Fix Maryland CTC phaseout per qualifying child

[PavelMakarchuk]

Fixes #9589.

Maryland's 2025 child tax credit phaseout was subtracted once per family, overstating credits for families with multiple qualifying children. Apply the reduction per eligible child: two children and $16,000 AGI now receive $900 instead of $950.

Maryland Worksheet 21C, lines 4–7 reduces the per-child credit before multiplying by qualifying children. Line 7 states: “Multiply Line 6 by the number of qualified children”. The existing two-child $18,000 test is corrected from $850 to $700; new tests cover $15,000/$15,001, $16,000/$16,001 and $24,000/$24,001. Also adds negative-AGI, 2024 cutoff, and credit-to-final-tax integration cases. Single-child and pre-2025 calculations are preserved.

Validation: make format and git diff --check passed. Isolated execution of the current formula against all 23 CTC YAML cases passed; the initial regression run demonstrated five failures against the original formula. Full local YAML execution was blocked before tests by the existing environment's missing spm_calculator.policyengine_adapter dependency; CI must validate full-model execution.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (909176a) to head (da8e351).
⚠️ Report is 37 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9591   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            4         1    -3     
  Lines           76        30   -46     
  Branches         2         2           
=========================================
- Hits            76        30   -46     

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]

Program Review

Source Documents

• PDF: 2025 Maryland resident tax booklet, Worksheet 21C (60 pages; file page 27 / printed 19).
• Corroboration: Enacted Chapter 604, §10-751 and effective-year provision.
• Year: 2025 onward; historical regression checked.
• Scope: Incremental test additions and prior unresolved A1/A2. Formula, parameters and references unchanged.
• Reviewed head SHA: da8e351
• Mode: incremental from 71ccc33
• Integrity: Baseline ancestry verified; prior report preserved; PDF SHA256 17bc9f77e055c0570daee8840921e635b1b0952a39d54f28e33516699b9b6ba5 revalidated; all 60 existing 300 DPI screenshots present.

Critical (Must Fix)

None.

Should Address

None remaining.

• [A1] RESOLVED — Refundable-credit integration: Added test calculates the $900 CTC from two eligible children and $16,000 AGI, asserts $900 refundable credits and −$800 final Maryland tax against $100 before refundable credits. All other registered refundable credits are explicitly zeroed. The CTC and aggregation are not overridden.
• [A2] RESOLVED — Negative-AGI and 2024 regression: Added 2025 two-child negative-AGI credit of $1,000 and 2024 two-child AGI $16,000 credit of $0.

Suggestions

None.

PDF Audit Summary

Category	Count
Confirmed correct comparisons, reused	8
Mismatches	0
Unmodeled items in changed scope	0
Added anchor verification, reused	1 correct

The prior complete audit's legal evidence remains applicable because no formula, parameter or reference changed. The worksheet supports per-child phaseout. No outstanding cross-reference, external-source request or numerical mismatch requires further verification; queue is NONE.

Validation Summary

Check	Result
Regulatory accuracy / references	Prior findings reused; no semantic change
Code patterns	All ten incremental categories audited; no issue
Test coverage	A1 and A2 resolved; 23 committed cases
Isolated formula execution	23/23 actual-formula/eligibility checks passed
Integration arithmetic	Actual credit-list arithmetic independently confirmed $900 / −$800
Additional probes	Prior 210 isolated checks remain applicable; formula unchanged
Current-head CI	No checks reported yet when refreshed for this SHA

The isolated checks use actual formula ASTs, dated parameters and lightweight entity stubs. They are not full-framework tests. No local full-model test success is claimed: a missing dependency prevents that local runtime, and heavy installs/imports were avoided for RAM/disk constraints. Prior-head CI passed selective tests but cannot validate this new head.

Review Severity: APPROVE

No unresolved review findings. Request Ziming's review; wait for current-head CI before merging. This recommendation does not perform a merge or formal GitHub approval.

Run Metrics

Initial full review: 3 unique delegates, 60 pages rendered at 300 DPI, 0 cache hits, 6.9 minutes before initial consolidation. Incremental pass reuses the same delegate and 60 checksum-valid page renders (no new rendering). Exact incremental duration not recorded. Initial captured branch was 0 commits behind base.

Next Steps

Request Ziming's review and monitor current-head CI. No further fix identified.

[hua7450]

The per-child phaseout fix is correct. I checked it against § 10-751 as amended by HB 352 and against the 2025 Worksheet 21C, and I worked out every new or changed expected value by hand; all of them match. I'm requesting changes only for the small cleanups below.

What I checked

• Worksheet 21C, lines 4–7: line 6 subtracts the reduction from $500, and line 7 multiplies the result by the number of qualified children. The formula now does the same. Flooring the family total at $0 gives the same result as a per-child floor, because every qualified child gets the same $500, including a child with a disability.
• § 10-751(c) as amended by HB 352 (Ch. 604): the credit drops by $50 for each $1,000 "or fraction thereof" of federal AGI over $15,000. The ceil step matches worksheet line 4 ("round up to the next whole number").
• The 2024 hard cap of "$15,000 or less" (Ch. 604, struck text) and refundability under § 10-751(d) both match the new 2024 and integration tests.

Requested changes

1. References: md_ctc.py cites casetext and Justia's 2022 edition of § 10-751. The 2022 edition only has the pre-2025 hard cap, not the phaseout. Please add the current official text: https://mgaleg.maryland.gov/mgawebsite/Laws/StatuteText?article=gtg&section=10-751&enactments=false. md_ctc_eligible.py has the same references and could get the same update.
2. Number formatting in tests: the new cases write 15000, 24001 and so on. The rest of md_ctc.yaml uses underscore grouping (15_000). Please match it.
3. Derivation comments in tests: the existing cases show their arithmetic; the new ones don't. One line per case would make the per-child math easy to check, for example # 2 × ($500 − ceil($1,001 / $1,000) × $50) = 2 × $400 = $800.
4. Edited $18,000 case: the comment Phase-out reduction: ($3,000 / $1,000) * $50 = $150 should say "per child", now that the reduction applies to each child.