Subtract Michigan premiums once and exclude the first $300 of gambling winnings and gifts (MI-1040CR lines 20, 24, 30 and 31)

[MaxGhenis]

Summary

Two fixes to Michigan total household resources (mi_household_resources), the income measure for the homestead property tax credit and the home heating credit.

1. Premiums were subtracted twice for a self-employed claimant. Line 30 subtracts the U.S. Schedule 1 adjustments, which include the self-employed health insurance deduction. Line 31 then subtracted every member's premiums again. The MI-1040CR says line 31 leaves out "any insurance premiums deducted on lines 21 or 30". Example: a self-employed claimant with $50,000 of Schedule C income and $5,000 of premiums lost $10,000 of household resources instead of $5,000. The model gave 36,467.61; the form gives 41,467.61.
2. The $300 exclusions in MCL 206.510(1)(a) and (b) were missing. Gambling winnings counted in full, where the form counts winnings "over $300" (line 20). Gifts from relatives and friends (financial_assistance) were not counted at all, where the form counts "the value over $300 in gifts" (line 24).

Merge after #9796 (household = claimant and spouse), which itself follows #9790, #9788 and #9801. This branch merges #9796's head 8d494dedec. Until those PRs merge, the diff here includes their commits. This PR's own change is git diff 8d494dedec 7fa23e51f6.

The exclusions apply once to the claimant's and spouse's total, as each form line is one amount for them.

Coordination with #9795. #9795 replaces line 30's list with household_resources_adjustments. Whichever lands second should keep #9795's list as the deductions variable here. Line 31 keeps the premiums deducted on line 30 only while self_employed_health_insurance_ald is in that list, and #9795's list includes it.

The law and the forms

MCL 206.510(1) (https://www.legislature.mi.gov/Laws/MCL?objectName=mcl-206-510):

"Income" means the sum of federal adjusted gross income as defined in the internal revenue code plus all income specifically excluded or exempt from the computations of the federal adjusted gross income. Also, a person who is enrolled in an accident or health insurance plan may deduct from income the amount that person paid in premiums in the tax year for that insurance plan for the person's family. Income does not include any of the following:
(a) The first $300.00 of gifts in cash or kind from nongovernmental sources.
(b) The first $300.00 received from awards, prizes, lottery, bingo, or other gambling winnings.
...
(h) Amounts deducted from monthly social security or railroad retirement benefits for medicare premiums.

The $300 amounts are fixed in the statute and not indexed. The section was last amended effective January 1, 2012 (2011 PA 38 and 172).

2025 MI-1040 book (https://www.michigan.gov/taxes/-/media/Project/Websites/taxes/Forms/IIT/TY2025/MI-1040-Book.pdf). The 2022, 2023 and 2024 books have the same text for each line below, on page 32 or 33:

• Page 27, what total household resources do not include: "The first $300 from gambling, bingo, lottery, awards, or prizes", "The first $300 in gifts of cash or merchandise received, or expenses paid on your behalf (rent, taxes, utilities, food, medical care, etc.) by parents, relatives, or friends", and "Amounts deducted from Social Security or Railroad Retirement benefits for Medicare premiums".
• Page 32:
  • Line 20: "Other taxable income includes: awards, prizes, lottery, bingo, and other gambling winnings over $300".
  • Line 21: "Medicare premiums reported on your Social Security or Railroad Retirement statement should be deducted."
  • Line 24: "Enter the value over $300 in gifts of cash or merchandise received, or expenses paid on your behalf (rent, taxes, utilities, food, medical care, etc.) by parents, relatives, or friends."
  • Line 30: "Enter total adjustments from your U.S. Form 1040, Schedule 1." The list includes the "Self-employed health insurance deduction".
• Page 33, line 31: "Do not include any insurance premiums deducted on lines 21 or 30".

The statute doesn't say whether premiums already deducted in federal AGI can be deducted again. Treasury's line 31 instruction says they can't, and this PR follows the form.

Line mapping

I mapped each model input to its MI-1040CR line before changing anything:

model input	before	after	why
gambling_winnings	line 20, in full	line 20, over $300	MCL 206.510(1)(b)
miscellaneous_income	line 24 (the comment said "gifts")	line 20, in full	It is in federal gross income (gov.irs.gross_income.sources; 26 CFR 1.61-14), so it is taxable "other income", not a gift (IRC 102 excludes gifts). The $300 in (b) covers only awards, prizes, lottery, bingo and gambling, so it doesn't apply.
financial_assistance ("Cash financial assistance from friends or relatives outside the household")	not counted	line 24, over $300	These are gifts in cash from nongovernmental sources (MCL 206.510(1)(a)). Gifts are excluded from federal AGI, and total household resources include excluded income.
self_employed_health_insurance_ald (line 30) and health_insurance_premiums (line 31)	both subtracted in full	line 31 leaves out the head's and spouse's line 30 premiums	page 33
Medicare premiums withheld from benefits (line 21)	gross social_security on line 21, premiums on line 31	unchanged	The form nets them on line 21 and leaves them off line 31. The model keeps gross benefits on line 21 and the premiums on line 31, so they still come off once. A YAML case pins this.

The fix

• mi_household_resources groups gambling winnings and gifts, like the business and rental groups on lines 16 and 17, and adds each group's amount over its exclusion.
• New parameters gov.states.mi.tax.income.household_resources_gambling_exclusion and household_resources_gift_exclusion ($300 each, from 2021) cite the statute and the form lines.
• household_resources.yaml adds financial_assistance (line 24) and moves miscellaneous_income to line 20.
• Line 31 subtracts every member's premiums, less the self-employed health insurance deduction that line 30 subtracted for the head and spouse, read through the same tax_unit_non_dep_add call. If a reform removes that deduction from the federal list, line 31 keeps the premiums.

Invariants

These hold for every household:

1. Each premium dollar is subtracted once. Zeroing every member's premiums raises household resources by between 0 and the total premiums, and by exactly that total when household resources were positive.
2. Gambling winnings. Household resources are non-decreasing in winnings. They don't change while the claimant's and spouse's winnings total at most $300. Above that, when household resources are positive, they rise by the change in winnings over $300.
3. Gifts. The same as 2.
4. Line 30 or line 31 doesn't matter. A reform that drops the self-employed health insurance deduction from gov.irs.ald.deductions moves those premiums from line 30 to line 31 and leaves household resources unchanged. Before the fix, it changed them by $5,000 in the example.
5. Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's form reference (in test_mi_household_resources_members.py) still holds with the new line 20, 24 and 31 rules. is_self_employed is now drawn, so line 30's premium deduction is exercised.

Tests

• 14 new YAML cases in mi_household_resources.yaml:
  • the $50,000/$5,000 self-employed example;
  • premiums above the deduction's earnings limit;
  • a joint return;
  • Medicare premiums withheld from Social Security;
  • gambling winnings and gifts below, at and above $300, and for a couple;
  • both exclusions together;
  • miscellaneous_income counted in full.
• New policyengine_us/tests/test_mi_household_resources_exclusions.py: deterministic tests plus Hypothesis properties for invariants 1 to 3 and a reform test for invariant 4 (Hypothesis is a dev extra).
• test_mi_household_resources_members.py (from Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796): the form reference uses the new rules, and financial_assistance (line 24) is among the drawn inputs.

Before the fix, run on #9796's head:

• YAML: 10 of the 14 new cases fail (the 3 premium cases, 4 gambling, 2 gifts, 1 combined). The 4 that pass are the Medicare case, gifts below and at $300 (gifts were not counted at all), and miscellaneous_income.
• New property file: all 7 tests fail. Hypothesis shrinks the failures to Schedule C income of $101 with $100 of premiums, which loses $101 of household resources, and to $1 of gambling winnings, which adds $1.
• The updated members test fails twice (test_spouse_counts_like_the_head, test_properties).

After the fix:

• YAML: all 53 cases in mi_household_resources.yaml and all 643 in gov/states/mi pass (head 7fa23e51f6).
• Python: the new property file (8) and the members and losses files (10) pass. tests/code_health (138) passed on c57e5b6216; nothing it checks has changed since.
• test_exclusions_read_the_parameters raises both exclusions to 500 by reform. With a hard-coded 300 in the formula it fails, as I checked by mutation.
• ruff format --check and ruff check are clean.

CI cost (local, macOS). test_mi_household_resources_exclusions.py takes 42.5 s wall including import (25.8 s in pytest) at a 2.31 GB peak RSS. It is a top-level tests/ file, so it runs in the Rest suite's remaining group. It covers what YAML can't: for-all-households properties (premiums subtracted once, monotonicity and flatness up to $300) and a reform of gov.irs.ald.deductions.

• A separate stress run passed with 400 Hypothesis examples for each new property and 150 for the members properties.

Impact

Real microsimulations on the default dataset (populace_us_2024.h5@populace-us-2024-spm-20260909): #9796's head b10ac14936 against this branch at c57e5b6216. The later commits change only tests, comments and #9796's documentation, not a formula. The runs are record by record, with ~/reviews/pe-us-merge-backlog-2026-10-02/impact/lockrun.sh and compare.py, one run at a time under the shared heavy lock.

variable	2024 change	2026 change	records changed (2024 / 2026)
mi_household_resources	+$308.95M	+$316.92M	26 up (82,992 weighted) / 26 up (84,361)
mi_homestead_property_tax_credit	~$0	~$0	1 down, weight below 1
mi_home_heating_credit	−$2.72M	−$3.03M	7 down (23,651 / 24,041)
mi_standard_home_heating_credit	−$5.22M	−$5.05M	7 down
mi_alternate_home_heating_credit	0	0	0
mi_refundable_credits	−$2.72M	−$3.03M	8 down
mi_income_tax	+$2.72M	+$3.03M	8 up
household_net_income	−$2.72M	−$3.03M	8 down (23,651 / 24,042)

The largest change in household resources is $9,600 in 2024 and $10,199 in 2026: a premium that was subtracted twice.

Why the homestead credit doesn't move. Only 4 of the 26 changed records get a homestead credit at all (31,427 weighted tax units in 2024). Three of those are at the 2024 credit cap ($1,800; $1,900 from 2025), and none of the three crosses a phase-out step, so higher household resources don't reduce them. The fourth has weight 0.07: its credit falls from $357.59 to $253.91.

Follow-ups from review r1, not in this PR:

• Line 31 reads only the direct health_insurance_premiums input. In the default dataset most premiums sit in the split premium variables, so line 31 may miss Medicare and other premiums in the microsimulation.
• Line 31 doesn't net out the premium tax credit (2025 book, page 33).

Both are in a separate audit task. Two input gaps are documented in the parameters: awards and prizes have no input of their own (entered as miscellaneous_income, they count in full), and line 24's gifts in kind have no input (financial_assistance is cash only).

All of the change comes from the premium fix. The dataset has no gambling_winnings or financial_assistance column, so both are zero in the microsimulation. miscellaneous_income is counted the same way before and after.

axiom: TheAxiomFoundation/rulespec-us#1498 queued (MCL 206.510(1) has no rulespec-us module; the corpus row us-mi/statute/206.510 exists; the signed encoder run needs approval)

🤖 Generated with Claude Code

[DTrim99]

PR Review: Subtract Michigan premiums once and exclude the first $300 of gambling winnings and gifts (MI-1040CR lines 20, 24, 30 and 31)

PR #9840 (author: MaxGhenis, head 3e7a8bf0ae, 30 commits ahead of main). The PR's own work is three commits: d94846ecda, c57e5b6216 and 7fa23e51f6. The other 17 non-merge commits come from #9796, #9790, #9788 and #9801, which have all merged since. This review covers the PR's own diff against #9796's branch tip 546c6011d4, and checks how that diff sits on current main.

Summary

The PR makes two fixes to mi_household_resources, the income measure for the homestead property tax credit and the home heating credit:

1. Line 31 no longer subtracts premiums that line 30 already subtracted. Line 30 carries the U.S. Schedule 1 adjustments, which include the self-employed health insurance deduction. Line 31 now subtracts every member's premiums less that deduction. The netting applies only while the deduction is on the line-30 list.
2. The $300 exclusions in MCL 206.510(1)(a) and (b) are applied.
   • Gambling winnings count only above $300 on line 20.
   • financial_assistance (cash gifts from relatives or friends) now counts above $300 on line 24. Before, it was not counted at all.
   • Each exclusion applies once to the claimant's and spouse's combined total.
   • miscellaneous_income moves from line 24 to line 20 and still counts in full, so its value doesn't change.

The law is applied correctly, and I found no wrong result for a household in any year the PR covers.

• All 53 cases in mi_household_resources.yaml pass on the extracted head tree, as do the 6 deterministic tests in the new property file.
• I hand-checked 10 expectations and reproduced the headline examples in 2023 and 2025.
• CI is green (35/35).

These fixes are still needed on current main: a probe of main still subtracts the $50,000/$5,000 claimant's premiums twice (36,467.61) and counts $250 of winnings in full. #9796 and #9795 do not duplicate either fix.

There is one should-address item (S1). The branch now conflicts with main in six files. One conflict needs a judgment call: the merged formula must keep #9795's line-30 list, and the netting check must use that list. Separately, the merge silently re-adds #9796's changelog fragment, which main has already consumed. I applied the resolution I recommend to main in a scratch tree, and every test passes.

Critical (must fix)

None.

Should address

S1. The branch conflicts with main in six files. Line 30 must keep #9795's list, and the merge re-adds #9796's consumed changelog fragment.

• Where:
  • policyengine_us/variables/gov/states/mi/tax/income/mi_household_resources.py:125-133 (head) against main's :119-135;
  • policyengine_us/parameters/gov/states/mi/tax/income/household_resources.yaml and household_resources_all_members.yaml;
  • policyengine_us/tests/policy/baseline/gov/states/mi/tax/income/mi_household_resources.yaml;
  • policyengine_us/tests/test_mi_household_resources_members.py;
  • policyengine_us/variables/gov/irs/income/taxable_income/adjusted_gross_income/irs_gross_income/filer_loss_limited_net_capital_gains.py;
  • changelog.d/mi-household-resources-claimant-spouse.fixed.md.
• What's wrong:

  • Why git sees conflicts. GitHub reports the PR as CONFLICTING. Two PRs landed on main after this branch last merged Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's branch:

    • Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 was squash-merged as c6633c69a1 (2026-10-10) from its later head 08f20a99e9.
    • Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 (f8bad3fe18, 2026-10-08) replaced line 30's federal list with household_resources_adjustments.

    This branch still carries Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's earlier commits, up to 546c6011d4. Git therefore sees both sides editing the same hunks.

  • Mechanical conflicts. Four of the six conflicts are Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 against Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796: household_resources_all_members.yaml, the members test, filer_loss_limited_net_capital_gains.py, and the variable's reference block. Taking main's side is correct for each.

  • The semantic conflict is lines 30 and 31.

    • At head, deductions is every entry in gov.irs.ald.deductions except loss_ald, with include_dependents (:125-133).
    • On main, line 30 reads household_resources_adjustments intersected with ald.deductions, without include_dependents.
    • Taking the head side reverts Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795. It puts us_bonds_for_higher_ed, qualified_adoption_assistance_expense, specified_possession_income and puerto_rico_income back on line 30.
    • Taking main's side leaves deductions undefined in the new line-31 code, which raises a NameError.
    • One more variant is easy to fall into: checking the netting condition against ald.deductions instead of the line-30 list. Then a reform that drops the deduction from Michigan's list leaves the premiums subtracted on neither line. In a probe on the resolved tree ($50,000 Schedule C, $5,000 premiums, 2025), that variant gives 46,467.61. The correct check gives 41,467.61.

  • The changelog fragment.

    • main's release commit fe0c3a651c ("Update PolicyEngine US") moved Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's fragment into CHANGELOG.md (line 179) and deleted the file.
    • The merge base has no such file, and this branch adds it. git merge-tree therefore re-adds it with no conflict.
    • If it is not deleted, the next release will list Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's fix a second time.
• Source:
  • The 2025 MI-1040 book, line 31 (PDF p.33): "Do not include any insurance premiums deducted on lines 21 or 30". Only premiums actually deducted on line 30 may leave line 31.
  • On main, line 30 is the household_resources_adjustments list, and self_employed_health_insurance_ald is on it (household_resources_adjustments.yaml:15, "line 17").
  • The PR body anticipated this: "Whichever lands second should keep Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795's list as the deductions variable here." Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 has landed, so this PR is second.
• Fix: rebase onto main (or merge it in), then:
  1. Take main's side for household_resources_all_members.yaml, filer_loss_limited_net_capital_gains.py, the members test, and the variable's reference block. Keep main's 2025 MI-1040CR-7 reference and add the PR's #page=27 href.
  2. In the formula, keep main's line-30 comment and list, and add the PR's winnings and gifts grouping:

     deductions = [
         deduction
         for deduction in p.household_resources_adjustments
         if deduction in ald.deductions
     ]
     adjustments = tax_unit_non_dep_add(tax_unit, period, deductions)
     ...
     premiums = add(tax_unit, period, ["health_insurance_premiums"])
     premiums_on_line_30 = (
         tax_unit("self_employed_health_insurance_ald", period)
         if "self_employed_health_insurance_ald" in deductions
         else 0
     )
     line_31 = max_(premiums - premiums_on_line_30, 0)
     return max_(0, total - adjustments - line_31)

  3. The rest of the PR's own diff applies to main unchanged with git apply: the two new parameters, the household_resources.yaml hunks, the 14 YAML cases (at a +259-line offset), the members-test hunks and the new property file.
  4. Delete changelog.d/mi-household-resources-claimant-spouse.fixed.md, so that the PR adds only its own fragment.
  5. Refresh the PR body:
     • Drop "Merge after Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 ... Until those PRs merge, the diff here includes their commits" and the Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 coordination paragraph, since both are now done.
     • Re-run the impact table on the rebased branch. It was measured against Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's head b10ac14936, before Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 changed line 30.
• Verified: I applied this resolution to main (2e9eea7e6e) in a scratch tree and ran it with policyengine-core 3.33.2:
  • 62/62 cases in mi_household_resources.yaml pass (48 from main plus the 14 new ones).
  • 9/9 deterministic tests in the exclusions and members files pass.
  • A reform that removes the deduction from household_resources_adjustments leaves household resources at 41,467.61.

Suggestions

1. Derandomize the two Hypothesis properties.

   • Where: policyengine_us/tests/test_mi_household_resources_exclusions.py:399 and :404.
   • Why: without derandomize=True, each CI run draws new examples, so a float-tolerance counterexample could fail an unrelated PR.
     • Most of the repo's invariant files pin the seed, for example tests/core/test_dependent_above_the_line_deductions_invariants.py:395 and tests/test_form_4952_election_worksheet.py:258.
     • The sibling Michigan files from Net Michigan business and rental losses in household resources (MI-1040CR lines 16, 17 and 30) #9790 and Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 don't pin it either, so this is optional.
   • Fix: add derandomize=True to these settings. Optionally, do the same in the losses and members files.

2. A supplied self-employed health insurance deduction can net out another member's premiums.

   • Where: mi_household_resources.py:148-155.
   • What happens: line 31 nets the tax unit's deduction against every member's premiums.
     • When the deduction is computed, it can't exceed the self-employed filer's own premiums. self_employed_health_insurance_premiums adds health_insurance_premiums (and is defined for is_self_employed), and the deduction is min_(earnings, premiums).
     • When it is supplied as an input, it can absorb a spouse's unrelated premiums. Supplying it is a supported path: the existing case "A supplied self-employed health insurance deduction comes off line 30" uses it.
   • Head probe (2025):
     • Inputs: head wages $40,000 and a supplied deduction of $5,000; spouse wages $30,000 and premiums of $2,000.
     • Result: 65,000. The spouse's $2,000 is netted away.
     • If the $5,000 was the head's own plan, which was not entered as health_insurance_premiums, the form gives 63,000: line 31 keeps the spouse's $2,000.
   • Fix (either):
     • Cap the netting at the filers' self-employed premiums: min_(tax_unit("self_employed_health_insurance_ald", period), tax_unit_non_dep_add(tax_unit, period, ["self_employed_health_insurance_premiums"])).
       • This gives 63,000 here and keeps the existing supplied case at 9,000.
       • It is identical whenever the deduction is computed.
       • Its own failure mode: someone who supplies the deduction and enters those premiums but leaves is_self_employed false would have them subtracted twice.
     • Or keep the code and state the assumption in the comment: line 31 treats the line-30 premiums as part of health_insurance_premiums.

3. The comments state PolicyEngine's federal limit, not the form's.

   • Where:

     • The comment in mi_household_resources.yaml:870: "limited to the 3,000 net earnings = 3,000".
     • The assertion at test_mi_household_resources_exclusions.py:269, [5_000, 3_000].

   • What's wrong: both pin the federal formula in self_employed_health_insurance_ald_person.py, which is the minimum of premiums and net self-employment income. Federal law limits the deduction to earned income under section 401(c), which is net of the deductible half of self-employment tax (26 U.S.C. 162(l)(2)(A)).

     • Form 7206 (2025), PDF p.1, line 7 says "Multiply Schedule 1 (Form 1040), line 15, deductible part of self-employment tax, by the percentage on line 6", and line 8 says "Subtract line 7 from line 4".
     • The limit is therefore 3,000 − 211.94 = 2,788.06.

   • The expected value is still right. Following the form:

     • line 30 = 211.94 + 2,788.06 = 3,000.00;
     • line 31 = 5,000 − 2,788.06 = 2,211.94;
     • line 33 = 43,000 − 3,000.00 − 2,211.94 = 37,788.06, the same as the PR.

     The PR's own design guarantees this: every premium dollar comes off once.

   • Fix: reword the comment to say these are PolicyEngine's federal amounts (a pre-existing simplification), and that line 33 is the same under Form 7206. The same note applies to row 2 of rulespec-us#1498 (see the axiom section).

4. Read the tax-unit deduction directly.

   • Where: mi_household_resources.py:148-153.
   • What happens: the code calls tax_unit_non_dep_add on one TaxUnit-level variable. For a tax-unit variable, that function returns tax_unit(variable, period) (tools/general.py:38-39).
   • Fix: tax_unit("self_employed_health_insurance_ald", period) says the same thing more plainly.

5. Optional earlier-year case.

   • Why: every new case uses 2025. The parameters start 2021-01-01 and cite the 2022 to 2024 books, which have the same line 20, 24 and 31 text. The section was last amended effective January 1, 2012.
   • Head probes: they give identical results in 2023 and 2025 (winnings $400 → 20,100; the $50,000/$5,000 case → 41,467.61).
   • Fix: add one 2022 or 2023 case to pin the earlier years.

6. Optional reference consolidation at rebase.

   • What: each new parameter cites the 2025 book twice (#page=32 and #page=27). main's CLAUDE.md (Document the reference page-citation convention #10009) now says "URLs that differ only by #page= are one source".
   • Fix: one entry with href #page=27 and a quoted title ending #page 27,32 would follow it. The variable's five 2025-book hrefs predate this PR, and the same applies to them.

Verified correct

• The $300 exclusions match the statute and the form.
  • MCL 206.510(1): "Income does not include any of the following: (a) The first $300.00 of gifts in cash or kind from nongovernmental sources. (b) The first $300.00 received from awards, prizes, lottery, bingo, or other gambling winnings." The section was last amended by 2011 PA 38 and 172, effective January 1, 2012.
  • Line 20: "awards, prizes, lottery, bingo, and other gambling winnings over $300".
  • Line 24: "Enter the value over $300 in gifts of cash or merchandise received, or expenses paid on your behalf ... by parents, relatives, or friends".
  • Both line texts are identical on PDF p.32 of the 2022, 2023, 2024 and 2025 books.
  • The 2025 list of what total household resources "do NOT include" (PDF p.27) has "The first $300 from gambling, bingo, lottery, awards, or prizes" and "The first $300 in gifts of cash or merchandise".
  • The home heating credit uses the same "total household resources" (MCL 206.527a).
• Each exclusion applies once per claim, and the boundaries are exact. Lines 20 and 24 are each one amount for the claimant and spouse.
  • Head probes: winnings of 300 → 0 counted; 301 → 1.
  • Winnings and gifts of 250 each → 0 counted. Of 400 each → 200 counted, since each has its own $300.
  • A couple's winnings of 250 + 250 → 200 counted.
  • A dependent's winnings or gifts neither count nor use up the claimant's $300. This is consistent with Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796's "household = claimant and spouse".
• Premiums come off once.
  • Line 31's "Do not include any insurance premiums deducted on lines 21 or 30" appears on PDF p.33 of all four books.
  • Line 30's list includes the "Self-employed health insurance deduction" (2025 p.32; 2022 to 2024 p.33).
  • Head probes:
    • The self-employed claimant gets 41,467.61 (main: 36,467.61).
    • The joint return gets 69,467.61.
    • A claimant flagged self-employed with a $5,000 Schedule C loss has no deduction, so all $4,000 of premiums stay on line 31 (26,000).
    • An unflagged $50,000 Schedule C also gives 41,467.61: the premiums simply move to line 31.
  • The else 0 branch is tested by the reform that drops the federal deduction. The line-31 floor is tested by the existing supplied-deduction case: 9,000, which would be 11,000 without the floor.
• Medicare premiums. The form nets them on line 21 ("Medicare premiums reported on your Social Security or Railroad Retirement statement should be deducted") and leaves them off line 31. MCL 206.510(1)(h) excludes them from income. The model keeps gross benefits on line 21 and the premiums on line 31. The result is the same: 20,000 − 2,220 = 17,780.
• Line mapping.
  • miscellaneous_income is in gov.irs.gross_income.sources (sources.yaml:35), so it is taxable "other income" on line 20, in full. It is not a gift, and its value doesn't change.
  • financial_assistance ("Cash financial assistance from friends or relatives outside the household") is a cash gift from a nongovernmental source, which line 24 counts.
• Tests and arithmetic. 53/53 YAML cases pass at head, and 6/6 deterministic Python tests pass locally. Hypothesis is not installed locally; CI's Rest suite ran the two properties. Hand-checked:
  • $50,000 Schedule C: 50,000 × 0.9235 × 0.153 / 2 = 3,532.39, so 50,000 − 3,532.39 − 5,000 = 41,467.61.
  • Wages $40,000 plus $3,000 Schedule C: 3,000 × 0.9235 × 0.153 / 2 = 211.94, so 43,000 − 211.94 − 3,000 − 2,000 = 37,788.06.
  • Joint return: 80,000 − 8,532.39 − (7,000 − 5,000) = 69,467.61.
  • Medicare case: 17,780.
  • Winnings 1,000 → 20,700. A couple's 200 + 200 → 20,100.
  • Gifts 250 + 250 → 20,200.
  • Winnings 500 and gifts 500 → 20,400.
  • The parameter-reform test: 400/600 at $300 → 20,400; at $500 → 20,100.
  • The dependent test: 20,000 and 20,400.
• Parameters. Descriptions follow the house template ("Michigan excludes this amount of ..."), with currency-USD, year, labels and statute titles. Every #page anchor shows the cited text. Test periods are plain (2025).
• Callers. Nine consumer variables read mi_household_resources (five homestead, four home heating), and its interface is unchanged. The two new parameters are read only here. Adding financial_assistance to the Michigan list changes no other program.
• Changelog. changelog.d/mi-household-resources-premiums-gambling.fixed.md is top-level, matches the branch name, and fixed is the right type.

Out of scope (pre-existing, not counted)

• Federal gross income omits gambling winnings (Federal gross income omits gambling winnings but still deducts wagering losses; OBBBA 90% wagering loss limit (2026+) not modeled #9637, open). In a head probe, $400 of winnings gives federal AGI of 0. Line 20 reads gambling_winnings directly, so Michigan is unaffected.
• Line 31's other rules. Line 31 says "You must reduce an insurance premium by the federal premium tax credit" and excludes "amounts paid by an employer with pre-tax payroll contributions" (2025 p.33). It also reads only the direct health_insurance_premiums input. The PR body lists these as follow-ups.
• Dependent filers' support on line 24. Line 24 also "includes the amount of financial support you received if you are claimed as a dependent on someone else's return". There is no input for this.

Axiom line assessment

axiom: TheAxiomFoundation/rulespec-us#1498 queued is valid.

• The issue is OPEN and labelled pe-parity. It was created 2026-10-06 and has no comments.
• It records the blocker: the signed, billed encoder run needs approval.
• It is dispatch-ready under the Add AMT parameters for 2020-2023 #1467 standard:
  • module path us-mi/statutes/206.510.yaml;
  • corpus row us-mi/statute/206.510, with the lock, blob and sha256;
  • the verbatim law in (1), (a), (b), (h) and (i);
  • four required outputs (gift_exclusion, gambling_exclusion, health_insurance_premium_deduction, income);
  • 14 companion tests taken from the statute and the 2025 MI-1040CR lines, plus properties.
• All 14 rows match the PR's current behavior.
• Two points to update when convenient:
  • Row 2's intermediate values follow PolicyEngine's federal deduction: premiums in AGI 3,000, AGI 39,788.06 and a premium deduction of 2,000. Form 7206 gives 2,788.06, 40,000.00 and 2,211.94. The income of 37,788.06 is the same either way (Suggestion 3).
  • The "Related" note about stacking on Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796, Net Michigan business and rental losses in household resources (MI-1040CR lines 16, 17 and 30) #9790 and Net capital gain distributions inside the capital loss limit in AGI #9788 is stale, since all three have merged.

CI status

All 35 checks pass at 3e7a8bf0ae (gh pr checks 9840, run of 2026-10-06), including Lint, Check changelog fragment, Full Suite - Rest (Python + variables), Microsimulation and Household API Partners. That run predates #9795 and #9796 landing on main. CI has not run on a merge with current main.

Branch status

• Position: 30 commits ahead of main and 198 behind (main at 2e9eea7e6e; merge base d3978dfc31).
• GitHub: reports CONFLICTING (merge state DIRTY).
• git merge-tree --write-tree PolicyEngine/main 3e7a8bf0ae exits 1 with six conflicts:
  • household_resources.yaml (content);
  • household_resources_all_members.yaml (add/add);
  • mi_household_resources.yaml tests (content);
  • test_mi_household_resources_members.py (add/add);
  • filer_loss_limited_net_capital_gains.py (content);
  • mi_household_resources.py (content).
• Silent re-add: the merge also re-adds changelog.d/mi-household-resources-claimant-spouse.fixed.md with no conflict.
• Resolution: see S1. With the resolution described there, the merged tree passes all affected Michigan tests.

Verdict: REQUEST_CHANGES

[DTrim99]

The policy is right. The $300 exclusions match MCL 206.510(1)(a)/(b) (once per claim, strictly over $300, one each for gifts and winnings, dependents left out), and premiums now come off once per line 31's instruction. The tests pass and rulespec-us#1498 is dispatch-ready. One should item (see the comment above): the branch conflicts with main in 6 files. Take main's side for the four #9796 conflicts. In the line-30/31 formula, build deductions from main's household_resources_adjustments ∩ ald.deductions (#9795) and test the self-employed health insurance deduction against that list; the head side reverts #9795 and main's side raises a NameError. Also delete the re-added #9796 changelog fragment.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (909176a) to head (b95613e).
⚠️ Report is 493 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9840   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         1    -3     
  Lines           76        44   -32     
  Branches         2         6    +4     
=========================================
- Hits            76        44   -32     

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.