[MaxGhenis]

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

Bug

wa_cascade_care_savings_eligible compared the truncated aca_magi_fraction (np.floor(100 * magi / fpg) / 100) with <= 2.5. Truncation maps every income from 250.00% to 250.99% of the poverty line to 2.50, so all of them passed. WAHBE's PY2026 Cascade Care Savings Policy, Section 4(1)(c) makes an individual an eligible enrollee who "Has income up to 250% of the Federal Poverty Level"; the PY2027 policy keeps the same words. RCW 43.71.110(4)(a)(ii) leaves the threshold to the appropriation or the exchange. No source prescribes truncating the percentage.

"Up to 250%" leaves exactly 250% open to either reading. The model already treats it as inclusive and this PR keeps that. Income above 250% is not "up to 250%" on any reading, so the truncation sliver is unambiguous.

Executed on main (645fcd3; the file is unchanged at 740e1eb), single filer in Washington, 2026, 2025 one-person guideline $15,650, Group 1 member:

Income	Exact % of FPL	aca_magi_fraction	Eligible on main	Eligible here
$39,125	250.000%	2.50	yes	yes
$39,125.47	250.003%	2.50	yes	no
$39,280.56	250.994%	2.50	yes	no
$39,281.50	251.000%	2.51	no	no

The end-to-end run (wages only, nothing supplied) gave the same results, since the federal credit continues to 400%.

Fix

Decide eligibility on the exact ratio max(aca_magi, 0) / tax_unit_fpg(period.last_year), the pattern of #10305. Nothing else in Cascade Care Savings reads the fraction. The fpl_limit comment now says the limit applies to the exact ratio.

Invariants

For every input, with the program in effect and a Group 1 or Group 3 member:

1. Band = policy. Eligible if and only if r <= 2.5, 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 (2.50, 2.51), so wa_cascade_care_savings can only fall. Deduced: floor(100r)/100 <= 2.5 holds exactly when r < 2.51.

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. The repo's test-design rules keep such cases in YAML.

Tests

• All cases in wa_cascade_care_savings_eligible.yaml, wa_cascade_care_savings.yaml and integration.yaml that supplied aca_magi_fraction now supply aca_magi against a prior-year tax_unit_fpg of $10,000, so each still drives the gate (the gate no longer reads the fraction; nothing else in these files did). The derived fraction outputs are unchanged.
• New income-derived cases with the real guideline: $39,124 (249.994%) and $39,125 (250%) eligible; $39,125.47 (250.003%) and $39,280.56 (250.994%) ineligible.
• Mutation check: main's gate fails exactly the two new above-limit cases.
• 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 58 mismatches, all in the truncation sliver.
• Local run: the Cascade Care Savings YAML folder, 73 passed.

Impact

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

Coordination

#10041 (draft) edits wa_cascade_care_savings.yaml and two other files in this folder 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