Remove Montana property tax rebate from non-refundable credit list

[DTrim99]

Summary

The Montana Property Tax Rebate (up to $675 for TY2022 and TY2023) is a standalone rebate — homeowners claim it on a separate application to the Department of Revenue (tap.dor.mt.gov, window Aug 15 – Oct 1, 2024 for TY2023). It is not a Form 2 nonrefundable income-tax credit and does not reduce income tax liability.

It was listed in gov.states.mt.tax.income.credits.non_refundable (the 2022-01-01 entry), so mt_income_tax_before_refundable_credits_joint subtracted it. Two compounding defects:

1. Misclassification — a standalone direct-payment rebate booked as a Form 2 line-19 credit.
2. Household multiplication — mt_non_refundable_credits is a Person variable that adds the list; mt_property_tax_rebate is a TaxUnit variable ($675), so PolicyEngine projected it onto every member of the tax unit ($675 × 3 = $2,025 for a family of three).

Net effect: Montana income tax understated by $675 per household member.

Change

Remove mt_property_tax_rebate from the 2022-01-01 credit list, leaving [mt_capital_gain_credit]. The rebate variable still computes its $675 value — it simply no longer reduces income tax. (The 2024-01-01: [] entry already drops it thereafter; mt_income_tax_rebate in the 2021 entry is untouched, as the emulator's srebate mechanism relies on it.)

Testing

Verified on the issue household (MT joint 2023, ages 26/25, 1 dependent, $114,039 wages, $204 interest, $2,755 property tax): mt_non_refundable_credits goes from [675, 675, 675] → [0, 0, 0] and mt_income_tax from 3,683.16 → 5,708.16, matching the TaxAct Form 2 (line 18 = line 20 = $5,708, credit schedule blank). Added a regression test to the Montana integration suite.

Found via PolicyEngine-TAXSIM discrepancy #1152 (and its 2022 counterpart #1150), reported by @feenberg.

Sibling of #8970 (@PavelMakarchuk), which scopes the Montana income tax rebate — same subsystem, different rebate, no file overlap.

Fixes #9347.

🤖 Generated with Claude Code

[PavelMakarchuk]

PR 9348 — Consolidated review report

"Remove Montana property tax rebate from non-refundable credit list" — DTrim99 (fixes #9347; found via TAXSIM discrepancies policyengine-taxsim#1150/#1152)

Source Documents

• Reviewed head SHA: 3066fdce840330ef55ddab14dab4e30046e9fd4f
• Mode: full (delegate roles: regulatory [with pdf-collector/audit folded in], references, code-pattern, tests, verification-planner)
• Snapshot: /tmp/policyengine-command-runs/2b3d452ec8fd/pr9348-pr-snapshot
• Role reports: pr9348-review-context.md, pr9348-review-regulatory.md, pr9348-review-references.md, pr9348-review-code.md, pr9348-review-tests.md, pr9348-review-verification-queue.md (all under RUN_ROOT)
• Archived primary sources (Wayback, byte-identical to originals): pr9348-mt-form2-2022-instructions.pdf (1,403,508 B), pr9348-mt-form2-2023-instructions.pdf (1,412,574 B)

Branch Status

• BEHIND=0, AHEAD=2 — current with main, no rebase needed.
• CI: all 31 checks pass.
• Diff (63 lines, 3 files): policyengine_us/parameters/gov/states/mt/tax/income/credits/non_refundable.yaml (removes mt_property_tax_rebate from the 2022-01-01 list), policyengine_us/tests/policy/baseline/gov/states/mt/tax/income/integration.yaml (1 new case), changelog.d/mt-property-tax-rebate.fixed.md.

Verified substance (why the removal is correct)

The core claim is source-confirmed against both years' primary documents:

1. Not a Form 2 credit. The 2023 MT Form 2 Instructions (archived copy, PDF p.5) describe the Property Tax Rebate (HB 222/HB 816, TY2022–2023, lesser of property tax paid or $675) as claimed separately via the TransAction Portal or Form MPTR23 in fixed claim windows, and state the rebate "is exempt from Montana income tax" — it sits entirely outside the income tax calculation. The complete Nonrefundable Credits Schedules for both TY2022 (PDF p.16, pp.43–44) and TY2023 (PDF p.18, pp.46–49) were enumerated line by line: the Property Tax Rebate appears in neither. The 2022 instructions (published Dec 2022, pre-HB 222) never mention it at all.
2. Tripling mechanism confirmed in code. mt_non_refundable_credits is a Person-entity variable with a parameter adds list; mt_property_tax_rebate is TaxUnit-entity. policyengine-core projects the tax-unit value onto every member in a Person-level adds, so a 3-member household carried 3 × $675 = $2,025 of phantom credit (verified in core map_to/project semantics).
3. Arithmetic re-derived exactly. Issue-9347 household (MT joint, 2023): tax before non-refundable credits $5,708.16; pre-fix $3,683.16 = 5,708.16 − 2,025; post-fix $5,708.16, matching the PR body's TaxAct verification. Both new test assertions (mt_property_tax_rebate: 675, mt_non_refundable_credits: [0, 0, 0]) were independently hand re-derived, and the case is confirmed discriminating (pre-PR code yields [675, 675, 675]).
4. No collateral damage. mt_property_tax_rebate is not orphaned (still consumed via gov.states.household.state_property_tax_credits → taxsim_state_property_tax_credit, plus its own unit test). The remaining list entries — mt_capital_gain_credit (2021/2022) and mt_income_tax_rebate (2021) — are both Person-entity, so no analogous projection defect remains in this list.

Critical

None in this PR's diff. The one-line parameter change is source-confirmed correct for both affected years, the regression test is sound and discriminating, and CI is green.

Disposition note on the references role's C1 (dead 2022 href): the mtrevenue.gov links were dead before this PR (MT DOR site migration; this PR adds no hrefs), the cited content was successfully retrieved and corroborated from archived and live-mirror copies, and verified replacement URLs exist. Per the pre-existing-issue discipline it does not block; it is carried as A1 below because the file is being edited anyway.

Pre-existing findings (outside this diff — follow-up issue material)

The code role's repo-wide entity-projection sweep (6,040 variables AST-parsed; 817 adds lists resolved) found 8 live instances of the same bug class this PR fixes — a Person-entity adds list containing a group-entity variable whose projected value is then summed or subtracted per person. All 8 pre-date this PR, none is touched by its diff, and per canonical rules none blocks it. Recommend opening a tracking issue (one umbrella or eight singles) citing verified core semantics: group→person adds projects the FULL group value onto each member.

#	Person-entity variable (file)	Offending group-entity entry	Consumer / mechanism
P1	policyengine_us/variables/gov/states/mt/tax/income/deductions/itemized/general/mt_misc_deductions.py	casualty_loss_deduction [TaxUnit]	mt_itemized_deductions_joint.py:24 sums per member (× household size); mt_itemized_deductions_indiv.py:29 doubles on separate-on-same-return. Same MT subsystem as this PR.
P2	policyengine_us/variables/gov/local/ca/la/general_relief/la_general_relief_gross_income.py	tanf [SPMUnit]	la_general_relief_net_income.py:16 — TANF counted once per member; exact analogue of #9347
P3	policyengine_us/variables/gov/states/ms/tax/income/ms_agi_adjustments.py	self_employed_health_insurance_ald, health_savings_account_ald, self_employed_pension_contribution_ald [TaxUnit]	ms_agi.py:20 subtracts full unit ALD from each non-dependent → double-counted for couples
P4	policyengine_us/variables/gov/states/oh/tax/income/deductions/medical_exepenses/oh_unreimbursed_medical_care_expense_deduction_person.py	oh_insured_unreimbursed_medical_care_expenses [TaxUnit]	oh_unreimbursed_medical_care_expense_deduction.py sums members → × member count
P5	policyengine_us/variables/gov/states/wv/tax/income/subtractions/senior_citizen_disability/wv_senior_citizen_disability_deduction_total_modifications.py	us_govt_interest [TaxUnit]	wv_senior_citizen_disability_deduction_person.py:22 — deduction understated for couples
P6	policyengine_us/variables/gov/local/ca/riv/general_relief/income/earned/ca_riv_general_relief_earned_income_deductions.py	state_withheld_income_tax, additional_medicare_tax [TaxUnit]	ca_riv_general_relief_net_earned_income.py — deductions double-counted per earner
P7	policyengine_us/variables/gov/local/ca/riv/general_relief/property/ca_riv_general_relief_countable_property_value.py	spm_unit_cash_assets, ca_riv_general_relief_countable_vehicle_value [SPMUnit]	ca_riv_general_relief_property_eligible.py:14 auto-sums members → assets × member count → wrongly fails eligibility
P8	policyengine_us/variables/gov/states/il/dhs/aabd/income/expense/il_aabd_expense_exemption_person.py	state_withheld_income_tax [TaxUnit]	il_aabd_earned_income_after_exemption_person.py:17 — exemption double-counted for couples

Additional pre-existing follow-ups (should-address grade, outside this diff):

• P9 — Rebate now reaches household net income through no channel at all (TY2022–2023). After this PR, mt_property_tax_rebate's only consumer is the TAXSIM reporting aggregate, which feeds nothing in household_net_income. Per the 2023 instructions p.5 the rebate is real money (direct payment, up to $675/yr, income-tax-exempt); faithful modeling is a benefit-side transfer for those historical years. Removing the wrong channel first is correct — track the missing right channel as a follow-up.
• P10 — 2024+ leakage into the TAXSIM aggregate. parameters/gov/states/mt/tax/income/credits/rebate/property/amount.yaml:3-4 has no sunset (comment says "only offered in the 2022 and 2023 tax years"), and parameters/gov/states/household/state_property_tax_credits.yaml still lists mt_property_tax_rebate at 2024/2025/2026 (lines 131, 156, 186) — a phantom $675 MT "property tax credit" in years the program no longer exists. Companion fix: add 2024-01-01: 0 and drop the 2024+ aggregate entries.
• P11 — 2021 mt_income_tax_rebate likely carries the same misclassification (the 2023 instructions describe the HB 192 income tax rebate as "automatically issued", i.e. another standalone rebate; the 2021 Line-19 list does not name it). This is open PR #8970's territory — flag there rather than expanding this PR.

Should Address

(Actionable within this PR; none blocking.)

• A1 — Refresh the dead mtrevenue.gov hrefs while the file is open (from references C1+S1). 4 of 5 cited hrefs 301→404 after the MT DOR site migration, including the one backing the edited 2022 entry. Verified working replacements at revenuefiles.mt.gov are supplied in pr9348-review-references.md (C1/S1 tables), with page anchors re-verified — note the 2022 anchor corrects #page=17 → #page=16 ("Line 19–Nonrefundable Credits" is on file p.16). While there: the 2025 reference title should name its section ("Schedule III – Tax Credits, Part I – Nonrefundable Credits") and line 10/13 trailing whitespace can be dropped.
• A2 — Cite the removal itself in non_refundable.yaml (references S2). The "not a Form 2 credit" rationale lives only in the PR body and a test comment. Add the verified 2023 What's New citation (.../2023_Montana_Individual_Income_Tax_Return_Form_2_Instructions.pdf#page=5, "claimed via TransAction Portal or Form MPTR23... exempt from Montana income tax"). Do not add an MCA statute cite without verifying the section number (candidate 15-1-2302 unverified).
• A3 — TY2022 is the untested half of the change (tests SA-1). The removed entry governed TY2022 and TY2023; only 2023 is tested. Add a period: 2022 sibling case asserting mt_property_tax_rebate: 675 and mt_non_refundable_credits: [0, 0, 0].
• A4 — No end-to-end mt_income_tax assertion (tests SA-2, code SA-1, regulatory suggestion). The headline TaxAct-verified figure (3,683.16 → 5,708.16) is untested; a regression re-booking the rebate downstream of mt_non_refundable_credits would pass the new case. Add mt_income_tax: 5_708.16 to the issue-9347 case (run the model to confirm reproduction; the neighboring issue-9012 case already follows this pattern).
• A5 — Nothing pins that the rebate was rerouted, not dropped (tests SA-3). Add an aggregate test (pattern exists at tests/.../gov/states/household/state_property_tax_credits/*_in_aggregate.yaml for vt/pa/nd/in): MT household, 2023, assert taxsim_state_property_tax_credit: 675.

Suggestions

• S1 — New case uses state_fips: 30 while every other case in integration.yaml uses state_code: MT (lines 541, 563); the spm_units block is unnecessary for the asserted outputs. Cosmetic consistency.
• S2 — No case covers the 2024-01-01 [] band of the non-refundable list (e.g. 2024 MT filer with capital gains → mt_non_refundable_credits: 0). Cheap guard on the third date band.
• S3 — Pre-existing nits in tests/.../credits/rebate/mt_property_tax_rebate.yaml (untouched here): second case ("676 → 675") is mislabeled "Property tax less than rebate amount"; the exact boundary (675 → 675) is untested.
• S4 — Pre-existing: mt_property_tax_rebate.py:13-19 models none of the statutory conditions (principal residence, 7-month ownership and occupancy, taxes billed and paid, claim filing) — worth a comment or follow-up if the rebate gains a benefit-side channel (P9).
• S5 — Pre-existing: ma_mbta_enrolled_in_applicable_programs.py and tx_dart_reduced_fare_program_eligible.py use the Person-adds-group projection intentionally (boolean membership checks, never summed) — semantically correct today but fragile; a clarifying comment would inoculate them against the P1–P8 cleanup.

PDF Audit Summary

• Scope adaptation: pdf-collector and pdf-audit were folded into the regulatory role.
• 2 archived MT DOR Form 2 instruction PDFs (TY2022, TY2023) fetched from the Wayback Machine (captures 2025-04-14 / 2025-05-15, byte-identical to the originals), saved under RUN_ROOT, and quoted verbatim with #page= cites; the 2025 instructions verified at the live revenuefiles.mt.gov host.
• Key audited facts: $675 lesser-of cap, TY2022/TY2023 only, separate claim (TAP/Form MPTR23), income-tax-exempt, and absence from both years' Nonrefundable Credits Schedules — all corroborated (5/5 references corroborate on content; 4/5 hrefs dead → A1).
• The diff adds no #page= references (removal-only parameter edit + test + changelog), so the 5E page-anchor check is N/A; existing anchors were re-verified by the references role (one correction: 2022 → #page=16).

Validation Summary

Role	Result
Regulatory (incl. PDF collect/audit)	0 CRITICAL — removal source-confirmed in both years' primary documents; tripling mechanism and 3,683.16 → 5,708.16 re-derived exactly
References	1 CRITICAL reported (dead 2022 href) — judged pre-existing link rot, content corroborated via archives; carried as A1 with verified replacement URLs
Code patterns	0 CRITICAL in the diff; 8 pre-existing sweep criticals (P1–P8); 9/10 pattern categories clean, 1 N/A
Tests	0 CRITICAL — both asserted values hand re-derived; case confirmed discriminating (pre-PR → [675, 675, 675])
Verification queue	NONE — all cross-role questions settled in-queue; no EXT verification needed
Adjacency	Open PR #8970 touches the same MT credit wiring (2021 income tax rebate) — different rebate, no file overlap, no semantic conflict; can merge in either order. P11 flagged to it.

Review Severity

APPROVE (posting mode: approve-or-request-changes per user instruction).
Judged on THIS diff: zero critical findings; the change is minimal, source-confirmed against primary documents for both affected years, regression-tested with a discriminating case, and CI-green. The 8 sweep criticals are pre-existing and outside the diff — follow-up issue material, not blockers. The should-address items (A1–A5) are non-blocking hardening that would fit naturally in this PR or an immediate follow-up.

Next Steps

1. Merge-ready as-is; ideally fold in A1–A5 first (A1/A2 are copy-paste edits from the references report; A3–A5 are three small test cases).
2. Open a tracking issue for the entity-projection bug class (P1–P8, file:line list above) — prioritize P1 (mt_misc_deductions, same MT subsystem) while the team is in this code.
3. Open follow-ups for P9 (benefit-side channel for the TY2022–2023 rebate) and P10 (2024+ sunset + aggregate leakage).
4. Flag P11 (2021 income tax rebate misclassification) on PR #8970 rather than expanding this PR.

[DTrim99]

Review addressed — thanks @PavelMakarchuk

Applied all Should-Address (A1–A5) and Suggestion (S1–S4) items. No change to the substantive removal; no parameter value changed.

Should Address

• A1 — refreshed the dead mtrevenue.gov reference hrefs to live revenuefiles.mt.gov URLs (each verified HTTP 200), corrected the 2022 anchor #page=17 → #page=16, named the 2025 reference section ("Schedule III – Tax Credits, Part I – Nonrefundable Credits"), and dropped the trailing whitespace.
• A2 — cited the removal itself with the 2023 Form 2 "What's New" (p.5): the property tax rebate is claimed via the TransAction Portal / Form MPTR23 and is exempt from Montana income tax — not a Form 2 nonrefundable credit. (No MCA statute cite added, per your note.)
• A3 — added a period: 2022 sibling case (the removed entry governed TY2022 as well as TY2023) asserting mt_property_tax_rebate: 675, mt_non_refundable_credits: [0, 0, 0].
• A4 — added mt_income_tax: 5_708.16 to the issue-9347 case (the TaxAct-verified headline figure).
• A5 — added an aggregate test (mt_property_tax_rebate_in_aggregate.yaml) pinning that the rebate is rerouted to taxsim_state_property_tax_credit (675), not dropped.

Suggestions

• S1 — the new case now uses state_code: MT and drops the unnecessary spm_units block.
• S2 — added a period: 2024 case covering the empty ([]) non-refundable list band.
• S3 — relabeled the mislabeled "676 → 675" rebate case and added the exact 675 → 675 boundary.
• S4 — noted the unmodeled statutory eligibility conditions (principal residence, 7-month ownership/occupancy, taxes billed & paid, claim filing) on the rebate variable.
• S5 skipped — the clarifying comments on the unrelated ma_mbta_*/tx_dart_* files belong with the P1–P8 entity-projection cleanup, not this diff.

On the pre-existing findings (P1–P11): they're outside this diff, but the 8-instance Person-adds-group projection sweep (P1–P8) is worth a tracking issue — happy to open one (umbrella or per-instance), and to file P9/P10 (MT rebate benefit-side channel + 2024+ aggregate sunset). Just say the word.

I couldn't run the suite locally (broken 3.14-alpha venv), so the 5_708.16 and aggregate 675 values rely on CI to confirm.

Fixed with Claude Code assistance.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (724e795) to head (9792584).
⚠️ Report is 29 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9348   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            1         1           
  Lines           35        15   -20     
  Branches         2         0    -2     
=========================================
- Hits            35        15   -20     

Flag	Coverage Δ
unittests	100.00% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[DTrim99]

Program review — PR #9348: Remove Montana property tax rebate from the Form 2 non-refundable credit list

Program review

PR #9348 removes mt_property_tax_rebate from the Montana Form 2 non-refundable
income-tax credit list (parameters/gov/states/mt/tax/income/credits/non_refundable.yaml).
The Montana property tax rebate (HB 222/816, 2023; MCA 15-1-2304) is a standalone,
income-tax-exempt direct rebate claimed on a separate application (TransAction Portal /
Form MPTR23), not a Form 2 line item. It was incorrectly listed as a non-refundable
income-tax credit, which both (a) wrongly reduced Montana income tax and (b) was tripled
across household members because a TaxUnit-level $675 value was projected through a
Person-level adds list. The PR corrects the placement: the rebate is removed from the
income-tax credit list but retained in the property-tax-relief aggregate
(state_property_tax_credits.yaml → taxsim_state_property_tax_credit).

Source documents

• 2023 MT Form 2 instructions, "Legislation Affecting Tax Year 2023 / What's New —
  Property Tax Rebates" (PDF p.5): "House Bills 222 and 816 created property tax rebates
  ... equal to the lesser of the actual amount of property tax paid on a principal Montana
  residence, or $675 for each year. ... Taxpayers must file a claim for the rebate ... on
  our TransAction Portal or on the Form MPTR23. The rebate is exempt from Montana income
  tax but may be subject to federal taxation."
• 2023 MT Form 2 instructions, Nonrefundable Credits Schedule (PDF pp.39–43): "property
  tax rebate" appears zero times — negative corroboration that it is not a Form 2 credit.
• Four refreshed revenuefiles.mt.gov Form 2 instruction PDFs (2021, 2022, 2023, 2025),
  all resolving HTTP 200.
• MCA 15-1-2304 "Property tax rebate" (codified in Title 15 Ch. 1, outside the income-tax
  credit chapters; "Terminated. Sec. 10, Ch. 764, L. 2023") — suggested durable statutory
  cite, not yet in the code.

Branch status

Branch is 6 commits behind main. This is informational only, not a review finding; a
rebase/merge before merge is routine. CI is green (33/33).

Critical

None. The removal is regulatorily correct and fully corroborated by the added 2023
"What's New" citation and the negative corroboration from the credit schedule. The rebate
is not double-dropped: it still computes (~$675) and still flows into
taxsim_state_property_tax_credit via the untouched property-tax aggregate. The income-tax
delta (MT income tax rises 3,683.16 → 5,708.16 after the phantom credit is removed) is in
the correct direction and the magnitude is confirmed ($675 × 3 household members = $2,025).

Should address

Refresh the dead reference URL on mt_property_tax_rebate.py

• File: variables/gov/states/mt/tax/income/credits/rebate/mt_property_tax_rebate.py
  (reference = line, ~line 9).
• The variable's sole reference still points to
  https://mtrevenue.gov/wp-content/uploads/dlm_uploads/2023/12/Form_2_2023_Instructions.pdf#page=5,
  which is dead: mtrevenue.gov 301-redirects to revenue.mt.gov/..., which returns
  HTTP 404. This PR touched this file (added a 5-line comment) but left the stale link,
  even though it migrated the parameter file's hrefs off mtrevenue.gov to the resolving
  revenuefiles.mt.gov host in the same PR.
• This is a maintenance fix, not a blocking mis-corroboration: the substantive change is
  fully sourced by the live YAML citations plus the "What's New" source. But it is a dead
  reference on a file the PR modified, so it should be refreshed for consistency.
• Suggested fix (mirror the YAML refresh):

  reference = "https://revenuefiles.mt.gov/files/Forms/Montana-Individual-Income-Tax-Return-Form-2-Instructions/2023_Montana_Individual_Income_Tax_Return_Form_2_Instructions.pdf#page=5"
  

Suggestions

1. Add a durable statutory cite. Both files lack the enacting-law / codified citation.
   Form 2 instruction PDFs are annually re-issued and their paths have already migrated once,
   so they are the least durable source. Add HB 222 (2023) and/or MCA 15-1-2304 (live URL:
   https://mca.legmt.gov/bills/mca/title_0150/chapter_0010/part_0230/section_0040/0150-0010-0230-0040.html).
   MCA 15-1-2304 independently supports the removal: it sits in Title 15 Ch. 1, not the
   income-tax credit chapters.
2. Shorten the parameter reference title. non_refundable.yaml (~line 20) — the refreshed
   title embeds a long explanatory parenthetical ("rebate claimed via TransAction Portal or
   Form MPTR23; exempt from Montana income tax, not a Form 2 nonrefundable credit"). The
   href#page=5 is correct; consider moving the rationale into a YAML comment and keeping
   title: to the document/section name. Purely stylistic.
3. Name the elderly credit variable in the aggregate-test comment.
   mt_property_tax_rebate_in_aggregate.yaml (lines 24–25) — the comment says "elderly
   homeowner/renter credit = 0" without naming mt_elderly_homeowner_or_renter_credit.
   Minor clarity nit.
4. (Optional) Zero-property-tax edge in the aggregate test. No case with
   real_estate_taxes: 0 → mt_property_tax_rebate: 0 and taxsim_state_property_tax_credit: 0.
   The min_() formula makes this low-risk, but it would cheaply confirm the aggregate does
   not spuriously add.

Validation summary

• Regulatory: APPROVE. Removal correct per MT law; rebate is a separate-application,
  income-tax-exempt direct rebate, absent from the Form 2 Nonrefundable Credits Schedule.
  Not double-dropped — kept in the property-tax aggregate, removed only from income tax.
  Income-tax direction and magnitude confirmed.
• References: PASS on corroboration and on the four refreshed revenuefiles.mt.gov
  YAML links (all resolve). One dead .py link (Should Address) and one missing durable
  statutory cite (Suggestion).
• Code: 0 critical / 0 should / 3 suggestions. non_refundable.yaml list edit is clean
  (only mt_property_tax_rebate removed from the 2022 band; 2021 and 2024 bands untouched);
  .py change is comment-only, formula byte-for-byte identical; entity levels correct;
  changelog mt-property-tax-rebate.fixed.md present and top-level.
• Partner contract tests: NO IMPACT. Grep of tests/policy/baseline/partners/** found
  zero references to mt_income_tax, mt_non_refundable_credits, mt_property_tax_rebate,
  or state_code: MT. Consistent with green CI.
• Tests: Strong. All three regression effects are hard-pinned so a revert to the old
  behavior fails at least one assertion:
  1. No income-tax reduction / corrected higher tax — mt_income_tax: 5_708.16 (2023) and
     mt_non_refundable_credits: [0, 0, 0] (2022 & 2023) both flip from old values.
  2. Rebate still computed + in the aggregate — mt_property_tax_rebate_in_aggregate.yaml
     pins mt_property_tax_rebate: 675 and taxsim_state_property_tax_credit: 675
     (guards the aggregate-drop failure mode; passes under old behavior too, by design).
  3. No household multiplication — 3-person case pins [0, 0, 0] (was [675, 675, 675]).
     Boundary coverage improved (below cap 600, at cap 675, above cap 676→675); TY2022 band
     now covered. Minor gap: no zero-property-tax case (Suggestion Implement Section 8 Project-Based Rental Assistance #4). 24 affected YAML tests
     pass on the PR head.

Review severity

COMMENT. The change is correct, well-tested, and well-sourced. One should-address item
(refresh the dead .py reference link) plus four suggestions; no critical issues.
Counts: 0 critical / 1 should address / 4 suggestions.

Next steps

Run /fix-pr 9348 to apply the should-address and suggestions (refresh the dead .py
reference URL; optionally add the MCA 15-1-2304 / HB 222 statutory cite, shorten the
parameter title, name the elderly credit variable, and add the zero-property-tax edge case),
then rebase onto main.

🤖 Generated with Claude Code /review-program