CDCC coupling accuracy: SC § 21(e)(4), ID/GA reform gating, KY/ME/VT static conformity

[MaxGhenis]

Accuracy-fix lane for CDCC coupling and reform gating. Four issues adjudicated. Every verdict is anchored to a primary source actually read (page/section anchored); nothing was changed without a contradicting statute or form. Local test counts are per-tree policyengine-core test runs in the worktree venv.

#8882 — SC CDCC: honor the IRC § 21(e)(4) separated-taxpayer exception — FIXED

Verdict: CORRECT (fixed). Fixes #8882.

S.C. Code § 12-6-3380 (verbatim, official SC Legislature): "The credit is computed as provided in Internal Revenue Code Section 21, except that the term 'applicable percentage' means seven percent and is not reduced, and only expenses that are directly attributable to items of South Carolina gross income qualify for the credit." The section has no filing-status language of its own — it imports § 21 wholesale, carrying both the § 21(e)(2) joint-return requirement and the § 21(e)(4) separated-taxpayer exception. SC conforms to the IRC as of Dec 31, 2024 (§ 12-6-40), well after § 21(e)(4) existed.

sc_cdcc_potential previously applied a blanket filing_status != SEPARATE exclusion, denying the exception. It now reads cdcc_filing_status_eligible (the shared § 21(e)(2)/(e)(4) gate, on main since #8704, and the same gate the federal cdcc_potential uses). An ordinary MFS filer stays ineligible (form-faithful, per the SC1040 instructions' blunt "cannot claim if MFS"); a separated MFS filer living apart while maintaining a home for a qualifying individual now qualifies, matching the federal treatment.

Verified end-to-end: separated MFS filer with a qualifying child and $3,000 expenses → sc_cdcc = $210 (7% × $3,000); ordinary MFS filer → $0.

Source: S.C. Code § 12-6-3380 and § 12-6-40 (https://www.scstatehouse.gov/code/t12c006.php).
Tests: added ordinary-MFS ($0) and separated-MFS ($210) cases to sc_cdcc.yaml (11 pass).

#8899 — ID/GA CTC contributed reforms leak into pre-activation years — FIXED

Verdict: CORRECT (fixed). Fixes #8899. Reproduced end-to-end via the real Simulation structural-reform path (user reform setting in_effect true only from 2028, start_instant=2026):

variable	pre-fix @2026 (pre-activation)	post-fix @2026	@2028 (activation)
ga_refundable_ctc	250 (leak)	0	250
id_non_refundable_credits	410 (leak)	0	410

• GA: ga_refundable_ctc had no per-period gate. The factory installs the reform for the whole simulation whenever in_effect is true in any of the next five years, so an installed-but-inactive reform paid the refund every year. Added where(p.refundable.in_effect, refundable_credit, 0), mirroring the Idaho refundable pattern.
• ID: the refundable payment was already period-gated, but modify_parameters revived id_ctc in the ordered nonrefundable list from a hardcoded 2026-01-01 regardless of activation date. It now revives from the earliest instant either in_effect toggle is true (floored at the 2026 baseline-drop date).

Tests: added test_id_ga_ctc_reform_activation.py — YAML input cannot express a dated parameter, so this drives a multi-year Simulation with in_effect true only from 2028 and asserts inertness at 2026 and activation at 2028 (2 pass). All 10 pre-existing ID/GA YAML tests still pass unchanged.

#8883 — MD CDCC separate-filer phase-out start — VERIFIED CORRECT (recommend close, no code change)

Verdict: WRONG (the model is right; the alleged $20,500 reflects superseded pre-2019 law). Recommend closing #8883.

Current Md. Tax-Gen § 10-716 (verbatim, MD General Assembly compiled statute): (a)(4) defines "'Taxpayer'" as "(i) an individual filing an income tax return; or (ii) a married couple filing a joint income tax return." Subsection (d) then has only two categories: (d)(1) "a taxpayer filing an individual return ... exceeds $30,000 ... reduced by 1% for each $2,000"; (d)(2) "a married couple filing a joint income tax return ... exceeds $50,000 ... 1% for each $3,000." There is no married-filing-separately clause. By the (a)(4) definition an MFS filer is an "individual filing an income tax return," so it falls in the $30,000 / $2,000-increment bucket — exactly what the model encodes for SEPARATE. The 2025 Form 502CR confirms it (single "Individual taxpayer" phase-out from $30,001; MFS grouped with single at the $112,100 cap). The model's SEPARATE agi_cap ($92,000 in 2019, matching (b)(1)) and phase-out are both correct.

The alleged $20,500 was real only through tax year 2018: pre-2019 law gave MFS filers half the individual thresholds ($20,500 full-credit ceiling, $25,000 cutoff). HB 810 (2019 session, effective TY2019) restructured the credit and folded MFS into the general "individuals" category. The model's values start at 2019-01-01, so $20,500 is superseded and does not apply. No change.

Sources: § 10-716 (https://mgaleg.maryland.gov/mgawebsite/Laws/StatuteText?article=gtg&section=10-716); 2025 Form 502CR (https://www.marylandcomptroller.gov/content/dam/mdcomp/tax/forms/2025/502cr.pdf); HB 810 (2019) Fiscal Note (https://mgaleg.maryland.gov/2019RS/fnotes/bil_0000/hb0810.pdf); Comptroller Tax Tip #50 (https://www.marylandcomptroller.gov/content/dam/mdcomp/tax/legal-publications/tips/personal/tip50.pdf).

#8826 — State CDCC coupling to federal § 21 (static-conformity drift audit) — AUDIT + PARTIAL FIX

Scope: enumerate every state variable consuming federal cdcc / capped_cdcc / cdcc_potential; classify each state's conformity (static fixed-date vs rolling); fix any that silently drifted where conformity is static. The federal § 21 change that matters is the 2026 OBBBA rate increase, which the model encodes at 2026-01-01: phase_out.max 0.35→0.50, phase_out.min 0.20→0.35, and a new amended_structure second phase-out (all labeled H.R.1 / OBBBA). Confirmed live in the model — so the drift is real: a state that (a) multiplies the federal § 21 credit amount (which embeds the rate) and (b) statically conforms to a pre-OBBBA IRC would silently adopt the 2026 increase.

Key distinction: only states multiplying the federal credit amount embed the rate. States that use only the § 21 expense base (cdcc_relevant_expenses) × their own rate (SC, CA, VA, ID) do NOT drift on the rate — for them the concern is the § 129 base (below), not the rate.

Fixed — static-pre-OBBBA states that multiply the federal credit

Verified verbatim against current statute text (credit-coupling AND conformity-date provisions):

State	CDCC statute (coupling)	Conformity statute → date	Model read	Fix
KY	KRS 141.067: 20% of "the federal credit allowed under Section 21"	KRS 141.010(21) → Dec 31, 2024	live cdcc	now pre_obbba_cdcc for ≥2026
ME	36 M.R.S. § 5218: 25% of "the federal tax credit allowable for child and dependent care expenses"	36 M.R.S. § 111(1-A) → Dec 31, 2024	live cdcc	now pre_obbba_cdcc for ≥2026
VT	32 V.S.A. § 5828c: 72% of "the federal child and dependent care credit allowed to the taxpayer"	32 V.S.A. § 5824 → Dec 31, 2024	live capped_cdcc	now pre_obbba_capped_cdcc for ≥2026

Reproduction (KY, low-income joint, 2 children, $6,000 childcare): ky_cdcc_potential was $356 in 2026 (live OBBBA rate) vs $240 in 2025; now $240 in 2026 (pre-OBBBA rate held). Implemented via new federal variables pre_obbba_cdcc_potential / pre_obbba_cdcc / pre_obbba_capped_cdcc that recompute the § 21 credit on the pre-OBBBA rate schedule (only the rate is frozen at parameters("2025-01-01"); the § 21(b)–(e) expense base and § 26 cap are OBBBA-invariant and reused live). Tests: federal pre_obbba_cdcc.yaml (2025 == live, 2026 holds pre-OBBBA) + 2026 regression cases in KY/ME/VT (all pass; KY/ME/VT trees 1,402 pass).

Audited — NOT fixed (live read is correct)

State(s)	Consumer var reads	Conformity	Adopts OBBBA § 21?	Action
OH	cdcc_potential/cdcc	Static, but SB 9 (136th GA, eff. Mar 5, 2026) moved the date past OBBBA and was enacted to sweep in post-3/7/25 federal changes	Yes	none — live read correct (ORC § 5701.11 verbatim confirms)
AR	ar_federal_cdcc (pinned to 2013 IRC, #8885)	Static, Jan 2, 2013	No	none — already pinned; verified, not duplicated
CO	cdcc_potential (2026 redesign, #8924)	Rolling	Yes	none — correct; not regressed
IA, KS, DE, OK, RI, MD, NE, LA, NM	cdcc/cdcc_potential	Rolling (Tax Foundation FF864 Table 1 + statutes)	Yes	none
NJ, PA, MS	cdcc/cdcc_potential	Selective general conformity, but each CDCC statute ties to the current federal credit allowed/claimed	Yes	none
DC, KS (verified in #8704)	—	—	—	none

GA and WV are static-current states whose CDCC multiplies the federal credit; both moved their conformity date on/after the OBBBA window for TY2026 (GA via 2026 HB 1199 → Jan 1, 2026; WV via 2026 SB 400 → Dec 31, 2025), so both adopt OBBBA § 21 and the live read is correct — not pinned. (Georgia and West Virginia conformity dates confirmed against 2026 session legislation.)

Out-of-lane / noted, not fixed here

• WI (2024+ branch): wi_childcare_expense_credit_potential uses cdcc_rate directly, so it also embeds the 2026 rate; WI conforms statically at Dec 31, 2022. This branch is already flagged in Verify state CDCC coupling to federal IRC § 21 filing-status and deemed-income changes (follow-up to #8704) #8826's own verification comment for a filing-status bug (WI/NY/HI § 21(e)(2)/(4) gating) that is explicitly out of this lane; the rate drift should be fixed together with that gating change. Noted for the WI/NY/HI follow-up.
• The WI/NY/HI filing-status bugs and the AR refundability / VT TY2022 / expense-base under-inclusion items from Verify state CDCC coupling to federal IRC § 21 filing-status and deemed-income changes (follow-up to #8704) #8826's comment are pre-existing and out of this lane (not regressions of the OBBBA changes). Left for their dedicated follow-ups.

§ 129 dependent-care-assistance base (re: PR #8931, not yet merged) — RESEARCHED, fix pending rebase

PR #8931 (open, mergeable) makes federal cdcc_limit = max(0, max×count − dependent_care_assistance_exclusion) per IRC § 21(c) / Form 2441 Part III, flowing into cdcc_relevant_expenses. States reading cdcc_relevant_expenses (incl. SC) inherit the § 129 reduction automatically. But CA, ID, VA independently recompute the raw $/count limit and do NOT subtract § 129:

• ca_cdcc_relevant_expenses, id_cdcc_limit, va_child_dependent_care_deduction_cdcc_limit all compute p.max × count without the exclusion.
• Verified against the state forms that each should be net of § 129: CA FTB 3506 Part IV lines 26–33 reduce the cap by excluded benefits and cap at federal 2441 line 31; ID Form 39R Line 6 worksheet excludes employer amounts (line 1) and subtracts "excluded benefits from Part III of Form 2441" (lines 3–4); VA deducts "the amount on which the federal credit is based" = Form 2441 line 31, already net (Va. Code § 58.1-322.03(3); Form 760 code 101). For a family with $5,000 of employer dependent-care benefits, the raw-limit recomputation over-states the state benefit.

This fix depends on dependent_care_assistance_exclusion, which does not exist until #8931 merges, so it is not encoded here (referencing a non-existent variable would break the branch). Recorded for the rebase: once #8931 is on main, subtract dependent_care_assistance_exclusion from the CA/ID/VA independent limits (flooring at zero). Sources: FTB 3506 (2024) Part IV; ID Form 39R (2023) Line 6 worksheet; Va. Code § 58.1-322.03 + Form 760 (2023) code 101.

---

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 94.91525% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.26%. Comparing base (f9e58e7) to head (0d7c6d2).
⚠️ Report is 150 commits behind head on main.

Files with missing lines	Patch %	Lines
...s/vt/tax/income/credits/cdcc/vt_low_income_cdcc.py	0.00%	3 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##              main    #8933      +/-   ##
===========================================
- Coverage   100.00%   96.26%   -3.74%     
===========================================
  Files            3        6       +3     
  Lines           55      134      +79     
  Branches         0        5       +5     
===========================================
+ Hits            55      129      +74     
- Misses           0        5       +5     

Flag	Coverage Δ
unittests	96.26% <94.91%> (-3.74%)	⬇️

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.