The EPA Superfund account was cut 47 percent to 282.8 million dollars for FY2026, four years after a one-time 3.5 billion dollar infusion cleared the backlog of unfunded sites

What is happening?

The Superfund program was built in 1980 on a polluter-pays trust fund, filled by excise taxes on petroleum and chemical feedstocks and by a corporate environmental tax. That excise tax lapsed on 1995-12-31. For the next twenty-six years the program ran on annual discretionary appropriations from the general fund, and those appropriations fell from about 2.6 billion dollars in FY1999 to about 537 million dollars in FY2024.

The Infrastructure Investment and Jobs Act, enacted 2021-11-15, reversed both halves at once. It reinstated the excise tax, effective 2022-07-01 and scheduled to expire 2031-12-31, and it added 3.5 billion dollars to the Superfund Remedial program. EPA obligated 1.026 billion dollars in FY2022 and reported that the backlog of 49 sites waiting for construction money had been cleared. By 2024-10-16 the agency had spent 3.296 billion of the 3.5 billion.

Then the annual cycle resumed. The FY2026 regular appropriation, enacted as Public Law 119-74 on 2026-01-23, set the Superfund account at 282.8 million dollars, down 47.4 percent from 537.7 million dollars in FY2025. What the account has to cover did not move with it: 1,340 sites were active on the National Priorities List as of 2025-03-05, about 90 percent of them non-federal.

Whose problem is this?

The body that sets the appropriation is not the body that has to sequence the work, and neither of them lives near the sites. A cleanup that slips a year produces no event that anyone has to announce.

Where does this problem end?

The boundary here is the financing structure and not the chemistry of any one site. Two other documents in this corpus already cover cleanup obligations that arise under other statutes, and the point of separating them is that a bonding shortfall and a lapsed excise tax fail in different places.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state. No source opened here states an official target: not a funding level the program is meant to reach, not an acceptable number of sites waiting for construction money, not a date by which the current NPL inventory is meant to be worked down. The 1980 statute describes a mechanism, a trust fund fed by the industries that generate the waste, and the sources record when that mechanism was on and when it was off. None of them records what it was supposed to deliver by when.

How big is it?

Between 23.3 million and 77.7 million people. EPA published both figures for the same FY2022 site inventory of 1,881 Superfund sites, which includes deleted, proposed and federal facility sites: 23,268,560 people within one mile and 77,652,857 within three miles, about 7 percent and 23 percent of the national population, updated July 2023.

The width of that interval is a choice of radius, not statistical uncertainty. Nobody measured the same thing twice and got two answers. One number counts everyone within a mile of a site and the other counts everyone within three, and the sources opened here give no basis for preferring either boundary.

This count measures how many people live near a Superfund site and not how many are harmed in a given year. It also counts sites that are finished, because the 1,881-site inventory includes sites already deleted from the list. The active inventory is 1,340 sites as of 2025-03-05, and no source opened here gives a population figure for that subset. Scaling the 77.7 million by the ratio of active to total sites would assume every site has the same number of neighbors, which is exactly the assumption the source does not support.

A separate figure exists for the 157 federal facility NPL sites as of October 2025, roughly 13 million people within three miles and 3 million within one mile. That is a different population count on a different date for a subset of the inventory, and it cannot be added to the numbers above.

Under what conditions does it arise?

1. A permanent liability funded by a temporary tax will always outlive the tax that funds it. The excise tax ran from 1980 to the end of 1995 and then again from mid-2022, with an expiry already written for the end of 2031. The contamination it pays to remove has no expiry date at all.

2. The gap gets filled by the most contested money in the budget. For twenty-six years the program lived inside annual discretionary appropriations, and over that span the account fell by roughly four fifths, from about 2.6 billion dollars in FY1999 to about 537 million in FY2024.

3. A one-time infusion resets the visible symptom without changing the structure. The 3.5 billion dollars from the 2021 law cleared the backlog of 49 unfunded sites and funded 22 new construction starts in FY2023 and 25 in FY2024. Once about 3.296 billion of it had been spent by 2024-10-16, the program was back on the annual account, which was then cut by 47.4 percent.

4. Deferral is invisible in a way that cancellation is not. A site that does not get a construction start this year is not reported as a decision. The last published count of sites waiting for construction money is the 49 that were cleared in FY2022, and the fiscal year in which the cut takes effect is still running.

What has been tried?

Two of these are structural and the rest are episodic. The 1980 statute and the 2021 reinstatement both changed where the money comes from. Everything in between changed how much arrived in a given year, which is the variable this problem is about.

What was found?

Why is it still unsolved?

Cost structure — the program has a durable obligation and a revenue base that has been switched off once and is scheduled to switch off again.

The 1980 design answered the financing question directly: the industries whose products generate the waste pay into a fund, and the fund pays for the cleanups. That answer stopped operating at the end of 1995 and did not resume for twenty-six years. What replaced it was the annual appropriation, which is the part of the federal budget that has to be re-won every twelve months against every other claim on the same account. Across that span the amount arriving each year fell by roughly four fifths while the inventory of sites did not.

The 2021 law did not replace that structure so much as suspend it. It restored the tax with an expiry date already attached and attached a fixed sum to the remedial program, and the fixed sum behaved exactly as a fixed sum does. It cleared the visible queue, funded four dozen construction starts across two years, and then ran down to about 204 million dollars by October 2024. The FY2026 appropriation then landed at 282.8 million dollars, roughly half the 537 million the account carried in FY2024 while the infusion was still being drawn down.

The third part is why none of this produces pressure. An appropriation that falls by half does not announce which cleanups it removed from next year. There is no roster of deferred sites published on the day the bill is signed, the last named backlog figure is the one that was cleared four years earlier, and the agency can report completions and cost recoveries from work that a prior year already paid for. The measurement that would make the cost visible is a count of sites waiting for construction money, and that count is published irregularly enough that the years in between look the same whether the money was there or not.

What observation would mean it is solved?

Candidates — (a) the dedicated revenue mechanism is made permanent rather than carrying an expiry date, (b) the annual account stops moving by tens of percent between adjacent years, (c) the count of sites waiting for construction money is published every year and stays at or near zero.

(a) alone is not enough. A permanent excise tax fixes the revenue base but says nothing about the appropriation, because money in the trust fund still has to be appropriated out of it before it reaches a site. The program had a dedicated tax throughout the 1980s and early 1990s and still ran on annual appropriations decisions.

(b) alone measures the wrong stability. A flat account at a low level is stable and still slower than the inventory requires. Stability matters here because remedial construction runs multi-year, and a project planned against one year of funding is exposed when the next year halves. But a steady number is not by itself a sufficient number.

(c) is the closest to a real measurement and it is the one that is missing. A single year of higher funding would not settle this, because the pattern in question is measured in decades. Sites waiting for construction money is the quantity that translates a budget line into a physical fact, it was reported at 34 in FY2019 and at 49 cleared in FY2022, and it has not been published for the years since. A measure that appears only when the news is good cannot be used to detect when the news is bad.

What is it connected to?

Fills with researchcleanup obligations financed through other federal statutes and other site inventories, including orphaned oil and gas well plugging and coal ash impoundment closure, both already covered separately in this corpus under different laws; state mini-Superfund programs and how their revenue bases behave over the same period; the Brownfields program as a separately funded EPA line; the wider question of how multi-decade federal liabilities are financed through annual discretionary accounts; and the three-part EPA Office of Inspector General series on climate risk at federal facility sites. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• What the backlog is now. The named counts of sites waiting for construction money stop at 34 in FY2019 and 49 cleared in FY2022. No source opened here gives a figure for any period after the FY2026 appropriation was enacted on 2026-01-23. The fiscal year is still running, so this is closer to not yet measured than to measured and withheld.
• What happened to the remaining IIJA balance. About 204 million dollars of the 3.5 billion was unspent as of 2024-10-16. No source opened here updates that figure, says whether it survived into FY2026, or says when it runs out. Whether the FY2026 account is the whole of the available money for the year therefore cannot be determined from these sources.
• Whether the appropriation level explains anything about cleanup pace. No third-party quantitative analysis linking the funding cut to a slowdown was found. EPA stated on 2026-06-03 that it completed 290 cleanups since January 2025, and as of 2026-08-16 no audit body report contradicting that was located. The absence of a finding is not a finding of absence in either direction.
• Whether the flood review gaps have anything to do with resources. The EPA Office of Inspector General reported on 2026-03-25 that 16 of 37 five-year reviews did not address flood impacts as guidance directs. No source opened here states what the report attributes that to, and the fieldwork closed in February 2026, weeks after the FY2026 appropriation was enacted. Placing the two facts in one document does not connect them.
• Whether the FY2024 and FY2025 figures are the same money. GAO gives about 537 million dollars for FY2024 and the CRS summary gives 537.7 million dollars for FY2025. The two are close enough to be the same appropriation restated for a different year or two consecutive years at the same level, and neither source addresses the other.
• Whether the long-run figures are adjusted for inflation. The decline from about 2.6 billion dollars in FY1999 to about 537 million in FY2024 is reported without a stated basis in any source opened here. If the figures are unadjusted, the real decline is larger than the ratio suggests, and none of these sources says which it is.
• How the climate risk series adds up. The Office of Inspector General issued three reports on site risk, on sea level rise in June 2025, inland flooding in March 2026 and wildfire in March 2026. No source opened here combines them into a single count of sites or people, and how they treat sites appearing in more than one report was not established.
• Anything about the 1,183 non-federal NPL sites and climate risk. The inspector general series covers federal facility sites only. Whether its findings extend to the roughly 1,183 remaining active sites, which are about 90 percent of the list, is outside what these sources examined.

See the evidence

No primary legal or budget document was read in full. The two EPA publications, the inspector general report and the GAO product carry the figures attributed to them, and the appropriation figures come from a Congressional Research Service summary of Public Law 119-74 rather than from the enacted text. Three press rows could not be fetched and are left with a blank URL and the reason rather than removed, and no claim in this document rests on any of the three. Where sources overlap they agree: the lapse date of 1995-12-31 and the reinstatement effective date of 2022-07-01 appear in more than one opened source, and the 3.5 billion dollar infusion appears in both EPA sources. One near-collision is left unresolved rather than reconciled, the GAO figure of about 537 million dollars for FY2024 against the Congressional Research Service figure of 537.7 million for FY2025. The position of the agency that runs the program is included at the same weight as the findings about it, in the form of the completion and cost recovery figures it published on 2026-06-03, and it is noted that the release does not address the appropriation level. This is a Path A output, a research-based definition, so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 23,268,560–77,652,857 As of FY2022 site inventory, published 2023-07

Sensitivity The width of this interval is a choice of radius, not statistical uncertainty. EPA published both figures for the same site inventory on the same date, so nobody measured one quantity twice and got two answers. A reader who treats 23.3 million to 77.7 million as a confidence range will misread it. Two further corrections were identifiable and neither could be computed from the sources opened here, and they run in opposite directions. Downward: the 1,881-site inventory includes sites already deleted from the National Priorities List as well as proposed and federal facility sites, while the active list held 1,340 sites as of 2025-03-05, so an unknown share of the counted population lives near work that is finished. That share was not computed, because scaling by the ratio of active sites to total sites would assume every site has the same number of neighbors, and the source gives no basis for that assumption. Upward: the count is drawn from the federal Superfund inventory alone and excludes contaminated sites handled under state programs or under other federal statutes, which are outside the scope of this document but not outside the exposure. The figures also count proximity and not harm, and no source opened here converts either radius into a count of people exposed to a specific contaminant or into any health outcome. A separate EPA Office of Inspector General figure exists for the 157 federal facility sites as of October 2025, roughly 13 million within three miles and 3 million within one mile, and it is a different count on a different date for a subset of this inventory, so it is not added here and is not used to adjust either bound.

Regional breakdown No source opened here gives a state-level or region-level population count near Superfund sites. The nearest geographic detail is that the FY2019 backlog of 34 sites waiting for construction money spanned 17 states and Puerto Rico, which locates sites rather than people and covers a different year and a small subset of the inventory. Splitting the national figure by state population would place people near sites in proportion to where people live, and Superfund sites are distributed by industrial history rather than by population, so that split would be wrong in both directions at once.