Repeal all ACA premium subsidies. Saves ~$1.1T but ~19M lose subsidized coverage.
Ten-year deficit effect, conventional-$840.8Breduces the deficit
With macroeconomic feedback and debt service-$937.0BFRB/US-Lite (Federal Reserve calibrated)
Checked against
Published score -$1,100.0B; this estimate is 23.6% away (unfitted reconstruction). Unfitted reconstruction, 23.6% from -$1.10T (CBO estimate) — no constant is fitted to this figure.
No out-of-sample band: no pre-registered row scores a premium-tax-credit reform. The 44 pre-registered Tier 1 rows are the only tier that measures this model's accuracy, and none of them scores a policy of this kind.
The distribution engine cannot yet represent this kind of policy household by household (for example estate tax, tariffs, international or enforcement provisions). An all-zero table would look flat and be wrong, so none is shown.
Economic effects
difference from the CBO baseline
GDP, summed over the window-1.11percent-years
Jobs at the largest effect-149K
Deficit effect of the GDP change+$77.9Bless tax collected
Deficit effect of interest on the debt-$174.0Bless debt, less interest
Fiscal year
GDP level
Jobs
Unemployment
10-year rate
GDP change, deficit
Interest, deficit
2026
-0.233%
-149K
+0.12 pt
-0.01 pt
+$16.3B
-$3.7B
2027
-0.229%
-146K
+0.11 pt
-0.02 pt
+$16.0B
-$6.4B
2028
-0.183%
-117K
+0.09 pt
-0.02 pt
+$12.8B
-$9.0B
2029
-0.139%
-89K
+0.07 pt
-0.03 pt
+$9.8B
-$11.7B
2030
-0.104%
-67K
+0.05 pt
-0.04 pt
+$7.3B
-$14.8B
2031
-0.077%
-50K
+0.04 pt
-0.05 pt
+$5.4B
-$18.1B
2032
-0.057%
-36K
+0.03 pt
-0.06 pt
+$4.0B
-$21.7B
2033
-0.040%
-26K
+0.02 pt
-0.07 pt
+$2.8B
-$25.5B
2034
-0.029%
-18K
+0.01 pt
-0.08 pt
+$2.0B
-$29.4B
2035
-0.020%
-13K
+0.01 pt
-0.09 pt
+$1.4B
-$33.6B
FRB/US-Lite (Federal Reserve calibrated). Demand effects fade over five to seven years as monetary policy responds; there is no supply-side channel. The dynamic score is the conventional score, plus the deficit effects of the GDP change and of interest on the debt: -$840.8B + $77.9B - $174.0B = -$937.0B.
Other scoring models
exploratory
Model
Ten-year deficit effect
Method
CBO-Style
-$840.8B
Static + ETI behavioral response + optional FRB/US dynamic feedback
TPC-Microsim Pilot
Cannot score this
ACA premium tax credits need marketplace enrollment data the pilot microdata does not carry.
No pilot model returned distributional output; the comparison is aggregate-only.
CBO-Style is the model behind the headline above. The TPC-style microsimulation is a pilot: where the two disagree, the gap tells you how much the answer depends on method. It is not an official range.
Limitations
holdout labels follow the locked post-2026-05-02 regression protocol; they are not retroactive historical out-of-sample claims.
No pre-registered out-of-sample row scores a policy of this class, so this model's accuracy on one has not been measured.
The module's annual is fitted to the superseded $350B, so the residual against the revised target is a measurement rather than bookkeeping. Nothing was retuned to close it.
Reclassified rather than retuned: the PTC module's behavioural offset carried the opposite sign to the static effect, so this repeal booked 10% more saving than it statically raises, and the fitted annual was compensating for that. The sign was corrected 2026-09-05 and the row moved 0.3% -> 18.5%; no constant was moved to close it.
The -$1,100B target is an untraceable secondhand figure ('CBO estimate', no publication) on the section 6.2 item 5 list, so the residual is not a measurement against a document. The provenance pass that was queued here has since run (PR #122) and traced its likely origin to CBO/JCT pub. 51298's June 2024 baseline projection ($966B outlays + $176B revenue reductions = $1,142B over FY2025-2034), then recorded the verdict against adopting it in EXAMINED_NOT_REVISED: a baseline projection is not a repeal score. The target therefore stays, and stays unsourced.
VINTAGE AND WINDOW, not model error, are most of what is left. Since lane W7 (2026-09-06) the score removes the baseline vintage's own credit path, and the app's vintage is CBO's February 2026 baseline, in which the ARPA/IRA enhancement has already lapsed: the credit's two legs are $959B over FY2026-2035 against $1,143B over FY2025-2034 on the June 2024 vintage the target came from. Scored on that vintage and window with no coverage response the same mechanism returns $1,143B, 0.09% from pub. 51298's own figure - which is the demonstration that the target is a baseline projection rather than a validation of the model.
The coverage response is now PRICED CHANNEL BY CHANNEL rather than transferred as one ratio (lane HSD/H11, 2026-09-11, `planning/lanes/HSD_h11_ptc_coverage.md`), and the row moved 29.6% -> 23.6% with no constant retuned. Pub. 60437's own itemisation is divided by the coverage movement each line is a response to - $2,971.43 per person-year of employment-based coverage ($101B of compensation shifting plus $3B of employer penalties over a 3.5M decline), $4,200 per Medicaid/CHIP person-year, $57.97 per marketplace person-year for the Basic Health Program, section 1332 waivers and other outlays, and $0 for someone who becomes uninsured - and applied to the scored vintage's OWN subsidized marketplace enrolment from pub. 51298 Table 1 (11.06M a year over FY2026-2035, against 19.0M of uncited constant before). Handed 60437's own coverage vector the rates return its own $79B, which is the identity test. The resulting share is 12.32%, not 19.28%, and the whole of the difference is one line of arithmetic: the offsets scale with PEOPLE and the gross with DOLLARS, and a repeal's average enrollee holds an $8,671 credit where the extension's marginal enrollee holds the $5,370 pub. 60437 Table 3 prints.
WHAT IS STILL TRANSFERRED is the destination split and the channel rates, because no CBO or JCT score of a full repeal exists (searched by PR #122, again by lane H9, again here - pub. 61734 publishes outlays, revenues and the net for the same extension on the app's own window and NO channel itemisation, so the share cannot even be recomputed on FY2026-2035). Two things are known about which way the transfer errs and NEITHER is acted on, because both move the row toward the -$1,100B this repository refuses as a baseline projection. (1) Pub. 60437 Table 3 puts 3.5M of the 6.9M marginal enrollees above 400% FPL and the employment-based decline at 3.5M, and report p. 6 says that decline affects people with higher incomes - so on a vintage where the enhancement has lapsed and eligibility is capped at 400% FPL, the band CBO's largest channel lives in is absent from the population a repeal reaches. Zeroing that channel gives -$996.3B, 9.4% from the target. It is not zeroed: 'most' does not license a zero and 26 U.S.C. 36B(c)(2)(C) bars only an AFFORDABLE offer. (2) A repeal also pushes people INTO Medicaid for a reason the extension has no mirror of, which CBO prices nowhere. (3) Separately, 42 U.S.C. 18051(d)(3) ties Basic Health Program funding to 95% of the credit, so a repeal ZEROES it rather than reversing 60437's +$17B - an unpriced additional saving. Sensitivity on the aggregate: the letter's footnote 4 gives $325B excluding TCJA interactions, an offsetting share of 21.7% rather than 19.3%.