Investigations · Taxes & the Safety Net

“Paid For by Cuts to Medicaid and SNAP”?

Both parties are wrong about the same bill — and the proof for each comes from the same CBO documents. The 2025 tax law’s dueling narratives, checked against the government’s own scorekeepers.

Facts first. Every figure is sourced and carries its scoring vintage. Core claims passed a three-vote adversarial verification against CBO, JCT, Senate roll-call records, and peer-reviewed research — 25 of 25 confirmed. This page checks a left-coded claim AND the right-coded defense; neither survives intact.

The 2025 law really did pair top-skewed tax cuts with the deepest Medicaid and SNAP reductions in a generation — CBO’s own analysis says the poorest tenth of households lose about $1,200 a year while the richest tenth gain about $13,600. But “paid for by” is false as a description of the money: the safety-net cuts offset only about a quarter of the tax-side cost. The tax cuts were paid for mostly by borrowing — $3.4 trillion in added deficits. And the defense fails just as hard: CBO projects the Medicaid chapter alone leaves 7.5 million more people uninsured. The honest sentence nobody says: tax cuts tilted to the top, financed mostly by debt, with historic safety-net cuts layered on top.

−$1,200

per year — poorest tenth of households, CBO’s analysis of the enacted law

+$13,600

per year — richest tenth of households, same CBO analysis

~27%

share of the tax-side cost the safety-net cuts actually offset

7.5M

more uninsured in 2034 from the Medicaid chapter alone (CBO)

00The narrative being tested

Two framings of the 2025 reconciliation law (P.L. 119-21) circulate in direct opposition, and both are wrong in documentable ways:

Left-coded: “Republicans gave tax cuts to billionaires paid for by cuts to Medicaid and SNAP” — including candidate-level versions like the Maine ad claiming Sen. Susan Collins “sided with Trump” on it (PolitiFact: False).
Right-coded: the law “overwhelmingly benefits low- and middle-class families” (Senate Finance Chairman Crapo); “no one loses coverage — only waste, fraud, and abuse”; “it’s just work requirements for able-bodied adults.”

Three honesty guardrails

Vintage discipline. Nearly every viral number on this law scores a draft, not the enacted statute — the House-bill figures (10.9M uninsured, −$1,600 bottom decile, the JCX-23-25 millionaire tables) differ from the enacted-law figures (~10M, −$1,200, JCX-33-25). Every figure here carries its vintage.
All scores are conventional (static) estimates against CBO’s January 2025 baseline, excluding macroeconomic feedback and tariffs, in 2025 dollars.
Checking your own side counts double. This page exists because a left-coded claim earned a False rating — and because the rebuttal to it earns one too.

01What the law actually does — the government’s own numbers

Who gains, who loses — CBO’s distributional analysis of the enacted law

Average annual change in household resources, 2026–2034, in 2025 dollars

Source: CBO, Distributional Effects of Public Law 119-21 (pub. 61367, August 11, 2025) — enacted-law vintage. Bottom decile: −$1,200 (−3.1% of income); top decile: +$13,600 (+2.7%). Bars drawn to a common scale (~$32.4 per px): the loss bar is small in dollars and large in lives.

Deficits: +$3.4 trillion over 2025–2034, +$4.1 trillion including $718 billion in interest (CBO, Aug 4, 2025).
Medicaid: the chapter cuts $886.8 billion and leaves 7.5 million more uninsured in 2034. The work requirement is the single largest saver — $317.0 billion — and CBO is explicit the savings exist because people lose coverage: ~2.9 million for failing to demonstrate compliance, plus ~2.8 million more from added verification steps (CBO supplemental estimate, Oct 28, 2025).
SNAP: ~2.4 million fewer participants in an average month from expanded work requirements; a new state cost-share projected to cut benefits for ~300,000 more; remaining households lose $10–$100/month depending on provision, with the 2034 average benefit capped at $213 vs. $227 baseline (CBO, Aug 11, 2025).
Taxes: JCT’s tables show the dollar cuts flowing predominantly upward — in the House-draft score, $96.1 billion of the 2027 cut goes to ~1.18 million returns over $1 million — while the under-$15,000 group’s taxes rise by 2031 after the tips/overtime/senior provisions sunset (JCX-23-25, House vintage; the enacted-equivalent JCX-33-25 shows the same direction, smaller magnitudes).

02Why “paid for by” is false anyway

The money: what the cuts actually cover

CBO’s accounting of the enacted law, 2025–2034

Sources: CBO pub. 61367 (Aug 11, 2025): taxes and cash transfers +$3.3T to household resources; in-kind transfers −$900B, “primarily because federal spending on benefits provided through Medicaid and SNAP will be lower.” CBO pub. 61466 (Aug 4, 2025): +$3.4T deficits excl. debt service. Bar scale: $3.4T = 416 px.

The accurate version of the charge is “tax cuts tilted to the top, paid for by debt, with Medicaid and SNAP cuts stacked on top” — which is arguably a harsher indictment, and has the advantage of being true.

The claim

“Tax cuts for billionaires, paid for by cuts to Medicaid and SNAP.”

Kernel true, mechanism false

The same bill did both things — but the $900B in cuts offsets ~27% of the $3.3T tax-and-cash side. Three-quarters of the tax cuts were financed by borrowing, not by the safety net.

The claim

“Susan Collins sided with Trump — billionaire handouts paid for by Medicaid and SNAP cuts.” (Maine Senate ad)

The record

False (PolitiFact, June 25, 2026). Collins voted for the motion to proceed — warning it didn’t predict her final vote — then voted against final passage, citing “the harmful impact it will have on Medicaid.” A claim can be true about a bill and false about a senator.

The “16 million lose insurance” number, decomposed

Where the viral figure comes from (CBO letter, June 4, 2025 — House-passed vintage)

Source: CBO letter to Sens. Wyden, Neal & Pallone (pub. 61463, June 4, 2025). Attributing all 16M to the law is wrong — and overstating a true harm hands the other side its rebuttal. Bar scale: 26 px per million.

03The defense fails on the same documents

The claim

“The law overwhelmingly benefits low- and middle-class families.” (Senate Finance Chairman Crapo)

The record

Mostly False (PolitiFact, July 21, 2025). The claim uses percentage cuts in tax liability — 16.4% of an almost-zero liability is almost zero dollars. In dollars, the cuts flow predominantly upward, and CBO’s own bottom line is −$1,200/yr for the poorest tenth vs. +$13,600 for the richest.

The claim

“No one loses coverage — we’re only cutting waste, fraud, and abuse.”

The record

False. CBO scores the Medicaid chapter at 7.5 million more uninsured in 2034. Notably, the House Budget Committee’s rebuttal recharacterizes who loses coverage — it does not dispute the totals.

The claim

“It’s just work requirements for able-bodied adults.”

The record

Misleading. The $317B in savings exists because ~2.9M lose coverage — including people who meet the requirement but fail the paperwork — plus ~2.8M more from added verification steps (CBO). The one real-world US test says the quiet part: Arkansas.

18,000

Arkansans who lost Medicaid within months of the state’s 2018 work requirement — while researchers found no significant change in employment, more than a third of the target population unaware the policy existed, and losses driven by paperwork and confusion rather than job status (Sommers et al., New England Journal of Medicine). It is the only large-scale US natural experiment on the policy the 2025 law now takes national.

The claim

“The bill is fiscally responsible.”

The record

CBO: +$3.4 trillion in deficits over ten years; +$4.1 trillion counting interest. Whatever else the law is, it is not paid for.

04The kernels of truth — both sides’

The left’s kernel is large: the same bill genuinely pairs top-skewed tax cuts with the deepest Medicaid/SNAP reductions in a generation, and the government’s own scorer says the poorest tenth are net losers. The anger is data-backed; the “paid for by” mechanism and the senator-level attributions are what fail.
The right’s kernels: the bracket-percentage figures are real JCT numbers (the deception is the metric, not the digits); most enrollees subject to the work requirement will satisfy it or qualify for exemptions — the losses concentrate in the compliance-paperwork gap; and roughly 1.4M of the House-bill coverage losses involved people without verified immigration status (House vintage — flagged).
The shared hazard: vintage confusion. Draft-bill scores quoted as enacted-law facts are the single most common error in circulating versions of both narratives.

05The argument, assembled

The distribution is real: −$1,200/yr for the bottom tenth, +$13,600 for the top tenth — CBO’s own analysis of the enacted law.
The mechanism is false: the safety-net cuts cover ~27% of the tax side. The tax cuts were paid for by borrowing — $3.4T in added deficits.
The defense is false: 7.5M more uninsured from the Medicaid chapter alone; the working-class framing rests on a percentage-of-nothing metric; the one real-world work-requirement test produced coverage losses and zero employment gain.
Attribution matters: the bill-level truth does not transfer to a senator who voted against it — that is how a data-backed grievance earns a False rating.
Precision is the whole game: ~10M, not 16M; enacted-law scores, not draft scores; “financed by debt with cuts stacked on top,” not “paid for by.” The truth here is damning enough without rounding up.

06Sources

Method & sourcing standard. Research run July 11, 2026 through a five-angle deep-research process (CBO/JCT scores; Medicaid/SNAP coverage estimates; deficit financing; the Collins attribution; kernel-and-counterspin evidence). 25 top claims each faced three independent adversarial verification votes: 25 confirmed, 0 refuted (one House-vintage figure passed 2-1 and is flagged). JCT figures were verified digit-for-digit against the JCX-23-25 PDF and CBO figures verbatim against the publications; the Collins roll-call record and the Arkansas NEJM findings were verified by direct fetch. Vintage discipline: House-draft scores (JCX-23-25; 10.9M; −$1,600) are labeled as such wherever cited; enacted-law scores (JCX-33-25; ~10M; 7.5M; −$1,200/+$13,600) carry the headline claims. All estimates are conventional/static, in 2025 dollars, against CBO’s January 2025 baseline. Ideologically aligned organizations (KFF/CBPP left; Heritage/Paragon right) were used for discovery or corroboration only.

CBO & JCT (the law’s official scores)

CBO pub. 61466 (Aug 4, 2025) — enacted-law deficits: +$3.4T / +$4.1T with interest — cbo.gov
CBO pub. 61367 (Aug 11, 2025) — Distributional Effects of P.L. 119-21 — cbo.gov
CBO supplemental estimate, Medicaid chapter (Oct 28, 2025) — $886.8B; 7.5M; Sec. 71119 — cbo.gov (PDF)
CBO, Effects of P.L. 119-21 on SNAP (Aug 11, 2025) — cbo.gov (PDF)
CBO letter to Wyden/Neal/Pallone (June 4, 2025) — the 10.9M / 16M decomposition — cbo.gov
JCT JCX-23-25 (May 13, 2025; House-draft vintage) — distributional tables — jct.gov
CRS R48755 — summary of the law — congress.gov

The Collins attribution & the political claims

Senate roll call vote 372 (119th Congress) — Collins NAY on final passage — senate.gov
Sen. Collins, statement on her vote — collins.senate.gov
PolitiFact (June 25, 2026) — Platner/Collins ad: False — politifact.com
Senate Finance (Crapo) release (July 1, 2025) — the “overwhelmingly benefits” framing — finance.senate.gov

The work-requirement evidence & corroboration

Sommers et al., Medicaid Work Requirements — Results from the First Year in Arkansas, NEJM — nejm.org
Harvard Chan School summary of the Arkansas findings — hsph.harvard.edu
KFF, state allocation of the Medicaid reductions (left-leaning; corroboration only) — kff.org
Bipartisan Policy Center, cost explainer — bipartisanpolicy.org

Related on this site: Is Social Security Going Bankrupt? and Corporate vs. Public Welfare — companion investigations on who the safety-net math actually serves.