Net Michigan business and rental losses in household resources (MI-1040CR lines 16, 17 and 30)

[MaxGhenis]

Summary

mi_household_resources floored each business and rental source at zero on its own, then subtracted every federal above-the-line deduction, including the business part of loss_ald. That is the losses it had just floored away, so a business or rental loss came off household resources after all. Example: wages 50,000 and a rental loss of 30,000. The model gave 20,000; MI-1040CR line 33 is 50,000. #9788 stopped the capital part of the same double count. This PR removes the rest and follows the form's netting:

• Line 16 nets Schedule C, Form 4797 Part II, Schedule E Parts II and III, and Schedule F, then floors the total at zero.
• Line 17 nets Schedule E Parts I, IV and V (rents, royalties, REMIC income, which has no input, and farm rental), then floors the total at zero.
• Lines 16 and 17 count only the claimant's and spouse's items (MCL 206.508(3): "'Household' means a claimant and spouse"). A dependent's business and rental income and losses are on the dependent's own return, as in loss_ald and irs_gross_income.
• Line 30 is the U.S. Schedule 1 adjustments: the federal above-the-line deduction list without loss_ald.

Removing loss_ald alone would have been wrong in the other direction. With per-source floors, a Schedule C loss could not offset S corporation income; only loss_ald was bringing it back. For example, Schedule C −10,000 and S corporation income 30,000 give line 16 = 20,000. A loss_ald-only change gives 30,000 (80,000 in total with wages of 50,000, against the form's 70,000).

#9788 (this PR's prerequisite) is merged. This branch merges main at df70a46db2, so the diff here is this PR's own change only (5 files).

Dependents' self-employment deductions (until #9801). Line 30 still subtracts self_employment_tax_ald and self_employed_pension_contribution_ald, which sum every member on main. So a dependent's positive Schedule C or F income now lowers the claimant's household resources by the deductible half of the dependent's self-employment tax and any SEP deduction. Example (2025): claimant wages 50,000 and a dependent's Schedule C 20,000 give 48,587.05; the form gives 50,000. Before this PR the same household got 68,587.05, so the absolute error shrinks, but its sign flips. #9801 makes those two deductions sum the head and spouse only. #9796, which builds on this PR and merges #9801, pins the form's 50,000 with tests (a dependent's own Schedule C income, positive or negative, leaves household resources unchanged).

The law and the forms

MCL 206.508(4): "'Total household resources' means all income received by all persons of a household in a tax year while members of a household, ... and increased by the following deductions from federal gross income: (a) Any net business loss after netting all business income and loss. (b) Any net rental or royalty loss. (c) Any carryback or carryforward of a net operating loss as defined in section 172(b)(2) of the internal revenue code."

2025 MI-1040 book (page 26), "Total Household Resources": "They are AGI, excluding net business and farm losses, net rent and royalty losses, and any carryover of a net operating loss, plus all income exempt or excluded from AGI."

MI-1040CR line 16 (page 31): "Enter amounts to the extent included in AGI from: • U.S. Schedule C (Profit or Loss from Business). • Part II (Ordinary Gains and Losses) of the U.S. Form 4797. • Part II (Income or Loss from Partnerships and S Corporations) and Part III (Income or Loss from Estates and Trusts) of the U.S. Schedule E. • U.S. Schedule F (Profit or Loss from Farming). • Include income items reported as a distributive share. If the total is negative enter '0.'"

Line 17: "Enter amounts to the extent included in AGI from: • Part I (Income or Loss from Rental Real Estate and Royalties) of the U.S. Schedule E. • Part IV (Income or Loss from Real Estate Mortgage Investment Conduits (REMIC)) of the U.S. Schedule E (rents, royalties). • Part V (Net farm rental income or (loss) from Form 4835) of the U.S. Schedule E. If the total is negative enter '0.'"

Line 30 (page 32): "Enter total adjustments from your U.S. Form 1040, Schedule 1. Describe adjustments to income. These adjustments reduce total household resources and include some of the following: • Payments to IRAs, SEP, SIMPLE, or qualified plans • Student loan interest deduction • ... • Deduction for self-employment tax • Self-employed health insurance deduction • Penalty on early withdrawal of savings • Alimony paid if deductible on your U.S. Form 1040 • Jury duty pay you gave to your employer • Archer Medical Savings Account (MSA) deduction • Health Savings Account (HSA) deduction • Any other adjustments to gross income included on your 2025 U.S. Form 1040, Schedule 1." Business, rental and capital losses are Schedule 1 Part I income items (and Schedule C, D, E and F), not Part II adjustments.

The 2024 instructions have the same line 16, 17 and 30 text and the same page 26 definition.

Changes

• mi_household_resources:
  • sums the line 16 sources and floors the total at zero, and does the same for the line 17 sources;
  • adds every other source as before;
  • subtracts the federal above-the-line deductions other than loss_ald (line 30), computed from the list itself, so a reform that edits gov.irs.ald.deductions still lands right;
  • subtracts health insurance premiums (line 31), unchanged.
• Lines 16 and 17 sum each source over the head and spouse with tax_unit_non_dep_add. That helper sums a person-level source over non-dependents and takes a tax-unit-level one (other_net_gain) as is.
• gov.states.mi.tax.income.household_resources adds estate_income (Schedule E Part III) and other_net_gain (Form 4797 Part II) to line 16. Before, their losses came off through loss_ald, but their income never counted.
• The reference URL for the 2024 MI-1040CR-7 book goes back to main's archived copy. Net capital gain distributions inside the capital loss limit in AGI #9788's michigan.gov link now returns 404. The 2025 MI-1040CR-7 book's URL does too, and michigan.gov's 2025 forms page now links "MI-1040CR-7" to its home heating credit deadline page. The statute's Legislature page and the 2025 book pages 26 and 31 are added.

Tests

mi_household_resources.yaml, with MI-1040CR line comments and expected values worked out from the form:

• Changed expectations:

  • "Negative business income floored at 0" (and its SSTB twin): 50,000 becomes 100,000.
  • "Negative rental income floored at 0": 20,000 becomes 50,000. Its comment said loss_ald of 30,000 is subtracted.
  • "Farm operations income counted as household resources": the test zeroed above_the_line_deductions by input, which line 30 no longer reads. It now expects 50,000 less the deduction for self-employment tax (3,532.39), which line 30 lists: 46,467.61.
  • "Cannot be negative", "Reduced to 0" and "Partially reduced": these input above_the_line_deductions directly. They now use a real Schedule 1 adjustment (the early withdrawal penalty) and keep their expected values.

• 13 new cases:

  • netting within line 16: Schedule C, S corporation, estate and Form 4797 losses against income;
  • a net line 16 loss floored at zero;
  • netting within line 17: rental against farm rental;
  • a business loss that does not reduce rental income, and the reverse;
  • estate income and Form 4797 gains counted;
  • Schedule 1 adjustments with a business loss;
  • a dependent's Schedule C loss, rental loss and S corporation income, none of which reach the claimant's lines 16 and 17.

  Nine of the 24 cases fail on Net capital gain distributions inside the capital loss limit in AGI #9788's head (638c794a18) and pass here. Five pass on both: those netting cases match the form today only because loss_ald restores the offset. They fail under a loss_ald-only fix (checked with a reform that keeps per-source floors: 80,000 vs 70,000, 75,000 vs 65,000, 70,000 vs 65,000, 70,000 vs 62,000, and 69,076.12 vs 65,076.11).

tests/test_mi_household_resources_losses.py: a 2,187-household grid (every sign combination of the seven line 16 and 17 sources) and Hypothesis tests (one- and two-person units, amounts up to ±$500,000, capital gains, early withdrawal penalties, premiums). The grid fails on #9788's head. Invariants, for every input:

1. mi_household_resources = max(0, wages + max(0, line 16 sources) + max(0, line 17 sources) + max(−3,000, capital gains) − Schedule 1 adjustments − premiums). This is a differential test against a reference written from the form.
2. It does not depend on loss_ald: a reform that removes loss_ald from gov.irs.ald.deductions changes AGI but not household resources.
3. Losses stay in their own line. When every line 16 source is zero or negative, the result equals the result with them set to zero; the same holds for line 17.
4. 0 ≤ household resources ≤ the sum of positive income amounts.
5. More S corporation, estate, Form 4797, rental or farm rental income raises household resources by between 0 and the increase. Intended exception: Schedule C and F income can lower household resources when line 16 is floored. The extra income still raises the self-employment tax deduction on line 30. The form gives that result.
6. A dependent's business and rental losses, and S corporation, estate and rental income, leave household resources unchanged (test_dependents_business_and_rental_items). Positive Schedule C and F income for a dependent is left out until Keep dependents' above-the-line deductions off the filer's AGI #9801; see above.

Property 2 is checked on the grid, which builds the reform once.

Tolerances are one cent or eight float32 spacings at the household's total absolute amount: $0.25 at $500,000, and up to about $4 for the largest two-person households drawn. Hypothesis's first counterexamples were float32 rounding in the test's own arithmetic, not model errors. In one, subtracting a large loss_ald in the check lost 0.014. In the other, a change of 17,152 read as 17,152.031 near $530,000, where the float32 spacing is 0.0625. The adjustments check now sums line 30 directly, as the formula does.

On head 0a2ca21b4a, all 614 Michigan YAML tests and the 5 tests in test_mi_household_resources_losses.py pass (about 64 s, 2.0 GB peak). They pass again on head 14bfcbc7f8, which merges main after #9788 landed. That merge changes no Michigan code, so the microsimulation below (branch 0a2ca21b4a) still describes this PR.

Microsimulation (default dataset, 2026, policyengine-core 3.32.12)

Base: main plus #9788, which is #9788's head 638c794a18. Branch: 0a2ca21b4a. The two ran one at a time under the shared heavy-job lock and a 24 GB watchdog (peak 14.8 GB) on populace_us_2024.h5@populace-us-2024-spm-20260909: 79,729 tax units, 1,974 in Michigan.

	base	branch	change
mi_household_resources (weighted, Michigan)	$518.21B	$521.54B	+$3.33B (123 units: 115 up, 8 down)
mi_homestead_property_tax_credit	$830.6M	$801.9M	−$28.7M (8 units)
mi_home_heating_credit	$184.6M	$172.4M	−$12.2M (13 units)
mi_eitc	$617.9M	$617.9M	0
mi_refundable_credits	$1,633.1M	$1,592.2M	−$40.9M (19 units)
mi_income_tax / state_income_tax			+$40.9M (19 units)
adjusted_gross_income, loss_ald, above_the_line_deductions, income_tax			0 (bitwise identical)

• Household resources rise for 115 Michigan units, with a median rise of $7,098. Among Michigan units, 43 have a negative line 16 total, 66 a negative line 17 total, and 114 a positive loss_ald.
• They fall for 8 units, by $31,843 weighted in all.
  • Seven are the dependent change: a dependent's positive Schedule C or rental income no longer counts. These falls run from $0.42 to $1,086,852. The three largest (rental income of $1,086,852 and Schedule C income of $32,889 and $2,751) are in units with zero weight.
  • The eighth falls by $0.001, from float32 summation order. An earlier run without the dependent change (base b047c783ab, branch 1a54e8999f) had one fall of the same kind (−$11.93 weighted) and the same 115 rises.
• No tax unit outside Michigan changes.
• Higher household resources cut the two household-resources-tested credits, which are refundable, so Michigan income tax net of credits rises by the same $40.9M.
• The dataset has no other_net_gain (Form 4797) amounts.

Scripts, logs and the full per-variable table: ~/reviews/pe-us-mi-household-resources-2026-10-05/microsim/ (diff_2026.txt; the earlier runs are in r1/ and r2-stale/).

Not in this PR (follow-ups)

• Line 30 still subtracts federal exclusions that are not Schedule 1 adjustments (Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795): us_bonds_for_higher_ed (Form 8815; its interest already counts on line 15, "including nontaxable interest"), qualified_adoption_assistance_expense, specified_possession_income and puerto_rico_income.
• Lines other than 16 and 17 still sum every tax unit member, dependents included. Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 limits them to the claimant and spouse (MCL 206.508(3)), except the amounts received for others on lines 21, 22 and 27.
• Schedule E Part IV (REMIC) income has no input.

axiom: TheAxiomFoundation/rulespec-us#1491 queued (MCL 206.508(4) has no rulespec-us module; the corpus row us-mi/statute/206.508 exists; the signed encoder run needs approval)

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (909176a) to head (14bfcbc).
⚠️ Report is 228 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9790   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         1    -3     
  Lines           76        26   -50     
  Branches         2         3    +1     
=========================================
- Hits            76        26   -50     

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 14bfcbc7f8.

• Since the r2 REQUEST_CHANGES at b6c5f613ce:
  • Merged Net capital gain distributions inside the capital loss limit in AGI #9788's head, then main after Net capital gain distributions inside the capital loss limit in AGI #9788 landed as df70a46db2. The conflicts were in the two Michigan files, where main's version equals Net capital gain distributions inside the capital loss limit in AGI #9788's head; I resolved them to this branch's side.
  • git diff upstream/main is now this PR's own change only (5 files).
  • r2 finding 4: tax_unit_non_dep_add for lines 16 and 17.
  • r2 finding 2: changelog.
  • r2 finding 3: body rewritten, with a fresh 2026 microsim (base 638c794a18 = main + Net capital gain distributions inside the capital loss limit in AGI #9788, branch 0a2ca21b4a): household resources +$3.33B, homestead −$28.7M, heating −$12.2M, MI income tax +$40.9M; federal unchanged.
  • r2 finding 1 (a dependent's positive Schedule C/F income still lowers resources through the self-employment deductions until Keep dependents' above-the-line deductions off the filer's AGI #9801) is disclosed in the body. Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 pins the form value.
• Review: r3 (Opus 5.5) is APPROVE at 14bfcbc7f8 (~/reviews/pe-us-mi-household-resources-2026-10-05/review-r3.md). Its one LOW was a body sentence about the earlier microsim; it's corrected.
  • Nit left in code: test_mi_household_resources_losses.py:314-315 says a dependent's self-employment tax deduction "is not on this return either". That becomes true once Keep dependents' above-the-line deductions off the filer's AGI #9801 lands; Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796, which carries Keep dependents' above-the-line deductions off the filer's AGI #9801, is the place to reword it.
• Tests on 14bfcbc7f8: 614 MI YAML and 5 losses tests pass.
• Pending: CI on 14bfcbc7f8 (MERGEABLE, no failures so far).
• Order: no open dependency (Net capital gain distributions inside the capital loss limit in AGI #9788 is merged). Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 and Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 build on this branch. The reviewer asks that Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 land with or soon after Keep dependents' above-the-line deductions off the filer's AGI #9801, since it holds the test pinning the dependent self-employment case.

[MaxGhenis]

Merged by the US + core hub on its merge gates. Audit:

• Head: 14bfcbc7f8, handed off by the Michigan chain driver (51965fe9).
• Independent review: r3 (Opus 5.5) APPROVE at this head (~/reviews/pe-us-mi-household-resources-2026-10-05/review-r3.md). r2's findings 2–4 are fixed. Finding 1 (dependents' Schedule C/F through the self-employment deductions) is disclosed and depends on Keep dependents' above-the-line deductions off the filer's AGI #9801.
• CI: every check passed at this head.
• Impact: 2026 microsimulation, base main + Net capital gain distributions inside the capital loss limit in AGI #9788:
  • household resources +$3.33B;
  • homestead credit −$28.7M;
  • heating credit −$12.2M;
  • MI income tax +$40.9M;
  • federal unchanged.
• Order: first after Net capital gain distributions inside the capital loss limit in AGI #9788. Keep federally excluded income in Michigan household resources (MI-1040CR line 30) #9795 and Count only the claimant and spouse in Michigan household resources (MI-1040CR lines 14-27, 30 and 31) #9796 build on it.