[MaxGhenis]

What was wrong

A Wisconsin homestead credit household is "a claimant and an individual related to the claimant as husband or wife" (Wis. Stat. 71.52(4)). Household income is "all income received by all persons of a household" (71.52(5)), less $500 for each dependent. The 2024 Schedule H instructions add two points:

• "A household is a claimant and the claimant's spouse living in the claimant's homestead" (file page 8).
• "do not include your children's social security or SSI benefits, whether paid to them directly or to you on their behalf" (file page 13; the 2020 instructions say the same on file page 13). Tax 14.03(5)(a)6 agrees.

PolicyEngine got two things wrong:

• Household income. It added the homestead income sources over every tax unit member, so a dependent child's survivor benefits, SSI or 2020 excluded unemployment compensation raised the claimant's household income. The list's Social Security item, tax_exempt_social_security, is a tax-unit amount built from tax_unit_social_security, which includes dependents' benefits.
• Earned income test. wi_homestead_eligible counted a dependent's earnings toward the earned income test. The instructions say "Only one spouse must have earned income to qualify for the homestead credit. The spouse that meets this definition must be the claimant" (file page 10).

Engine output on main:

TY	Household	Variable	main	This PR
2024	Claimant 70, Social Security 18,000, pension 6,000, rent 9,000 (heat included); dependent grandchild 17 with survivor benefits 12,000	wi_homestead_income	35,500	23,500
		wi_homestead_eligible	false (over 24,680)	true
		wi_homestead_credit	0	82.88
2020	Claimant 66, Social Security 15,000, pension 5,000; dependent child with UC 8,000 (excluded under IRC 85(c)) and SSI 3,000	wi_homestead_income	30,500	19,500
2024	Claimant 40 with UC 6,000 and no earnings; dependent child with wages 4,000	wi_homestead_eligible	true	false

What changes

All of these follow from Wis. Stat. 71.52(4): the household is the claimant and the claimant's spouse.

• Household income. wi_homestead_income adds the sources with tax_unit_non_dep_add, so only the head's and spouse's person-level amounts count. tanf is the W-2 payment the claimant receives, an SPM unit amount, and is added as before.
• Social Security. The sources list takes filer_tax_exempt_social_security, the existing head-and-spouse version of tax_exempt_social_security. Wisconsin household income also preserves negative aggregate filer benefits as qualifying prior-year nontaxable repayments under 71.52(6) and the 2024 instructions, file page 12. Negative inputs are assumed to meet the prior-inclusion requirement; the model has no claim history.
• Capital loss add-back. wi_homestead_disqualified_losses adds back filer_loss_limited_net_capital_gains (Schedule 4 line 3). A dependent's gains and losses no longer move household income.
• Property taxes and rent. wi_homestead_property_tax counts only the head's and spouse's property taxes (71.52(7): a homestead "owned by the claimant or a member of the claimant's household") and rent (71.52(8): "paid … by a claimant and his or her household").
• Eligibility. wi_homestead_eligible tests each of the head and spouse as the claimant. The claimant must be 18 or older, and also 62 or older, disabled, or have their own earned income (71.53(2)(c), 71.54(1)(dm)–(gm); earned income is defined by 71.52(1g); 2024 instructions, file pages 4 and 10). Before, the code summed every member's earnings and split the age and earnings tests across people:
  • a dependent's wages qualified the unit;
  • a spouse's self-employment loss could cancel the other spouse's wages;
  • a 17-year-old earner with a 19-year-old non-earning spouse passed.

Invariants

• Leak removed. Adding any income, property tax or rent to a tax unit dependent leaves wi_homestead_income, wi_homestead_property_tax and the claimant's eligibility unchanged. The one exception is the dependent's effect on W-2 (TANF) eligibility, which counts the family's income.
• Unchanged without dependents. For a tax unit with no dependents, the credit is unchanged except for married units, where each spouse is now tested separately. Qualifying negative net Social Security repayments continue to reduce household income.

How it was checked

1. All-state household sweep, TY2020–2026. It is the same sweep as Leave dependents' unemployment compensation off California, Rhode Island and Wisconsin filer returns #10187, with a 17-year-old dependent who has UC 10,000 or survivor benefits 12,000. With this PR merged on top of Leave dependents' unemployment compensation off California, Rhode Island and Wisconsin filer returns #10187, Fix the 2020 unemployment compensation exclusion: cutoff test, per-spouse cap, dependents #10175 and Leave dependents' capital income out of state lines on the filers' return #9894/Leave dependents' own items out of DC, Delaware and Ohio subtractions #9939, every Wisconsin homestead credit decrease caused by the child's Social Security, SSI or 2020 UC goes away. That is 18 rows across 2020–2026.

   The rows that remain are for a 15,000 earner whose family loses W-2 (Wisconsin Works) when the child's income counts. W-2 is the claimant's household income (Tax 14.03(4)(b)3.b), so losing it raises the credit.

2. Tests. New cases come from the 2024 Schedule H:

   • wi_homestead_income Cases 6–7;
   • wi_homestead_eligible Cases 8–9;
   • wi_homestead_credit Case 5.

   4 fail on main and pass here, and the WI integration and income tax files pass. Round 2 adds 6 more cases:

   • wi_homestead_income Cases 8–9;
   • wi_homestead_property_tax Cases 2–3;
   • wi_homestead_eligible Cases 10–11.

   All 6 fail on main. Cases 10 and 11 also fail under the round 1 code. The credit case asserts the continuous formula's 82.88. Schedule H's Tables A and B round to bands and give 84, as for the existing credit tests.

Changes after review

• Added all three spouse scenarios requested by DTrim99 under both head/spouse labellings: only one earner, only one spouse age 65, and both spouses' Social Security plus SSI producing household income of $17,000. Removing the spouse earnings/age paths and restricting person-level income to the head makes the new spouse cases fail; restoring the formulas makes them pass.
• Added the eligibility statutes (71.53(2)(c) and 71.54(1)(dm)–(gm)) and corrected both eligibility-test comments.
• Documented that a dependent age 62 or older may make a separate claim under 71.53(2)(d), which this tax-unit model does not represent.
• Preserved qualifying negative aggregate filer Social Security repayments under 71.52(6), with regressions for single filers, spouses, mixed-sign benefits and a dependent's excluded benefits.
• Updated the queued parity dispatch in TheAxiomFoundation/rulespec-us#1726 for property taxes/rent, filer capital losses, per-claimant eligibility and qualifying repayments, with companion cases and statutory citations.
• Left the pre-existing S corporation earned-income treatment unchanged; proposed a separate follow-up under 71.52(1g).

Review

Round 1, at 4c0b18e1f1: REQUEST_CHANGES.

• Major:
  • a dependent's capital gains and losses still moved household income;
  • property taxes and rent counted dependents.
• Minor: eligibility summed earnings and split the age and earnings tests across spouses.
• Nit: the invariant wording for negative Social Security.

a576e21681 addresses all of them. Round 2: APPROVE at a576e21681.

The reviewer re-ran the round 1 probes and checked:

• that no dependent income, loss, rent or property tax moves household income or the property tax amount;
• single filers;
• negative wages;
• 62-year-old spouses and dependents.

It also confirmed that only wi_homestead_credit and wi_homestead_income read the changed variables. In round 2 it flagged that Case 11 lacked explicit head and spouse roles, so the 17-year-old was inferred as a dependent, and that the changelog was too narrow. 509e5c9d71 fixes both.

State SSI supplement (merge-delta review)

Main added wi_ssp (the regular state supplement, SSI-E and the Caretaker Supplement), but the homestead income source list only carried a commented-out placeholder, and wi_ssp reached household income through no other path. Wisconsin counts "cash public assistance" as income (Wis. Stat. 71.52(6)), and DOR has claimants report state SSI, SSI-E and Caretaker Supplement payments with their SSI (Publication 127, page 9). sources.yaml now lists wi_ssp; the existing dependent filter keeps the claimant's and spouse's payments, including a parent's Caretaker Supplement, and leaves out a dependent child's own supplement.

TY	Household	wi_homestead_income before	after	wi_homestead_credit before	after
2026	Claimant 70, federal SSI 11,928, regular supplement 1,072.40 and SSI-E 1,228.68, allowable property taxes 1,460	11,928	14,229.08	896.16	734.44
2026	SSI parent with a 3,000 Caretaker Supplement and SSI 9,000, one child	8,500	11,500
2026	Wage earner 20,000 with a dependent child on SSI 6,000 and supplement 1,000	19,500	19,500
2024	Cases 10–11: spouse's SSI 1,000 brings the regular supplement, 83.78 × 12 = 1,005.36	17,000	18,005.36

Credit: (14,229.08 − 8,060) × 0.08785 = 541.95 phase-out; (1,460 − 541.95) × 0.8 = 734.44, against (1,460 − 339.80) × 0.8 = 896.16 without the supplement.

axiom: TheAxiomFoundation/rulespec-us#1726 queued (rulespec-us us-wi has no homestead credit module; the issue is dispatch-ready with us-wi/statute/71.52 from the pinned release and companion tests from these YAML cases)

🤖 Generated with Claude Code

Hub microsimulation impact (PR head 4f5e4a6ff1 against main de3815f4cc)

A real microsimulation on the default dataset, main at the merge base de3815f4cc against this PR's then head 4f5e4a6ff1, for 2024 and 2026. These figures predate the current head. Counting the state SSI supplement in household income (above) and main's exclusion of SSI custodial parents from W-2 can both move households through the credit phase-out. Sums are weighted.

Output	2024 change	2026 change	2024 records changed (weighted)	2026 records changed (weighted)
wi_homestead_income	-1.35bn	-1.46bn	18 (105,621)	18 (107,363)
wi_homestead_eligible	+9,089	+22,041	6 (17,339)	7 (30,428)
wi_homestead_credit	+34.31M	+35.91M	4 (36,949)	4 (37,559)
wi_homestead_property_tax	-1,563	-1,675	4 (2)	4 (2)
wi_homestead_disqualified_losses	+0	+0	0 (0)	0 (0)
wi_refundable_credits	+34.31M	+35.91M	4 (36,949)	4 (37,559)
wi_income_tax	-34.31M	-35.91M	4 (36,949)	4 (37,559)
filer_tax_exempt_social_security	+0	+0	0 (0)	0 (0)
state_income_tax	-34.31M	-35.91M	4 (36,949)	4 (37,559)
income_tax	+0	+0	0 (0)	0 (0)
household_net_income	+34.31M	+35.91M	4 (36,949)	4 (37,559)
household_benefits	+0	+0	0 (0)	0 (0)

Run at 4f5e4a6ff1, which includes the review-round change that restores qualifying negative Social Security repayments (§71.52(6)). These figures replace the earlier runs at a576e21681 and 4c0b18e1f1.

---

Generated by Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (909176a) to head (4c7110c).
⚠️ Report is 991 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main    #10191   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         4           
  Lines           76        78    +2     
  Branches         2         1    -1     
=========================================
+ Hits            76        78    +2     

Flag	Coverage Δ
unittests	100.00% <100.00%> (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.

[MaxGhenis]

Hand-off to the US + core hub

Head: 509e5c9d71

Review. The Codex review lanes were held until 06:00Z and the Claude lanes were closed, so per the fallback rule an independent in-session Opus 5.5 reviewer reviewed this PR:

• Round 1 at 4c0b18e1f1: REQUEST_CHANGES.
  • a dependent's capital gains and losses still moved household income;
  • property taxes and rent counted dependents;
  • eligibility was split across spouses.
• Round 2 at a576e21681: APPROVE, after those were fixed.
• Round 3 at 509e5c9d71: APPROVE, after Case 11 got explicit head and spouse roles and the changelog was widened.

The verdict file is ~/reviews/dependent-uc-state-lines-2026-10-12/review/verdict_10191.md on the hub host.

Passed locally:

• the five homestead YAML files, 36 cases;
• the related WI, household and contrib files, 27 cases.

10 of the 11 new cases fail on main; eligibility Case 9 is a control.

CI. It is running on 509e5c9d71; when last checked, 1 check had passed and 25 were pending, with no failures.

Impact. One item is still owed, and the hub is asked to run it. In the body, 2024 is at a576e21681, whose formulas match the current head; 2026 is still the earlier run at 4c0b18e1f1. I withdrew the 2026 re-run before its PR side under the disk pause relayed by Fleet ops at 03:07Z; it held no lock.

When the Disk manager reports more than 60 GB free:

1. Run zsh ~/reviews/us-hub/scripts/impact_queue.sh ~/reviews/us-hub/fixes/impact-queue/spec-2026-10-13-10191b.txt (years 2024 and 2026).
2. Run python3 ~/reviews/us-hub/scripts/impact_body.py 10191 2024 2026 --apply, which replaces the hand-written section.

In 2024 the round 2 fixes moved no credit record, so I expect 2026 to be close to the earlier +35.9M. That is a guess; it has not been run.

Order. This PR is independent of #10187 and #10175.

Merge. The PR needs an approving review (REVIEW_REQUIRED), so land it on the gates with --admin --match-head-commit once CI is green and the impact section is current.

Axiom. TheAxiomFoundation/rulespec-us#1726 is queued.

Follow-ups. Other states' credit income lists still count dependents' Social Security through tax_exempt_social_security, which includes them (CO, ME, MI, MO, WV). That is proposed as a separate session.

[DTrim99]

PR Review: Leave dependents' income out of Wisconsin homestead credit household income

PR #10191 (author: MaxGhenis, head 509e5c9d71, merge-base de3815f4cc). 11 files (+358/−25), 4 commits. Axiom: TheAxiomFoundation/rulespec-us#1726 queued.

Summary

The PR applies Wis. Stat. § 71.52(4), under which the homestead credit "household" is the claimant and spouse, to four variables:

• wi_homestead_income:
  • The income sources are now added with tax_unit_non_dep_add, so a dependent's person-level amounts (SSI, tax-exempt interest, pensions and unemployment compensation) drop out.
  • The Social Security item becomes filer_tax_exempt_social_security, which already exists on main.
• wi_homestead_disqualified_losses: the Schedule 4 line 3 capital-loss add-back uses filer_loss_limited_net_capital_gains.
• wi_homestead_property_tax: only the head's and spouse's real_estate_taxes and rent count, following § 71.52(7) and (8).
• wi_homestead_eligible:
  • The head and the spouse are each tested as the claimant: 18 or older, and also 62 or older, disabled, or with their own earned income.
  • Previously the code pooled every member's earnings and split the age and earnings tests across people.
• Tests: 11 new YAML cases, plus a renamed input (filer_loss_limited_net_capital_gains) in two existing disqualified-loss cases.

The policy is right:

• Law. The statute, Tax 14.03(5)(a)6, and the 2020 and 2024 Schedule H instructions all exclude a dependent's income, rent and property taxes. The 2024 instructions also require the earner to be the claimant.
• No double counting. PolicyEngine's tax-unit AGI already leaves dependents out, so the new sums count each item once.
• Hand checks. I checked all 21 expectations in the 11 new cases by hand, and they all hold. Ten of the new cases fail on the merge-base; eligibility Case 9 is a control.
• Local runs. The five homestead files pass locally at the head (36/36) and on the merged tree (42/42, with integration.yaml and wi_income_tax.yaml).
• CI is green, and the merge into main is clean.

One should-fix item remains: no test covers the spouse side of the new head/spouse logic. I removed the spouse's earned-income and age-62 paths and restricted household income to the head, and every WI test still passed. There are also five suggestions.

Critical (must fix)

None.

Should address

1. No test covers the spouse side of the new head/spouse logic.

• Where:

  • variables/.../homestead/wi_homestead_eligible.py:28 (spouse_earnings) and :36-40 (spouse_qualifies)
  • variables/.../homestead/wi_homestead_income.py:28: tax_unit_non_dep_add must keep the spouse's person-level items

• The law:

  • 2024 Schedule H instructions, Question 1d (p. 10): "Only one spouse must have earned income to qualify for the homestead credit. The spouse that meets this definition must be the claimant."
  • Step 1(a)1 (p. 4): "Are age 62 or older as of December 31, 2024 (in order to meet this qualification you must be the claimant)?"
  • 2024 Schedule H (p. 1): "Household Income Include all 2024 income as listed below. If married, include the incomes of both spouses."

• What the tests cover:

  • Every new eligibility case turns on the head. In Case 10 the earner is person1, who is the head. In Case 11 the 19-year-old spouse has no earnings.
  • The only case where the spouse qualifies is the pre-existing unit test 6, where the spouse is disabled.
  • No income case gives the spouse SSI, tax-exempt interest, unemployment compensation or a pension.
  • Property tax Case 3 does cover the spouse's property taxes.

• Mutation check. On a copy of the head I made three changes:

  • removed | (spouse_earnings > 0) from spouse_qualifies
  • removed (spouse_age >= min_elderly_age) from spouse_qualifies
  • replaced the income sum with a head-only sum of the person-level sources

  All 41 cases still pass: the five homestead files, integration.yaml and wi_income_tax.yaml. So do the other 35 WI-related cases: wi_income_tax_before_refundable_credits.yaml, the two retirement-exclusion files, household_state_income_tax.yaml and the WI contrib reform test.

• Why it matters. A later refactor could deny every married couple in which only the spouse works or only the spouse is 62 or older, and CI would stay green. That is the case the instructions single out.

• Fix. Add three cases. I reproduced these values on the head, and the mutant gives the value in brackets. Set explicit roles, ssi: 0 for anyone without SSI, and tanf: 0 on the SPM unit, as Cases 8–9 do.

  • wi_homestead_eligible, 2024. Head 40 with no income; spouse 38 with wages 6,000. Expected true (mutant false). Household income is 6,000.
  • wi_homestead_eligible, 2024. Head 50 with no income; spouse 65 with Social Security 9,000. Expected true (mutant false).
  • wi_homestead_income, 2024. Head 70 with pension 6,000 and Social Security 6,000; spouse 64 with Social Security 4,000 and SSI 1,000.
    • AGI is 6,000: provisional income is 6,000 + 5,000 = 11,000, under the 32,000 joint base, so no benefits are taxable.
    • Household income = 6,000 + 10,000 + 1,000 = 17,000 (mutant 16,000).

Suggestions

1. Bring rulespec-us#1726 up to the PR's current scope.

• When it was written. The issue was filed at 00:19 UTC, during round 1. It lists three mistakes: dependents' tax-exempt items, dependents' Social Security, and dependents' earnings in the earned income test.
• Fixes it doesn't cover. The head (round 2, a576e21681) also fixes:
  • Property taxes and rent of household members only. § 71.52(7) covers taxes "levied on a homestead owned by the claimant or a member of the claimant's household", and § 71.52(8) covers rent "paid in cash or its equivalent by a claimant and his or her household". See property tax Cases 2–3.
  • The filers' capital loss add-back (§ 71.52(1e)(b); Schedule 4 line 3). See income Cases 8–9.
  • Per-claimant age and earned income tests. See eligibility Cases 10–11.
• Wrong section in the issue. It says "If the encoder also takes the 71.53 eligibility rule". Only the under-18 bar is in § 71.53(2)(c). The 62-or-older, disabled and earned-income tests are in § 71.54(1)(dm), (em) and (gm).
• Fix:
  • Add outputs for (7) and (8). They are in the same pinned us-wi/statute/71.52 row.
  • Name § 71.53(2)(c) and § 71.54(1)(dm)–(gm) for eligibility.
  • Add the round 2 companion tests. All of them come from the 2024 instructions (pp. 4, 7 and 10) and the statute.

2. Cite the eligibility statute and correct two test comments.

• Variable reference. wi_homestead_eligible.py:9-14 cites only the LFB paper and the 2024 instructions.
  • Add § 71.53. Subsection (2)(c): no claim if "The claimant was under 18 years of age at the close of the year to which the claim relates."
  • Add § 71.54.
    • (1)(dm): "no credit may be allowed ... unless the claimant or the claimant's spouse is over the age of 61"
    • (1)(em): "unless the claimant is disabled"
    • (1)(gm): "no credit may be allowed ... if the claimant had no earned income in the taxable year"
• wi_homestead_eligible.yaml:144-146 cites "Wis. Stat. 71.52(1g)" for "18 or older and 62 or older, disabled or have their own earned income". Subsection (1g) only defines earned income. Cite § 71.53(2)(c) and § 71.54(1)(dm)–(gm) instead.
• wi_homestead_eligible.yaml:99 says "The claimant or the claimant's spouse must have earned income". That describes the old pooled rule. The code and p. 10 require the earner to be the claimant. Suggested wording: "The claimant must have their own earned income; either spouse may be the claimant."
• Optional. The new statute URLs point at the whole section. The subsection anchors /52/4, /52/7 and /52/8 would match wi_homestead_disqualified_losses.py, which uses /52/1e.

3. Document that a dependent aged 62 or older can file their own claim, which PolicyEngine does not model.

• The law.
  • § 71.53(2)(d) bars a claimant who "was claimed as a dependent for federal income tax purposes by another person ... but this limitation shall not apply if the claimant was 62 years of age or older".
  • The 2024 instructions repeat this in Step 2(a) (p. 4).
• What happens now.
  • In property tax Case 2, the 70-year-old dependent's $2,000 of property taxes now goes to no one.
  • Leaving it off the filer's claim is correct under § 71.52(7), but PolicyEngine has no claim for the dependent.
  • Before the PR, the filer received it, which was also wrong.
• Fix: a comment in wi_homestead_property_tax.py (or a follow-up issue) noting the unmodeled claim. This isn't blocking.

4. Negative net Social Security now floors at zero.

• The PR body discloses this.
• The law.
  • § 71.52(6): "Amounts not included in adjusted gross income but added to 'income' under this subsection in a previous year and repaid may be subtracted from income for the year during which they are repaid."
  • 2024 instructions (p. 12): "Nontaxable income that was included in household income in a prior year and was required to be repaid in 2024 may be subtracted from household income".
• Effect.
  • Take a single filer with net Social Security of −1,000. tax_exempt_social_security was −1,000. filer_tax_exempt_social_security is 0, because tax_unit_social_security_for_taxability floors at zero.
  • So household income is now 1,000 higher (probe at the head).
  • This is rare in the data.
• Fix: optionally subtract the filers' negative net benefits, or add a comment in sources.yaml noting that the floor is deliberate.

5. Pre-existing, outside this diff: S corporation income counts as earned income.

• Where: eligible/earnings_sources.yaml lists partnership_s_corp_income.
• The law. § 71.52(1g) defines earned income as "wages, salaries, tips, and other employee compensation ... plus the amount of the claimant's net earnings from self-employment". An S corporation shareholder's pass-through share is not net earnings from self-employment.
• Why raise it here. The PR now tests this list per claimant and cites § 71.52(1g).
• Fix: a separate PR or issue.

Verified correct

• Statute (Wis. Stat. § 71.52):

  • (4): "'Household' means a claimant and an individual related to the claimant as husband or wife."
  • (5): "all income received by all persons of a household in a calendar year while members of the household, less $500 for each of the claimant's dependents".
  • (1): "When 2 individuals of a household are able to meet the qualifications for a claimant, they may determine between them as to who the claimant is." So either spouse may be the claimant.

• Instructions:

  • Household. 2024 instructions p. 8: "A household is a claimant and the claimant's spouse living in the claimant's homestead. If you are single, you are a household, whether you live alone or with others."
  • Children's benefits. p. 13: "do not include your children's social security or SSI benefits, whether paid to them directly or to you on their behalf."
  • 2020. The 2020 instructions have the same text on p. 8 and p. 13, which covers income Case 7.
  • Administrative code. Tax 14.03(5)(a)6 (PDF p. 4) excludes "Social security or SSI payments received on behalf of a claimant's children or the children of the claimant's household."

• Who is summed:

  • AGI. adjusted_gross_income already leaves dependents out:
    • irs_gross_income multiplies by ~is_tax_unit_dependent.
    • limited_business_loss, capital_losses_allowed_against_gains, limited_capital_loss and self_employment_tax_ald all exclude dependents.
    • So tax_unit_non_dep_add's fall-back to add for this tax-unit item is right.
  • Social Security. filer_tax_exempt_social_security = the filers' benefits − tax_unit_taxable_social_security. Together with AGI, that gives the filers' total benefits, which is Schedule H line 7 plus line 9b.
  • TANF. tanf still goes through add's SPM-to-tax-unit split, which is unchanged.
  • No overlap. No item is counted twice. The capital loss add-back uses the same filer-only Schedule D lines as AGI's loss deduction (Case 9).

• Eligibility matches the statute:

  • It combines § 71.53(2)(c) with § 71.54(1)(dm)–(gm).
  • (dm) lets the age test be met by "the claimant or the claimant's spouse". A spouse over 61 is also over 18 and can be the claimant, so testing each spouse as the claimant gives the same answer.
  • The instructions say the same on p. 4: "you must be the claimant".

• Probes at the head:

  • Own earnings net out. A head with wages 3,000 and their own self-employment loss of 5,000 is ineligible. This matches the instructions' Example 2 (p. 7): "$5,000 of wages and ($8,000) of net earnings from self-employment ... would not qualify".
  • Under-18 spouse. A 30-year-old head with no earnings whose 17-year-old spouse has wages is ineligible.
  • Disabled spouse. The unit is eligible.
  • Dependent's items. A dependent's tax-exempt interest, wages, self-employment loss, rental loss, 2024 unemployment compensation or tax-exempt pension leaves AGI, the add-back and household income unchanged.
  • Spouse's SSI is counted (5,000).

• Hand checks (all 21 expectations in the 11 new cases; the main ones are below):

  Case	Calculation	Expected
  Credit 5 (2024)	Household income 6,000 + 18,000 − 500 = 23,500. Rent 9,000 × 20% = 1,800, capped at 1,460. (1,460 − 15,440 × 8.785%) × 80% = 103.596 × 80%	82.88
  Credit 5, Schedule H tables	Table A, 23,470–23,585 → 1,359 (p. 25); 1,460 − 1,359 = 101; Table B, 100–110 (p. 26)	84, as the comment says
  Income 6 (2024)	Provisional income 6,000 + 9,000 < 25,000, so AGI is 6,000; 6,000 + 18,000 − 500	23,500
  Income 7 (2020)	Head: 5,000 + 15,000 − 500; the child's 8,000 unemployment compensation and 3,000 SSI are left out	19,500
  Income 8 (2024)	The child's −2,000 is not in AGI (6,000); add-back 0; 6,000 + 12,000 − 500	17,500
  Income 9 (2024)	AGI 6,000 − 3,000 = 3,000; add-back 3,000 (the child's +2,000 is left out); 3,000 + 3,000 + 12,000 − 500	17,500
  Property tax 2	The dependent's 2,000 is left out; credit 0	0 / 0
  Property tax 3	300 + 6,000 × 20%; the dependent's 1,000 rent is left out	1,500
  Eligible 8	Head 40 with unemployment compensation only, so no earned income (p. 7); the dependent's wages don't count	false
  Eligible 9	Head's wages 4,000; household income 3,500	true
  Eligible 10	Head's wages 5,000 > 0; household income −3,000 + 8,000 add-back = 5,000	true
  Eligible 11	Head 17 is under 18; spouse 19 has no earned income	false

• The tests are functional.

  • I ran the head's five test files against the merge-base code. Eleven cases fail: every new case except eligibility Case 9, plus disqualified-loss Case 1, whose input was renamed to filer_loss_limited_net_capital_gains.
  • Eligibility Case 9 passes on both, as the control.

• Callers.

  • wi_homestead_income is read only by wi_homestead_credit and wi_homestead_eligible.
  • wi_homestead_property_tax is read only by wi_homestead_credit.
  • The credit feeds wi/.../credits/refundable.yaml, retirement_income_exclusion_refundable.yaml and state_property_tax_credits.yaml.
  • The PR doesn't change filer_tax_exempt_social_security or filer_loss_limited_net_capital_gains, so their other users (the CRFB surtax, net investment income, Michigan household resources) are unaffected.

• Code-health ratchet (test_no_group_variables_in_person_adds.py): no interaction.

  • The ratchet scans adds/subtracts on Person variables.
  • The changed homestead/income/sources list is read by a TaxUnit formula through tax_unit_non_dep_add, and its new entry is a TaxUnit variable.
  • The earnings sources go through add(person, ...), and all of them are person-level.
  • KNOWN_VIOLATIONS is untouched.

• References and pages:

  • The # PDF pages 4, 10 and # PDF pages 8, 13 comments above single #page= hrefs follow the convention in CLAUDE.md (Document the reference page-citation convention #10009).
  • In the 2024 instructions, file pages equal printed pages, and each anchor opens the quoted text.

• Changelog. changelog.d/wi-homestead-dependent-income.fixed.md is top-level, typed fixed, and describes all four changes.

• Test periods are plain (2020, 2024).

• Code style:

  • The code is vectorized (&/| on arrays) and uses tax_unit.sum over role masks.
  • No policy value is hard-coded, and the reference is a tuple.

• Coverage. Codecov reports 100% of the changed lines covered.

Axiom line assessment

Valid queued.

• The line: axiom: TheAxiomFoundation/rulespec-us#1726 queued.
• The issue is OPEN and labelled pe-parity, created 2026-10-13 at 00:19 UTC. It has:
  • module path: us-wi/statutes/71.52.yaml
  • corpus citation: us-wi/statute/71.52 from the pinned release us-rulespec-2026-08-08-obbb-alien-snap
  • verbatim law: § 71.52(4)–(6)
  • required outputs: homestead_household_member, homestead_income_of_person and homestead_household_income
  • blocker: the billed signed-encoder run needs approval
  • companion tests A–D, taken from the 2024 Schedule H instructions and computed by hand rather than from PolicyEngine output
• Gap. The issue matches the round 1 behavior only. It doesn't cover the round 2 fixes: property taxes and rent under (7)–(8), the filers' capital loss add-back, and the per-claimant tests under § 71.53(2)(c) and § 71.54(1)(dm)–(gm). It also attributes the eligibility rule to "71.53" (Suggestion 1).

CI status

All 28 checks pass (gh pr checks 10191, run 38338762863 on 509e5c9d71).

• Where the homestead tests ran. Baseline states-shard-3 ran the five homestead files:
  • credit: 5 cases
  • disqualified losses: 8
  • eligibility: 11
  • income: 9
  • property tax: 3
• Partner tests. The PR touches no partner test file, and the Household API Partners job passes.
• Local run. policyengine-core 3.33.3 also passes these 36 cases at the head.

Branch status

• Position: 4 commits ahead of PolicyEngine/main (35a94c3b02) and 22 behind; merge-base de3815f4cc.
• Merge: git merge-tree --write-tree PolicyEngine/main 509e5c9d71 is clean (tree 8f661c2185).
• Changes on main since the merge-base that touch the same area:
  • Offset SSTB and non-SSTB Schedule C results in federal AGI #9995 splits SSTB and non-SSTB Schedule C losses in disqualified_loss_sources.yaml. It also adds disqualified-loss Case 9 and rewords a comment in wi_homestead_income.yaml Case 5.
  • Give each spouse their own above-the-line deductions in per-person AGI #9890 adds filer_share to tools/general.py.
  • None of these conflicts textually or semantically with this PR.
• Merged tree. It passes the five homestead files plus integration.yaml and wi_income_tax.yaml locally (42/42).

Verdict: REQUEST_CHANGES

[DTrim99]

The law supports the change: under Wis. Stat. 71.52(4) the household is the claimant and spouse, Tax 14.03(5)(a)6 excludes children's SS and SSI, and the Schedule H instructions agree. All 21 new expectations check by hand. One should item (see the comment above): nothing tests the spouse side. Removing the spouse's earnings and age-62 eligibility paths, and summing income for the head only, leaves all 76 WI cases passing. Add three cases: only the spouse earns (eligible), only the spouse is 65 (eligible), and spouse SS 4,000 + SSI 1,000 giving household income 17,000.

[DTrim99]

PR Review (round 2): Leave dependents' income out of Wisconsin homestead credit household income

PR #10191 (author: MaxGhenis, head 4f5e4a6ff1, last reviewed 509e5c9d71, merge-base de3815f4cc). Since round 1 there are four author commits (+340/−8) and no merges. Axiom: TheAxiomFoundation/rulespec-us#1726 queued.

Summary

All six round 1 findings are addressed: five are fixed, and the S corporation item is deferred with a reason.

• Should 1 (no test of the spouse side): fixed. Eligibility Cases 12–15 and income Cases 10–11 add the three scenarios I asked for, each with the head and spouse labels swapped. Each scenario fails under the round 1 mutants.
• Suggestions 1–4: fixed.
  • rulespec-us#1726 now covers the round 2 and round 3 scope.
  • The code cites the eligibility statutes, and the two test comments are corrected.
  • A comment documents the separate claim a dependent aged 62 or older can file.
  • Qualifying negative Social Security repayments now reduce household income. Five new income cases (12–16) test this.
• Suggestion 5 (S corporation income counted as earned income): deferred. It is pre-existing and outside this diff.

The new repayment rule follows the law.

• Statute. § 71.52(6) lets a household subtract "Amounts not included in adjusted gross income but added to 'income' ... in a previous year and repaid".
• Instructions. The 2024 Schedule H instructions on p. 12 allow a negative entry on the income line and tell married claimants to combine both spouses' income.
• Code. The code adds the filers' combined net Social Security when it is negative, so a spouse's positive benefits offset a repayment before anything is subtracted. A dependent's benefits stay out.
• Tests. Each of the five new cases fails under a plausible wrong variant: dropping the line, flooring each person separately, or including dependents.

Checks.

• Hand checks. I checked all 22 new expectations by hand.
• CI. All 29 checks pass at the head. CI ran on a merge with main as of 2026-10-19.
• Merge with current main. main is now 142 commits ahead, and the merge is clean.
  • None of the newer main commits touches any variable or parameter these tests depend on, so merging changes no result.
  • The Tie source-conflict markers to the register shared with Axiom #10287 source-conflict ratchet finds no conflict-sounding comments in the PR's files on the merged tree.

There are no open critical or should-address items. I have two optional suggestions: a pre-existing gap under § 71.53(2)(d), and a tracking issue for the deferred S corporation item.

Prior findings

Finding	Status	Evidence
Should 1. No test of the spouse side of the head/spouse logic	Fixed	wi_homestead_eligible.yaml Cases 12/13: only the spouse earns (6,000), run with each labelling; true, household income 6,000. Cases 14/15: only the spouse is 65 (Social Security 9,000); true, 9,000. wi_homestead_income.yaml Cases 10/11: the spouse's Social Security and SSI; 17,000. Under the round 1 mutants: Case 12 gives false without the spouse earnings path, Case 14 gives false without the spouse age path, and Case 10 gives 16,000 with a head-only income sum. So each mutant now fails.
Suggestion 1. Bring rulespec-us#1726 up to the PR's scope	Fixed	Issue comment (created 2026-10-19 00:36 UTC, edited 00:53 UTC) adds outputs for § 71.52(7)–(8), the filers' capital-loss add-back under § 71.52(1e)(b), and per-claimant eligibility under § 71.53(2)(c) and § 71.54(1)(dm)–(gm), with § 71.52(1g) for earned income. It adds the § 71.52(6) repayment rule and companion cases E–Q, all hand calculations from the statute and instructions. See the Axiom section.
Suggestion 2. Cite the eligibility statute; correct two test comments	Fixed	wi_homestead_eligible.py now cites /71/viii/53/2/c and /71/viii/54. wi_homestead_eligible.yaml cites § 71.53(2)(c) and § 71.54(1)(dm)–(gm) in place of § 71.52(1g). It now reads "The claimant must have their own earned income; either spouse may be the claimant". The optional subsection anchors are partly adopted (/53/2/c, /53/2/d); the income and property tax variables still cite the whole /52 page. That is fine.
Suggestion 3. Document the separate claim of a dependent aged 62 or older	Fixed	wi_homestead_property_tax.py comment: "A dependent age 62 or older may file a separate claim under 71.53(2)(d). This tax-unit model does not represent that separate claim, so their taxes and rent are not credited to anyone here." It adds the reference /71/viii/53/2/d, which resolves to that paragraph.
Suggestion 4. Negative net Social Security floored at zero	Fixed	wi_homestead_income.py adds min_(tax_unit_non_dep_add(tax_unit, period, ["social_security"]), 0). Regressions in income Cases 12–16. Verified below.
Suggestion 5. S corporation income counted as earned income (pre-existing)	Deferred, with a reason	PR body: "Left the pre-existing S corporation earned-income treatment unchanged; proposed a separate follow-up under 71.52(1g)." I found no tracking issue (new Suggestion 2).

New findings

Critical (must fix)

None.

Should address

None.

Suggestions

1. Pre-existing, outside this diff: § 71.53(2)(d) bars a claimant who was claimed as a dependent elsewhere, unless the claimant is 62 or older.

• The law. § 71.53(2)(d): no claim if "The claimant was claimed as a dependent for federal income tax purposes by another person during the year to which the claim relates but this limitation shall not apply if the claimant was 62 years of age or older". The 2024 instructions repeat this in Step 2(a) (p. 4).
• What the code does. wi_homestead_eligible.py:31-42 now tests each spouse as the claimant and cites § 71.53(2)(c), but it does not apply (2)(d).
  • Take a 20-year-old head with wages who is claimed on a parent's return. That head passes the earned-income path and can receive the credit.
  • head_is_dependent_elsewhere and spouse_is_dependent_elsewhere exist on main and at the merge-base.
  • Other renter and homestead credits already use them, including mn_homestead_credit_refund_eligible, mn_renters_credit_eligible and hi_tax_credit_for_low_income_household_renters_eligible.
• Why this isn't blocking. main behaves the same way, and the PR doesn't claim to fix it.
• Fix (here or as a follow-up). Add & (~head_is_dependent_elsewhere | (head_age >= min_elderly_age)) to head_qualifies, and the same for the spouse. Then add a case: a 20-year-old head with wages 5,000 and claimed_as_dependent_on_another_return: true is ineligible, and the same head at 62 is eligible.

2. File the S corporation follow-up as an issue.

• Status. The PR body says a follow-up under § 71.52(1g) was "proposed", but I found no issue in policyengine-us.
• Why it matters. eligible/earnings_sources.yaml still lists partnership_s_corp_income. § 71.52(1g) limits earned income to "wages, salaries, tips, and other employee compensation ... plus the amount of the claimant's net earnings from self-employment".
• Fix. Open a tracking issue so the item isn't lost when this PR merges.

Verified

• Repayment rule (§ 71.52(6); 2024 Schedule H instructions):

  • The statute (§ 71.52): "Amounts not included in adjusted gross income but added to 'income' under this subsection in a previous year and repaid may be subtracted from income for the year during which they are repaid."
    • Wisconsin excludes Social Security from Wisconsin AGI.
    • So every benefit was "not included in adjusted gross income", and the only condition the model can't check is a prior homestead claim.
    • The code comment and the PR body both disclose that assumption.
  • The instructions (p. 12, "Repaid amounts"):
    • "Nontaxable income that was included in household income in a prior year and was required to be repaid in 2024 may be subtracted from household income on your 2024 Schedule H. Subtract the amount repaid on the income line of Schedule H to which the repayment relates."
    • "If the subtraction results in a negative amount, fill in that amount as a negative number."
    • The same page also says "If married, combine the incomes of both spouses and fill in the totals for each category". That supports netting the spouses together before any negative amount is applied.
    • The file page matches the printed page 12.
  • The arithmetic is linear.
    • filer_tax_exempt_social_security is max(0, filers' net benefits) − tax_unit_taxable_social_security. Taxable benefits are 0 when the filers' net is ≤ 0, because gross_ss is 0.
    • Adding min_(filers' net, 0) therefore gives exactly the filers' net benefits minus the taxable part, for any sign.
    • It never double-subtracts: a per-person floor would, and Case 14 would catch it.
    • It never lets a dependent's positive benefits absorb a filer's repayment. tax_unit_social_security_for_taxability removes them, and the new sum uses tax_unit_non_dep_add; Case 16 tests both.
  • No-dependent units are unchanged. For a unit with no dependents, the result equals main's tax_exempt_social_security for either sign, so the PR body's claim that repayments "continue to reduce household income" is accurate.

• The new tests are functional.

  • Repayment cases. Removing the new line fails Cases 12, 13 and 16 (6,000, 6,000 and 5,500). A per-person min_ fails Case 14 (8,000). Summing net benefits with add instead of tax_unit_non_dep_add fails Case 16 (5,500).
  • Spouse cases. Each paired case fails under one of the round 1 mutants.

• Hand checks (all 22 new expectations). Every case is TY2024, with SSI and TANF at 0 where not listed. Max household income is 24,680, and the dependent exemption is 500.

  Case	Calculation	Expected
  Eligible 12	Spouse wages 6,000, so AGI is 6,000 and household income 6,000	6,000
  Eligible 12	Head 40: not 62, not disabled, no earnings, so fails. Spouse 38: 18 or older with own earnings 6,000 > 0, so qualifies. 6,000 < 24,680	true
  Eligible 13	Labels swapped; same arithmetic	6,000 / true
  Eligible 14	Spouse Social Security 9,000. Provisional income 4,500 < 32,000, so AGI is 0; filers' tax-exempt Social Security 9,000	9,000
  Eligible 14	Head 50 fails. Spouse 65 ≥ 62 qualifies. 9,000 < 24,680	true
  Eligible 15	Labels swapped	9,000 / true
  Income 10	AGI: pension 6,000. Provisional income 6,000 + 10,000 ÷ 2 = 11,000 < 32,000, so no Social Security is taxable	6,000
  Income 10	6,000 + 10,000 (both spouses' Social Security) + 1,000 (spouse's SSI)	17,000
  Income 11	Labels swapped	6,000 / 17,000
  Income 12	AGI 6,000; net benefits −1,000 aren't taxable	6,000
  Income 12	6,000 + 0 + min(−1,000, 0)	5,000
  Income 13	AGI 6,000; filers' net benefits 1,000 − 2,000 = −1,000	6,000
  Income 13	6,000 + 0 − 1,000	5,000
  Income 14	AGI 6,000; filers' net benefits −1,000 + 4,000 = 3,000; provisional income 7,500 < 32,000	6,000
  Income 14	6,000 + 3,000 + min(3,000, 0)	9,000
  Income 15	Labels swapped	6,000 / 9,000
  Income 16	AGI 6,000; the child's 3,000 of survivor benefits are not on the filer's return	6,000
  Income 16	Filer benefits for taxability are max(0, 2,000 − 3,000) = 0. 6,000 + 0 + min(−1,000, 0) − 500	4,500

• Comments and references:

  • The # PDF pages 8, 12-13 comment above the single #page=8 href in wi_homestead_income.py follows the convention in CLAUDE.md.
  • The /71/viii/53/2/c, /53/2/d, /54 and /52/6 URLs all resolve to the quoted text. § 71.54(1)(dm), (em) and (gm) say "the claimant or the claimant's spouse is over the age of 61", "unless the claimant is disabled", and "if the claimant had no earned income".

• Changelog. changelog.d/wi-homestead-dependent-income.fixed.md is still top-level and typed fixed. It now says "Preserve qualifying negative net Social Security repayments in household income."

• Code style.

  • income += min_(...) updates a fresh array that tax_unit_non_dep_add returns.
  • No policy value is hard-coded. The variable names in the list literal match the existing ["real_estate_taxes"] pattern.
  • Test periods are plain (2024).

• Callers. No variable outside the homestead files changed. wi_homestead_income is still read only by wi_homestead_credit and wi_homestead_eligible.

• Partner tests. The PR touches no file under tests/policy/baseline/partners/**, and the Household API Partners job passes.

Axiom line assessment

Valid queued. The line is axiom: TheAxiomFoundation/rulespec-us#1726 queued.

• State. The issue is OPEN, labelled pe-parity, and was last updated 2026-10-19 at 00:53 UTC, one minute after the head commit.
• The body has:
  • module path: us-wi/statutes/71.52.yaml
  • corpus citation: us-wi/statute/71.52 from the pinned release us-rulespec-2026-08-08-obbb-alien-snap
  • the verbatim § 71.52(4)–(6) text
  • required outputs: homestead_household_member, homestead_income_of_person and homestead_household_income
  • blocker: the billed signed-encoder run needs approval
  • companion tests A–D
• The 2026-10-19 comment adds:
  • outputs for (7) and (8)
  • the capital-loss add-back
  • corrected eligibility citations: § 71.53(2)(c), § 71.54(1)(dm)–(gm) and § 71.52(1g)
  • the § 71.52(6) repayment rule, with "Do not floor the household's qualifying net benefits at zero, or subtract a repayment twice"
• Companion cases E–Q are hand calculations from the statute and instructions. They match the head:
  • E–J match property tax Cases 2–3, income Cases 8–9 and eligibility Cases 10–11.
  • K–M match eligibility Cases 12–15 and income Cases 10–11.
  • N–Q match income Cases 12–16.
• Still unencoded. rulespec-us has no us-wi/statutes directory and no open PR that encodes it, so queued remains the right status.

CI status

All 29 checks pass (gh pr checks 10191, run 39184855705 on 4f5e4a6ff1, started 2026-10-19 01:19 UTC). I re-checked them at the end of this review.

• Where the homestead tests ran. Baseline states-shard-3 ran the five homestead files on the PR merge ref: credit 5, disqualified losses 9, eligibility 15, income 16 and property tax 3, so 48 cases in all.
• Coverage. Codecov patch and project pass.
• Timing. The run predates Tie source-conflict markers to the register shared with Axiom #10287 (the source-conflict ratchet) and the 2026-10-21 FL/SD LIHEAP ratchet breakage on main. See Branch status for the static ratchet check.
• No local runs. CI is green at the current head, and none of the newer main commits touches the tests' dependencies.

Branch status

• Position: 8 commits ahead of PolicyEngine/main (328d1721e2) and 142 behind; merge-base de3815f4cc.
• Merge: git merge-tree --write-tree PolicyEngine/main 4f5e4a6ff1 is clean (tree c37c7b943b), with no textual conflicts.
  • On the merged tree the five homestead files hold 48 uniquely named cases, the same set CI ran.
• Semantic check against newer main:
  • Unchanged dependencies. main hasn't changed any of the following since the merge-base: tax_unit_non_dep_add, filer_tax_exempt_social_security, tax_unit_social_security(_for_taxability), tax_unit_taxable_social_security, social_security, filer_loss_limited_net_capital_gains, the age, disability and role variables, adjusted_gross_income, or any homestead parameter other than disqualified_loss_sources.yaml.
  • Offset SSTB and non-SSTB Schedule C results in federal AGI #9995 changed disqualified_loss_sources.yaml. Round 1 reviewed it, and it was already in CI's merge ref.
  • Integrate modeled state unemployment insurance into income flows #9558 points tax_exempt_unemployment_compensation at total_unemployment_compensation. It is still a Person variable, so tax_unit_non_dep_add still leaves dependents out. The PR adds no raw unemployment_compensation reader, so the Integrate modeled state unemployment insurance into income flows #9558 guard is unaffected.
  • Leave dependents' unemployment compensation off California, Rhode Island and Wisconsin filer returns #10187 (dependents' unemployment compensation on WI returns) changes wi_unemployment_compensation_subtraction, which feeds Wisconsin AGI. Homestead income uses federal AGI, so it doesn't interact.
  • Make rules for a filer claimed elsewhere independent of which spouse is labelled head #10024 and Apply the federal rules for a filer who can be claimed as a dependent #10087 (filers claimed elsewhere) add dependent-elsewhere variables, which the homestead code doesn't read (Suggestion 1).
  • Commits after CI's merge ref that touch WI or federal gross income (Wire military retirement pay and survivors into federal and state tax (#10181) #10182, Count taxable employment disability benefits in federal gross income (IRC 105(a)) #10185, Cap Kentucky qualified residence interest at principal-residence interest from 2026 #10003, Support state-specific other additions and subtractions for remaining 30 states (Phase 2) #10302) add new zero-default sources or change WI AGI and the itemized deduction credit. None changes an input or variable in the PR's tests.
  • Make household state tax totals equal each state's own income tax #10039's new invariant (test_state_income_tax_accounting_invariants.py) checks an accounting identity, not homestead values.
  • Conclusion: merging changes no result and breaks no test.
• Source-conflict ratchet (Tie source-conflict markers to the register shared with Axiom #10287) on the merged tree:
  • No new comments. Run statically with the guard's regex, it counts 0 conflict-sounding comments in all five scanned files the PR touches, the same as on main and at the head.
  • No markers. No comment starts like a # Source conflict marker.
  • Entries untouched. KNOWN_UNMARKED lists homestead/income/disqualified_loss_sources.yaml: 1, which the PR does not touch.
  • Result: the ratchet passes on the merged tree.

Verdict: APPROVE

[DTrim99]

Round 2: Should 1 is fixed. The spouse-side cases now run with both labellings and catch the round-1 mutants. S1-S4 are fixed too, including negative net Social Security repayments reducing household income per s. 71.52(6). All 22 new expectations hand-check, and the merged tree is clean and passes the ratchet. Two suggestions: the pre-existing s. 71.53(2)(d) dependent-claimant bar, and filing the S corporation follow-up as an issue. 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]

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: 5c3716b10baeacc4396bfcb7968dba1ea2d0a6ef (fast-forwards from a953caf419; main still merges cleanly, so no new merge).

What the merge-delta review asked, and how it is resolved

P2: count Wisconsin's state supplement in homestead household income.

• policyengine_us/parameters/gov/states/wi/tax/income/credits/homestead/income/sources.yaml:24-28: the commented # wi_state_supplement placeholder becomes - wi_ssp (regular supplement, SSI-E and Caretaker Supplement), with Wis. Stat. 71.52(4)-(6) and DOR Publication 127, page 9, added as references.
• wi_ssp is a monthly Person variable. wi_homestead_income already sums its sources with tax_unit_non_dep_add, so the claimant's and spouse's payments count and a dependent child's own supplement does not. A parent's Caretaker Supplement is the parent's own wi_cts, so it counts.
• New cases in wi_homestead_income.yaml:
  • Case 17: the SSP integration Case 10 household (2026), federal SSI 11,928 plus supplement 2,301.08 = 14,229.08.
  • Case 18: a dependent child's SSI 6,000 and supplement 1,000 are left out: 20,000 − 500 = 19,500.
  • Case 19: an SSI parent's 3,000 Caretaker Supplement counts: 9,000 + 3,000 − 500 = 11,500.
• New case in wi_homestead_credit.yaml, Case 6: the same Case 10 household with 1,460 of allowable property taxes gets $734.44.
• Existing Cases 10-11 move from 17,000 to 18,005.36: the spouse's 1,000 of SSI brings the regular supplement, 83.78 × 12 = 1,005.36, exactly as the review's hand computation predicted. I updated both expectations with the arithmetic in a comment.

P3: the impact label. The body's heading now reads "PR head 4f5e4a6ff1 against main de3815f4cc", says the figures predate the current head, and says the impact is being rerun because this fix and main's SSI-parent W-2 exclusion can both move households through the credit phase-out. A new body section, "State SSI supplement (merge-delta review)", shows the before/after values.

Hand computations (2026 parameters: phase-out start 8,060, rate 8.785%, credit rate 80%, property tax cap 1,460)

• With the supplement: (14,229.08 − 8,060) × 0.08785 = 541.95; (1,460 − 541.95) × 0.8 = 734.44.
• Without it: (11,928 − 8,060) × 0.08785 = 339.80; (1,460 − 339.80) × 0.8 = 896.16 (overpayment 161.72).

Tests (local, one file per process)

File	Result
wi/tax/income/wi_homestead_income.yaml	19 passed
wi/tax/income/wi_homestead_credit.yaml	6 passed
wi/tax/income/wi_homestead_eligible.yaml	15 passed
wi/tax/income/wi_homestead_property_tax.yaml	3 passed
wi/tax/income/wi_homestead_disqualified_losses.yaml	9 passed
wi/dhs/ssp/integration.yaml	10 passed
wi/dhs/ssp/wi_ssp.yaml	17 passed
wi/dhs/ssp/wi_cts.yaml	55 passed
wi/dhs/ssp/wi_ssp_countable_for_hud.yaml	13 passed
code_health/test_source_conflicts.py	103 passed

That is 147 YAML cases, 0 failed.

Mutant (wi_ssp commented out of sources.yaml, restored after): wi_homestead_income.yaml fails 4 of 19 (Cases 10 and 11 at 17,000, Case 17 at 11,928, Case 19 at 8,500), and wi_homestead_credit.yaml fails Case 6 at household income 11,928. Case 18, the dependent child's supplement, passes either way, as it should.

CI

• The first push (d48646b614) failed in Quick Feedback and states shard 3, on Cases 12-16 of wi_homestead_income.yaml: "Reported Social Security benefits must be finite and nonnegative". Those are the PR's earlier 71.52(6) repayment cases, which give a negative social_security_retirement. Main (Add reported-benefit Social Security PIA ratio lever #10484) now treats that input as a reported benefit and rejects negative amounts, and CI tests the merge with main.
• 5c3716b10b gives those cases each filer's net benefits through social_security instead (Case 16's dependent gets social_security: 3_000), with the same expected amounts. They pass on the branch and on a scratch merge with current main (19 of 19 both ways).
• The Pull request run at 5c3716b10b passed 36 of 36 jobs.

Judgment calls

• The repayment feature now depends on supplying net social_security directly: main's component inputs can't be negative. A model user with a repayment recorded as a negative component would hit main's error; that's main's rule, not this PR's.
• Case 19 sets the parent's wi_cts (with wi_ssp_regular and wi_ssi_e at 0) instead of building a full Caretaker Supplement eligibility case, so it isolates the homestead rule from the CTS formula. Setting the child's wi_ssp in the same case would have marked wi_ssp as an input for every member and zeroed the parent's composed value, so the child's supplement isn't set.
• I updated Cases 10-11 rather than adding wi_ssp: 0 inputs, because the review's hand computation shows 18,005.36 is the correctly composed amount.
• Sources: this session's network couldn't reach docs.legis.wisconsin.gov or revenue.wi.gov, so the statute and Publication 127 are cited, not quoted. That they count state SSI, SSI-E and Caretaker Supplement payments rests on the review and the brief: unverified here.
• No microsimulation was run; the impact rerun is the hub's.

---

Generated by Claude Code

[MaxGhenis]

Hub review in progress

Reviewing head 5c3716b10baeacc4396bfcb7968dba1ea2d0a6ef, unchanged since the hand-off. These are interim findings; the final review follows as a new comment.

• Defect reproduces at main and is fixed. At current main 25c4253483, the SSP integration Case 10 household (2026, age 70, SSI-E certified) has ssi 11,928, wi_homestead_income 11,928 and, with 1,460 of allowable property tax, wi_homestead_credit 896.16. On a merge of this head into that main, the values are 14,229.08 and 734.44.
• Hand values agree. (14,229.08 − 8,060) × 0.08785 = 541.95, and (1,460 − 541.95) × 0.8 = 734.44. Cases 10–11: 6,000 + 10,000 + 1,000 + 83.78 × 12 = 18,005.36.
• Head tests: 147 YAML cases passed and 0 failed across 9 files, and test_source_conflicts.py passed 103, one file per process.
• Mutant A (the hand-off's): with wi_ssp removed from sources.yaml, income Cases 10, 11, 17 and 19 fail and credit Case 6 fails. That matches the hand-off.
• CI: the Pull request run at the head passed 36 of 36 jobs.
• Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104 overlap: wi_homestead_eligible.yaml Cases 8–15 are byte-identical (203 lines, same SHA-256). A trial merge conflicts in both wi_homestead_eligible.py and wi_homestead_eligible.yaml.
• Sources: this VM cannot reach docs.legis.wisconsin.gov or revenue.wi.gov. The rule check rests on the hub's 71.52(4)–(6) text.

Still to do: mutant B (each tax unit member's supplement), the body checks, and the final write-up.

---

Generated by Claude Code

[MaxGhenis]

Ready for hub review
VERDICT: APPROVE
5c3716b
METHODOLOGY_CHOICE: yes

This is an independent delta review of a953caf419..5c3716b10b (3 commits: ee0916e8aa, d48646b614, 5c3716b10b). The head has not moved since the hand-off. The methodology choices carried over from earlier rounds are unchanged:

• negative net Social Security is treated as a qualifying 71.52(6) repayment without claim history;
• a child's benefits paid to the parent on the child's behalf are treated as the child's, following the Schedule H instructions.

This delta adds no new reading choice.

No P1 findings. The code change is correct and its tests kill both mutants. The remaining items are body, citation and test-strength fixes, plus one follow-up.

Findings

P2: the body has no impact: line

• Where: PR body. It has an axiom: line (body line 100) but no line starting impact:.
• Command: gh api repos/PolicyEngine/policyengine-us/pulls/10191 --jq .body | grep -n '^impact:' returns nothing.
• Why it matters: the diff moves baseline outputs. On the default dataset, every Wisconsin SSI recipient who is a homestead claimant or spouse now has wi_ssp in household income, so credits fall and some units cross the 24,680 income limit.
• Fix: the brief says this line is the hub's to add after it reruns the impact.
• Population check needed: a rerun of wi_homestead_income, wi_homestead_eligible and wi_homestead_credit at this head for 2021–2026, not only 2024 and 2026. The supplement now reaches 2021–2025 too (see the P3 follow-up below).

P3: cite 71.52(6) for the supplement, not 71.52(5) alone

• Where: policyengine_us/parameters/gov/states/wi/tax/income/credits/homestead/income/sources.yaml:26-27, and the body's "State SSI supplement" paragraph.
• The problem: both cite "71.52(5)" as the provision that counts state SSI payments. Under the hub's text, (5) defines household income as "all income received by all persons of a household". It is (6) that makes "cash public assistance" part of "income".
• Replacement for lines 26–27:

      # State SSI, SSI-E and Caretaker Supplement payments are cash public
      # assistance (Wis. Stat. 71.52(6); Publication 127, page 9). Like ssi, a
      # dependent child's own supplement stays out of household income
      # (71.52(4)-(5)).

P3: Case 19 does not test the dependent exclusion

• Where: policyengine_us/tests/policy/baseline/gov/states/wi/tax/income/wi_homestead_income.yaml:509.
• What it tests: the parent's wi_cts is set directly at 3,000 and the child's supplement composes to 0. The case shows that a CTS amount recorded on the parent counts.
• What it can't test: it can't tell tax_unit_non_dep_add from tax_unit.sum. Mutant B, below, passes it. It also doesn't show that main's wi_cts formula puts the payment on the parent.
• Composed probe: I ran this one-off scratch case at the head.
  • Inputs: 2026; parent age 40 with ssi 11,928; child age 8 with parent_1_id set; one family and household in Dane County, WI.
  • Results: wi_cts is [3,300, 0], which is 275 × 12 on the parent. wi_ssp is [4,372.40, 0], which is 3,300 + 4 × 83.78 + 8 × 92.16. wi_homestead_income is 11,928 + 4,372.40 − 500 = 15,800.40.
  • Run: python -m policyengine_core.scripts.policyengine_command test <scratch>.yaml -c policyengine_us gave 1 passed.
• Suggestion: add that composed case, or give the child in Case 19 wi_ssp_regular: 1_000. The hand-off says setting wi_ssp on the child blocks the parent's composed value; setting the component avoids that. Either way the expected value stays 11,500.

P3: wording in the impact section

• Where: body line 106 says the supplement change is "(below)", but the "State SSI supplement" section is above it.
• Fix: replace "(below)" with "(above)".

Follow-up, not this PR's defect: the supplement amounts are back-extrapolated before 2026

• wi_ssp_regular, wi_ssi_e and wi_cts read parameters whose first values are dated 2026-01-01:
  • amount/individual.yaml: 83.78;
  • exceptional_expense/individual.yaml: 95.99;
  • cts/first_child.yaml: 275.
• Their only references are 2026 sources, so every earlier year uses the 2026 January amounts.
• Witness: income Cases 10–11 (TY2024) expect 83.78 × 12 = 1,005.36. Under mutant B, Case 7 (TY2020) gives the child a 1,005.36 supplement.
• Main already counts wi_ssp in household_state_benefits for 2023 onward on the same values, so the extrapolation isn't new. This PR newly carries it into the 2021–2025 homestead credit.
• Proposed follow-up issue:
  • Title: "Add pre-2026 history to Wisconsin SSP, SSI-E and Caretaker Supplement amounts"
  • Description: The WI SSP and CTS amount parameters start on 2026-01-01, so 2021–2025 benefits and the homestead credit use 2026 amounts. Add the sourced history from DHS operations memos, or gate the earlier years.

Checked and sound

• The rule. I checked it against the hub's text of 71.52(4)–(6).

  • (6): "income" includes "cash public assistance", excluding only homestead and farmland credits and s. 46.27 amounts. The code cites s. 49.77 (wi_ssp_eligible) for the regular supplement and SSI-E, and s. 49.775 (wi_cts) for the Caretaker Supplement. Neither falls in those exclusions, so all three are cash public assistance that counts.
  • (4)–(5): the household is the claimant and the claimant's spouse, and household income counts what "persons of a household" receive. So the claimant's and spouse's payments count, a dependent child's own payments do not, and a parent's CTS is the parent's own payment.
  • The code: wi_homestead_income.py:28 sums sources.yaml with tax_unit_non_dep_add. wi_ssp is a Person variable, so it goes through tax_unit_non_dep_sum, which multiplies by not_(is_tax_unit_dependent) (tools/general.py:82-86). The variable is MONTH and is read for a YEAR, and the cases confirm the annual sum. The same rule applies in every year and filing status, because it doesn't depend on either.

• The defect at main. I reproduced it with a Python probe of the SSP integration Case 10 household (2026, age 70, wi_ssi_e_certified, SNAP take-up off).

  Tree	ssi	wi_homestead_income	wi_homestead_credit (1,460 allowable property tax)
  main 25c4253483	11,928	11,928	896.16
  this head merged into main 25c4253483	11,928	14,229.08	734.44

• Hand computations, done before reading the expected values:

  Case	Computation	Result
  Income Case 17	federal SSI 994 × 12 = 11,928; regular 4 × 83.78 + 8 × 92.16 = 1,072.40; SSI-E 4 × 95.99 + 8 × 105.59 = 1,228.68; 11,928 + 1,072.40 + 1,228.68	14,229.08
  Credit Case 6	(14,229.08 − 8,060) × 0.08785 = 541.95; (1,460 − 541.95) × 0.8	734.44
  Credit without the supplement	(11,928 − 8,060) × 0.08785 = 339.80; (1,460 − 339.80) × 0.8	896.16
  Income Cases 10–11	pension 6,000 + Social Security 10,000 (nontaxable: 5,000 + 6,000 is under the 32,000 joint base) + spouse SSI 1,000 + 83.78 × 12 = 1,005.36	18,005.36
  Income Case 18	20,000 wages − 500, with the child's SSI 6,000 and supplement 1,000 excluded	19,500
  Income Case 19	9,000 + 3,000 − 500	11,500

  • The 2026 parameters are phase-out start 8,060, rate 0.08785, credit rate 0.8 and property tax cap 1,460, read from the parameter files.
  • In Cases 10–11 the head has no SSI, so the spouse gets the individual rate.
  • Every value matches the YAML.

• The CI fix (5c3716b10b). Cases 12–16 test the same rule as before.

  • wi_homestead_income.py reads net social_security through tax_unit_non_dep_add(..., ["social_security"]), and social_security is only the sum of its four components. The old negative social_security_retirement inputs reached the formula only through that sum, so supplying the sum directly exercises the same branch.
  • Expected values are unchanged: 5,000, 5,000, 9,000, 9,000 and 4,500.
  • On main, a negative component input now raises "Reported Social Security benefits must be finite and nonnegative". My probe got that error at main and on the merge: a 2024 claimant with pension 6,000 and social_security_retirement −1,000.
  • A repayment entered as net social_security: -1000 still reaches the rule on main: wi_homestead_income is 5,000 on both main and the merge.
  • Population data has no negative benefits, so this branch has no baseline effect.

• Tests at the head, one file per process.

  File	Result
  wi_homestead_income.yaml	19 passed
  wi_homestead_credit.yaml	6 passed
  wi_homestead_eligible.yaml	15 passed
  wi_homestead_property_tax.yaml	3 passed
  wi_homestead_disqualified_losses.yaml	9 passed
  dhs/ssp/integration.yaml	10 passed
  dhs/ssp/wi_ssp.yaml	17 passed
  dhs/ssp/wi_cts.yaml	55 passed
  dhs/ssp/wi_ssp_countable_for_hud.yaml	13 passed
  code_health/test_source_conflicts.py	103 passed

  • That is 147 YAML cases with 0 failed.
  • On a merge with current main 25c4253483, the five homestead files pass 19, 6, 15, 3 and 9.
  • The new cases assert the claim itself, not a weaker one. They check exact amounts at a 0.01 margin, and the delta has no generated ranges.

• Mutants. Each ran in its own process on a scratch worktree, and git status was clean afterwards.

  Mutant	Change	Income file	Credit file	Killed by
  A (the hand-off's)	- wi_ssp commented out of sources.yaml	4 failed of 19: Cases 10 and 11 (17,000), 17 (11,928) and 19 (8,500)	1 failed of 6: Case 6 (income 11,928)	Cases 10, 11, 17, 19 and credit Case 6
  B (mine): count every member's supplement	tax_unit_non_dep_add(..., include_dependents=("wi_ssp",)) at wi_homestead_income.py:28	2 failed of 19: Case 18 (20,500) and Case 7 (2020: 20,505.36, because the child's composed supplement now counts)	6 passed	Cases 18 and 7

• Readers. sources.yaml is read only by wi_homestead_income. wi_homestead_income is read by:

  • wi_homestead_eligible (the < max_income test);
  • wi_homestead_credit, which feeds wi/.../credits/refundable.yaml, retirement_income_exclusion_refundable.yaml and household/state_property_tax_credits.yaml.

  Each reader moves in the direction 71.52(6) requires. The one pre-existing limitation is listed above with its witness.

• Overlap with Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104 (head a35d9295a3, open).

  • Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104's wi_homestead_eligible.yaml lines 417–619 and this head's lines 103–305 (Cases 8–15) are byte-identical: 203 lines each, both with SHA-256 3c5e7ab5…63fb23.
  • A trial git merge pr-10104 onto this head merged into main conflicts in wi_homestead_eligible.py and wi_homestead_eligible.yaml.
  • What the second lander should keep:
    • Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104's per-person formula, which tests each head or spouse alone and adds the 71.53(2)(d) claimed-dependent gate. It matches this PR's head and spouse tests: head_is_disabled is is_disabled for the head.
    • Both PRs' reference entries: the 2024 and 2025 Schedule H instructions, 71.53(2)(c), 71.54 and 71.52(1g).
    • Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104's rewritten unit tests 1–7 and its nine 2025 claimed-dependent cases.
    • One copy of the shared Cases 8–15.
  • Fix the 2020 unemployment compensation exclusion: cutoff test, per-spouse cap, dependents #10175 matched the search only by text. It shares no file with this PR.

• Body checks.

  • The "State SSI supplement" table: rows 1 and 4 match my runs and hand computations, and rows 2 and 3 are income Cases 19 and 18.
  • The impact heading is relabelled to 4f5e4a6ff1 against de3815f4cc and says the figures predate this head.
  • The changelog sentence matches the code.
  • The axiom: line is present.

• CI. The Pull request run at 5c3716b10b (run 41739391616, attempt 1) completed with success. All 36 jobs succeeded, and none was skipped or cancelled. They include all four baseline states shards, Quick Feedback, Household API Partners and Microsimulation. Main moved to 25c4253483 after that run; my local merge with it passes the homestead files above.

• Scope. No file under .github/ or policyengine_us/tests/policy/baseline/partners/ is in the PR's 11 files.

What I couldn't check

• Sources. From this VM, docs.legis.wisconsin.gov and revenue.wi.gov are blocked, both to curl and to the fetch tool. govinfo.gov, uscode.house.gov, ecfr.gov, dhs.wisconsin.gov, ssa.gov and law.cornell.edu are blocked too. As a result:
  • The rule check rests on the hub's quotes of 71.52(4)–(6).
  • I couldn't read Publication 127 page 9, ss. 49.77 and 49.775, or the DHS memo behind the 2026 supplement amounts.
  • That 83.78, 95.99 and 275 also applied before 2026 is unverified.
• Population impact. I ran no microsimulation, so the impact rerun is the hub's.
• Peak memory of the test runs.

---

Generated by Claude Code

[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: dbe35eb8bcecb6942ecbe6d927f2cf6529ba6dcd

Built on 5c3716b10b, the head the round 4 review approved. No one else pushed in between. I didn't merge main: git merge-tree against origin/main (25c4253483) is clean, and no check needed it.

Items

Item 5, the citation in sources.yaml. Lines 26-30 now carry the brief's comment at the file's indentation (commit 89b8dc0e8b).

The old comment ran over lines 26-28, not 26-27. Its third line, "supplement stays out of household income.", is restated by the new comment's last sentence. Replacing only 26-27 would have left that fragment dangling, so I replaced all three lines.

Sources:

• DOR Publication 127, fetched from revenue.wi.gov. On page 9 (file page 9), under "SOCIAL SECURITY, SSI, AND RAILROAD RETIREMENT", it says "SSI" includes "the additional 'State SSI' payments", "SSI-E" payments and "caretaker supplement" payments. The same page tells claimants to report SSI on Schedule H line 9b.
• 2025 Schedule H instructions, file page 5. Among household income sources they list "Social security benefits (including federal and state SSI, SSI-E, SSD, and caretaker supplement payments)" and "other cash public assistance".
• Wis. Stat. 71.52: I couldn't read it.
  • docs.legis.wisconsin.gov reset the connection to curl and returned 503 to the fetch tool.
  • The Justia and FindLaw copies returned 403, and web.archive.org is blocked.
  • So the 71.52(6) "cash public assistance" wording rests on the hub's reading.

Item 6, Case 19. wi_homestead_income.yaml:509-541 (commit dbe35eb8bc):

• The child now has wi_ssp_regular: 1_000 (line 525).
• With that input the child's wi_ssp composes to 1,000, so the case's wi_ssp output had to change from [3_000, 0] to [3_000, 1_000] (line 538). The comment now says the child's 1,000 stays out.
• wi_homestead_income stays 11,500.

Mutants, each in its own process, with the code restored and git status clean after each:

Mutant	Income file	Killed by (run vs expected)
B, the review's: include_dependents=("wi_ssp",) at wi_homestead_income.py:28	3 failed, 16 passed	Case 7 (20,505.36 vs 19,500), Case 18 (20,500 vs 19,500), Case 19 (12,500 vs 11,500)
B2, the brief's wording: add(tax_unit, period, sources), a tax_unit.sum over every member for every source	3 failed, 16 passed	Case 7 (31,505.36), Case 18 (26,500), Case 19 (12,500)
B, against Case 19 alone, before and after this change	before: 1 passed; after: 1 failed	—

Item 7, the body.

• Replaced the 71.52(5) sentence with the brief's 71.52(6) sentence, and changed "(below)" to "(above)".
• I left the impact figures, and the missing impact: line, for the hub.
• The readback matches what I sent except one blank line before the footer, which the server dropped.
• Case 19's change leaves the "State SSI supplement" table true. Row 2 (8,500 before, 11,500 after) holds because the child's supplement is left out both before and after.

Item 8, runs at this head, one file per process:

File	Result
wi_homestead_income.yaml	19 passed
wi_homestead_credit.yaml	6 passed
wi_homestead_eligible.yaml	15 passed
test_source_conflicts.py (its comment ratchet reads parameter files)	103 passed

Hand computation

Case 19:

• The parent counts: SSI 9,000 + wi_ssp 3,000 (wi_cts 3,000; regular supplement and SSI-E 0).
• Less the $500 dependent exemption: 9,000 + 3,000 − 500 = 11,500.
• The child's SSI is 0, and the child's 1,000 supplement is left out.
• Under B: 9,000 + 3,000 + 1,000 − 500 = 12,500.

CI

Pull request run 41860003710 at dbe35eb8bc hadn't finished 60 minutes after the push (15:39Z). At 16:41Z it was still in_progress: 35 of 36 jobs had concluded success, and none had failed or been cancelled. The one still running was Full Suite - Contrib (other-shard-3), on its YAML tests since 16:08Z. The jobs that had finished include every Baseline shard, every Python shard, Microsimulation, Household API Partners, Quick Feedback and PR validation.

Judgment calls

• Comment lines. I replaced lines 26-28 rather than 26-27, for the reason above.
• Case 19's output. I changed its wi_ssp output to match the new input.
• A second form of mutant B. I ran the brief's form as well as the review's.
• Text that predates this round. The body still names two local paths: ~/reviews/... in "How it was checked" and fixes/10191-impact/ in the impact section. Its "Review" section also names a model. I left all three: they predate this round, and the impact section is the hub's. The hub may want to drop them before landing.
• Ruff. This round changed no Python file. The locked ruff 0.15.5 passes ruff format --check and ruff check on the tree.

Follow-ups

None new. The review's proposed issue still stands: "Add pre-2026 history to Wisconsin SSP, SSI-E and Caretaker Supplement amounts".

---

Generated by Claude Code

[MaxGhenis]

Hub review in progress

Done: the round 5 review at dbe35eb8bc is posted at #10191 (comment) (APPROVE).

---

Generated by Claude Code

[MaxGhenis]

Ready for hub review
VERDICT: APPROVE
dbe35eb
METHODOLOGY_CHOICE: yes

This is an independent round 5 delta review of 5c3716b10b..dbe35eb8bc, two commits:

• 89b8dc0e8b rewrites the citation comment in sources.yaml;
• dbe35eb8bc gives Case 19's child a supplement.

The head hasn't moved past dbe35eb8bc. The delta changes one parameter-file comment and one test case, so it can't move any output. The methodology choices from earlier rounds carry over unchanged:

• negative net Social Security is treated as a qualifying 71.52(6) repayment;
• a child's benefits paid on the child's behalf are treated as the child's.

There are no P1 findings. The round 4 items for #10191 (items 5–7 of the assignment) are done as asked, Case 19 now kills mutant B, and CI is green. What remains is the hub's impact: line and one optional wording fix.

Findings

P2 (the hub's, carried from round 4): the body still has no impact: line

• Where: the PR body. It has an axiom: line (body line 98), but no line starts with impact:.
• Command: gh api repos/PolicyEngine/policyengine-us/pulls/10191 --jq .body | grep -nE '^(impact|axiom):' returns only the axiom: line.
• Why it matters: the PR as a whole moves baseline outputs. On main, 10 of the PR's 19 income cases fail (see "Checked and sound"), and the impact table was run at 4f5e4a6ff1, before wi_ssp joined the source list.
• Fix: the assignment leaves this line to the hub. The population check it calls for is a rerun of wi_homestead_income, wi_homestead_eligible and wi_homestead_credit (and so wi_income_tax and household_net_income) at this head. Run 2021–2026, because wi_ssp now reaches 2021–2025 household income through 2026 parameter values extrapolated backward (the round 4 follow-up). The round 5 delta itself needs no impact run.

P3: the new comment's "which" can be read as pointing at the statute

• Where: policyengine_us/parameters/gov/states/wi/tax/income/credits/homestead/income/sources.yaml:27-28: "cash public assistance under Wis. Stat. 71.52(6), which DOR has claimants report with SSI".
• The problem: the nearest noun before "which" is "Wis. Stat. 71.52(6)", but what DOR has claimants report is the payments. Lines 26–30 otherwise read as one comment and say what the code does: wi_ssp is listed, and tax_unit_non_dep_add at wi_homestead_income.py:28 keeps a dependent's supplement out. This is the brief's own text, so the fix is optional.
• Replacement for lines 26–30:

      # Wisconsin counts state SSI, SSI-E and Caretaker Supplement payments as
      # income: they are cash public assistance under Wis. Stat. 71.52(6), and
      # DOR has claimants report them with SSI (Publication 127, page 9). Like
      # ssi, a dependent child's own supplement stays out of household income
      # (71.52(4)-(5)).

Checked and sound

• The assignment's Leave dependents' income out of Wisconsin homestead credit household income #10191 items.

  • Item 5: the comment is in place at the file's indentation. git diff 5c3716b10b dbe35eb8bc shows the old comment took three lines (26–28), not two. Its third line was "supplement stays out of household income.", so replacing all three, as the session did, was right: replacing only 26–27 would have left that fragment behind.
  • Item 6: Case 19's child has wi_ssp_regular: 1_000 (wi_homestead_income.yaml:525), and the expected wi_homestead_income stays 11,500. The wi_ssp output changed from [3_000, 0] to [3_000, 1_000], which follows from the new input.
  • Item 7: body line 87 has the brief's 71.52(6) sentence word for word, and line 104 says "(above)". The "State SSI supplement" section (lines 85–96) sits above the impact section (lines 102–121), so "(above)" is correct. No "(below)" and no "71.52(5)" citation for the supplement remain in the body. The body is unchanged since I first read it.
  • Item 8: see the test table below.

• Case 19 by hand, done before running.

  • The parent counts SSI 9,000 plus wi_ssp 3,000 (regular 0 + SSI-E 0 + CTS 3,000), with no AGI. Less the $500 dependent exemption, that gives 9,000 + 3,000 − 500 = 11,500.
  • The child's 1,000 supplement stays out. If it counted, the result would be 12,500.
  • That matches the YAML and the run.
  • Note: the child has ssi: 0 but a state supplement, a combination that wouldn't occur in practice. It's an input that isolates the dependent filter, and Case 18 covers the realistic child with SSI 6,000 plus supplement 1,000.

• Mutants. Each ran in its own process on a scratch worktree of the head, with the code restored and git status clean after each.

  Mutant	Change	Income file (19)	Credit file (6)	Killed by
  B (round 4's)	include_dependents=("wi_ssp",) at wi_homestead_income.py:28	3 failed	6 passed	Case 7 (20,505.36 vs 19,500), Case 18 (20,500 vs 19,500), Case 19 (12,500 vs 11,500)
  B2 (the session's)	add(tax_unit, period, sources) at line 28	3 failed	6 passed	Case 7 (31,505.36), Case 18 (26,500), Case 19 (12,500)
  B against Case 19 alone	extracted Case 19 at 5c3716b10b and at dbe35eb8bc	old: 1 passed; new: 1 failed (12,500)	—	the new Case 19
  C (mine): only the CTS counts	- wi_ssp changed to - wi_cts in sources.yaml	3 failed	1 failed	income Cases 10 and 11 (17,000 vs 18,005.36), Case 17 (11,928 vs 14,229.08); credit Case 6

  Case 19 passes under C, as designed: it holds only a parent's CTS. Every number the hand-off gives for B and B2 matches my runs.

• The defect at main (25c4253483, run in its own environment because main needs a newer policyengine-core than this branch's lock).

  • The PR's income file at main: 10 failed, 9 passed. The failures are Cases 6 (35,500 vs 23,500), 7 (30,500 vs 19,500), 8, 9, 10, 11, 16, 17, 18 and 19.
  • The first two values are the body's main column. Case 19 gives 8,500, the "before" value in row 2 of the body's supplement table.

• Tests, one file per process.

  File	At head	Head merged into main 25c4253483
  wi_homestead_income.yaml	19 passed	19 passed
  wi_homestead_credit.yaml	6 passed	6 passed
  wi_homestead_eligible.yaml	15 passed	15 passed
  wi_homestead_property_tax.yaml	3 passed	3 passed
  wi_homestead_disqualified_losses.yaml	9 passed	9 passed
  code_health/test_source_conflicts.py	103 passed	—

  The merge with main is clean. Case 19 asserts the claim itself, an exact 11,500 at a 0.01 margin. The delta adds no generated ranges.

• Readers.

  • gov.states.wi.tax.income.credits.homestead.income.sources is read only by wi_homestead_income.py:28.
  • wi_homestead_income is read by wi_homestead_eligible.py:46 and wi_homestead_credit.py:23. wi_homestead_credit feeds wi/tax/income/credits/refundable.yaml, retirement_income_exclusion_refundable.yaml and household/state_property_tax_credits.yaml.
  • The delta changes no reader's inputs, because the comment and the test case aren't read by any formula.

• CI. The Pull request run 41860003710 at dbe35eb8bc (attempt 1) completed with success at 16:43Z. All 36 jobs concluded success, and none was skipped or cancelled. Contrib other-shard-3, which the hand-off left running, succeeded. The jobs include all four Baseline states shards, Household API Partners, Microsimulation, Quick Feedback and PR validation.

• Scope. The PR's 11 files include nothing under .github/ or policyengine_us/tests/policy/baseline/partners/.

• Overlap with open PRs. I searched open PRs for each changed file name and compared file lists.

  • Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104 shares wi_homestead_eligible.py and wi_homestead_eligible.yaml. Round 4 already recorded that conflict and what the second lander should keep; the delta doesn't touch either file.
  • Fix the 2020 unemployment compensation exclusion: cutoff test, per-spouse cap, dependents #10175, Model the Medicare Part D low-income subsidy as a CBO means-tested transfer #10127, Apply the Form 4684 $100 per-casualty reduction and drop dependents' losses in the federal, NY, AL and HI casualty deductions #9912, Zero a separate filer's whole standard deduction and exclude dependent care benefits from wages #10579, Add West Virginia LIEAP regular heating assistance #10618 and Add Hawaii LIHEAP regular heating eligibility and subsidized-dwelling benefit #10614 matched by text only and share no file.

What I couldn't check

• Sources. This session can't reach docs.legis.wisconsin.gov or revenue.wi.gov: the proxy refuses the CONNECT, and the fetch tool can't resolve either host. So:
  • I couldn't read Wis. Stat. 71.52(4)–(6), DOR Publication 127 page 9, or the Schedule H instructions. The comment's 71.52(6) "cash public assistance" reading and the Publication 127 citation rest on the hub's reading and the session's.
  • The hand-off says the 2025 instructions list SSI-E and caretaker payments on file page 5. The pre-existing reference in sources.yaml:9 points to #page=6. I couldn't open the PDF to say which page is right.
• Population impact. I ran no microsimulation. The impact: line and its 2021–2026 rerun are the hub's.
• Peak memory of the test runs.

---

Generated by Claude Code

[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: 4c7110c3b5fc41fff018eebf352d11ad8c1aa607

Built on dbe35eb8bc, the head the item named; no other commit landed on the branch. No merge of origin/main was needed.

Items

7. P3, the comment's "which". policyengine_us/parameters/gov/states/wi/tax/income/credits/homestead/income/sources.yaml:26-30 now carries the item's five-line comment word for word, at the file's indentation, above - wi_ssp. The change touches only this file, as the item said; Fix Vermont student loan interest, Utah's at-home parent credit and Wisconsin's homestead claimant #10104 doesn't carry it. The P2 impact: line is left to the hub.
8. Runs, one file per process at this head:

File	Result
wi_homestead_income.yaml	19 passed
wi_homestead_eligible.yaml	15 passed
wi_homestead_credit.yaml	6 passed

Mutants and hand computations

None. The change is a YAML comment, so it can't move an output, and no Python file changed.

Body

I read the body. It paraphrases the 71.52(6) and Publication 127 point in its own words (the "State SSI supplement" section) and doesn't quote the comment, so no sentence became false. It has no CI sentence, so I made no PATCH.

CI

Pull request run 41945112558 at 4c7110c3b5 concluded success: all 36 jobs completed with success, none cancelled or skipped.

Judgment calls

• None.

---

Generated by Claude Code