Wire NY itemized-deduction phase-out into IT-196 Line 40 (pre-TCJA Pease limitation)

[MaxGhenis]

Fixes #8909

Problem

gov.states.ny.tax.income.deductions.itemized.phase_out.start was an orphaned parameter: it held values (2021-2025) but no formula consumed it. It stores the pre-TCJA federal "applicable amount" from 26 U.S.C. § 68(b) — the income threshold above which New York limits itemized deductions on Form IT-196 Line 40 — which NY Tax Law § 615 conforms to and continues to inflation-adjust. Without a consumer, high-income NY itemizers were not having their itemized deductions limited as the statute requires.

Statutory basis

NY Tax Law § 615 requires itemized deductions to be computed using federal rules as they existed immediately before TCJA (Public Law 115-97). Those pre-TCJA rules include the IRC § 68 "Pease" overall limitation on itemized deductions, which TCJA zeroed out for 2018-2025 federally but which NY still applies via its own conformity.

Form IT-196's Line 40 "Total itemized deductions worksheet" implements this limitation: reduce itemized deductions by the lesser of

• 3% of federal AGI over the applicable amount (worksheet lines 5-8, IRC § 68(a)(1)), or
• 80% of the deductions subject to the limitation (worksheet lines 1-4, IRC § 68(a)(2)),

where the deductions not subject to the limitation (IRC § 68(c)) are medical, investment interest, and casualty/theft losses.

This is distinct from Line 46, which is New York's own § 615(f)/(g) higher-income itemized-deduction adjustment — that one is keyed on NY AGI (thresholds $100k/$475k/$525k/$1M/$10M) and is already implemented in ny_itemized_deductions_reduction.py. Line 40 (federal-AGI-gated) and Line 46 (NY-AGI-gated) are sequential steps in the same worksheet chain (Line 40 → Lines 41-45 → Line 46 → Line 47), not alternatives. A high-income filer completes both. Confirmed by reading the IT-196 form and instructions directly (Line 40 worksheet, Table 1) and NY Tax Law § 615.

Approach — clean line-level mapping (not an approximation)

The Line 40 worksheet's line-total inputs map cleanly onto existing PolicyEngine variables:

• Worksheet line 1 (L1) = form lines 4, 9, 15, 19, 20, 28, 39 = total itemized deductions before the limitation = ny_itemized_deductions_max minus ny_college_tuition_deduction (the NY-only college tuition deduction is form line 48/43, added after Line 40, so it is excluded from the Pease base).
• Worksheet line 2 (L2) = form lines 4, 14, 16a, 20, 29, 30, 37 = the § 68(c) exclusion set = medical_expense_deduction + investment_interest_expense + ny_casualty_loss_deduction.
• L3 = L1 − L2 = mortgage interest + real estate taxes + charitable gifts + 2%-floor misc — exactly the § 68(c) base. Investment interest appears in both L1 (inside interest_deduction) and L2, so it cancels out of L3, as intended.

The mapping is gated on federal AGI (adjusted_gross_income = IT-201 line 19, "Federal amount column"), matching the worksheet's line-7 "if not over, stop" check — correctly not ny_agi.

Simplifying assumptions (honest disclosure)

The mapping is exact for every deduction PolicyEngine models. Two IT-196 line items are not modeled anywhere in PolicyEngine and are therefore treated as zero:

• Form line 39 "other itemized deductions" (gambling losses, estate tax on income in respect of a decedent, amortizable bond premium, federal disaster loss, etc.) — absent from ny_itemized_deductions_max regardless, so no new gap is introduced.
• Form line 16a "qualified contributions" subset (PolicyEngine does not split charitable into qualified/non-qualified).

These affect L1/L2 only for filers who have those specific (rare, currently-unmodeled) deductions; for all modeled cases the computation is exact.

Changes

New parameters (both cite pre-TCJA IRC § 68(a) and the 2025 IT-196 Line 40 worksheet):

• .../itemized/phase_out/agi_excess_rate.yaml = 0.03 (3% of AGI-over-threshold limb)
• .../itemized/phase_out/deduction_limited_rate.yaml = 0.80 (80%-of-limited-deductions cap)

New variables:

• ny_itemized_deductions_phase_out — the Line 40 lesser-of reduction.
• ny_itemized_deductions_phase_out_applies — federal-AGI gate on phase_out.start.

Modified variable:

• ny_itemized_deductions — inserted ny_itemized_deductions_phase_out into subtracts before ny_itemized_deductions_reduction, matching the form's Line 40 → Line 46 sequence.

New tests (9 household-level cases):

• ny_itemized_deductions_phase_out.yaml (7): SINGLE/JOINT boundaries at the 2025 applicable amounts ($340,700 / $408,850); both the 3% and 80% limbs binding; the § 68(c) exclusions (medical/investment interest/casualty); the college-tuition exclusion from L1.
• ny_itemized_deductions_phase_out_integration.yaml (2): end-to-end from raw household inputs, confirming Line 40 stacks with Line 46.

Test results

• New phase-out tests: 9 passed.
• Full NY state tree (gov/states/ny): 326 passed, 0 failed — no regressions. Existing high-income NY itemizer fixtures (e.g. the IT-196 MFJ TAXSIM case in ny_income_tax.yaml) are unaffected because their federal AGI is below the Line 40 applicable amount.
• Cross-state itemized (state_itemized_deductions.yaml): 16 passed.
• test_ny_itemized_phase_out_start pinned parameter test: unchanged, still passing (parameter values are untouched — only a consumer was added).
• Structural/integrity (test_parameter_files, test_system_import, test_build_metadata, test_ctc_itemizing_branch_cycle): 16 passed, 1 skipped — the CTC itemizing-branch cycle test confirms no circular dependency was introduced.

ruff format and ruff check pass on all touched files.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (f9e58e7) to head (29d0a8f).
⚠️ Report is 127 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #8917   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         3           
  Lines           55        52    -3     
=========================================
- Hits            55        52    -3     

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.