Air Quality Policy Outcomes at U.S. Coal Power Plants

This dissertation investigates the effects of SO2 regulations on U.S. coal power emissions, focusing on how policy implementation, plant operator behavior, and electricity market structure interact with and influence these effects. Each study leverages high resolution public data on coal power plant operations and econometric methods to estimate the effect of air quality policy on its first-order targets—polluters—and later chapters model the associated health impacts. In doing so, this research offers insight into how future policy might be designed consciously around pre-existing policy and market incentives to most effectively reduce pollution in the context of SO₂ and beyond.

Chapter 2 estimates how the implementation process of the 2010 SO₂ National Ambient Air Quality Standards (NAAQS) revision affected emissions from coal-fired power plants. The goal of the revision was to respond to new research showing the acute health impacts of SO₂ by replacing preexisting limits on annual and daily SO₂ concentrations with a peak hourly limit. This analysis considers not only whether the NAAQS was associated with SO₂ emissions reductions, but whether reductions were greater at peak emissions hours relative to median hours. Using data on electricity generating unit (EGU) characteristics, SO₂ emissions, fuel prices, and PM_2.5 NAAQS classifications from 2001-2019 from public data sources, we estimate that, after a county was classified under the 2010 SO2 standard, EGUs were found to have reduced SO2 emissions during daily maximum hours by 31.6% (95% CI [−0.381,−0.247]) at the 99th percentile and 34.4% (95% CI [−0.424,−0.253]) at the 50th percentile. After a nearby ambient SO₂ monitor was added to assess NAAQS compliance, hourly emissions fell by 12.4% (95% CI [−0.188,−0.054]) at the 99th percentile and 14.4% (95% CI [−0.228,−0.050]) at the 50th percentile. Furthermore, plants in counties prioritized for early designation were found to be less responsive to the policy change. These results suggest that the 2010 SO₂ NAAQS change may have reduced emissions, but may not have had an outsized impact on peak emissions as originally intended.

Chapter 3 describes how generator decisions can cause actual SO₂ emissions to deviate from predicted emissions when installing SO₂ control technologies as a response to policy. Predictive policy models often assume that plants will continue generating SO₂ at the same rate after scrubber installation, and will remove SO2 equivalent to the equipment’s rated removal rate. Meanwhile, plants often have financial incentives to (a) burn cheaper, higher sulfur coal after installing SO₂ scrubbers, and (b) operate outside of the standard conditions under which pollution control equipment is rated. This study compares the effects of each behavior by decomposing SO2 emissions as a function of coal sulfur content and SO₂ removal efficiency across plants from 2008 to 2021. Coal plants that installed SO2 controls over this time period are found to have increased their rate of SO₂ generation from coal by 58% after installation. Furthermore, total SO₂ tonnage captured was 2.4% lower than predicted from the rated removal rates, resulting in 740 additional deaths per year on average as estimated by the Air Pollution Emissions Experiments and Policy version 4 (AP4) model.

Chapter 4 considers how electricity market structure affects the operation of SO₂ control technologies and how this effect interacts with technology-forcing policy. U.S. electricity market restructuring in the 1990s shifted some firms from the regulated utility model with a rate of return on generating assets to the nonregulated model in which capital costs must be recovered through electricity markets. This study tests whether nonregulated units have lower rates of SO2 control adoption and SO₂ removal, and whether the effect of market type changes in the presence of a New Source Review (NSR) consent decree mandating regulation under the New Source Performance Standards (NSPS). While nonregulated EGUs are not found to have statistically significantly different SO₂ control installation rates than regulated units, nonregulated EGUs do remove significantly less SO₂, removing 9.94 percentage points less than comparable regulated units. Nonregulated units under NSR consent decree, however, achieve an average removal rate that is only 1.4 percentage points lower than the regulated removal rate, while an active consent decree does not impact regulated unit performance. These results demonstrate that electricity market deregulation may have exacerbated negative externalities through reduced SO₂ removal rates, but this externality might have been partially addressed through strict mandates control technology operation.