[MaxGhenis]

Fixes #8947

What and why

Four verified model bugs from the statute-first backlog (statute governs, forms fill gaps; ruled 2026-10-19). Each section gives the source, the change, a witness, and the years.

1. NYC school tax credit rate reduction amount: statutory second-band bases (us-nyc-school-tax-credit-rate-reduction-base, 2017+)

• Source. N.Y. Tax Law § 606(ggg)(4-b)(A)-(C), enacted by L.2017, ch. 59, Part C, § 2 (S.2009-C, PDF pages 8-9); § 7 applies it to tax years beginning on and after January 1, 2017 (PDF page 14). "Over $14,400 but not over $500,000: $25 plus 0.228% of excess over $14,400." The IT-201-I line 69a tables give the same values (2025, PDF page 21).
• Before. Five marginal-rate scales (0.171% below the threshold, 0.228% above), so the second band started from the first band carried forward: 36.936 joint and surviving spouse, 24.624 head of household, 20.52 single and separate.
• Change. The scales become rate_reduction/amount/{threshold,base} (by filing status) and {first_band_rate,excess_rate}, from 2017-01-01. The formula is 0.171% x T for T not over the threshold, else base + 0.228% x (T - threshold). Income equal to the threshold stays in the first band. Eligibility and the $500,000 limits are unchanged.
• Witness. Head of household, 2017, city taxable income $20,000: $25 + 0.228% x 5,600 = $37.768 (main: $37.392). Joint 2025, $30,000: $56.152 (main $56.088). Single 2025, $20,000: $39.24 (main $38.76).
• Tests. The 2022 upper-branch cases lose their $1 tolerances; new 2017 and 2025 cases cover every filing status, both sides of each threshold, negative income and a mixed vector. income_tax_elimination/integration.yaml re-enables the nyc_school_tax_credit and nyc_income_tax assertions that were commented out because of this bug. The differential invariant test's tolerance drops from 0.5 to 0.001. Tests that had pinned the old carry-in to 0.01 now use the statutory bases: nyc_school_tax_credit.yaml (single $441,000: 998.64 to 999.12; single $272,000: 613.32 to 613.80; joint $482,950: 1,088.81 to 1,088.878), the two NYC school tax credit reform files under contrib/local/nyc/stc/, and spm_unit_local_tax.yaml (single, $100,000 wages: NYC tax down $0.48).

2. Indiana supplemental homestead credit: qualification, homestead base, referendum exclusion (us-in-supplemental-homestead-proxies, 2026+)

• Source. IC 6-1.1-20.6-7.7(b)-(d), added by SEA 1 (2025), SECTION 74 (enrolled act, PDF page 124). (b): a person who qualifies for the IC 6-1.1-12-37 homestead standard deduction (or is married at death to a deceased spouse who did) is entitled to the credit "against the person's property tax liability on the same homestead". (c): the lesser of 0.1 x that liability or $300. (d): taxes approved by referendum or local public question "shall not be considered". DLGF's June 18, 2025 FAQ, questions 26-30 and 33 (PDF pages 7-9), agrees and orders this credit after the Over 65 and Blind/Disabled credits.
• Before. min(10% x all tax-unit real_estate_taxes, $300), with no qualification gate, no homestead-only base and no referendum exclusion.
• Change. Three new tax-unit variables: in_supplemental_homestead_credit_eligible (7.7(b) qualification), in_homestead_property_tax_liability (the homestead's liability after earlier credits and before this one) and in_homestead_referendum_property_tax. Credit = eligible x min(0.1 x max(liability - referendum, 0), $300). The first two have overridable defaults:
  • qualification defaults to real_estate_taxes > 0, the Over 65 credit's homeowner proxy (7.7(e): no application is required);
  • the liability defaults to real_estate_taxes less the Over 65 and Blind/Disabled credits, floored at 0, the order IC 6-1.1-20.6-3(3) (SEA 1 SECTION 73) sets.
  • The referendum tax defaults to 0.
• Witness.
  • 2026, real_estate_taxes $3,000, qualification set to false: $0 (main: $300).
  • Eligible, homestead liability $2,000, referendum $1,000: $100 (main: $200 on $2,000 of taxes).
  • Defaults: $2,000 of taxes, under 65 and not disabled → $200. Disabled, $1,000 of taxes → 1,000 − 125 = 875 → $87.50 (main: $100).
• Microsimulation. Before the defaults, the credit was $0 for every household. With them, from 2026 an Indiana tax unit with real_estate_taxes gets min(10% × (taxes − Over 65 − Blind/Disabled credits), $300): main's amount, less 10% of the earlier credits where they apply. The referendum exclusion stays at 0 until the data supplies it.

3. Virginia Commonwealth Savers deduction: per-account cap and carryforward (us-va-529-deduction-owner-and-line, 2024+)

• Source. Va. Code § 58.1-322.03(7)(a): "limited to $4,000 per prepaid tuition contract or college savings trust account"; the excess "may be carried forward and subtracted in future taxable years until ... fully deducted". (7)(b): an owner aged 70 or older deducts "the full amount ... less any amounts previously deducted". The 2024 Form 760 instructions, code 104 (PDF page 28) agree.
• Before. Owner allocation and placement after VAGI were already fixed (Allocate Virginia's 529 deduction to the owner of record and deduct it after VAGI #9875). Under 70 the model pooled the cap as min(total contributions, $4,000 x beneficiary count). It had no carryforward.
• Change. New Person inputs va_529_plan_deduction_capped_amount (sum over owned accounts of min($4,000, contributions + undeducted carryforward)) and va_529_plan_carryforward.
  • The default. When account-level amounts are not supplied, the capped amount defaults to min(contributions + carryforward, $4,000 x accounts): an even split across accounts, the largest deduction any split of the same totals allows. An account holding only a carryforward counts as one account.
  • The deduction. Under 70 it is the capped amount, bounded by contributions plus carryforward. From 70 it is contributions plus carryforward.
  • The instructions. They agree on the per-account cap. Their age-70 sentence speaks only of the year's contributions and is silent on a carryforward, so (7)(b) governs it.
  • Owner and dependent masks are unchanged.
• Witness.
  • 2024, age 45, $7,000 and $1,000 to two accounts (capped amount $5,000): $5,000 (main: $8,000).
  • $3,000 carried forward into one account, no new contribution: $3,000 (main: $0), with or without a beneficiary count.
  • A supplied capped amount of $9,000 against $2,500 available: $2,500.
  • Age 70, $1,000 plus $6,000 carried forward: $7,000.

4. Washington senior and disabled persons property tax exemption: statutory disability and veteran pathways (us-wa-pte-disability-proxy, 2008+)

• Source. RCW 84.36.381(3)(a)(i): at least the age threshold, "or ... retired from regular gainful employment by reason of disability" (RCW 84.36.383(6) incorporates 42 U.S.C. 423(d)(1)(A)); unchanged since 2008 c 6 § 706 (PDF page 85). (3)(a)(ii): a veteran "entitled to and receiving compensation" from the VA at a total disability rating (2015 3rd sp.s. c 30, collection years 2016+), or a combined rating of 80% or more (2019 c 453 §§ 1, 9, PDF pages 3, 13; 2020+), lowered to 40% (2025 c 200 §§ 1, 3, PDF pages 3, 5; 2027+). WAC 458-16A-130 agrees. A 100% schedular evaluation is a total rating (38 CFR 3.340(a)(2)).
• Before. age >= threshold | is_disabled: a working 45-year-old with is_disabled qualified, and a compensated 80% veteran without is_disabled did not.
• Change. New Person inputs wa_pte_retired_due_to_disability, wa_pte_receives_va_disability_compensation, wa_pte_va_service_connected_rating and wa_pte_va_total_disability_rating; parameters veteran/combined_rating_threshold (none before 2020, 0.8 from 2020, 0.4 from 2027) and veteran/rating_requirement_in_effect (from 2016). Every VA pathway requires is_veteran.
  • Overridable defaults. wa_pte_retired_due_to_disability defaults to SSDI receipt (social_security_disability > 0), SSA's finding under the 42 U.S.C. 423(d)(1)(A) definition that RCW 84.36.383(6) adopts. wa_pte_receives_va_disability_compensation and wa_pte_va_total_disability_rating default to is_fully_disabled_service_connected_veteran. The combined rating defaults to 0.
  • Ratchet. The wa_pte_categorical_eligible.py entry leaves the source-conflict ratchet's KNOWN_UNMARKED, since the "in practice" comment is gone.
• Witness.
  • 2024, age 45, is_disabled, not retired, $5,000 tax on a $500,000 home: not eligible, $0 (main: $3,600). Retired by reason of disability: $3,600.
  • Compensated veteran rated 40%: not eligible in 2026, eligible in 2027.
  • Defaults: age 45 with SSDI → eligible. A fully disabled service-connected veteran → eligible.
• Before collection year 2016 (PolicyEngine ruling). The veteran pathway then admitted veterans with a total service-connected disability under the Social Security definition, 42 U.S.C. 423(d)(1)(A). PolicyEngine reads that as a disability meeting the Social Security definition that is wholly service-connected. The model has no service-connection input for that definition, so is_disabled is the proxy.
  • The new input. wa_pte_wholly_service_connected_disability makes the proxy visible and overridable: it defaults to is_disabled or is_fully_disabled_service_connected_veteran.
  • Its limit. The proxy over-includes a veteran whose disability isn't service-connected; setting the input to false removes them.
  • The non-veteran disability-retirement rule applies to every year from 2008.
• Microsimulation. Before the defaults, only the age pathway was open. With them, the disability pathway is open to SSDI recipients. The veteran pathway is open to people the data marks is_fully_disabled_service_connected_veteran (from 2016; and, through is_disabled, to disabled veterans before 2016). The combined-rating route below a total rating stays closed until the data supplies ratings.

Invariants

• NYC: amount >= 0; amount = 0.171% x T at T <= threshold; continuous except for the statutory step at the threshold (base - 0.171% x threshold = $0.064 joint, $0.376 HoH, $0.48 single); matches the § 606(ggg)(4-b) tables for every filing status (differential test, tolerance 0.001).
• IN: 0 <= credit <= min($300, 0.1 x homestead liability); zero when not qualified or outside Indiana.
• VA: deduction >= 0; under 70, deduction = capped amount <= contributions + carryforward when the default applies; from 70, deduction = contributions + carryforward; dependents get 0 (existing test_va_529_invariants.py).
• WA: eligibility is monotone in each pathway input; is_disabled alone never qualifies a non-veteran.

Test plan

• Review of 2026-10-21 answered (defaults, the pre-2016 input, VA guard and wording, NYC candidate); see the hand-off comment.
• YAML tests from the statute and agency examples, written failing first. The "main" witness values above were computed on main ce9377ee80 before the fix.
• Run here at b403e928f1 (main merged), one file per process:
  • wa_pte_categorical_eligible.yaml: 26 passed (5 new default and pre-2016 cases);
  • WA integration.yaml: 14 passed; wa_pte_eligible.yaml: 9; wa_senior_disabled_property_tax_exemption.yaml: 18;
  • in_supplemental_homestead_credit.yaml: 14 passed (3 new default cases);
  • in_supplemental_homestead_credit_in_aggregate.yaml and in_blind_disabled_property_tax_credit_in_aggregate.yaml: 2 and 2 passed (liability now derived: 1,850 and 1,725);
  • va_529_plan_deduction.yaml: 18 passed (2 new); VA plan_529/integration.yaml: 3;
  • nyc_school_tax_credit_rate_reduction_amount.yaml: 16; nyc_school_tax_credit.yaml: 5; income_tax_elimination/integration.yaml: 6; spm_unit_local_tax.yaml: 4;
  • the two contrib/local/nyc/stc/ files: 9 and 2;
  • test_nyc_school_tax_credit_income_test_invariants.py: 6; test_va_529_invariants.py: 2; test_source_conflicts.py: 103.
• Source conflicts: none of the four items is one. The adjacent NYC fixed amount ($62.50 in § 606(ggg)(4-a)(B), $63 on the IT-201-I) is now a candidate, us-ny:nyc-school-tax-credit-fixed-amount-rounding.
• Removed parameter paths: the five rate_reduction/amount/{single,joint,separate,head_of_household,surviving_spouse} scales are gone, as the changelog now says. Saved policies that edit them need the new threshold/base parameters.
• ruff format / ruff check on changed files.
• Changelog fragment in changelog.d/.
• No microsimulation was run here; the hub runs the impact.

Proposed follow-up (not in this PR)

• WA age threshold before 2020. age_threshold.yaml uses 62 before 2020, but the 2008 and 2015 texts of RCW 84.36.381(3)(a) already say sixty-one.

axiom: us-ny:policies/tax/it-201-instructions/nyc-school-tax-credit-rate-reduction#nyc_school_tax_credit_rate_reduction_amount encoded-correct (TY2025+ amount tables only; companion tests joint_return_first_band, single_second_band, head_of_household_second_band, above_final_bounded_row); TheAxiomFoundation/rulespec-us#1452 queued, extended with a comment (TY2017-2024 statutory amount tables; dated-source ingestion and re-pin); TheAxiomFoundation/rulespec-us#1510 queued, extended with a comment (per-account caps and carryforward; encoder approval and form-source re-pin); TheAxiomFoundation/rulespec-us#1862 queued (IN 2026 liability-source ingestion and re-pin); TheAxiomFoundation/rulespec-us#1863 queued (WA dated-source ingestion and re-pin).

---

Generated by Claude Code

[DTrim99]

PR Review: Follow the statute for the NYC school tax rate reduction, IN supplemental homestead, VA 529 and WA senior/disabled property tax exemption

PR #10330 (author: MaxGhenis), head 29778a23d8. This is a first review. The PR makes four model changes:

• NYC school tax credit, rate reduction amount (2017+). The five marginal-rate scales become rate_reduction/amount/{threshold,base} by filing status plus first_band_rate and excess_rate. Above the threshold, the amount is now the statute's fixed base ($37 / $25 / $21) plus 0.228% of the excess, instead of the first band carried forward at 0.171%.
• Indiana supplemental homestead credit (2026+). Three new tax-unit inputs: homestead deduction qualification, the homestead's liability after earlier credits, and its referendum taxes. The credit is now eligible × min(10% × max(liability − referendum, 0), $300), instead of 10% of all real_estate_taxes.
• Virginia Commonwealth Savers deduction. Two new Person inputs: a per-account capped amount and an undeducted carryforward. Under 70 the deduction is the capped amount; from 70 it is contributions plus carryforward.
• Washington senior/disabled property tax exemption (2008+). is_disabled no longer qualifies anyone by itself. The disability pathway needs the new wa_pte_retired_due_to_disability input. The veteran pathway needs is_veteran, receipt of VA compensation and a qualifying rating, dated by two new parameters. Before 2016, veterans keep is_disabled.
• Tests and housekeeping. 15 YAML files and one Python test change. The WA file leaves the ratchet's KNOWN_UNMARKED, and there is a fixed changelog fragment.

The PR has no comments or reviews yet. It touches no partner test file.

Path prefixes:

• IN/ = policyengine_us/variables/gov/states/in/tax/property/supplemental_homestead_credit/
• VA/ = policyengine_us/variables/gov/states/va/tax/income/deductions/plan_529/
• WA/ = policyengine_us/variables/gov/states/wa/dor/property_tax_exemption/senior_disabled/
• WAT/ = policyengine_us/tests/policy/baseline/gov/states/wa/dor/property_tax_exemption/senior_disabled/

Sources read for this review (all fetched 2026-10-21):

• New York:
  • S.2009-C / A.3009-C (L.2017, ch. 59), Part C §§ 2 and 7, file pages 7-9 and 14;
  • current Tax Law § 606(ggg);
  • IT-201-I line 69a tables for 2017 (file page 30), 2018 (file page 31) and 2025 (file page 21).
• Indiana:
  • SEA 1 (2025), enrolled, SECTIONS 73 and 74 (file pages 123-124);
  • the DLGF June 18, 2025 FAQ, questions 26-30 and 33 (file pages 7-9).
• Virginia:
  • Va. Code § 58.1-322.03(7);
  • the Form 760 instructions, code 104, for 2024 and 2025 (file page 28 of each).
• Washington:
  • 2008 c 6 § 706 (file page 85);
  • the RCW 84.36.381 supplement with 2015 3rd sp.s. c 30 § 2 and the § 4 application note;
  • 2019 c 453 §§ 1 and 9 (file pages 3 and 13);
  • 2025 c 200 §§ 1 and 3 (file pages 3 and 5);
  • current RCW 84.36.381 and RCW 84.36.383(6);
  • WAC 458-16A-130(2), as amended effective 2/20/26.
• Repository: source_conflicts/candidates.yaml and the ratchet test on main; the CI job logs; siblings Fix WA property tax exemption crash for households outside Washington #10113, Guard county-keyed lookups in VT renter credit, MN CCAP, WI Shares and MD CCS #10116, Keep claimable filers' own exemptions under the HI and VA dependent-exemption repeal #10271, Follow state statutes where forms differ: NM medical exemption, VT CDCC, NY/NYC household credit, PR EITC #10300, Fix four statute-first benefit bugs: MO TANF rounding, SNAP GU/VI limits, HI CCAP TANF income, SSI food ISM date #10332 and Point the NM-WY source-conflict candidates at their code on main #10352.
• rulespec-us: main at cea8379e0b, including us-ny/policies/tax/it-201-instructions/nyc-school-tax-credit-rate-reduction.yaml and us-va/statutes/58.1/58/1-322/03.yaml, and issues What program participation can affect one's eligibility for the Affordable Connectivity Program? #1452 and Allow data tests to work without a policyengine_us package being installed #1510.

Summary

What is right. All four changes follow the governing statute in the years the PR claims. None of them is a statute-versus-form conflict: in each case the agency's form or guidance agrees with the statute, and the old model matched neither.

• 1. NYC (2017+): statute and forms agree on fixed bases.
  • The statute. § 606(ggg)(4-b), as amended by L.2017, ch. 59, Part C, § 2 (file pages 8-9):
    • (A) joint and surviving spouse: "Over $21,600 but not over $500,000: $37 plus 0.228% of excess over $21,600";
    • (B) head of household: "Over $14,400 but not over $500,000: $25 plus 0.228% of excess over $14,400";
    • (C) single and separate: "Over $12,000 but not over $500,000: $21 plus 0.228% of excess over $12,000".
  • Years. (4): "For taxable years beginning after two thousand sixteen, the credit shall equal the 'fixed' amount ... plus the 'rate reduction' amount". § 7: "shall apply to taxable years beginning on and after January 1, 2017" (file page 14). The current § 606 text is unchanged.
  • The forms. IT-201-I 2025 (file page 21) prints "$ 37 plus 0.228% of the excess over $21,600", "$ 21 plus 0.228% of the excess over $12,000" and "$ 25 plus 0.228% of the excess over $14,400". The 2017 (file page 30) and 2018 (file page 31) tables are identical, with ".228%".
  • No conflict. The old test comments ("The $37 amount is rounded in the instructions") were wrong: the bases are in the statute itself. The jump at each threshold ($0.064 joint, $0.376 head of household, $0.48 single) is statutory.
• 2. Indiana (2026+): statute and DLGF agree.
  • The credit. IC 6-1.1-20.6-7.7 (SECTION 74, file page 124):
    • (a) "applies for property taxes first due and payable in calendar years beginning after December 31, 2025";
    • (b) a person who "qualifies for a standard deduction from the assessed value of the person's homestead under IC 6-1.1-12-37 (or is married at the time of death to a deceased spouse who qualifies ...) is also entitled to receive a supplemental homestead credit ... against the person's property tax liability on the same homestead";
    • (c) "the lesser of: (1) ... the property tax liability first due and payable on the homestead property for the calendar year; multiplied by ... one-tenth (0.1); or (2) three hundred dollars ($300)";
    • (d) "Property taxes imposed after being approved by the voters in a referendum or local public question shall not be considered".
  • The order of credits is statutory. IC 6-1.1-20.6-3(3), as amended by SECTION 73 (file pages 123-124), defines the liability for section 7.7 as "determined after application of all credits and deductions under this article or IC 6-3.6, including the credit granted by section 7 or 7.5 of this chapter, but not including the credits granted under section 7.7 or 8.5". So the over-65 and blind/disabled credits come first.
  • DLGF agrees. Question 27: "The credit should be calculated based on the tax liability attributable to the homestead property that is receiving the Homestead Deduction." Question 29: it "would be calculated after the application of the new Over 65 and Blind/Disabled Credits." The question 33 table lists them as steps 5 and 6.
  • Not the deduction. The brief for this review framed this item as the IC 6-1.1-12-37.5 supplemental homestead deduction. The PR touches only the 7.7 credit; DLGF question 26 stresses that they are different. PE-US doesn't model the 37.5 deduction.
• 3. Virginia: statute and instructions agree on the cap and carryforward.
  • (7)(a): "the amount deducted on any individual income tax return in any taxable year shall be limited to $ 4,000 per prepaid tuition contract or college savings trust account ... If the purchase price or annual contribution to a college savings trust account exceeds $ 4,000, the remainder may be carried forward and subtracted in future taxable years until the purchase price or college savings trust contribution has been fully deducted". The transferee "shall succeed to the transferor's tax attributes ..., including ... carryover".
  • (7)(b): an owner who "has attained age 70 shall not be subject to the limitation ... Such taxpayer shall be allowed a deduction for the full amount paid for the contract or contributed to a college savings trust account, less any amounts previously deducted."
  • The instructions (2024 and 2025, code 104): "enter the lesser of $4,000 or the amount contributed during the taxable year to each Commonwealth Savers account ... If you contributed more than $4,000 per account during the taxable year, you may carry forward any undeducted amounts until the contribution has been fully deducted."
  • The age-70 sentence differs slightly (Suggestion G2), but it is silent rather than contrary, and the statute controls.
  • Keep claimable filers' own exemptions under the HI and VA dependent-exemption repeal #10271 doesn't interact. It changes only the HI/VA dependent-exemption repeal reform and its tests. The 529 deduction still masks to head and spouse, and the two PRs share no file or variable.
• 4. Washington: statute, session laws and WAC agree.
  • 2008 text (2008 c 6 § 706, file page 85), RCW 84.36.381(3): "(a) sixty-one years of age or older on December 31st of the year in which the exemption claim is filed, or must have been, at the time of filing, retired from regular gainful employment by reason of disability, or (b) a veteran of the armed forces of the United States with one hundred percent service-connected disability as provided in 42 U.S.C. Sec. 423 (d)(1)(A) as amended prior to January 1, 2005."
  • 2015 (3rd sp.s. c 30 § 2): "(ii) A veteran ... entitled to and receiving compensation from the United States department of veterans affairs at a total disability rating for a service-connected disability". § 4: "This act applies to taxes levied for collection in 2016 and thereafter."
  • 2019 (c 453 § 1, file page 3): "(A) A combined service-connected evaluation rating of eighty percent or higher; or (B) A total disability rating for a service-connected disability without regard to evaluation percent." § 9 (file page 13): "applies for taxes levied for collection in 2020 and thereafter."
  • 2025 (c 200 § 1, file page 3): "A combined service-connected evaluation rating of ((80)) 40 percent or higher". § 3 (file page 5): "applies to taxes levied for collection in 2027 and thereafter."
  • Disability. RCW 84.36.383(6): "'Disability' has the same meaning as provided in 42 U.S.C. Sec. 423(d)(1)(A)". WAC 458-16A-130(2)(b): "retired from regular gainful employment and unable to engage in any substantial gainful activity by reason of any medically determinable physical or mental impairment".
  • WAC timing. WAC 458-16A-130(2)(c) already prints 40 percent. The amendment took effect 2/20/26, for claims filed in 2026 for 2027 collections, so it is consistent with § 3.
  • The parameters (.inf for 2016-2019, 0.8 from 2020, 0.4 from 2027; the rating requirement from 2016) match these dates, keyed by collection year.
• Tests. I hand-checked 22 expectations, listed under Verified correct. CI passed every changed file on a tree that contains Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334.
• Batch checks.
  • Ratchet. It passed in CI, and statically.
  • NH Add New Hampshire state supplementation for 2026 #10156. No state-gated key.
  • CI cost. No new Python test or synthetic reform.

There are no critical findings and three should-address items:

• S1. Invalid Axiom line. axiom: pending: hub files parity is not an allowed form, as on Fix the 2021 CTC split and surviving-spouse phaseout threshold #10306, Fix pre-2018 AMT child exemption and income base #10308, Fix Arkansas tax indexing, Kentucky 2021 CDCC, and WV 529 coordination #10324 and Fix Montana disability exclusion repeal, elderly credit receipt timing and occupancy tests #10331. rulespec-us already encodes the NYC amount correctly for TY2025+, and Allow data tests to work without a policyengine_us package being installed #1510 covers Virginia. Indiana, Washington and NYC 2017-2024 need queued issues.
• S2. The new inputs default to "off". The Indiana credit becomes $0, and Washington's disability and veteran pathways close, for every household that doesn't set them. That includes all microsimulation households. The model already has statute-grounded defaults: an owner-occupancy proxy and the earlier credits for Indiana, and SSDI receipt for 42 U.S.C. 423(d)(1)(A).
• S3. Pre-2016 Washington veterans still qualify on generic is_disabled. Both readings the PR gives for the 2008 text require a service-connected disability. New Case 20 pins a result that both readings reject, and an existing input (is_fully_disabled_service_connected_veteran) fits both.

There are also six suggestions: Virginia default and guard, Virginia age-70 wording, Indiana citation and pre-2026 gating, NYC follow-ups including an unrecorded $62.50/$63 conflict, the Washington age threshold, and the PR body.

Status.

• CI: 27 checks pass and 3 are skipped (umbrella jobs of superseded runs). The green run started at 14:33 UTC on a base that contains Tie source-conflict markers to the register shared with Axiom #10287 and Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334.
• Branch: 7 commits ahead of main and 11 behind, and it merges cleanly.

Critical (must fix)

None.

Should address

S1. The Axiom line axiom: pending: hub files parity is not a valid form

• Where: the last line of the PR body.
• The rule (CLAUDE.md, "Axiom Parity"): a PR that adds, updates or fixes policy needs one of encoded-correct, <rulespec PR> encoded, <rulespec issue> queued or n/a: <reason>. pending is not one of them. We flagged the same form on Fix the 2021 CTC split and surviving-spouse phaseout threshold #10306, Fix pre-2018 AMT child exemption and income base #10308, Fix Arkansas tax indexing, Kentucky 2021 CDCC, and WV 529 coordination #10324 and Fix Montana disability exclusion repeal, elderly credit receipt timing and occupancy tests #10331.
• Why n/a doesn't fit. All four provisions are statutory, and the PR changes results in 2017+ (NYC), 2026+ (IN), every modelled VA year, and 2008+ (WA).
• What rulespec-us has (main cea8379e0b):
  • NYC. us-ny/policies/tax/it-201-instructions/nyc-school-tax-credit-rate-reduction.yaml, rule nyc_school_tax_credit_rate_reduction_amount, computes ..._second_band_base_amount + ..._excess_rate * max(0, nyc_taxable_income - ..._first_band_upper_bound) with bases 37, 21 and 25 and rates 0.00171 and 0.00228.
    • TY2025+ is encoded correctly. That is this head's formula, so encoded-correct is available for TY2025+.
    • TY2017-2024 is not. The module starts at 2025-01-01 and cites the IT-201-I, not § 606(ggg)(4-b). What program participation can affect one's eligibility for the Affordable Connectivity Program? #1452 (open, pe-parity) re-encodes this module for the income tests but stays at TY2025.
  • Virginia. us-va/statutes/58.1/58/1-322/03.yaml holds only college_savings_annual_deduction_limit = 4000 and college_savings_no_limit_age = 70.
    • Allow data tests to work without a policyengine_us package being installed #1510 (open, pe-parity, blocker recorded: billed-run approval and form corpus) asks for per-account deductions with "the $4,000 annual limit separately to each account", college_savings_carryforward_for_account, and the age-70 rule "less amounts previously deducted". That matches this head. Its companion cases come from Allocate Virginia's 529 deduction to the owner of record and deduct it after VAGI #9875 and have zero carryforward.
  • Indiana and Washington. There is no module or issue for IC 6-1.1-20.6-7.7 or RCW 84.36.381.
• Severity: should. An invalid Axiom line is a merge blocker, not a model error.
• Fix: change the line to axiom: us-ny:policies/tax/it-201-instructions/nyc-school-tax-credit-rate-reduction#nyc_school_tax_credit_rate_reduction_amount encoded-correct; TheAxiomFoundation/rulespec-us#<ny> queued; TheAxiomFoundation/rulespec-us#1510 queued; TheAxiomFoundation/rulespec-us#<in> queued; TheAxiomFoundation/rulespec-us#<wa> queued. Alternatively, cite encoder PRs as encoded. Before relying on these:
  • <ny>. Extend What program participation can affect one's eligibility for the Affordable Connectivity Program? #1452, or file a new issue, for TY2017-2024 under § 606(ggg)(4-b), for example us-ny/statutes/TAX/606/ggg.yaml. Cite L.2017 ch. 59 Part C §§ 2 and 7. Companion cases from the statute's tables:
    • 2017 head of household, $20,000 → 37.768;
    • 2017 surviving spouse, $21,601 → 37.00228;
    • 2017 separate, $12,000 → 20.52;
    • 2017 joint, $21,600 → 36.936.
  • Allow data tests to work without a policyengine_us package being installed #1510. Comment with this PR's carryforward and per-account cases, which follow from (7)(a)-(b):
    • under 70, $7,000 and $1,000 to two accounts → 5,000, with 3,000 carried forward;
    • a $3,000 carryforward and no new contribution → 3,000;
    • age 70, $1,000 plus a $6,000 carryforward → 7,000.
  • <in>. A new issue modelled on Add AMT parameters for 2020-2023 #1467:
    • Target: for example us-in/statutes/6-1.1-20.6-7.7.yaml, citing 7.7(a)-(d) and § 3(3) verbatim (quoted above).
    • Required outputs: qualification, the homestead base net of referendum taxes, and the credit.
    • Companion cases:
      • liability 2,000 → 200; 3,000 → 300; 5,000 → 300;
      • not qualified → 0;
      • liability 2,000 with 1,000 of referendum tax → 100;
      • gross homestead tax 2,000 after the $150 over-65 and $125 blind/disabled credits → 1,725 → 172.50.
  • <wa>. A new issue:
    • Target: for example us-wa/statutes/84/84.36/84/36/381.yaml, with the 2008, 2015, 2019 and 2025 texts and their application sections.
    • Companion cases:
      • age 45, retired by reason of disability → eligible; generic disability and not retired → ineligible;
      • compensated 40% veteran → ineligible for 2026 and eligible for 2027;
      • 2024, 80% → eligible; 70% → ineligible;
      • 2019, 90% without a total rating → ineligible; with a total rating → eligible;
      • a total rating without receipt of compensation → ineligible.
    • Blocker: each new issue must name it.

S2. The new inputs default to "off", so the Indiana credit and Washington's disability and veteran pathways vanish unless a user sets them

• Where:
  • IN/in_supplemental_homestead_credit_eligible.py and IN/in_homestead_property_tax_liability.py: inputs with no formula, defaulting to false and 0;
  • WA/wa_pte_retired_due_to_disability.py, WA/wa_pte_receives_va_disability_compensation.py, WA/wa_pte_va_service_connected_rating.py and WA/wa_pte_va_total_disability_rating.py: inputs defaulting to false and 0.
• What changes for users who don't set them:
  • Indiana (2026+). A homeowner with real_estate_taxes gets $0. The PR's own Case 7 shows it: $3,000 of taxes gives 0, where main gave $300. Yet 7.7(e) says "A person is not required to file an application for the credit", and the county auditor applies it to every homestead with the IC 6-1.1-12-37 standard deduction.
  • Washington. A 45-year-old SSDI recipient who owns a home gets nothing unless wa_pte_retired_due_to_disability is set. So does a 100% service-connected veteran unless three new inputs are set. The household calculator and the microdata supply is_disabled, not these.
  • Microsimulation. The PR body says so: "this credit is zero in microsimulation until the data supplies them" and "only the age pathway remains".
• The statute does give a basis for defaults:
  • Washington disability. RCW 84.36.383(6) adopts 42 U.S.C. 423(d)(1)(A), the Social Security disability definition. An SSDI award (social_security_disability > 0) is SSA's finding under that definition and requires inability to engage in substantial gainful activity, which is WAC 458-16A-130(2)(b)'s test.
  • Washington veterans. is_fully_disabled_service_connected_veteran already exists ("a fully disabled veteran who became so as a result of an injury during service"). It implies a total rating.
  • Indiana. The sibling credits in the same umbrella still use real_estate_taxes > 0 as the homestead proxy (in_over_65_property_tax_credit_eligible) and real_estate_taxes as the base. Section 3(3) fixes the order: liability after the over-65 and blind/disabled credits. The PR's aggregate tests compute exactly that by hand ($2,000 − $150 − $125 = $1,725).
• Reach. Neither output feeds household net income. The Indiana credit is read only by taxsim_state_property_tax_credit (via state_property_tax_credits.yaml), and wa_senior_disabled_property_tax_exemption has no consumer. Both outputs are now zero or understated for most households that would qualify under the law.
• Severity: should. Input defaults give wrong results under ordinary use where the model already has statute-grounded values (compare Fix Montana disability exclusion repeal, elderly credit receipt timing and occupancy tests #10331 S3).
• Fix: keep the inputs as overrides, but give them formulas:
  • in_supplemental_homestead_credit_eligible: add(tax_unit, period, ["real_estate_taxes"]) > 0, the over-65 credit's homeowner proxy.
  • in_homestead_property_tax_liability: max_(real_estate_taxes − in_over_65_property_tax_credit − in_blind_disabled_property_tax_credit, 0), citing § 3(3).
  • wa_pte_retired_due_to_disability: social_security_disability > 0. Optionally add adult is_ssi_disabled without earnings above SGA, since SSI uses the same adult definition.
  • wa_pte_va_total_disability_rating: is_fully_disabled_service_connected_veteran.
  • Tests. The explicit-input cases still pass. Add one default case per state, and let the IN aggregate cases assert the derived 1,725 and 1,850.
  • Caveat. As on Fix Montana disability exclusion repeal, elderly credit receipt timing and occupancy tests #10331 S4, a value supplied for one unit gives other units the default array, not the formula. If that matters, split each override into its own input.

S3. Before 2016, Washington veterans still qualify on generic is_disabled, which both of the PR's readings reject

• Where:
  • WA/wa_pte_categorical_eligible.py:43: earlier_veteran = person("is_disabled", period);
  • WAT/wa_pte_categorical_eligible.yaml:347, Case 20 (2015, age 45, is_disabled, is_veteran, not retired): expected true;
  • the PR body, "Left as is (more than one defensible reading)".
• The 2008 text (file page 85): a veteran "with one hundred percent service-connected disability as provided in 42 U.S.C. Sec. 423 (d)(1)(A)".
• The PR's two readings: "a 100% VA rating, or ... a wholly service-connected disability meeting the Social Security definition".
  • Both require service connection, and the first also needs a rating.
  • Generic disability isn't enough. is_disabled establishes neither: the PR itself says it covers non-service-connected and still-working cases.
• The result. Case 20's veteran is under 61, is not retired by reason of disability (so (3)(a) fails), and has no service-connected disability on record (so (3)(b) fails). The model says eligible, and a new test now pins that.
• A fitting input exists. is_fully_disabled_service_connected_veteran is common to both readings: 100% and service-connected. A person fully disabled in that sense also meets 423(d)(1)(A).
• Severity: should. The reach is limited to veterans under 61, not retired by reason of disability, in collection years 2008-2015. But the deferral's stated reason doesn't support keeping is_disabled, and the PR adds a test that pins it.
• Fix:
  • Use is_fully_disabled_service_connected_veteran for earlier_veteran.
  • Change Case 20 to expect false with is_disabled alone.
  • Add a 2015 case with the service-connected input → true.
  • Update the PR body's "Left as is" paragraph.

Suggestions

G1. Virginia: the default ignores carryforward-only accounts, and a supplied capped amount is not bounded

• Where: VA/va_529_plan_deduction_capped_amount.py:34-35: min_(contributions + carryforward, p.cap * accounts), with accounts = count_529_contribution_beneficiaries.
• The gap. That input is an int with no default (0), labelled "Number of beneficiaries to 529 college savings plan contributions". An owner with a carryforward and no contribution this year naturally has 0 beneficiaries of contributions, so the default gives min(3,000, 0) = 0.
  • The witness depends on it. The PR body's "$3,000 carried forward into one account, no new contribution: $3,000" holds only because the test supplies count 1 and the capped amount (va_529_plan_deduction.yaml:195).
• No bound. Under 70, a supplied va_529_plan_deduction_capped_amount above contributions plus carryforward passes straight through (VA/va_529_plan_deduction_person.py:35).
• Fix (optional):
  • In the default, count at least one account when the carryforward is positive, or document that accounts holding only a carryforward must be counted.
  • Bound the under-70 branch by contributions + carryforward.
  • Add a carryforward-only case with count 0.

G2. Virginia: the instructions' age-70 sentence covers only this year's contributions

• The instructions (2024 and 2025, code 104, file page 28): "However, if you are age 70 or older on or before December 31 of the taxable year, you may deduct the entire amount contributed during the taxable year."
• The statute (7)(b): "the full amount paid for the contract or contributed ..., less any amounts previously deducted", which includes an undeducted carryforward.
• Which source controls. The statute does: it is explicit, and the instructions are silent on carryforward at 70 rather than contrary. The PR follows the statute (contributions plus carryforward).
• The wording. The PR body and the YAML header say the instructions "agree" on (7)(b).
• Fix (optional):
  • Say that the instructions are silent on the age-70 carryforward.
  • If you have evidence that Tax limits age-70 filers to the current year, add a candidates.yaml entry (follows: statute).

G3. Indiana: cite the statute for the order, and gate the credit before 2026

• The citation. IN/in_homestead_property_tax_liability.py cites DLGF questions 27, 29 and 33 for "after the credits applied before the supplemental homestead credit". IC 6-1.1-20.6-3(3), as amended by SECTION 73 (file pages 123-124, quoted under Summary), is the statutory source. Add it.
• The pre-2026 credit. 7.7(a) starts the credit with taxes first due in 2026, and the new formula comment says so. But rate.yaml and cap.yaml extrapolate back, so the standalone variable is nonzero in 2025 when the inputs are supplied. Only the umbrella list excludes it (the aggregate file's 2025 case says so). This predates the PR.
• Fix (optional): add an in-effect gate, or zero values before 2026.

G4. NYC follow-ups

• (a) Pre-2017 extrapolation. The new parameters start 2017-01-01 and extrapolate back. § 606(ggg)(4) starts the rate reduction amount "for taxable years beginning after two thousand sixteen"; for 2016, (3) gave only a flat $125 / $62.50. This predates the PR, because the old scales also extrapolated back. Consider a zero before 2017.
• (b) Removed parameter paths. The PR deletes five parameter paths (rate_reduction/amount/{single,joint,separate,head_of_household,surviving_spouse}). Saved policies that edit them would break. Say so in the PR body or changelog.
• (c) An unrecorded conflict in the same credit.
  • The statute: § 606(ggg)(4-a)(B) sets the fixed amount for unmarried, head-of-household and separate filers at "sixty-two dollars and fifty cents".
  • The form and the model: fixed/amount.yaml uses the IT-201-I's $63, and its own comment says "IT-201-I Table 1 rounds the statutory $62.50 ... to $63".
  • The record. This statute-versus-form conflict is in neither candidates.yaml nor the register.
  • Why mention it here. It is outside this diff, but it is the mirror of item 1: there the PR rejects a form rounding in favour of the statute.
  • Fix: record it as a candidate, or fix it in a follow-up.

G5. Washington age threshold

• The text. The PR body notes that age_threshold.yaml uses 62 before 2020 although the 2008 and 2015 texts say sixty-one.
• The timing. The test is "on December 31st of the year in which the exemption claim is filed", and the exemption applies to taxes "due and payable in the year following the year in which a claim is filed". So in a collection-year period, which is how the new veteran parameters are keyed, a claimant is 62 by the end of the collection year.
• Fix (optional): file the follow-up, and settle the period convention in it, since the 2020+ value of 61 may then be the one that's off.

G6. PR body

• Stale test plan. It lists files "Not run here ... CI covers these". CI has now run all of them (see CI status).
• Bill names. The reference titles call the 2019 and 2025 bills "SSB 5160" and "HB 1106". The session laws are ESSB 5160 and EHB 1106.
• Fix (optional): update both, and fix the Axiom line (S1).

Verified correct

• Hand-checks (22 expectations).
  • NYC:
    1. 2017 head of household, $20,000: 25 + 0.00228 × 5,600 = 37.768.
    2. 2025 joint, $30,000: 37 + 0.00228 × 8,400 = 56.152.
    3. nyc_school_tax_credit.yaml, single $441,000: 21 + 0.00228 × 429,000 = 999.12.
    4. 2017 separate at $12,000: "not over", so 0.00171 × 12,000 = 20.52. Joint at $21,600: 36.936.
    5. Mixed vector: 0.00171 × 10,000 = 17.1; 25 + 0.00228 = 25.00228; 37 + 0.00228 × 78,400 = 215.752.
    6. spm_unit_local_tax.yaml Case 1:
       • Taxable income. 100,000 − 8,000 = 92,000.
       • Amount. 21 + 0.00228 × 80,000 = 203.40, up $0.48.
       • NYC tax. 3,175.1702 − 0.48 = 3,174.6902.
       • SPM net income. It is the float32 output (62,902.109 against 62,902.113 by hand).
    7. Contrib Case 5, joint $263,950: 37 + 0.00228 × 242,350 = 589.558, plus 125 = 714.558.
    8. Elimination-credit Cases 1 and 3:
       • Case 1: 63 + 25 + 0.00228 × 8,400 = 107.152.
       • Case 3: 125 + 37 + 0.00228 × 18,896 = 205.08288, so nyc_income_tax = 1,375.71552 − 192.25944 − 205.08288 = 978.3732.
  • Indiana (2026):
    9. Case 9: min(0.1 × (2,000 − 1,000), 300) = 100.
    10. Case 10: max(1,000 − 1,200, 0) = 0, so 0.
    11. Case 6, joint, liability 4,000: min(400, 300) = 300.
    12. Blind/disabled aggregate:
    - Liability. 2,000 − 150 − 125 = 1,725.
    - Credit. 172.50.
    - Umbrella. 150 + 125 + 172.50 = 447.50.
    13. Over-65 aggregate: 0.1 × 1,850 = 185; umbrella 335.
  • Virginia (2024):
    14. Uneven accounts, capped amount supplied (5,000): 5,000; main gave min(8,000, 8,000) = 8,000.
    15. Default with a $2,000 carryforward and one account: min(3,000 + 2,000, 4,000) = 4,000.
    16. Age 70: 1,000 + 6,000 = 7,000, with the supplied cap ignored.
    17. Spouses: [7,000, 500 + 2,500 = 3,000], so 10,000. Dependent owner: [0, 0].
  • Washington:
    18. Case 8 (2024): generic disability and not retired, so ineligible and 0. Main gave 1,500 + 300,000 × 0.007 = 3,600.
    19. Case 9: 5,000 × 0.30 + max(60,000, 300,000) × (3,500 / 500,000) = 3,600.
    20. Integration Case 2 (King County lookup): 4,800 × 0.30 = 1,440; 300,000 × 0.00672 = 2,016; total 3,456.
    21. Case 10: 0.4 < 0.8 in 2026, so false; 0.4 ≥ 0.4 in 2027, so true. Cases 16-18 (2019): 0.9 against .inf without a total rating → false; rating 1 → total → true; total flag → true.
    22. Case 21 vector:
    - 61 ≥ 61 → true;
    - generic disability only → false (is_veteran is veterans_benefits > 0, which is 0);
    - retired → true;
    - compensated veteran at 0.8 → true.
• References.
  • NYC. S.2009-C file pages 8-9 hold the (4-b) tables and file page 14 holds § 7, so the title "#page 8-9,14" follows CLAUDE.md. IT-201-I 2017 p. 30, 2018 p. 31 and 2025 p. 21 show line 69a.
  • Washington. The 2008 session law p. 85 shows (3). 2019 c 453 file pages 3 and 13 show § 1(3)(a)(ii) and § 9. 2025 c 200 file pages 3 and 5 show "((80)) 40" and § 3.
  • Indiana. SEA 1 p. 124 shows 7.7, and the DLGF FAQ pp. 7-9 show questions 26-33. The variable tuples carry # PDF pages 7-9 comments.
  • Virginia. File page 28 shows code 104.
• Code.
  • NYC. p.threshold[filing_status] and p.base[filing_status] cover all five statuses. max_(nyc_taxable_income, 0) handles negative income. where(taxable_income > threshold, ...) keeps the threshold itself in the lower band, which is the statute's "not over". The $500,000 limits stay in the unchanged eligibility variable.
  • Washington. where on the scalar rating_requirement_in_effect broadcasts. The .inf threshold makes rating >= threshold false for 2016-2019. rating >= 1 implements 38 CFR 3.340(a)(2).
  • Vectorization. Everything is vectorized, and the mixed NYC and WA cases exercise it.
• Callers at the head.
  • No orphaned paths. No code reads the deleted NYC scales; the two NYC STC reforms read the variable, not the parameters.
  • in_supplemental_homestead_credit is read only through state_property_tax_credits.yaml (2026 block).
  • wa_pte_categorical_eligible is read only by wa_pte_eligible.
  • va_529_plan_deduction_person is unchanged downstream.
• Batch lessons:
  • Source-conflict ratchet (Tie source-conflict markers to the register shared with Axiom #10287).
    • In CI. test_conflict_sounding_comments_do_not_grow passed in Rest (Python, shard-1) on merge commit 78ef47c3, which merges the head into fec98b87c8 (Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334).
    • Statically. I applied main's CONFLICT_COMMENT_RE to the 24 scanned files the PR touches:
      • the only match at the merge base was wa_pte_categorical_eligible.py ("subsumed under is_disabled in practice");
      • it has 0 at the head;
      • the PR deletes its KNOWN_UNMARKED entry, as the guard requires.
    • Test YAML is not scanned.
  • Source conflicts. None of the four items is a statute-versus-form conflict (see Summary), so no marker or candidate is required. The PR body's "not applicable" is right, apart from G2's wording and G4(c)'s adjacent conflict.
  • NH Add New Hampshire state supplementation for 2026 #10156 (state-gated keys): not applicable.
    • The keys. The only parameter keys are filing_status, which is not gated.
    • The new Washington and Indiana inputs have no defined_for.
    • va_529_plan_deduction_capped_amount is gated to VA but is a float used in arithmetic.
  • County lookups. The PR doesn't touch the county-median income tests. Integration Case 2 still runs the King County lookup. The open crash fix Fix WA property tax exemption crash for households outside Washington #10113 merges cleanly with this head, and Guard county-keyed lookups in VT renter credit, MN CCAP, WI Shares and MD CCS #10116 has no Washington file.
  • Integrate modeled state unemployment insurance into income flows #9558, Document disjoint health premium inputs and correct consumer treatment #10046 and OK Add Oklahoma state supplementation for 2026 #10165: not applicable.
  • CI cost. The PR adds no Python test, fixture or reforms: case. The only Python edit is the NYC invariant test's tolerance (0.5 → 1e-3).
  • Partner tests: untouched. The Household API Partners check passed.
• Changelog. changelog.d/statute-bugs-nyc-in-va-wa.fixed.md is a top-level fragment, and fixed is the right type.

Axiom line assessment

Invalid: axiom: pending: hub files parity (S1).

• Recommended form: axiom: us-ny:policies/tax/it-201-instructions/nyc-school-tax-credit-rate-reduction#nyc_school_tax_credit_rate_reduction_amount encoded-correct; TheAxiomFoundation/rulespec-us#<ny> queued; TheAxiomFoundation/rulespec-us#1510 queued; TheAxiomFoundation/rulespec-us#<in> queued; TheAxiomFoundation/rulespec-us#<wa> queued.
  • encoded-correct covers the NYC amount for TY2025+ only. The module's $37 / $25 / $21 bases match this head.
  • <ny> extends What program participation can affect one's eligibility for the Affordable Connectivity Program? #1452, or is a new issue, for TY2017-2024 under § 606(ggg)(4-b).
  • Allow data tests to work without a policyengine_us package being installed #1510 (open, pe-parity, dispatch-ready, blocker recorded) already requires per-account caps and carryforward. Add this PR's three carryforward cases.
  • <in> and <wa> are new open, pe-parity-labelled, dispatch-ready issues with the module targets and statute-derived companion cases listed under S1.
• Not n/a. These are statutory provisions with changed results in claimed years.

CI status

27 pass, 3 skipping, 0 pending, 0 failed. I re-checked gh pr checks 10330 at the end of this review, and the head is still 29778a23d8.

• The green run. Run 39617825642 (pull_request, created 2026-10-21 14:33 UTC) tested merge commit 78ef47c3, which merges the head into fec98b87c8 (Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334).
  • Correction to the expected run. The run that started at 02:32 UTC (39533622018) was skipped, and run 39608362524 (13:28 UTC) was cancelled. Their umbrella jobs are the three skipped checks.
  • What the base contains. CI's tree contains Tie source-conflict markers to the register shared with Axiom #10287 and Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334, and the ratchet passed. The known FL/SD failure doesn't apply.
• Where the changed tests ran:
  • Rest-a: the NYC files (nyc_school_tax_credit_rate_reduction_amount.yaml 16 cases, nyc_school_tax_credit.yaml, the elimination integration.yaml 6, and ..._eligible.yaml).
  • States-shard-1: the Indiana unit file (11) and both aggregate files.
  • States-shard-2: the Washington files (categorical 21, integration 14, eligible, exemption 18).
  • States-shard-4: the Virginia files (va_529_plan_deduction.yaml 16, integration.yaml 3).
  • Household: spm_unit_local_tax.yaml.
  • Contrib other-shard-1: both NYC STC reform files.
  • Rest (Python + variables): the NYC invariant tests and test_va_529_invariants.py.
  • Full suites. Microsimulation and Household API Partners passed.
• Locally: I ran no tests or model probes. CI is green at the current head. The expectations and the S2/S3/G1 traces were checked by hand from the code, parameters and sources. The ratchet regex was applied statically.

Branch status

• Position: 7 commits ahead of main (ed71efe294) and 11 behind. The merge base is ce9377ee80 (Tie source-conflict markers to the register shared with Axiom #10287).
• Merge: clean.
  • git merge-tree --write-tree PolicyEngine/main 29778a23d8 exits 0 (tree f9436a768e).
  • main since the merge base touched only test_source_conflicts.py among the PR's files, through Count the FL and SD LIHEAP comments in the source-conflict ratchet baseline #10334, which CI's tree includes.
  • Merged after CI's base: Apply the federal 110% poverty floor to Indiana EAP, Maryland MEAP and Riverside County LIHEAP #10281, Support state-specific other additions and subtractions for remaining 30 states (Phase 2) #10302, Count gross unemployment compensation and Social Security in South Carolina TANF #9687 and Keep claimable filers' own exemptions under the HI and VA dependent-exemption repeal #10271. None touches a PR file. Support state-specific other additions and subtractions for remaining 30 states (Phase 2) #10302's new other-subtraction inputs default to 0, and Keep claimable filers' own exemptions under the HI and VA dependent-exemption repeal #10271 changes only the dependent-exemption repeal reform.
• Siblings (note only). Each merges cleanly with this head as text:
  • Fix four statute-first benefit bugs: MO TANF rounding, SNAP GU/VI limits, HI CCAP TANF income, SSI food ISM date #10332 edits a different hunk of KNOWN_UNMARKED.
  • Point the NM-WY source-conflict candidates at their code on main #10352 edits candidates.yaml, which this PR doesn't touch. It would be the file for G2 or G4(c).
  • Follow state statutes where forms differ: NM medical exemption, VT CDCC, NY/NYC household credit, PR EITC #10300 edits KNOWN_UNMARKED and the NYC household credit.
  • Fix WA property tax exemption crash for households outside Washington #10113 (the Washington crash fix) also edits wa_senior_disabled_property_tax_exemption.yaml.

0 critical, 3 should and 6 suggestions are open at this head.

Verdict: REQUEST_CHANGES

[DTrim99]

All four changes follow the governing statute in the claimed years (NYC 606(ggg)(4-b), IN 6-1.1-20.6-7.7, VA 58.1-322.03(7), WA 84.36.381(3)(a)), and 22 expectations hand-check. Three should-address items: (1) the 'pending: hub files parity' axiom line is invalid; (2) the new IN and WA inputs default to off, so the IN credit is $0 and the WA disability and veteran pathways close for every household (microsim included) unless they're set, so derive the defaults; (3) pre-2016 WA veterans still qualify on generic is_disabled. Six suggestions. Full report in the comment above.

[MaxGhenis]

Picking this up for the US hub. I'll post a hand-off comment here when it's ready.

---

Generated by Claude Code

[MaxGhenis]

Ready for hub review

Head: b403e928f160730e92868be2891f74513e55125e

Picked up at 29778a23d8 (unchanged since 2026-10-21). Two commits on top:

• 433fccbb1e: a clean merge of main.
• b403e928f1: the review changes.

What the item asked for

2. The new IN and WA inputs default to off

Each input is now a formula that a supplied value overrides, with the default documented in its documentation.

Indiana:

• Qualification. in_supplemental_homestead_credit_eligible.py:28-29 defaults to real_estate_taxes > 0, the homeowner proxy in_over_65_property_tax_credit_eligible uses. The docstring (:8-18) notes the credit needs no application (7.7(e)) and that the model has no input for the IC 6-1.1-12-37 standard deduction.
  • It now has defined_for = StateCode.IN.
• Liability. in_homestead_property_tax_liability.py:32-39 defaults to max_(real_estate_taxes − in_over_65_property_tax_credit − in_blind_disabled_property_tax_credit, 0). IC 6-1.1-20.6-3(3), as amended by SEA 1 SECTION 73, orders the 7.7 base after the section 7 and 7.5 credits.
  • That citation is now in the reference tuple, merged with SECTION 74 as one SEA 1 entry (#page=123, # PDF pages 123-124), per the one-entry-per-document rule. It is also on the property credit (G3).
• Referendum tax. in_homestead_referendum_property_tax stays an input with default 0.

Washington:

• Disability retirement. wa_pte_retired_due_to_disability.py defaults to social_security_disability > 0. RCW 84.36.383(6) adopts 42 U.S.C. 423(d)(1)(A), and an SSDI award is SSA's finding under that definition.
• VA compensation and total rating. wa_pte_receives_va_disability_compensation.py and wa_pte_va_total_disability_rating.py both default to is_fully_disabled_service_connected_veteran.
• Combined rating. wa_pte_va_service_connected_rating stays an input with default 0.

Microsimulation-facing behaviour, before → after (no microsimulation run here; from the formulas and which inputs the household and microdata supply):

• Indiana supplemental homestead credit, 2026+.
  • Before: $0 for every tax unit that didn't set the inputs, so all of microsim.
  • After: min(10% × (real_estate_taxes − Over 65 credit − Blind/Disabled credit), $300) for every Indiana tax unit with real_estate_taxes > 0. That is main's amount (min(10% × real_estate_taxes, $300), which I read on main), less 10% of any earlier credit.
  • The referendum exclusion stays at $0 until data supplies it.
• Washington exemption.
  • Before: only the age pathway was open.
  • After, the disability pathway: open to anyone with social_security_disability > 0, which the household calculator and the CPS-based data supply.
  • After, the veteran pathway: open from collection year 2016 to anyone marked is_fully_disabled_service_connected_veteran. I didn't check whether the microdata populates it; if it doesn't, that pathway stays closed in microsim.
  • Combined ratings below a total rating stay unmodelled unless supplied.
  • Before 2016 see item 3.
• Reach. Neither output feeds household net income: the IN credit feeds taxsim_state_property_tax_credit via state_property_tax_credits.yaml, and the WA exemption has no consumer (per DTrim99's caller check, which I didn't repeat).

3. Pre-2016 WA veterans on generic is_disabled

I kept the hub's ruling and made the assumption visible.

• The new input. wa_pte_wholly_service_connected_disability.py is a Person bool used only with is_veteran and only before collection year 2016. It defaults to is_disabled | is_fully_disabled_service_connected_veteran.
• Its documentation states the ruling's reading, that the proxy over-includes a veteran whose disability isn't service-connected, and that setting it to false removes that veteran.
• The pathway. wa_pte_categorical_eligible.py:38-44 reads it for earlier_veteran. It isn't closed: Case 20 (2015, is_disabled veteran) still expects true.
• New cases. Case 25 (2015, is_fully_disabled_service_connected_veteran only → true) and Case 26 (2015, is_disabled with the input set to false → false).

Suggestions

• G1, Virginia: taken.
  • va_529_plan_deduction_capped_amount.py:36-39 counts at least one account when a carryforward remains.
  • va_529_plan_deduction_person.py:35-38 bounds the under-70 deduction by contributions plus carryforward.
  • Two new cases: carryforward-only with no beneficiary count → 3,000; supplied $9,000 against $2,500 available → 2,500.
• G2, the VA age-70 wording: taken. The va_529_plan_deduction.yaml header (:171-180) and the PR body now say the instructions agree on the per-account cap but are silent on the age-70 carryforward, so (7)(b) governs. No candidate, since there is no evidence Tax limits age-70 filers to the current year.
• G3, the IN citation: taken. The pre-2026 gate: skipped. It predates the PR, and the umbrella list already excludes the credit before 2026.
• G4.
  • (a) Pre-2017 NYC zero: skipped (predates the PR).
  • (b) Removed parameter paths: taken. The changelog and body name the five removed scales.
  • (c) The $62.50 / $63 conflict: taken as a candidate, us-ny:nyc-school-tax-credit-fixed-amount-rounding (follows: conflicting-source, unreviewed), in source_conflicts/candidates.yaml.
• G5, WA age threshold: skipped. It's a follow-up, below.
• G6.
  • The body is updated: test plan, defaults, the pre-2016 ruling, the VA wording.
  • The bill names in veteran/combined_rating_threshold.yaml:18,20 ("SSB 5160", "HB 1106") are not changed. I couldn't reach the legislature site to confirm the ESSB / EHB designations, and the session-law file is 5160-S.SL.pdf. unverified.

Hand computations (new expectations)

• IN Case 12 (default, under 65, $2,000 of taxes): no earlier credit, liability 2,000, credit 200.
• IN Case 13 (default renter): not eligible, 0.
• IN Case 14 (disabled, $1,000 of taxes): Blind/Disabled credit 125; liability 1,000 − 125 = 875; credit 10% = 87.50.
• IN aggregates, now derived:
  • Over-65 case: 2,000 − 150 = 1,850, credit 185, umbrella 335.
  • Blind/disabled case: 2,000 − 150 − 125 = 1,725, credit 172.50, umbrella 447.50.
• WA Case 22 (2024, age 45, SSDI 12,000): retired by reason of disability, eligible.
• WA Case 23 (is_disabled, no SSDI): not eligible.
• WA Case 24 (fully disabled service-connected veteran): compensated, total rating, eligible.
• WA Cases 25 and 26 as above.
• VA: min(0 + 3,000, 4,000 × 1) = 3,000; min(9,000, 2,000 + 500) = 2,500.

Tests run (one file per process, at this head)

• wa_pte_categorical_eligible.yaml: 26 passed (5 new).
• WA integration.yaml: 14 passed.
• wa_pte_eligible.yaml: 9 passed.
• wa_senior_disabled_property_tax_exemption.yaml: 18 passed.
• in_supplemental_homestead_credit.yaml: 14 passed (3 new).
• in_supplemental_homestead_credit_in_aggregate.yaml: 2 passed.
• in_blind_disabled_property_tax_credit_in_aggregate.yaml: 2 passed.
• va_529_plan_deduction.yaml: 18 passed (2 new).
• VA plan_529/integration.yaml: 3 passed.
• nyc_school_tax_credit_rate_reduction_amount.yaml: 16 passed.
• nyc_school_tax_credit.yaml: 5 passed.
• income_tax_elimination/integration.yaml: 6 passed.
• spm_unit_local_tax.yaml: 4 passed.
• contrib/local/nyc/stc/..._and_eligibility_by_children.yaml: 9 passed.
• contrib/local/nyc/stc/..._joint_override.yaml: 2 passed.
• test_nyc_school_tax_credit_income_test_invariants.py: 6 passed.
• test_va_529_invariants.py: 2 passed.
• test_source_conflicts.py: 103 passed.

Mutants (each in its own process, restored after)

Mutant	Result
M1 IN liability ignores earlier credits	killed (1 failed)
M2 IN default qualification off	killed (2)
M3 WA SSDI default off	killed (1)
M4 WA total-rating default inverted	killed (2)
M5 WA pre-2016 back to bare is_disabled	killed (2)
M6 VA carryforward-only account not counted	killed (1)
M7 VA under-70 bound removed	killed (1)

7 of 7 killed.

CI

Pull request run 41418693987 at b403e928f1: success, all 34 jobs passed, including Microsimulation and Household API Partners.

Judgment calls

1. Defaults that read other variables. As DTrim99 notes (from Fix Montana disability exclusion repeal, elderly credit receipt timing and occupancy tests #10331), supplying one of these inputs for one unit gives other units the default array, not the formula. I left them as single variables, as he allows.
2. The pre-2016 default also accepts is_fully_disabled_service_connected_veteran. That veteran meets the ruling's reading (wholly service-connected, under the Social Security definition), so it only adds people the statute admits.
3. Indiana defined_for. I added defined_for = StateCode.IN to the two IN defaults, so they read Indiana-only credits and are 0 elsewhere. The credit itself was already Indiana-only.
4. Unverified. The statutory phrasings in the new docstrings (7.7(e), § 3(3), RCW 84.36.383(6) and the 2008 text) come from DTrim99's report and the PR's references. I couldn't fetch iga.in.gov or leg.wa.gov here.

Proposed follow-up issue

Washington senior/disabled exemption: settle the age threshold and its period before 2020. age_threshold.yaml uses 62 before 2020, though the 2008 and 2015 texts of RCW 84.36.381(3)(a) say sixty-one, tested at December 31 of the filing year. Since the exemption applies to the following collection year, decide whether the model's period is the filing or the collection year, and set the pre- and post-2020 values on that basis.

Reply to DTrim99

Thanks for the review. Taking each item in turn:

1. S1, the axiom line. The hub's.
2. S2, defaults. Every new input now has an overridable default, as you suggested.
   • IN qualification: real_estate_taxes > 0.
   • IN liability: real_estate_taxes less the Over 65 and Blind/Disabled credits (§ 3(3)).
   • WA disability retirement: SSDI receipt.
   • WA VA compensation and total rating: is_fully_disabled_service_connected_veteran.
   • New default cases: IN Cases 12–14, WA Cases 22–24.
   • The IN aggregates now derive 1,850 and 1,725.
   • The before and after microsim behaviour is above.
3. S3, pre-2016 veterans. The hub ruled on this: before collection year 2016 the statute means Social Security's definition, wholly service-connected. The model has no service-connection input for that, so is_disabled is the proxy, which keeps the pathway open rather than closing it for every veteran.
   • So I kept the reading but made it explicit. wa_pte_wholly_service_connected_disability defaults to is_disabled (or is_fully_disabled_service_connected_veteran), documents that it over-includes non-service-connected cases, and can be set to false.
   • Case 20 still expects true. Case 25 covers the service-connected input and Case 26 the override.
4. G1. Both changes taken, with two cases.
5. G2. Wording fixed in the YAML header and the body.
6. G3. Citation added. The pre-2026 gate is skipped: it predates the PR, and the umbrella excludes the credit.
7. G4. (a) skipped (predates the PR). (b) and (c) taken, (c) as a candidate.
8. G5. Proposed as a follow-up issue.
9. G6. The body is updated. The bill names are unchanged, because I couldn't verify ESSB / EHB from here.

---

Generated by Claude Code