Fix West Virginia low-income exclusion double count

[PavelMakarchuk]

Summary

• remove the low-income earned-income exclusion from West Virginia Schedule M subtractions
• preserve its single application between West Virginia adjusted gross income and taxable income
• add an integrated 2026 regression covering the intermediate values and pre-credit tax

Statutory and form basis

West Virginia IT-140 lines 1-7 compute West Virginia adjusted gross income from federal AGI plus Schedule M additions minus Schedule M subtractions, then subtract the low-income earned-income exclusion and personal exemptions to reach taxable income. This matches West Virginia Code sections 11-21-10 through 11-21-12 and prevents the exclusion from being applied twice.

Checks

• focused wv_taxable_income YAML: 3 passed
• full baseline WV income-tax YAML directory: 109 passed
• WV contrib/reform YAML: 16 passed (test assertions completed; Python 3.14 process was interrupted during teardown)
• household marginal-tax-rate YAML: 5 passed
• make format (including Ruff): passed
• git diff --check: passed
• independent review-fix-review cycle: CLEAN at ad56b65

No partner contract files are changed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (4c6ac7e) to head (ad56b65).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9166   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         1    -2     
  Lines           49        14   -35     
=========================================
- Hits            49        14   -35     

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.

[PavelMakarchuk]

@MaxGhenis This is ready for maintainer review: exact reviewed head ad56b658abc3 is 33/33 green. The independent review-fix-review cycle corrected the public WV AGI structure (single low-income exclusion application) and ended CLEAN; full WV income-tax checks are 109/109.

[DTrim99]

Program Review — PR #9166 (Fix West Virginia low-income exclusion double count)

Author: PavelMakarchuk · Bug fix

Source Documents

• WV IT-140 (2024) form order of operations · W. Va. Code §§ 11-21-10, -11, -12

Branch Status

⚠ 6 commits behind main (1 ahead). Minor; review scoped to the merge-base diff.

Summary

Correct fix for a real, confirmed double-count bug. 0 critical, 0 blocking. The WV low-income earned-income exclusion was being subtracted twice on main — once inside the Schedule M subtractions list (federal AGI → WV AGI) and again between WV AGI and taxable income. The PR removes exactly the Schedule M listing, preserving the single correct application. CI green (33/33). Verified correct against the statute and IT-140.

The bug was real (confirmed against IT-140 + §11-21-11)

IT-140's front-of-form order of operations is decisive:

Line	Item
4	WV AGI = federal AGI + Schedule M additions − Schedule M subtractions
5	Low-Income Earned Income Exclusion (worksheet)
7	WV Taxable Income = Line 4 − Lines 5 & 6, floored at 0

The exclusion appears only on Line 5 — it is not among the Schedule M decreasing modifications (lines 29–52). On main it was in both places:

• subtractions.yaml listed it → wv_subtractions → subtracted in wv_agi (path 1)
• wv_taxable_income.py subtracted it again (path 2)

Net effect: over-stated subtractions, under-stated taxable income, under-stated tax for eligible low-income earners — contradicting §11-21-11 and IT-140.

The fix is correct

• Removes wv_low_income_earned_income_exclusion from subtractions.yaml (it leaves WV AGI). ✅
• Keeps the single subtraction in wv_taxable_income.py (IT-140 Line 5 / §11-21-11) — not removed entirely. ✅
• Grep-confirmed the exclusion's only remaining consumer is wv_taxable_income.py (applied exactly once). ✅
• Direction/magnitude correct: raises taxable income for affected earners to the statutory amount. ✅
• Reference updated §11-21-4E → §11-21-11 (the correct "WV taxable income = WV AGI less exemptions" section). ✅

Worked example (new regression test, 2026 SINGLE, fed AGI 10,000, earned 7,000): exclusion = 7,000; wv_subtractions: 0; wv_agi: 10,000 (exclusion NOT in AGI — pins the fix); wv_taxable_income = max(10,000 − 7,000 − 2,000, 0) = 1,000 (applied once); tax = 21.10. Under the old double-count, taxable income would have been 0 — so the test asserts a strictly-higher value and fails on multiple lines if the duplicate returns. Non-vacuous, well-constructed.

Blast radius — clean

The Schedule M subtractions list feeds only wv_subtractions → wv_agi → wv_taxable_income. No other WV variable relied on the exclusion being in that list; notably the LIFTC AGI variable builds its own base from adjusted_gross_income and is unaffected. Removing the list entry breaks nothing.

Critical / Should Address

None.

Suggestions (optional)

• Add a second integration case for a JOINT/SEPARATE filer (cap 10,000 vs 5,000) or one with a genuine non-zero Schedule M subtraction alongside the exclusion — to lock in that the exclusion no longer leaks into wv_subtractions when other subtractions are present.
• Minor: subtractions.yaml metadata still references §11-21-12 (Schedule M) while the variable now cites §11-21-11 — both correct for their respective scopes, worth a glance for consistency.

Validation Summary

Check	Result
Regulatory Accuracy	Double-count confirmed real via IT-140 (Line 5, not Schedule M) + §11-21-11; fix removes exactly the duplicate, keeps single application; direction/magnitude correct
Code Patterns	0 critical / 0 should — correct add()-list de-dup mechanism, no hardcoded values, all TaxUnit entities, changelog fixed correct, reference corrected
Test Coverage	Strong non-vacuous regression (old→taxable 0, new→1,000; pins intermediate WV AGI = exclusion excluded, and single application)
Blast Radius	Clean — LIFTC and all other consumers unaffected
CI Status	Passing (33/33)

Review Severity: APPROVE

A correct, well-scoped bug fix backed by the IT-140 form and §11-21-11: the exclusion was genuinely double-applied and is now applied exactly once, with a rigorous end-to-end regression test and a clean blast radius. No blocking or should-address items. Rebase off the 6-commit lag before merge.

Review generated with Claude Code via /review-program