Coal plant closures cut local jobs and tax revenue sharply and persistently, but the damage concentrates in the affected sector and fiscal line item, not necessarily total county headcount

Summary

Research on coal mine and coal power plant closures consistently finds large, persistent local damage that grows over the first 5-10 years, even though the plant itself may employ only a few hundred people. The most rigorous evidence (a 2026 Philadelphia Fed working paper studying 81 Appalachian counties that lost coal mines in 2011-2016) finds payroll employment down 4% at year 5 and 6% at year 10, wages and salaries down 8% then 10%, county GDP down 14% then 17%, and the unemployment rate up almost a full point — with an estimated job-loss multiplier of about 2 (one additional job lost elsewhere in the local economy for every coal job lost) by year 10. On the tax-base side, case studies of power-plant closures show county/school property-tax revenue losses in the millions to tens of millions of dollars per year (e.g., roughly $8.5M/yr in Adams County, OH; up to an estimated $60M/yr for Navajo County, AZ from the Cholla plant), and a national NBER analysis estimates coal-reliant counties could see total local revenue fall on the order of 20% as coal disappears. An original check of BLS county employment data for Adams County, OH around its 2018 coal-plant closures shows the mechanism clearly: direct utility-sector jobs collapsed to a BLS-suppressed near-zero and only partially recovered, while total county employment across all industries stayed roughly flat and then grew — illustrating that the damage is concentrated in the closed sector and the local tax rolls, and can be invisible in a county's aggregate jobs count.

1. The best-identified estimate: Appalachian coal mine closures, 2011-2016 shock

Nason, Scavette & Stephens (Federal Reserve Bank of Philadelphia Working Paper 26-30, published June 2026) is the strongest causal design available: a difference-in-differences event study comparing 81 Appalachian counties that had at least one net coal-mine closure during the 2011-2016 shale-gas-driven coal bust against 39 similar coal counties that did not. Pre-trends are flat for a decade before the shock, supporting a causal read. Results (average effect on treated counties, relative to controls):

The paper's job-loss multiplier — non-coal jobs lost per coal-mining job lost — rises from 1.6 at 5 years to 2.0 at 10 years, roughly 50% higher than the 1.35 multiplier found for the 1980s coal bust in the same region (Black, McKinnish & Sanders 2005). The authors attribute the larger multiplier to a much bigger local coal wage premium: coal jobs paid 53% more than the local average in 1986 and 102% more by 2011, so each lost coal job removes far more local spending power than an average lost job would. Sectorally, the job losses fan out beyond mining itself into administrative/support services, transportation and warehousing, finance/insurance, and retail trade — i.e., the businesses that served mine payrolls. Effects are roughly 50% larger in counties where coal was a bigger share of employment before the shock (75th percentile exposure vs. the average county).

2. Coal-fired power plant closures specifically: the Appalachian Ohio case

Michaud, Jolley, Khalaf & Sandler (2023, Regional Science Policy & Practice) studied the closure of two Dayton Power & Light coal-fired plants (Killen and J.M. Stuart stations) and an associated training center in Adams County, Ohio, completed in 2018. They document: 370 direct jobs and roughly 761 additional (indirect/induced) jobs at risk regionally, and about $8.5 million per year in lost tangible-personal-property tax revenue to county government and school districts — because Ohio law exempts decommissioned power plants from the tangible personal property tax that previously applied. The study frames post-closure adjustment as needing a "multi-pronged recovery effort," since displaced power-plant workers moving into the next-best-matching local occupations (tourism, rural healthcare) face real wage cuts.

3. The tax-base mechanism: how big a hole, and how fast

Morris, Kaufman & Doshi (NBER Working Paper 27307, and companion University of Chicago EEPE chapter) analyze county finances across coal-reliant U.S. counties and find coal-related revenue (chiefly property and severance taxes) can fund a third or more of the budget in the most coal-dependent counties. Extrapolating their regression of 27 coal-reliant counties, they estimate the eventual disappearance of coal could lower these counties' own-source revenue by roughly 20% — and they explicitly note this likely understates the true hit, because it does not capture the secondary erosion of the broader tax base (falling home values, closed businesses) as the dominant local industry collapses.

Individual plant closures illustrate the range: a coal plant in Wise County, VA has historically provided about $8.5 million/year in property tax revenue (about 15% of the county budget); Navajo County, AZ was reported to be at risk of losing nearly $60 million in property-tax revenue as the Cholla coal plant fully closes, while neighboring San Juan County, NM draws about 10% of its property-tax revenue ($3.8M) from a coal mine and power plant. In West Virginia's Boone County, coal-company bankruptcies alone left $8 million in uncollected property tax in a single year (2015), forcing the closure of three elementary schools that year.

4. An original data check: does the county's total job count actually move?

To see the mechanism directly rather than relying only on published multipliers, we pulled BLS Quarterly Census of Employment and Wages (QCEW) data for Adams County, OH (area_fips 39001) around its 2018 coal-plant closures. Two series, indexed to 2018=100:

• Utility-sector employment (NAICS 2211, the plants themselves and directly related utility payroll): ran at roughly 104-123% of the 2018 level from 2013-2017, then collapsed after the 2018 closures — BLS suppresses the 2019-2020 cells as disclosure-risk (a near-single-employer cell), which itself confirms employment fell to a handful of workers — before a partial recovery to about 40-57 workers (roughly a third of pre-closure levels) by 2021-2025, likely decommissioning-related work.
• Total county employment, all industries: was essentially flat through the closure (100 in 2018, 97 in 2019, 96 in the pandemic year 2020) and then grew steadily to 106% of the 2018 level by 2023 and further to about 113-114% by 2025.

This is the empirical shape the literature's "job-loss multiplier" language can obscure if read casually: even a multiplier of 2 applied to a few hundred direct jobs is a small percentage of a county's total employment base (Adams County's total covered employment is roughly 6,000-6,900), so it can be swamped by unrelated county-wide growth in the aggregate all-industries series even as the specific sector, the specific workers, and specific tax lines (which are concentrated on a handful of large taxable parcels, not spread across the whole economy) take a much sharper and more persistent hit. The Nason/Scavette/Stephens Appalachian sample — smaller, more coal-dependent counties with fewer alternative industries to backfill the loss — is the setting where the aggregate county numbers actually do move by the double-digit percentages reported above; a more economically diversified county can absorb a similarly sized plant closure with little visible dent in its total headline employment count, while still taking the full fiscal hit.

Caveats and what this does and does not show

• The Philadelphia Fed paper studies coal MINE closures in Appalachia specifically, not coal power PLANT closures nationwide; the mechanisms (lost high-wage jobs, local demand multiplier, tax base erosion) are closely analogous but the magnitudes are not a direct read-out for a plant closure in a different, more diversified region.
• This corpus's EIA power-plant table (energy.eia_power_plants) does not currently carry any populated retired/cancelled (operating_status RE/CN) rows or a working retirement_year column despite the table description claiming that coverage. We attempted to build a nationwide before/after employment comparison by identifying still-operating coal-plant counties and EXCLUDING any county that also appears with a retired coal unit (predicate: county_fips NOT IN (SELECT county_fips FROM energy.eia_power_plants WHERE operating_status='RE' AND technology ILIKE '%coal%')). That exclusion subquery matched zero rows — because no RE-status rows exist in this table at all — so the exclusion removed nothing and the resulting list is simply every county with an operating coal plant in 2023, not a genuine never-treated control group. We abandoned that approach, reported the data gap, and instead used a single documented case (Adams County, OH) with closure dates taken from published research rather than from this table.
• The Adams County BLS series is a single case, not a controlled comparison; it is presented to illustrate the sector-vs-aggregate mechanism, not as a replacement for the causal Appalachian-mine-closure estimates above.
• Tax-revenue figures for individual plants ($8.5M Adams Co., up to $60M Navajo Co., $3.8M San Juan Co.) come from case reporting and one peer-reviewed study, not from a single consistent national dataset, so they are illustrative of the range rather than a representative average.

Every query behind this report

5 warehouse calls ran in this session, in order. Each is reproducible against the same snapshot.

SELECT county_fips, state_abbr
FROM energy.eia_power_plants
WHERE technology ILIKE '%coal%' AND operating_status = 'OP' AND report_year = 2023
AND county_fips NOT IN (
  SELECT county_fips FROM energy.eia_power_plants WHERE operating_status='RE' AND technology ILIKE '%coal%'
)
GROUP BY county_fips, state_abbr
LIMIT 50

SELECT * FROM energy.eia_power_plants WHERE operating_status='RE' FETCH FIRST 3 ROWS ONLY

SELECT operating_status, COUNT(*) FROM energy.eia_power_plants GROUP BY operating_status ORDER BY 2 DESC

SELECT "year", annual_avg_emplvl, total_annual_wages, annual_avg_wkly_wage
FROM econ.county_qcew
WHERE area_fips = '39001' AND own_code = '0' AND industry_code = '10'
ORDER BY "year"

SELECT "year", annual_avg_emplvl
FROM econ.county_qcew
WHERE area_fips = '39001' AND industry_code = '2211' AND own_code = '5'
ORDER BY "year"