfederal, ROW, and total. The Hotelling-style price solution mechanism leads to a
smoothed reduction in emissions that is relatively flat over the time horizon, because
the percentage increase in prices is the same across the time horizon, which evenly
spreads the demand reductions across periods. This has the interesting effect of reducing
emissions before the policy is fully phased in through an announcement effect. For
example, an announced gradual phaseout of federal drilling (i.e., a leasing moratorium)
reduces expected future supply, which leads to an immediate rise in prices due to a
standard no-arbitrage condition, which in turn reduces consumption immediately. As
a result, global emissions reductions can exceed the US emissions reductions in the few
years after the policy is announced but before it is fully phased in.
62
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
18.7
5% O
nsho
re R
oyal
ty R
ate
(RR
)
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
25%
Ons
hore
RR
MMTCO2e per year
Glo
bal
US
Fed
eral
US
Non
fede
ral
RO
W
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
25%
RR
, Ons
hore
and
Offs
hore
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
$50
Car
bon
Add
er (
2%)
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
$50
Car
bon
Add
er (
2%)
+ 2
5% R
RMMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
Mor
ator
ium
MMTCO2e per year
Fig
ure
A.3
:E
mis
sion
sE
ffec
ts,
Ove
rT
ime
(Bas
eE
last
icit
ySce
nar
io)
Notes:
“RO
W”
rep
rese
nts
glob
alch
ange
inem
issi
on
sm
inu
sch
an
ge
inem
issi
on
sass
oci
ate
dw
ith
oil
an
dgas
pro
du
ctio
nfr
om
US
sou
rces
.
63
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
18.7
5% O
nsho
re R
oyal
ty R
ate
(RR
)
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
25%
Ons
hore
RR
MMTCO2e per year
Glo
bal
US
Fed
eral
US
Non
fede
ral
RO
W
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
25%
RR
, Ons
hore
and
Offs
hore
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
$50
Car
bon
Add
er (
2%)
MMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
$50
Car
bon
Add
er (
2%)
+ 2
5% R
RMMTCO2e per year
2020
2025
2030
2035
2040
2045
2050
−600−400−2000200400
Mor
ator
ium
MMTCO2e per year
Fig
ure
A.4
:E
mis
sion
sE
ffec
ts,
Ove
rT
ime
(Hig
hE
last
icit
ySce
nar
io)
Notes:
“RO
W”
rep
rese
nts
glob
alch
ange
inem
issi
on
sm
inu
sch
an
ge
inem
issi
on
sass
oci
ate
dw
ith
oil
an
dgas
pro
du
ctio
nfr
om
US
sou
rces
.
64
D.3 Sensitivity: High Oil and Gas Prices
2020 2030 2040 2050
020
4060
8010
0
Oil
Pric
e ($
/bar
rel)
02
46
810
Gas
Pric
e ($
/mm
btu)
WTI (left)Henry Hub (right)
Figure A.5: High Oil (WTI) and Gas (Henry Hub) Prices (2020$) used in SensitivityCase
65
Tab
leA
.9:
Pol
icy
Impac
tson
Oil
and
Gas
Pri
ces,
CO
2e
Em
issi
ons,
and
Roy
alty
&C
arb
onR
even
ues
(Annual
Ave
rage
s20
20–2
050)
,B
asel
ine
vers
us
Hig
hO
ilan
dG
asP
rice
Sen
siti
vit
y
Em
issi
ons
Ch
ange
(∆M
MT
CO
2e/
year
),P
rice
2020
-50
Ave
rage
Roy
alty
Ch
ange
(%)
US
Lea
kage
&C
arb
onO
ilG
asF
eder
alN
onfe
der
alR
OW
Glo
bal
rate
Rev
enu
e(∆
$b)
Bas
eO
ilan
dG
asP
rice
sin
Fig
ure
A.2
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
18.7
5%O
nsh
ore
Roy
alty
Rat
e(R
R)
0.1%
0.2%
-16
48
-473
.8%
$1.1
25%
RR
,O
nsh
ore
only
0.1%
0.3%
-31
815
-873
.8%
$2.0
25%
RR
,O
nsh
ore
and
Off
shor
e0.
2%0.
3%-3
79
18-1
073
.4%
$3.0
$50
Car
bon
Ad
der
(ris
ing
at2%
)1.
2%1.
6%-2
1652
106
-58
73.2
%$7
.1$5
0C
arb
onA
dd
er(2
%)
&25
%R
R1.
3%1.
7%-2
3857
117
-64
73.1
%$6
.5M
orat
oriu
m1.
9%1.
9%-3
1473
155
-85
72.9
%-$
5.5
Hig
hO
ilan
dG
asP
rice
sin
Fig
ure
A.5
(IE
A)
18.7
5%O
nsh
ore
RR
0.1%
0.2%
-24
810
-676
.9%
$2.4
25%
On
shor
eR
R0.
2%0.
4%-4
716
20-1
176
.9%
$4.3
25%
RR
,O
nsh
ore
and
Off
shor
e0.
2%0.
4%-5
619
24-1
376
.6%
$6.2
$50
Car
bon
Ad
der
(ris
ing
at2%
)1.
1%1.
8%-2
7192
115
-64
76.5
%$1
2.9
$50
Car
bon
Ad
der
(2%
)+
25%
RR
1.3%
1.9%
-303
103
129
-71
76.4
%$1
3.6
Mor
ator
ium
2.4%
2.3%
-460
154
196
-110
76.1
%-$
11.1
Not
es:
RO
W=
rest
ofw
orld
,R
R=
roya
lty
rate
.A
llva
lues
are
rela
tive
toth
eb
usi
nes
s-as
-usu
alsc
enar
io.
All
pol
icie
sex
cep
tfo
rth
efi
rst
two
app
lyb
oth
onsh
ore
and
offsh
ore
wel
ls.
Th
eca
rbon
add
ergr
ows
at2
per
cent
annu
ally
,in
lin
ew
ith
IWG
esti
mat
es.
Oil
and
gas
pri
cep
erce
nta
gech
ange
sar
ere
lati
veto
WT
Ian
dH
enry
Hu
bre
spec
tive
ly.
Col
um
n(6
)eq
ual
sth
esu
mof
colu
mn
s(3
),(4
),an
d(5
).C
olu
mn
(7)
equ
als
[(4)
+(5
)]/|
(3)|.
Bas
elin
ed
eman
del
asti
citi
esar
eu
sed
.
66
Resources for the Future iii