[MaxGhenis]

North Dakota pays its renter's refund on its own application (SFN 24777), outside the income tax, yet the refund reached no household net income, SPM resources or poverty measure. nd_renters_refund sat only in state_property_tax_credits, which feeds taxsim_state_property_tax_credit and its deprecated alias and nothing else. Five other members of that list had the same gap: Minnesota's homestead credit refund, Montana's property tax rebate, Pennsylvania's property tax or rent rebate, Vermont's renter credit and the renter part of Utah's homeowner's and renter's relief.

Changes

• gov.states.household.state_property_tax_refunds lists the six cash refunds by year, and the tax-unit variable state_property_tax_refunds adds them. Each refund enters in the year it enters state_property_tax_credits, so the new list stays inside that one in every year. The list opens with an empty block on that list's first date (1977). Without it, the model would copy the 2021 block back to 2015, its first modeled year, and count the Vermont renter credit it computes for 2015 to 2020 from its 2021 parameters.
• household_state_benefits (2023 to 2026 blocks) and spm_unit_benefits, with its four reform copies, count state_property_tax_refunds, so each refund raises both net incomes by the same dollar.
• ut_homeowner_renter_relief now adds two new variables. The Tax Commission pays ut_renters_credit from the General Fund (Utah Code 59-2a-205(5)), and counties apply ut_homeowners_credit against property taxes accrued (59-2a-302(2)(b) and 59-2a-304). The total keeps its place in state_property_tax_credits, and its amounts don't change.
• vt_renter_credit now pays only tax units that paid rent. It gave every low-income Vermont tax unit a credit based on fair market rent, homeowners included, and only the TAXSIM comparison read that amount until this PR. Vermont gives the credit to a claimant "who rented the homestead" and requires six calendar months of renting (32 V.S.A. 6066(b)(1) and (c)(3)). The model has no months-rented input, so the new vt_renter_credit_eligible tests for rent above zero. A tenant whose subsidy covers all the rent gets nothing, because Vermont pays a subsidized tenant a share of the gross rent paid under (b)(1)(D). The 28 YAML cases that set rent: 0 while checking the credit amount now set rent, with the same expected amounts.
• vt_renter_credit now runs for every household. Counting it in net income puts its formula on every household's path, and core runs a formula over the whole population whenever any row passes defined_for. The formula indexes tables by tax unit size and county, so tax units of nine or more and counties outside Vermont's tables raised ParameterNotFoundError. This PR takes the Vermont lookup guard from Guard county-keyed lookups in VT renter credit, MN CCAP, WI Shares and MD CCS #10116 word for word. Tax units above eight people use the eight-person values, as the Department of Taxes' calculator does (IF(Household_Size>8,8,Household_Size)), and a county outside Vermont's tables falls back to Addison. Vermont adopts HUD's limits in 32 V.S.A. 6061(19) and (20), and HUD adjusts them for larger families under 42 U.S.C. 1437a, so the eight-person cap stands in until the model encodes HUD's schedule.

2026 impact

We estimate these refunds add $525.8 million to household net income and to SPM net income in 2026, on the default Populace release (populace-us-2024-spm-20260909). Main raises an error when it computes vt_renter_credit for every household, so our baseline adds only the Vermont lookup guard to main at 9960edee31, and the branch at dc5083e7da runs against it. Main's net income never reads that variable, so the baseline's net income matches main's.

	Net income change	Households gaining	Sample households	Mean gain
United States	$525.8M	556,700	244	$945
Minnesota	$340.0M	245,100	123	$1,387
Pennsylvania	$142.7M	258,000	60	$553
Vermont	$28.2M	24,100	35	$1,168
Utah	$9.1M	18,700	5	$487
North Dakota	$5.7M	10,700	21	$534
Montana	$0.0M	0	0	$0

Each household's and SPM unit's net income moves by its refunds to within $0.02, the rounding of the model's 32-bit totals. No household's net income falls, market income doesn't change, and every other refund and bill credit keeps its baseline amount. marginal_tax_rate reads household_net_income, so each income-tested refund's phase-out now shows in household marginal tax rates.

We estimate SPM poverty stays at 13.75% nationally and doesn't change at two decimal places in any of the six states. The sample holds 245 SPM units with a refund, 36 of them in poverty, and the one that crosses the line has a weight under 0.001. We estimate national deep poverty falls by 1,811 people, all from one Vermont record.

None of the six programs has a take-up variable, so the model pays every eligible tax unit. North Dakota's refund, Pennsylvania's rent rebate, Utah's renter credit and Vermont's credit read rent, and this release predates PolicyEngine/microcosm#1444, which places each rent draw on its own household's head. For example, we estimate 19% of Vermont's renter tax units and 16% of its owner tax units carry rent, and owner tax units receive 59% of Vermont's credit.

We estimate vt_renter_credit at $28.2 million for 24,100 tax units with the rent test and $178.3 million for 166,700 without it, an amount only the TAXSIM comparison read. The Department of Taxes reports 12,010 recipients and $7.2 million for the credits it paid in 2024 on 2023 income, and our 2026 estimate exceeds those by 101% and 290%. The Vermont lookup guard moves no Vermont amount in this release: every Vermont household has a Vermont county, and the largest Vermont tax unit has seven people.

In Act 169, Vermont raised its rate to 12.5% and its cap to $3,250 for claim year 2027, when it takes claims on 2026 income under 32 V.S.A. 6061(9) and (12). Main applies them to 2027, so the Vermont figure here uses the 10% rate and the $2,500 cap.

Every member of state_property_tax_credits

Route	Member	Basis
Cash refund, now in both benefit totals	nd_renters_refund	N.D.C.C. 57-02-08.1(2)(b) and (d): the Tax Commissioner pays the refund
	mn_homestead_credit_refund	Minn. Stat. 290A.07, subd. 1, and the M1PR instructions: the commissioner pays claims by check or direct deposit
	mt_property_tax_rebate	Montana DOR rebate report (May 2024) and SB 542 sec. 3: the department pays the rebate on its own claim; $0 from tax year 2025
	pa_property_tax_or_rent_rebate	PA-1000 booklet, Line 20: the Department of Revenue deposits or mails the rebate
	vt_renter_credit	Vermont pays the credit on Form RCC-146, and Schedule IN-112 leaves it out of the refundable credits it carries to Form IN-111
	ut_renters_credit, part of ut_homeowner_renter_relief	Utah Code 59-2a-205(5)
Property tax bill reduction, in neither total	nd_primary_residence_credit	N.D.C.C. 57-02-08.10(1): the county auditor applies it on the property tax statement
	dc_senior_disabled_property_tax_relief	D.C. Code 47-863(b)(1): a 50% deduction in computing real property tax liability
	in_blind_disabled_property_tax_credit, in_over_65_property_tax_credit	IC 6-1.1-51.3-0.5 (SEA 1, 2025): applied to property tax liability
	in_supplemental_homestead_credit	IC 6-1.1-20.6-7.7(e): the county auditor applies it to property tax liability
	ut_homeowners_credit, part of ut_homeowner_renter_relief	Utah Code 59-2a-302(2)(b) and 59-2a-304
State income tax credit, already in both net incomes	az_property_tax_credit, ct_property_tax_credit, dc_ptc, il_property_tax_credit, ma_senior_circuit_breaker, me_property_tax_fairness_credit, mi_homestead_property_tax_credit, mn_renters_credit, mo_property_tax_credit, mt_elderly_homeowner_or_renter_credit, nj_property_tax_credit, nm_property_tax_rebate, ny_real_property_tax_credit, ok_ptc, ri_property_tax_credit, wi_homestead_credit, wi_property_tax_credit, wv_homestead_excess_property_tax_credit	Each state's refundable or non-refundable credit list, or mt_refundable_credits and nj_refundable_credits

Neither net income total subtracts property tax, so a credit that lowers the bill moves neither. household_net_income counts utility bill discounts (CARE, FERA, EZ-SAVE, SHARE) as benefits, and we leave open whether property tax bill credits should follow.

Invariants

Tests execute each of these for every input they draw:

1. Each member of state_property_tax_credits takes exactly one route in every year from 2023 to 2026, and the state lists named for income tax credits hold them (test_every_property_tax_credit_has_one_documented_route, test_property_tax_credit_routes_have_no_stale_entries). state_property_tax_refunds holds nothing outside state_property_tax_credits in any year either list changes (test_refunds_stay_inside_the_credits_list_in_every_year).
2. At $500,000 of wages, each earner owes enough state income tax to use a non-refundable credit in full. There, a dollar of any member moves household_net_income and spm_unit_net_income by a full dollar for a refund or income tax credit and by nothing for a bill reduction. test_property_tax_credits_move_both_net_incomes_together draws the amount with Hypothesis and runs one simulation with a pair of households for every member.
3. A dollar of a cash payment adds a dollar to SPM resources and to household market income plus benefits. test_a_dollar_of_cash_adds_a_dollar_to_both_totals draws the payment from CASH_PAYMENTS, which now holds the six refunds.
4. Utah's two parts add to its total, and the benefit lists still differ only where HOUSEHOLD_ONLY and SPM_ONLY document it. Every reform copy of spm_unit_benefits keeps state_property_tax_refunds (SHARED_CASH_PROGRAMS in test_aggregate_list_integrity.py). The same file checks that each member of the new list names a defined variable with a state gate, and that no later year block drops a member.
5. Vermont pays nothing to a tax unit with no rent, in any county, family size or income the test draws, and an unsubsidized renter's credit stays the same at any rent above zero (test_a_household_that_pays_no_rent_gets_no_credit, test_unsubsidized_credit_does_not_depend_on_rent_paid).

New YAML cases in gov/states/tax/property/state_property_tax_refunds.yaml put each refund in both totals, run North Dakota's worked example, start Pennsylvania's rebate in 2025 and Vermont's renter credit in 2021, split a Utah owner who also rents, and keep North Dakota's and Indiana's bill credits out of both totals. vt_renter_credit_eligible.yaml covers a renter, a homeowner, a tenant whose subsidy covers all the rent, a couple where the spouse pays the rent, and a renter outside Vermont. vt_renter_credit_lookups.yaml runs out-of-state, unknown-county and nine-or-more households in one simulation. Utah's relief cases now assert both parts.

Related PRs

• Keep every baseline benefit in SPM resources under reforms that pay a benefit #10662 moves spm_unit_benefits into a parameter. Whichever PR merges second adds state_property_tax_refunds to that parameter.
• Leave federally excluded income out of North Dakota renter's refund income #10697 and Count SSI and TANF in North Dakota renter's refund income #10724 change how the model computes North Dakota's renter's refund income, and those changes now reach both net incomes. The worked example sets nd_renters_refund_income as an input, and each test household that computes it fails the age and disability test, so neither PR moves an expectation here.
• Apply the RI, SC, UT, VT, WV and WI rules for a filer who can be claimed as a dependent #10044 adds Vermont's dependent bar under 32 V.S.A. 6066(c)(2), which this PR leaves to it. It rewrites the same county and fair market rent lines of vt_renter_credit, so whichever PR merges second keeps safe_county and the exemptions index.
• Guard county-keyed lookups in VT renter credit, MN CCAP, WI Shares and MD CCS #10116 holds the same Vermont lookup guard along with Minnesota's, Wisconsin's and Maryland's. Once this PR merges, Guard county-keyed lookups in VT renter credit, MN CCAP, WI Shares and MD CCS #10116 can drop its Vermont changes, because vt_renter_credit_lookups.yaml carries its four Vermont cases with rent set.

Flaky test on main

test_cash_income_never_lowers_spm_net_income can fail on main when Hypothesis draws Connecticut with a small amount, because $1 of partnership income at $500,000 of wages raises Connecticut income tax by $50.07 through a recapture step.

axiom: TheAxiomFoundation/rulespec-us#1810 queued (32 V.S.A. 6066(b)(1) and (c)(3) rent test); TheAxiomFoundation/rulespec-us#1678 queued (household sizes above the eight-person tables); n/a for the refund lists, Utah's split and the county fallback: aggregation and lookup plumbing, no provision's rules change

🤖 Generated with Claude Code

[MaxGhenis]

This PR needed admin rights to merge at dc5083e7da, because the repo requires one approving review and GitHub doesn't let an author approve their own PR. All 36 jobs in run 42598692963 passed, and the three skipped lines in gh pr checks come from the draft run of the same commit. An independent Claude Opus 5.5 reviewer agreed with the code and this description at that head.