The roles of energy markets and environmental regulation in reducing coal‐fired plant profits and electricity sector emissions

• Date: 01 December 2019
• Published date: 01 December 2019
• Author: Kristen McCormack,Joshua Linn
• DOI: http://doi.org/10.1111/1756-2171.12294

RAND Journal of Economics

Vol.50, No. 4, Winter 2019

pp. 733–767

The roles of energy markets and

environmental regulation in reducing

coal-fired plant profits and electricity sector

emissions

Joshua Linn∗

and

Kristen McCormack∗∗

Between 2005 and 2015, US electricity sector emissions of nitrogen oxides and sulfur dioxide,

which harm human health and the environment,declined by two thirds, and many coal-fired power

plants became unprofitable and retired. Intense public controversy has focused on these changes,

but the literature has not identified their underlying causes. Using a new electricity sector model

of the US eastern interconnection that accurately reproduces unit operation, emissions, and

retirement, we find that electricity consumption and natural gas prices account for nearly all the

coal plant profitability declines and resulting retirements. Environmental regulations had little

effect on these outcomes.

1. Introduction

Electricity sector emissions of nitrogen oxides (NOx) harm human health and the envi-

ronment by raising ambient concentrations of ozone and particulates. The United States began

regulating electricity sector NOxemissions in the 1970s, and emissions declined gradually and

steadily from then until around 2000, after which emissions declined sharply. Between 2000

and 2015, emissions declined at a rate four times greater than between 1990 and 2000, and

emissions in 2015 were just one fifth of 1990 emissions. Coinciding with these changes are the

tightening stringency and broadening scope of NOxemissions caps that the US Environmental

Protection Agency (EPA) administers. Likewise, emissions of other pollutants, such as sulfur

dioxide, have declined dramatically since 2000. During the same period, many coal-fired plants

became unprofitable and about one third of coal-fired plants prepared to retire.

In the political debate over electricity sector policy, two views have emerged about the cause

of the decline in electricity sector emissions and the retirement of coal-fired plants. The first

∗University of Maryland and Resources for the Future; linn@rff.org.

∗∗Har vardUniversity; kmccormack@g.harvard.edu.

C2019, The RAND Corporation. 733

734 / THE RAND JOURNAL OF ECONOMICS

view credits technological innovation and pro-renewables policies for reducing costs of natural

gas-fired plants and renewables and causing a shift from coal to lower-emitting sources. Many

adherents of this view favor tightening emissions caps and other regulations in light of their

benefits and lower-than-expected compliance costs. The second view is that by raising the costs

of coal-fired power plants relative to other technologies, emissions regulations have excessively

harmed coal-fired plant profits, jobs, local communities, and the reliability of electricity supply.

Some adherents of this view call for weakening regulations to end the “war on coal.” These two

views crystalized during the 2016 presidential election, and favoring the second view, the Trump

administration has begun to weaken regulations on the coal sector.

The economics literature suggests that on the margin, lownatural gas prices reduce coal-fired

generation, but it does not provide direct evidence for the aggregate effects of natural gas prices

and other market forces and does not identify the causes of the retirements. Several recent articles

examine the statistical relationships among natural gas prices, wind generation, fossil fuel-fired

generation, and emissions (e.g., Cullen and Mansur, 2017; Holladay and LaRiviere, 2017; Fell

and Kaffine, 2018; Linn and Muehlenbachs, 2018; Johnsen, LaRiviere, and Wolff, forthcoming).

However, because of their reduced-form approach, these articles focus only on the short-run and

marginal effects of natural gas prices and wind generation; long-run responses may differ. The

long run includes entry and exit decisions that depend on fixed costs, whereas short-run responses

depend only on variable costs. For example, low gas prices may lead to an increase in natural

gas plant investment, potentially compounding the short-run effects on coal plant profits and

retirements. In addition, when emissions caps are binding, low natural gas prices may reduce

emissions credit prices, lowering costs for coal plants and opposing the short-run effects of gas

prices on coal plant profits. Thus, the long-run effects of market shocks may differ positively or

negatively fromt he short-run effects that havebeen the focus of the literature, leading to different

conclusions about the historical effects of these shocks.1

Moreover, although several studies have compared expected and realized costs of sulfur

dioxide emissions reductions under the Acid Rain Program (ARP) (e.g., Carlson et al., 2000;

Ellerman et al., 2000), the literature has not compared the effects of market forces with the effects

of regulation. Most electricity sector NOxemissions in the East are covered by EPA emissions

caps. When they were established, the EPA expected them to cost the sector at least $3 billion

per year (2005 dollars).2These costs fall largely on the coal-fired fleet (Linn, 2010), implying

roughly a 10% cost increase for those units, with most of the costs incurred by the older and higher-

emitting coal-fired units. The Mercury and Air Toxics Standards (MATS) were expected to cost

substantially more than the NOxcaps. Thus, environmental regulation may have substantially

reduced profitability of coal-fired units, leading to their retirement. Fowlie and Muller (2013)

analyze the costs of achieving NOxemissions caps, but they consider only the early portion of

the program and do not evaluate whether either viewof the electricity sector trends is cor rect. We

are not aware of any ex post analysis of MATS.

In this article, we use a newcomputational operational and investment model of the electricity

system and quantify the effects of market shocks and emissions regulations on emissions, profits,

and retirements of coal-fired plants. We focus on the eastern United States, which accounts for

about 90% of electricity sector NOxemissions.3

1Houser, Bordoff, and Marsters (2017) compare the effects of electricity consumption, natural gas prices, and

renewables on coal consumption between 2006 and 2016 and conclude that natural gas prices werethe most impor tant

factor, followed by electricity consumption. Theydo not analyze the effects of these factors on emissions or coal plant

profits. The Department of Energy (DOE) (2017) argues that natural gas prices are the most important factor explaining

coal plant retirements but provides little evidence.

2These costs represent a large share of overallestimated costs of federal environmental regulations of the electricity

sector. Between 2003 and 2015, the EPA implemented the emissions caps in three phases (see Section 2). The agency

reports costs of complying with each phase (EPA, 1998,2005, 2011). For the latter two phases, the costs are combined

with the costs of achieving the sulfur dioxide caps. In the main text, weuse only the cost estimate from EPA (1998).

3The United States has three major interconnections, across which there is little availabletransmission. Throughout

the article, East refers to the eastern interconnection, which spans the Great Plains to the East Coast (see Figure A1).

C

The RAND Corporation 2019.

LINN AND MCCORMACK / 735

The model includes 3500 generation units in the eastern United States and characterizes unit

construction, retirement, emissions abatement, and hourly operation. Weapproximate uncertainty

in consumption, uncertainty in unit availability, and constraints on unit operation by extending the

approach of Davis and Hausman (2016). The model accurately predicts observed hourlyoperation

and emissions and coal plant retirements. Weshow that a conventional economic dispatch model,

which is constructed using the same underlying data but omits these features, would overpredict

the effects of changing natural gas prices on coal-fired plants.

We model three market shocks: natural gas prices, renewables generation, and electricity

consumption. Largely because of the rise in production from shale formations, natural gas prices

in 2015 were 30% lower than the level projected in 2005. Improved wind generator performance

and subsidies caused wind generation in 2015 to be 10 times higher than had been expected.

Because of the 2008–2009 economic recession and other factors, 2015 electricity consumption

was 20% below 2005 expectations. For convenience, we refer to differences between the 2005

projections and the 2015 realized outcomes as energy market shocks, noting that policies have

contributed to them.

We model two environmental regulations: NOxemissions caps that were adopted between

2005 and 2015, and MATS.4The emissions caps require that 2015 emissions be about half of

2005 levels. MATS requires plants to reach specific emissions standards for mercury and other

pollutants. Using 2005 projections for electricity consumption, wind generation, and fuel prices,

we estimate that without shocks, NOxabatement costs would have been about $2.9 billion per

year, which roughly agrees with ex ante EPA assessments.5Note that we model explicitly the

compliance decisions for the NOxabatement caps, whereas we estimate MATS costs based on

observed decisions. We make this distinction because the caps applied throughout the period of

analysis, 2005 through 2015, whereas the initial compliance period of MATS falls near the end

of the period, after the market shocks occurred.

The market shocks explain 80% of the coal-fired plant retirements observed between 2005

and 2015.6After accounting for these shocks, the emissions caps had a small effect on coal-fired

plant profits and retirements. The three shocks collectively reduced regulatory costs from $2.9

billion to $0.4 billion per year (86%) and reduced coal-fired plant profits by 89%. We find that,

after accounting for the shocks, MATShad a small effect on retirements and profits. These results

confirm the first of the two views, that factors other than environmental regulation explain most

of the decline in the profits of coal-fired plants and the resulting retirements.

4We do not model sulfur dioxide emissions caps because emissions credit prices were close to zero in 2015,

indicating that the caps were not binding. The EPA’s Clean Power Plan, finalized in 2015, established carbon dioxide

emissions standards for fossil fuel-fired generators. Linn, Burtraw, and McCormack. (2016) see two reasons why it is

unlikely to have caused any coal plant retirements in 2015: (i) it would not have taken effect until 2022, and therefore

would not have affected power plant operational decisions in 2015, and (ii) compliance decisions for MATSwere made

prior to 2015. The only plants that would retire in 2015 because of the Clean Power Plan are those for which the firm

needed to make a life-extending investment in 2015, and which would have been unprofitable under the Clean Power

Plan. Our data do not appear to contain any such plants.

5We cannot compare our estimated costs directlywith EPA estimates. Although the agency reports costs of com-

plying with each of three regulatory phases, the costs are estimated relative to differentbaselines, making it inappropriate

to add the three cost estimates.

6There is some disagreement about retirements in the Energy Information Administration (EIA) and EPA data.

According to the Energy Information Administration (EIA), between 2005 and 2015, firms announced the retirements of

about 88 gigawatts (GW) of coal-fired plant capacity.Of this amount, about 20 GW did not operate in 2005, according

to EPA data. According to EPA data, about 41 gigawatts stopped operating by 2015. Much of the remaining 27 GW

of retirements in the EIA data appears to have stopped operating since 2015. Because we model market shocks and

regulation through 2015, we focus on units that stopped operating by 2015 and the percentage of retirements cited in the

text includes only those plants that stopped operating between2005 and 2015, according to EPA data. The model predicts

low profits for the units that stop operating after 2015, indicating that many of those units would be predicted to retire

after 2015 if we model market shocks after 2015.

C

The RAND Corporation 2019.