[MaxGhenis]

Draft. Done: the fix and YAML cases. Pending: full-file impact, Axiom parity line, GPT-6.1 Sol review.

Bug

md_premium_assistance_eligible compared the truncated aca_magi_fraction (np.floor(100 * magi / fpg) / 100) with <= 4.0. Truncation maps every income from 400.00% to 400.99% of the poverty line to 4.00, so all of them passed. MHBE's adopted PY2026 parameters say "No state subsidy above 400% FPL", and the PY2027 parameters say "Eligibility: 400% FPL or below, eligible for APTC, and ineligible for Medicaid". No source I found prescribes truncating the percentage.

Under current law the gate also requires is_aca_ptc_eligible, whose income test ends at 400% from 2026, so incomes above 400% already failed end to end. The Maryland ceiling decides alone when a reform keeps the federal credit above 400%.

Executed on main (645fcd3; the file is unchanged at 740e1eb), single filer in Maryland, 2026, 2025 one-person guideline $15,650:

Income	Exact % of FPL	aca_magi_fraction	Gate alone on main (is_aca_ptc_eligible supplied)	Here
$62,600	400.000%	4.00	yes	yes
$62,600.47	400.003%	4.00	yes	no
$62,755.56	400.994%	4.00	yes	no
$62,756.50	401.000%	4.01	no	no

Fix

Decide the ceiling on the exact ratio max(aca_magi, 0) / tax_unit_fpg(period.last_year), the pattern of #10305. The target-contribution schedule still interpolates on aca_magi_fraction; only the eligibility edge moves.

Invariants

For every input, with the program in effect and an APTC-eligible member:

1. Band = design. Eligible if and only if r <= 4, r = max(MAGI, 0) / FPG(prior year).
2. Only the sliver changes. Relative to main, eligibility can only be lost, and only for r in (4.00, 4.01). Under current law that set is already federally ineligible, so baseline results should not move; the impact run is the check.

No Hypothesis test is added: the gate is one comparison on a ratio, and the YAML cases hit the limit, 0.003 and 0.994 points above it, plus a reform case. The repo's test-design rules keep such cases in YAML.

Tests

• md_premium_assistance_eligible.yaml: cases that supplied aca_magi_fraction now supply aca_magi against a prior-year tax_unit_fpg of $10,000. New income-derived cases: $62,599 (399.994%) and $62,600 (400%) eligible; $62,600.47 (400.003%) and $62,755.56 (400.994%) ineligible. Case 18 sets gov.aca.ptc_income_eligibility[2].amount: true (federal credit above 400%): at $62,600.47 the household is federally eligible and Maryland-ineligible.
• md_premium_assistance.yaml: seven cases supplied a fraction that disagreed with their own aca_magi, checked by re-deriving each fraction from its MAGI. With the gate reading MAGI, each would have judged eligibility and the target percentage on different incomes. None crossed 400%, so none changed result here; the inputs now agree, with expected values recomputed by hand:
  • Case 16, "225% FPL", used $35,000 (223.6% of $15,650): now $35,212.50; $1,061.50 → $1,052.77.
  • Case 17, a two-person unit "at 300% FPL", used $46,950 (222% of $21,150): now $63,450 with a self-consistent APTC of $2,680.38; $929.39 → $1,256.31.
  • Case 19, 2027 "250% FPL", used $39,125 (245% of the 2026 guideline $15,960): now $39,900; $66.425 → $18.22.
  • Case 21, "175% FPL", used $27,000 (172.5%): now $27,387.50 with APTC $4,630.625; $1,350 → $1,369.375.
  • Case 9 (2025), Case 10 (2028) and Case 12 (zero MAGI) are outside the program or have no income; their inputs are made consistent and their $0 / $2,500 results are unchanged.
• Mutation check: main's gate fails exactly the three new above-limit cases (16, 17, 18); the recomputed amount cases pass on main too.
• Local differential sweep (a script, not committed, per the repo's test-design rules): 256 tax units per year, 60% with income in cents within $2 of an edge and the rest uniform, guidelines $10,000-$80,000 including $20 multiples, comparing the eligibility gate with an exact-rational reading of the source on the float32 values the model receives. 0 mismatches here; on main the same draws give 93 mismatches, all in the truncation sliver.
• Local run: the Maryland Premium Assistance YAML folder, 102 passed.

Impact

Pending: full-file microsimulation via the US hub impact queue, 2026 and 2027.

Coordination

#10041 (draft) edits md_premium_assistance.yaml and changes aca_magi_fraction's denominator to a new aca_fpg; if it lands, this gate's exact ratio should use the same denominator. Part of the aca_magi_fraction audit (table: #10329) begun with #10305; federal edge in #10312.

axiom: pending

🤖 Generated with Claude Code