The US tax code provides nearly $1 billion annually in tax credits to subsidize the use of “refined” coal, which is supposed to reduce emissions of nitrogen oxides (NOₓ), sulfur dioxide (SO₂), and mercury (Hg). Such coal accounts for a rising share of coal consumption from the power sector, reaching 20 percent in 2017. To be eligible for the tax credit (currently set at $7 per ton, which is about 10–15 percent of the price of eastern coal and about 35 percent of the price of lignite), refiners must demonstrate that compared to standard coal, refined coal emits 20 percent less NOₓ and 40 percent less SO₂ or Hg per unit of thermal energy. Firms typically demonstrate eligibility through a laboratory test, which can diverge from actual operational conditions. We use boiler-level data on power plant emissions in a panel regression framework to estimate whether burning refined coal actually achieves the emission reductions required for eligibility. Our estimates suggest that plants do not achieve the purported emissions reduction targets in practice, implying a misuse of billions of dollars in taxpayer funds. A cost–benefit analysis backs up the inefficiency of this legislation as applied. Because the tax credit is up for reauthorization in 2021, our work has immediate policy relevance.