Skip to content

Decide Colorado Premium Assistance's 400% ceiling on the exact FPL ratio - #10318

Merged
DTrim99 merged 3 commits into
mainfrom
co-premium-assistance-exact-fpl-ratio
Nov 19, 2026
Merged

DTrim99 merged 3 commits into
mainfrom
co-premium-assistance-exact-fpl-ratio

Conversation

@MaxGhenis

@MaxGhenis MaxGhenis commented Oct 20, 2026 •

Copy link
Copy Markdown
Contributor

Still pending: full impact run, Axiom parity line, GPT-6.1 Sol review. Done: the fix and YAML cases, on current main.

Update onto current main (2026-11-18)

  • Merged upstream/main with no conflicts. Max's commit stays; nothing was rebased or force-pushed.
  • The ceiling now divides by aca_fpg, the poverty line aca_magi_fraction and the federal credit have used since Keep a filer who can be claimed as a dependent out of the premium tax credit family, and apply each state premium program's rule #10041: the prior-year guideline for the section 36B tax family. The division runs in float64.
  • The 47 fixtures that pinned tax_unit_fpg for the prior year now pin aca_fpg: 10_000 for the coverage year. No expected value changed.
  • New eligibility Cases 18 and 19 cover a joint return with one claimable spouse. The ceiling follows the one-person tax family: $62,600 is eligible, and $70,000 is not (447% of $15,650, though 331% of the whole unit's $21,150).
  • The sources don't settle the choice of aca_fpg. Amended Regulation 4-2-78, Section 4.F gives "Federal Poverty Level" the meaning in C.R.S. 10-16-1203(4), which adopts the "poverty line" of 42 U.S.C. 9902(2). Section 5.A names household size as a factor without defining it, and Section 5.C.1 leaves identifying eligible enrollees to the Exchange. None of these sets the guideline year or the family counted. aca_fpg is a modeling choice: it is the line of the federal credit this gate already requires.
  • Modal run pa-10318-a: all 102 cases pass across the three Colorado YAML files (61), healthcare_benefit_value.yaml (22) and household_health_benefits.yaml (19), and all 14 tests in test_claimable_filer_ptc_invariants.py pass.
  • Mutation checks on Modal: restoring the old denominator (pa-10318-b) fails amount Case 7 and eligibility Cases 5 and 19. Restoring main's truncated gate (pa-10318-c) fails eligibility Cases 15, 16 and 17.
  • Colorado's 2027 amounts and 2028 switch-off are left for a later change.
  • Not changed here: Cases 9 and 10 of household_health_benefits.yaml still supply aca_magi_fraction: 2.50, which this gate no longer reads. They supply no income, so they check that the amount flows through, not the income test.
  • Not rerun: the differential sweep below, and the $62,756.50 row of the table below. The table's other three rows are covered by the eligibility cases at $62,600, $62,600.47 and $62,755.56.

Bug

co_premium_assistance_eligible compared the truncated aca_magi_fraction (np.floor(100 * magi / fpg) / 100) with <= fpl_limit.upper (4.0). Truncation maps every income from 400.00% to 400.99% of the poverty line to 4.00, so all of them passed. Amended Regulation 4-2-78 (3 CCR 702-4), Section 4.C defines an "eligible enrollee" as one "whose household income is no greater than 400% of the Federal Poverty Level". (Section 5.B's heading reads "Households Below 400% FPL"; the definition, not the heading, sets the test.) No source prescribes truncating the percentage.

Under current law the gate also requires is_aca_ptc_eligible, whose income test ends at 400% (2026 onward), so incomes above 400% already failed end to end. The Colorado ceiling decides alone when a reform keeps the federal credit above 400%, such as extending the enhanced credit.

Executed on main (645fcd3; the file is unchanged at 740e1eb), single filer in Colorado, 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) / aca_fpg, divided in float64, the pattern of #10305 as it stands on main. Nothing else in Colorado Premium Assistance reads the fraction.

Invariants

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

  1. Band = regulation. Eligible if and only if r <= 4, r = max(MAGI, 0) / aca_fpg; no floor.
  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

  • Every case in the three Colorado Premium Assistance YAML files that supplied aca_magi_fraction now supplies aca_magi against an aca_fpg of $10,000, so it still drives the gate.
  • New income-derived cases with the real guideline: $62,599 (399.994%) and $62,600 (400%) eligible; $62,600.47 (400.003%) and $62,755.56 (400.994%) ineligible.
  • Case 17 sets gov.aca.ptc_income_eligibility[2].amount: true (federal credit above 400%) and computes is_aca_ptc_eligible: at $62,600.47 the household is federally eligible and Colorado-ineligible.
  • Mutation check: main's gate fails exactly Cases 15, 16 and 17.
  • Local differential sweep, run before the move to aca_fpg (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 83 mismatches, all in the truncation sliver.
  • Local run before the move to aca_fpg: the Colorado Premium Assistance YAML folder, 53 passed. Current results are in the update section above.

Impact

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

Coordination

#10041, merged 11/13, moved aca_magi_fraction's denominator to aca_fpg, and this gate's exact ratio now uses the same denominator. #10312 makes exactly 400% federally eligible; the two are independent (the new cases here supply is_aca_ptc_eligible or sit above 400%). Part of the aca_magi_fraction audit (table: #10329) begun with #10305.

axiom: pending

🤖 Generated with Claude Code

Amended Regulation 4-2-78 Section 4.C defines an eligible enrollee by
household income "no greater than 400% of the Federal Poverty Level". The
gate compared the truncated aca_magi_fraction with 4.0, so every income
from 400.00% to 400.99% passed. The federal PTC gate ends at 400% under
current law, so the error showed only when a reform extends the federal
credit; test the exact ratio of nonnegative ACA MAGI to the prior-year
guideline anyway, so the state ceiling holds on its own.

Tests that supplied aca_magi_fraction now supply aca_magi against a
prior-year guideline of $10,000; new cases use the real 2025 guideline
at the edge and a reform that extends the federal credit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
DTrim99 and others added 2 commits November 18, 2026 13:47
Divide ACA MAGI by aca_fpg, the prior-year guideline for the section 36B
tax family that the federal credit has used since #10041, instead of the
whole tax unit's guideline. Divide in float64 so a ratio just above the
limit can't round down onto it. Pin the test fixtures to aca_fpg and add
a joint return with a claimable spouse, whose ceiling follows the
one-person tax family.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@DTrim99
DTrim99 marked this pull request as ready for review November 18, 2026 19:07
@DTrim99
DTrim99 requested a review from hua7450 November 18, 2026 19:07

@hua7450 hua7450 left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks, David. Colorado's rule checks out: Amended Regulation 4-2-78 Section 4.C requires household income "no greater than 400% of the Federal Poverty Level" in both the PY2026 and PY2027 versions, no source rounds the percentage, and the floor comes from the federal credit through is_aca_ptc_eligible. The 2026 test incomes are right (4 × $15,650 = $62,600). Approving, with one test fix before merge.

Before merge

Duplicate case numbers. co_premium_assistance_eligible.yaml already has Case 13 (line 264) and Case 14 (line 286); the seven appended cases restart at "Case 13" (line 310) and run to "Case 19". Please renumber them 15 to 21. No expected values change.

Optional

  • References. Cite C.R.S. 10-16-1203(4) and (7), which define household income by the federal credit and support the aca_fpg pairing, in the comment at lines 39-45; and add the PY2027 version of Regulation 4-2-78 to the ceiling's references, since the limit carries into 2027.
  • Tests. Several new cases repeat existing equivalence classes.
  • Zero guard (see below).
  • 2027 amounts (separate PR). The PY2027 regulation sets $70 for the first member and $25 for each additional member; the model still carries $80 and $29 into 2027.

Series-wide (follow-up, not for this PR). The exact ratio max_(aca_magi, 0) / aca_fpg is now computed inline in each PR of this series, plus two California copies on main, and only 10312 guards a zero poverty line (territories, where aca_fpg is 0; the formula runs before the state mask, so a mixed simulation emits NumPy warnings, though outputs are unaffected). One guarded tax-unit variable for the untruncated ratio, read by every gate, would be a good follow-up PR once these merge.

@DTrim99
DTrim99 merged commit dd4b7e7 into main Nov 19, 2026
46 checks passed
@DTrim99
DTrim99 deleted the co-premium-assistance-exact-fpl-ratio branch November 19, 2026 03:58
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants