A few weeks ago, I released a new working paper on the impact of ending new oil and gas leasing on federal lands. Just a week prior to that, the American Petroleum Institute (API) had released the results of another study about a similar topic. My study and API’s are based on different models and come to starkly different conclusions about the potential effects on carbon emissions. While I estimate that reducing federal oil and gas production through a leasing ban would decrease emissions, API’s study projects an increase in emissions.
In this blog post, I drill into the details of API’s analysis to try to reconcile these very different findings. In short, API’s counterintuitive findings stem from a series of questionable assumptions that lead them to overestimate the effects that a federal leasing ban would have on oil and gas production and carbon emissions.
Under existing law, the US Department of the Interior issues leases to private companies to produce oil and gas located on federal lands and waters. In exchange, the government collects royalties on the revenues from oil and gas produced on that land. Emissions associated with that production has increasingly attracted policymakers’ attention. A bill to ban new leases was introduced in the House last year, and the Biden campaign has also endorsed the policy. A leasing ban would mean that the government would cease issuing new leases, although companies would still be allowed to drill and produce from wells on existing leases.
Because a leasing ban is emerging as an increasingly popular climate policy, my recent paper estimates the impacts that such a policy would have on greenhouse gas emissions. To make these estimates, I custom built a detailed economic model of the US upstream oil and gas industry. The methodology underlying this paper is based on my past peer-reviewed academic literature on the US oil and gas supply, but also includes new features to differentiate between the economics of federal versus nonfederal oil and gas production.
The model accounts for a number of important features of current policy proposals, including the potential for “leakage,” whereby reductions in federal oil and gas production could be offset by increases elsewhere. It also accounts for time lags between changes in leasing policy and realized oil and gas production to capture the fact that production from existing federal leases would be untouched by an end to new leases. The model also accounts for the dramatic drop in oil prices that occurred this year, which has led to a large contraction in US oil supply absent any policy change. In short, my model is purpose-built for the particular question it answers.
Ultimately, I estimate that a ban on new federal leasing could reduce global greenhouse gas emissions by 85 to 147 million tons of CO₂ equivalent per year on average, even after accounting for the likelihood that production on state and private lands and abroad would increase as a result of a leasing ban. By contrast, API estimates a 58-million-ton increase in CO₂ emissions.
Why are the results so different? It is immediately difficult to know for sure because the PowerPoint slides API provided to summarize the report’s findings are just that—slides—and are thus light on many important details. As my colleague at Resources for the Future, Marc Hafstead, noted when describing a different study, the public simply cannot know how credible the results are, without a full study that lays out its assumptions in detail. So, I dug into the information that API made public, and then followed up with API staff themselves to get further clarity on what exactly their study assumed.
To perform its analysis, API contracted with OnLocation, a consulting firm with longstanding experience in running the in-house model used by the Energy Information Administration (EIA). API’s study took that model, known as the National Energy Modeling System (NEMS), modified it, and ran it assuming an end to new federal oil and gas leasing and development (not an end to just leasing alone, an important distinction that I return to below). The results then found an increase in US carbon emissions, driven by an unlikely comeback of coal in the power sector.
Ultimately, I estimate that a ban on new federal leasing could reduce global greenhouse gas emissions by 85 to 147 million tons of CO₂ equivalent per year on average.Brian Prest
This estimated increase in carbon emissions is surprising, because it differs from EIA’s own analysis that uses the same model as API, NEMS. In particular, EIA’s 2020 Annual Energy Outlook includes “side cases” in which it assumes lower US oil and gas supply in NEMS. While NEMS is not specifically designed to model federal versus nonfederal oil and gas production in the detailed way that my model can, the effects of reducing oil and gas supply in the standard NEMS model can nonetheless be informative as a comparison to API’s modified version of NEMS.
EIA’s results reveal that cutting the oil and gas supply in NEMS reduces emissions, both in the power sector and in the economy as a whole. EIA’s estimated reductions in the power sector occur because reduced gas supply gets replaced increasingly by renewables, rather than by coal. Similarly, the peer-reviewed academic literature has consistently ruled out the possibility that decreased natural gas supply would substantially increase power-sector emissions, generally finding that reduced gas supply has no substantial effect on emissions—or could even decrease emissions.
So, how can we explain emissions increases in API’s modified version of NEMS, versus emissions decreases in both my model and EIA’s core version of NEMS? My investigation has revealed several highly questionable assumptions in API’s analysis that lead to its finding of an increase in emissions.
Saving the Coal Industry, by Assumption?
The first questionable assumption in API’s analysis is that coal-fired power plants currently planning to retire will suddenly decide to scrap those plans and go back to the drawing board, but only in the “policy case” where the federal ban on new leases is implemented. In the “reference case,” in which no leasing ban is implemented, those plants are assumed to close on schedule, regardless of the economics. (For reference, EIA’s most recent projection using the NEMS model estimates that nearly half of the current coal capacity will retire by 2030.) This modification of NEMS results in higher estimates of coal plants online under a federal leasing ban largely by assumption, which leads to more emissions in its model.
But those coal plants have already announced their retirement dates. Further, since API’s analysis lets those coal plants consider reopening only under a federal leasing ban—but not in the alternative scenario—the estimated increase in emissions reflects both the impacts of the leasing ban and the assumption that those coal retirements will not actually happen as planned. Because the comparison is not ceteris paribus, there is no way to know for sure the extent to which API’s estimates are driven by this assumption or the underlying economics. By contrast, EIA’s modeling correctly conducts a ceteris paribus analysis and comes to the opposite conclusion. One way or the other, accurately measuring the effect of a leasing ban requires making the same assumption in both the reference and policy cases—not picking and choosing.
API may counter that its analysis predicts gas prices would increase by 4.5 percent under a leasing ban, which then would prompt many coal plants to renege on their retirement plans. For context, a 4.5 percent change in the price of gas is equivalent to the price changing from $2.00 to $2.09 per unit energy (million British thermal units, or MMBtu)—a relatively small shift, given the typical fluctuation in gas prices. Natural gas prices go up or down by more than 4 percent every day, on average. A 4.5 percent increase in gas prices certainly could reduce gas consumption on the margin somewhat, but the change is not big enough to justify the assumption that all coal plants that have announced retirements will suddenly scrap those plans.
It’s worth noting that this assumption isn’t just arbitrary—it’s also contrary to repeated underestimates by NEMS of coal plant retirements and the growth of renewables. In 2010, NEMS projected that coal capacity would grow. In fact, coal capacity shrunk by a third. In 2015, NEMS projected coal capacity would stabilize at 260 gigawatts (GW) in 2020, and then the retirements would stop. In reality, many more coal plants retired. Now, in 2020, NEMS projects that coal plant retirements will slow in 2025 and capacity will stabilize at 130 GW. In short, NEMS consistently under predicts coal plant retirements to begin with. Instead of correcting this bias, the API study amplifies it.
The API study also appears to assume that a leasing ban would prompt no net change in emissions associated with oil. Nearly all of API’s environmental results are focused only on the power sector, where it estimates an increase in emissions.
But the graphs indicate that total emissions appear to increase by the same amount as the increase in power sector emissions. The only way this math works is if API finds no net change in emissions from all sectors outside of electricity generation—including emissions from changes in oil consumption, a major driver of emissions, along with emissions from natural gas use in the industrial, residential, and commercial sectors. And on federal lands, oil production currently exceeds gas production in terms of both energy and emissions, so finding no change in emissions from oil is puzzling. Indeed, in my model, emissions reductions from oil amount to about two-thirds of the total impacts.
Although API does not present the effects of a leasing ban on global oil consumption and related emissions, we can approximate the effects through a back-of-the-envelope calculation using standard economic tools. API’s study reports an approximately 2.5 percent average increase in the price of oil. The academic literature provides estimates of the price elasticity of oil demand ranging from -0.2 to -0.5 (see, for example, here, here, and here). That is, for each 1 percent increase in the price of oil, we would expect a reduction in demand in the range of 0.2 percent to 0.5 percent. Applied to API’s estimated 2.5 percent price increase, these estimates suggest a reduction in global oil demand of about 0.5 percent to 1.25 percent. Global crude oil consumption is typically around 100 million barrels per day (MMb/d), which would imply reductions in oil consumption of 0.5 to 1.25 MMb/d in this case. At an emissions factor of 0.43 tons CO₂ per barrel, these results imply that emissions reductions from oil alone in API’s model would amount to 78 to 196 million tons of CO₂ annually, which would swamp even API’s (problematic) estimated 58-million-ton increase for the power sector. Yet, crucially, this increase in emissions does not appear in API’s results.
Timing is Everything
API’s analysis also overstates the impacts of an end to new federal leasing because it does not, in fact, model an end to new leasing. While API’s tables and figures repeatedly suggest that the results reflect a policy of “No Federal leasing,” API in fact models an end to both new leasing and development—which is very different than a ban on leasing alone. Even a more aggressive policy of ceasing to issue new drilling permits (rather than leases) would not result in an end to development. For example, many companies operating on federal lands report already having enough drilling permits to last years, even if no new ones are issued. Yet API’s analysis assumes an even stronger policy: an immediate end to all drilling on federal lands and waters, including on existing leases.
In reality, the government can’t simply end development on existing leases by decree. Indeed, at least one oil industry CEO operating on federal lands thinks such a policy would be extremely unlikely. A substantial time lag would occur before any change in leasing policy would have any substantial economic impact, because companies would continue drilling on leases they already hold. The Trump administration issued nearly 5,000 new federal oil and gas leases covering nearly 10 million acres between 2017 and 2019. As oil and gas leases typically last 10 years (and can be further extended beyond that), companies holding those already issued leases can develop them as late as 2029, even if new leasing ended today. Since oil and gas companies typically do wait until near the end of that 10-year period before producing from those leases, any change in leasing policy today wouldn’t have much effect on federal oil and gas production until nearly a decade into the future.
Because API’s analysis assumes an immediate end to drilling even on existing leases, it calculates large, immediate reductions in oil and gas production. These estimated reductions in production for 2025 are more than ten times larger than my estimates, and even by 2030, they are still too high by a factor of three. This accelerated policy implementation could also explain why API estimates large job losses as soon as 2021, but only minimal impacts on unemployment beyond 2024, after the economy has had time to adjust. (Somewhat surprisingly, API finds that a shutdown of federal oil and gas development would slightly reduce the unemployment rate after 2025.) In reality, the largest effects of a leasing ban mostly would not be felt until after the end of this decade, which would give the economy more time to adjust than API’s modeling suggests.
Another odd result in the API analysis is that it projects that oil and gas production will either decline or remain unchanged across all but one of the production categories it reports. (The one exception is the small enhanced oil recovery category, which is projected to increase very slightly from 0.45 to 0.46 MMb/d.) For example, API projects that tight oil production, which is predominantly located on private land, will decline by 0.9 MMb/d by 2030. (This result is particularly implausible, given that total onshore federal production today is less than 0.8 MMb/d, and due to low oil prices, my model projects it to remain below 0.9 MMb/d through 2030, even without a leasing ban.) Similarly, shale and tight gas, which is also predominantly located on state and private land, is projected to decline by a combined 1.3 trillion cubic feet. But economics tells us to expect the opposite result—that production on state and private land would rise as operators shift their focus away from federal lands and toward other sources, particularly if higher oil and gas prices make drilling elsewhere more profitable.
The NEMS model includes price-responsive supply, so it’s surprising that this leakage effect is not apparent in API’s results. One reason for this surprising result could be that the implicit supply elasticities in NEMS do not reflect the rise of the shale boom, which has increased the price responsiveness of oil and gas supply. My model explicitly captures this phenomenon, finding that nearly 25 percent of reduced federal production would be replaced by increased production elsewhere in the United States.
Putting all these observations together, we can see that API’s analysis involves a number of features that appear to conflict with the standard version of NEMS used by EIA.
- It assumes that coal plants currently scheduled to retire will suddenly reconsider, if—and only if—the federal government imposes a leasing ban.
- It appears to predict no net additional emissions from oil, despite the fact that oil contributes the highest proportion of emissions from oil and gas production on federal lands.
- It models the impacts of an end to all new leasing and development, even though policymakers are considering an end to new leases only.
- It features no apparent offsetting increases in oil and gas production on nonfederal land.
As a result, API’s projections of US oil and gas production highly overstate the effect of a federal leasing ban, particularly in its assessments of macroeconomic impacts over the next few years. Further, the effects on carbon emissions run contrary to all conventional estimates; indeed, EIA’s own analysis earlier this year, which also uses NEMS to model an analogous reduction in oil and gas supply, finds that emissions would decrease. In summary, the API study does not reliably indicate the effects of a proposed federal leasing ban on the industry, nor on the economy as a whole.