Skip to content

Exclude dependents' deducted health premiums from Schedule A - #10831

Merged
MaxGhenis merged 13 commits into
mainfrom
fix-dependent-deducted-health-premiums
Nov 16, 2026
Merged

MaxGhenis merged 13 commits into
mainfrom
fix-dependent-deducted-health-premiums

Conversation

@MaxGhenis

@MaxGhenis MaxGhenis commented Nov 12, 2026 •

Copy link
Copy Markdown
Contributor

Exclude medical premiums a dependent deducts through self-employed health insurance from the claiming filers' Schedule A medical expenses.

Parents may count only the medical expenses they paid under 26 U.S.C. 213(a) and 26 CFR 1.213-1(a)(1). A dependent who deducts their own premiums paid them, so those premiums leave the parents' Schedule A. Taxpayers also exclude their own deducted premiums from Schedule A under section 162(l)(3).

For example, a dependent deducts $3,000 of medical premiums and $480 of eligible long-term care premiums against $10,000 of Schedule C or partnership earnings. With $50,000 of other medical expenses, the parents' medical subtotal falls from $53,000 to $50,000. The dependent's $3,480 deduction stays unchanged. The IRS sets the 2025 long-term care limit in Rev. Proc. 2024-40, section 2.28.

Invariants and tests

For computed subtotals with derived exclusion helpers and deductions within their corresponding paid premium pools:

  • Schedule A premium use plus the filers' deduction and every dependent's deduction equals medical premiums plus eligible long-term care premiums within $0.01, and never exceeds premiums paid within that tolerance.
  • Each dependent's exclusion consumes only their own premiums. Non-premium expenses remain intact, and filers' premiums leave the subtotal once.
  • Filers without a dependent medical exclusion retain the existing computed subtotal. Supplied deductions and subtotals retain their existing precedence.

For supplied amounts:

  • A subtotal supplied at its computed value, with the computed applied exclusion and long-term care term, gives the computed Montana and New Jersey deductions.
  • For a computed subtotal, the North Dakota and New Mexico formulas add the filers' exclusion and the dependents' exclusion back, whether the federal formula derives the filers' exclusion or a caller supplies it.

We keep undeducted dependent-paid premiums in the parents' subtotal, as main does. #10270 addresses payer attribution more broadly.

Hypothesis tests check conservation, earnings caps, premium splits, separate tax units, and state paid-cost bases. test_mt_medical_premium_deduction_invariants.py checks Montana's supplied-amount equality on a seeded population for 2021, 2022 and 2023. One New Jersey YAML case checks the same equality, and no property test does. YAML cases test zero and premium-pool boundaries, excess supplied deductions, sibling isolation and a self-employed filer alongside a dependent. They also test supplied subtotals and exclusions in Montana and New Jersey, and a supplied exclusion in North Dakota and New Mexico. The Montana cases use 2022 and 2023, and the seeded test covers 2021 to 2023, so neither runs Montana's 2017 to 2020 years.

Pull request run 43914459007 at 44fb51ba82 passed all 37 jobs on GitHub's merge with main 8e1837790b.

State effects

The New Jersey formula removes a dependent's deducted premiums once, from a computed or a supplied subtotal. For 2017 to 2023, the Montana formula keeps other medical expenses in line 1 and keeps supplied exclusions.

With a supplied subtotal, a supplied applied exclusion holds every premium that subtotal leaves out. The Montana and New Jersey formulas read the part of it above the filers' own computed exclusion as the dependents' deducted medical premiums, up to the dependents' computed exclusion. The Montana formula takes the rest from the filers' medical premiums first. The New Jersey formula subtracts only the dependents' deducted premiums that part does not cover. An exclusion that covers the filers only gives main's result in both states.

For example, a New Jersey parent with $30,000 of wages pays $10,000 of other medical costs, and a dependent pays and deducts $3,000 of medical premiums. The formulas compute a $10,000 subtotal and a $3,000 exclusion. Those two amounts supplied as inputs give the computed deduction of $10,000 − 2% × $30,000 = $9,400.

With a computed subtotal, a supplied applied exclusion covers the filers' premiums only. The Montana, North Dakota and New Mexico formulas each add the dependents' computed exclusion to it, as the derived path does. North Dakota takes out of renter refund income the medical expenses paid "by the applicant or the applicant's dependent" (NDCC 57-02-08.1(5)(c)). New Mexico's Taxation and Revenue Department lets filers aged 65 or older count expenses for the care of "any combination of you, your spouse, or dependents" toward the medical care credit (PIT-RC line 23). Its exemption instructions cover the filer's dependents the same way (PIT-ADJ line 18).

For example, a North Dakota parent with $40,000 of wages pays $2,000 of other medical costs, and a dependent pays $3,000 of premiums and deducts $1,000 of them. The federal subtotal leaves out the deducted $1,000 and comes to $4,000. With a supplied zero filer exclusion, we still add the $1,000 back, and renter refund income stays at $40,000 − $5,000 = $35,000. In New Mexico, with the same supplied zero, a dependent with $28,001 of premiums and a $1,000 deduction leaves a 65-year-old filer a $27,001 subtotal. We add the $1,000 back to reach $28,001, so the filer keeps the $2,800 credit and the $3,000 exemption.

Affected filers may deduct less for medical expenses in states that use the federal subtotal or deduction. When calculating Missouri's subtraction, we use the corrected subtotal in Form 5695, lines 10–12, so affected families' premium subtraction can shift.

Methodology choices

  • Remove only premiums the dependent deducts. Preserve the existing inclusion of undeducted dependent-paid premiums pending Count only filer-paid premiums in Schedule A medical expenses #10270.
  • Reuse the existing person deductions and medical/long-term care split. Bound each dependent's medical exclusion to their own paid premiums and deduction after its long-term care part.
  • Preserve the existing supplied-input convention. With a computed subtotal, a supplied exclusion covers filers only, and the Montana, North Dakota and New Mexico formulas add the dependents' computed exclusion to it. The computed exclusion now holds the dependents' part. A caller who supplies that full amount beside a computed subtotal counts the part twice in North Dakota and New Mexico, and in Montana up to the filers' undeducted medical premiums. With a supplied subtotal, the North Dakota and New Mexico formulas restore the complete supplied exclusion.
  • With a supplied subtotal, the Montana and New Jersey formulas split a supplied exclusion between the filers and the dependents by the computed amounts. No source sets that split, so we keep main's result wherever the exclusion covers the filers only.
  • Preserve the head/spouse pooled medical-first allocation. State paid-cost add-backs restore the newly removed dependent medical premiums and retain the existing dependent long-term care treatment.
  • Attribute premiums to the person who paid them, including family coverage paid by a filer. Supplied deductions above the corresponding paid premium pools fall outside the conservation invariant.
  • New Mexico's Taxation and Revenue Department asks for costs "you paid". As main does, we count premiums a dependent paid in New Mexico's paid-cost base; Count only filer-paid premiums in Schedule A medical expenses #10270 addresses that payer question.

Overlap

#10270 counts only the head's and spouse's premiums in the federal subtotal and adds dependents' own premiums back in North Dakota. Whichever PR lands second must apply every item below, so dependent premiums leave the subtotal once and no state adds them back twice:

  1. Keep Count only filer-paid premiums in Schedule A medical expenses #10270's filer-only premium sum and drop itemized_medical_expenses_excluded_dependent_premiums from itemized_medical_expenses. After items 1 to 5 no formula reads that variable, so remove it with the 20 YAML outputs that assert it, or keep it for reporting alone.
  2. Drop the dependent term from itemized_medical_expenses_applied_exclusion.
  3. In Montana, take premiums_in_subtotal from the filers' premiums only, and drop every dependent term in mt_medical_expense_deduction_joint: the computed path's exclusion, the add-back beside a supplied exclusion, and the dependents' part of an exclusion supplied with a subtotal.
  4. Drop New Jersey's supplied-subtotal dependent subtraction, with its offset against a supplied exclusion.
  5. Drop the supplied-exclusion dependent add-backs in nd_renters_refund_income, nm_medical_expense_credit and nm_medical_expense_exemption. Under Count only filer-paid premiums in Schedule A medical expenses #10270, the North Dakota formula adds every dependent's premiums back on all paths, and the New Mexico formulas count no premiums a dependent paid.
  6. Restate this PR's conservation equality and the YAML cases that keep undeducted dependent premiums for filer-paid premiums. For example, the earnings-capped case moves from $52,550.65 to $50,480, and the New Mexico dependent-premium cases move to a $0 credit and a $0 exemption. The Montana and New Jersey cases that supply a subtotal with a dependent's deducted premiums need new expectations too. Drop the dependents' medical part from the applied-exclusion, subtotal and conservation assertions in test_long_term_care_premium_properties.py, and from the three YAML outputs that expect an applied exclusion with a dependent's part.
  7. Keep main's long-term care term.
  8. Rerun both PRs' attribution and conservation cases together.

Microsimulation impact

We estimate that this branch changes no tax unit's federal or state income tax and no household's benefits or net income in 2026 on the default dataset. Schedule A medical expenses before the 7.5% floor fall by $151,839, for 9 sampled tax units (6,200 weighted). The medical expense deduction falls by $689, for 6 sampled tax units (1 weighted). The exclusion that the North Dakota and New Mexico formulas add back, itemized_medical_expenses_applied_exclusion, rises by $151,839, and none of the Montana, New Jersey, North Dakota or New Mexico outputs moves. We compared main at 61c743cb61 with this branch at a91ff3cb76; the head adds one commit that changes the wage input of two North Dakota test cases and no formula or parameter.

Output Main total Change Records changed (weighted)
itemized_medical_expenses $976.0bn -$151,839 9 (6,200)
itemized_medical_expenses_applied_exclusion $17.35bn +$151,839 9 (6,200)
medical_expense_deduction $408.5bn -$689 6 (1)
mt_medical_expense_deduction_joint $1.35bn $0 0 (0)
nj_medical_expense_deduction $25.66bn $0 0 (0)
nd_renters_refund_income $44.56bn $0 0 (0)
nm_medical_expense_credit $30.4m $0 0 (0)
nm_medical_expense_exemption $32.6m $0 0 (0)
mt_income_tax $2.32bn $0 0 (0)
nj_income_tax $29.14bn $0 0 (0)
nm_income_tax $2.26bn $0 0 (0)
state_income_tax $576.5bn $0 0 (0)
income_tax $2,594.9bn $0 0 (0)
household_benefits $2,060.2bn $0 0 (0)
household_net_income $17,246.6bn $0 0 (0)

axiom: TheAxiomFoundation/rulespec-us#1654 queued (26 U.S.C. 213(a) and 162(l)(3) with 26 CFR 1.213-1(a)(1), extended with a comment for a dependent's deducted premiums; the pinned corpus release lacks section 162(l)(3) and 26 CFR 1.213-1, so the signed encoder can't run yet)

🤖 Generated with Claude Code

MaxGhenis and others added 8 commits November 12, 2026 00:53
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With a computed subtotal, a supplied applied exclusion covers the filers
only. North Dakota and New Mexico now add back the dependents' computed
exclusion, as Montana does, so their paid-cost bases match the derived
path. The govinfo copy of 26 U.S.C. 162 replaces the Cornell link.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
No source-conflict register entry cites these captures, and the source
captures folder holds texts that register entries cite.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@MaxGhenis
MaxGhenis marked this pull request as ready for review November 14, 2026 20:51
MaxGhenis and others added 4 commits November 14, 2026 16:54
The computed subtotal leaves out the medical premiums dependents deduct,
and the computed applied exclusion holds them. Beside a supplied
subtotal, the Montana formula capped the exclusion's medical part at the
filers' premiums, so it took those dependent premiums out of line 1 a
second time. It now caps the medical part at the filers' premiums plus
the dependents' computed exclusion. A subtotal and exclusion supplied at
their computed values then give the computed deduction, as
test_mt_medical_premium_deduction_invariants.py requires.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Beside a supplied subtotal, the Montana formula now reads the part of
the supplied exclusion above the filers' own computed exclusion as the
dependents' deducted premiums, up to the dependents' computed exclusion.
An exclusion that covers the filers only then leaves the dependents'
premiums in the subtotal, as on main, and an exclusion supplied at its
computed value still gives the computed deduction.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
A caller can supply an exclusion above what the return's people
deducted. Beside a supplied subtotal, the Montana formula reads no more
of that exclusion as a dependent's premiums than the dependent deducted.
The new case gives 7,500 in place of 7,000 when the formula drops that
cap.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
…tal once

The computed subtotal leaves out the medical premiums dependents deduct,
and the computed applied exclusion holds them. Beside a supplied
subtotal, the New Jersey formula subtracted the dependents' computed
exclusion again, so a subtotal and exclusion supplied at their computed
values gave 6,400 where the computed path gives 9,400.

The formula now reads the part of the supplied exclusion above the
filers' own computed exclusion as the dependents' deducted premiums, up
to the dependents' computed exclusion, and subtracts only the rest. A
subtotal supplied alone, or with an exclusion that covers the filers
only, still loses the dependents' deducted premiums, as on main.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Since #10724, main counts SSI and TANF in North Dakota renter's refund
income. The two North Dakota cases supplied the parent's $40,000 of wages
as employment_income and left employment_income_before_lsr at zero. The
TANF formulas read earnings from the second input, so the model paid the
family the full standard of need, 9 x $704 + 3 x $739 = $8,553, and refund
income came to $43,553 against the expected $35,000.

Both cases now supply the wages as employment_income_before_lsr, as main's
North Dakota cases do. The family's countable earnings, $1,216.67 a month
or more, exceed the $704 and $739 standards, so the family gets no TANF
and every expectation holds as written.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@MaxGhenis MaxGhenis closed this Nov 16, 2026
@MaxGhenis MaxGhenis reopened this Nov 16, 2026
@MaxGhenis MaxGhenis closed this Nov 16, 2026
@MaxGhenis MaxGhenis reopened this Nov 16, 2026
@MaxGhenis
MaxGhenis merged commit fba7abe into main Nov 16, 2026
75 of 81 checks passed
@MaxGhenis

Copy link
Copy Markdown
Contributor Author

US + core hub merge audit, us#10831 at 44fb51ba828c570ad7d22633060777b066cdf75e (squash): Exclude dependents' deducted health premiums from Schedule A.

  • Reviews:
    • The PR has no GitHub reviews. Three independent review rounds ran outside GitHub, with a fix round after each of the first two. After the third round, one commit changed the wage input of two test cases, and a check read it.
    • The first round, at 1210c47fa9, read 26 U.S.C. 162(l)(3) and 213(a) on govinfo, 26 CFR 1.213-1(a)(1) on ecfr.gov and Rev. Proc. 2024-40 on irs.gov. It ran ten households on the head and on the merge base 935b24ff4f. For the description's example, the subtotal fell from $53,000 to $50,000 and federal income tax rose by $360, as its hand figures gave.
    • In that round the new federal YAML file's 18 cases, New Jersey's 17 and both property files passed on the head. On main's code, 17 of the 18 federal cases failed.
    • It found the law and the numbers right, and it requested changes to the description, which gave no impact estimate and narrated process.
    • It listed five minor findings. The PR's entry in the source-conflict register rested the change on a cross-return reading of 162(l)(3), where 213(a)'s "paid" and 1.213-1(a)(1) settle it. Beside a supplied exclusion, the Montana formula added the dependents' part, and the North Dakota and New Mexico formulas did not. A mutant that dropped the new variable's dependent test passed all 18 new cases. Count only filer-paid premiums in Schedule A medical expenses #10270 edits the same files, and the description left out Missouri.
    • The first fix round, in 440cc863f1 and 3dd4f8e6f2, removed that register entry and cited 213(a) and 1.213-1(a)(1) in the variables. It added a case with a self-employed filer beside a self-employed dependent, and two cases that pinned the North Dakota and New Mexico results. It recorded the Count only filer-paid premiums in Schedule A medical expenses #10270 overlap in the conservation test and the description, and it changed no formula.
    • The second round, at 3dd4f8e6f2, read the code and computed each case by hand, and it ran no test. It found no defect in the federal change. The two new state cases pinned results the sources don't support. The North Dakota case expected $36,000 where NDCC 57-02-08.1(5)(c) gives $35,000, and the New Mexico case moved the credit and exemption from $2,800 and $3,000 to $0.
    • It also found two source captures that no entry cited, a law.cornell.edu link, and a Montana year range that main had since moved to 2017. It checked the Count only filer-paid premiums in Schedule A medical expenses #10270 overlap against both diffs. A merge that kept both PRs' lines would give $47,000 for the description's example, where each PR alone gives $50,000.
    • The second fix round ran from 40afcd9766 to a91ff3cb76. It ran the test files the second round left owed, and each passed. The dependent-test mutant failed the filer-and-dependent case, 8,000 against 3,000. The North Dakota and New Mexico formulas now add the dependents' computed exclusion beside a supplied filer exclusion. The two cases expect $35,000 again, and the $2,800 credit and the $3,000 exemption. The round deleted the captures, replaced the link and merged main at 61c743cb61.
    • CI first started the full suite at the merged head 3067406b8c, since every job skips on a draft. The seeded Montana test failed in three of its cases before the next push cancelled the run. This PR had added the dependents' part to the applied exclusion, and Montana's supplied-subtotal branch still capped the medical part at the filers' premiums. e7a4603e56, 2455d61eae and f49d9056cb fixed Montana and pinned the cap on the dependents' part.
    • That round then found the same defect in New Jersey, where no test covered it. A computed $10,000 subtotal and $3,000 exclusion, supplied back, gave $6,400 where the computed path gives $9,400. a91ff3cb76 fixed it and added three cases.
    • Each of seven further mutants failed a case. One of them, Montana's dependent part without its cap, passed Montana's own YAML file and failed the federal file's new case.
    • The third round, at a91ff3cb76, raised no blocking finding. It read the six new commits and the merge, recomputed every changed expectation by hand and matched each fix-round result to its log. It fetched NDCC 57-02-08.1 from ndlegis.gov and New Mexico's 2025 PIT-RC and PIT-ADJ instructions. It confirmed that the merge of main dropped no case and weakened no assertion.
    • It read run 43650033673, which passed all 37 jobs at a91ff3cb76 merged into main at 94054aa57e. The run covered 25 federal cases, 20 in New Jersey, 29 in Montana, 39 in North Dakota, and 24 and 17 in New Mexico's credit and exemption files. It also covered the conservation and long-term care property tests and the seeded Montana test.
    • It listed three minor findings, and each stays open below. A caller who supplies the formulas' own exclusion beside a computed subtotal counts the dependents' part twice. One YAML case checks New Jersey's round trip, and no property test does. A documentation string in itemized_medical_expenses_applied_exclusion.py still says "Montana's 2021-2023 schedule".
    • The third round listed one description edit, and I made it before merge. I also named the years the seeded Montana test covers, recorded the double count under Methodology choices and extended the Overlap list.
    • A later run at a91ff3cb76, 43891134822, tested that head merged into main at fd0a748698. In two of its 29 jobs, the federal YAML file's two North Dakota cases failed: nd_renters_refund_income came to $43,553 against the expected $35,000. Of the other 27 jobs, 26 passed and the Contrib job skipped.
    • Count SSI and TANF in North Dakota renter's refund income #10724 had added SSI and TANF to the income North Dakota counts. Both cases supplied the parent's $40,000 of wages as employment_income, and the YAML test runner left employment_income_before_lsr at zero. The TANF formulas read earnings from that input, so the model paid the family the full standard of need, 9 × $704 + 3 × $739 = $8,553.
    • 44fb51ba82 supplies the wages as employment_income_before_lsr in both cases, as Count SSI and TANF in North Dakota renter's refund income #10724 did in two of main's North Dakota cases. It changes two lines of the federal YAML file and no expectation, formula or parameter.
    • A check of that commit parsed the file before and after it. The 25 cases keep their names, periods and expectations, and the two North Dakota cases differ in the parent's wage input alone. The check recomputed both cases' expectations by hand on the head merged into main at fd0a748698. The family's countable earnings, $1,216.67 a month or more, exceed the $704 and $739 standards, so the model pays no TANF and refund income comes to $35,000. No independent review round followed, since the commit changes test inputs alone.
    • In run 43900129556, which tested 44fb51ba82 merged into main at fd0a748698, the Quick Feedback and Baseline (irs) jobs each passed all 25 cases of the federal YAML file. Apply New Jersey, Ohio, Oklahoma and Pennsylvania rules for a filer whom another taxpayer can claim as a dependent #10043 then changed North Dakota and New Jersey files on main, and I restarted CI at 44fb51ba82 on main at 8e1837790b. The restart cancelled that run's last two jobs, both Contrib shards, after its other 35 had passed.
  • Methodology choices for maintainers:
    • The PR removes only the premiums a dependent deducts. Premiums a dependent paid and did not deduct stay in the parents' subtotal, as on main, though 213(a) and 26 CFR 1.213-1(a)(1) let the parents count only expenses they paid. The conservation test pins that treatment, and Count only filer-paid premiums in Schedule A medical expenses #10270 removes it.
    • With a supplied subtotal, the Montana and New Jersey formulas split a supplied exclusion by the computed amounts. They read the part above the filers' own computed exclusion as the dependents' deducted medical premiums, up to the dependents' computed exclusion. No source sets that split. It returns the computed deductions when a caller supplies the computed subtotal and exclusion, and main's result when the exclusion covers the filers only. A maintainer could choose another split.
    • New Mexico's Taxation and Revenue Department asks for expenses "you paid" on PIT-RC line 23 and PIT-ADJ line 18. As on main, the model counts premiums a dependent paid in New Mexico's paid-cost base. In the supplied-zero case, a 65-year-old filer who paid nothing keeps the $2,800 credit and the $3,000 exemption on a dependent's $28,001 of premiums. A maintainer who holds the filer to "you paid" would expect $0 for both, as Count only filer-paid premiums in Schedule A medical expenses #10270 does.
    • Beside a computed subtotal, a supplied applied exclusion covers the filers only, and the Montana, North Dakota and New Mexico formulas add the dependents' computed exclusion to it. The variable's documentation states that convention. A caller who supplies the formulas' full computed exclusion there counts the dependents' part twice in North Dakota and New Mexico, and in Montana up to the filers' undeducted medical premiums. In the description's North Dakota household, the computed $1,000 supplied back gives $34,000 against the computed $35,000.
    • The state paid-cost add-backs restore a dependent's deducted medical premiums and leave out the long-term care premiums the same deduction took, as on main. Under NDCC 57-02-08.1(5)(c), North Dakota takes medical expenses the applicant's dependent paid out of income, so a maintainer may want its base to restore that part too.
  • Gates checked live at merge:
  • Impact: we estimate that this branch changes no tax unit's federal or state income tax and no household's benefits or net income in 2026 on the default dataset. Schedule A medical expenses before the 7.5% floor fall by $151,839 (9 tax units, 6,200 weighted), and the medical expense deduction falls by $689 (6 tax units, 1 weighted). Main 61c743cb61 against a91ff3cb76, to which the head adds one commit that changes the wage input of two North Dakota test cases and no formula or parameter.
  • Next:
    • Count only filer-paid premiums in Schedule A medical expenses #10270 counts only the head's and spouse's premiums in the federal subtotal, and it already conflicts with main in seven files. After this squash it lands second. Its merge of main must take each step below, so a dependent's premiums leave the subtotal once:
      • keep Count only filer-paid premiums in Schedule A medical expenses #10270's filer-only premium sum, and drop itemized_medical_expenses_excluded_dependent_premiums from itemized_medical_expenses and from itemized_medical_expenses_applied_exclusion;
      • remove that variable and the 20 YAML outputs that assert it, or keep it for reporting alone, since no formula then reads it;
      • in mt_medical_expense_deduction_joint, take premiums_in_subtotal from the filers' premiums and drop every dependent term: the computed path's exclusion, the add-back beside a supplied exclusion, and the dependents' part of an exclusion supplied with a subtotal;
      • drop New Jersey's supplied-subtotal dependent subtraction, with its offset against a supplied exclusion;
      • drop the supplied-exclusion add-backs in nd_renters_refund_income, nm_medical_expense_credit and nm_medical_expense_exemption;
      • restate the conservation equality for filer-paid premiums, and drop the dependents' medical part from three assertions of test_long_term_care_premium_properties.py;
      • give new expectations to the YAML cases that keep undeducted dependent premiums, supply a subtotal with a dependent's deducted premiums, or expect an applied exclusion with a dependent's part;
      • keep main's long-term care term, and run both PRs' attribution and conservation cases together on the result.
    • File an issue for Montana's years. Six variable files, a generator comment and mt_separate_return.py still describe Montana's own schedule as 2021-2023, where state_specific_deduction_applies starts in 2017. This PR's Montana cases and the seeded test run 2021 to 2023 alone.
    • File an issue to replace the law.cornell.edu references that main's itemized medical and self-employed health insurance variables still carry, 14 links in ten files under gov/irs, with the govinfo pages this PR cites. self_employed_health_insurance_plan_business.py and self_employed_long_term_care_premiums.py carry one more each.
    • File an issue for a seeded round-trip test of New Jersey's medical deduction: supply the computed subtotal and exclusion back and require the computed deduction. One YAML case checks that today.
    • The third round suggested the supplied-subtotal split beside a computed subtotal too. The formulas' own exclusion, supplied back, would then return the computed results in Montana, North Dakota and New Mexico. That needs an issue.
  • Axiom: Amend IRC 213 medical deduction to exclude premiums deducted under IRC 162(l) TheAxiomFoundation/rulespec-us#1654 queued. A comment there records that this head takes a dependent's deducted premiums out of the parents' section 213 expenses, and gives the law, the readings, the leaf inputs and twelve companion rows. The signed encoder can't run yet: the release rulespec-us pins lacks 26 U.S.C. 162(l)(3) and 26 CFR 1.213-1, and the corpus holds no copy of NDCC 57-02-08.1.

🤖 Generated with Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant