Fix AR TEA below-trigger payment and AL TANF work expense rate; verify MT TANF standards

[MaxGhenis]

Summary

Source-first review of three TANF/TEA parameters flagged by Axiom encoding. Two were genuine defects (fixed here); one was already correct (verified, references dated).

Per-issue verdicts

Arkansas TEA below-trigger payment — FIX (Fixes #8841)

The below-trigger branch subtracted countable income from the maximum grant, which contradicts the TEA Policy Manual Section 2362 ("Reduced Payment – Gross Income Trigger", effective 01/01/2023, manual p. 93).

The manual defines a flat grant, not a gap-filled benefit. Income eligibility (countable income ≤ $513) is a separate gate:

• Section 2362 Example Basic prototype #1: family of 3, $150 monthly gross income → "their payment is the maximum grant for a family size of three (3) or $204."
• Section 2362 Example Run black formatting on docs #2: family of 2, $1,200 monthly gross income (≥ $1,026 trigger) → "the Browns' payment is 50% of the maximum for a two-person family, or $81."

Corrected ar_tea to return the full maximum grant below the trigger and 50% of the maximum at/above it. Updated the unit and integration tests, anchoring cases directly to Examples #1 and #2.

Alabama TANF earned-income work expense — FIX (Fixes #8840)

PolicyEngine used a 20% work-expense rate from the DHR Public Assistance Payment Manual (2022). The current Ala. Admin. Code r. 660-2-2-.30(2) (chapter revised 2/14/2025; this section's amendment effective 05/15/2022, source) provides earned-income deductions of:

(a) disregard all earnings from new employment that is timely and accurately reported in the first twelve months in which wages are received, then deduct the following work expenses thereafter; (b) the first 30% of gross earnings; and (c) the cost … of child care …

The Administrative Code is the promulgated regulation and controls over the sub-regulatory payment manual (and is the newer source). Changes:

• work_expense_rate: added 2022-05-15: 0.3 (kept the prior 2002-01-01: 0.2), with Admin Code reference.
• Documented the first-12-month new-employment full earnings disregard (660-2-2-.30(2)(a)) as a known limitation in al_tanf_countable_earned_income — it depends on months-of-current-employment, which is not in the input data, so established (>12-month) employment is assumed. Not fabricating an input we cannot populate.
• Recomputed all Alabama TANF tests for the 30% rate.

Confirmed separately that al_tanf (payment standard − countable income) matches r. 660-2-2-.32 ("The grant amount is computed by subtracting rounded total net income from the payment standard"), so the gap-fill structure is correct.

Montana TANF standards source hierarchy — VERIFIED CORRECT (Addresses #8842)

No value change. ARM 37.78.420 on Cornell/LII carries fixed tables from the 2011 amendment based on the FY2007 FPL (2-person payment standard $401). But Montana's own current DPHHS TANF 001 Monthly Income Standards (effective 07/01/2025, which cites ARM 37.78.420 as its authority, source) publishes standards derived from the current FPL:

AU size	GMI	NMI	Benefit std	Payment std
1	859	464	365	425
2	1,162	628	493	575
3	1,465	792	622	725

The manual states these are Payment = 35% of 2023 FPL, Benefit/GMI/NMI = 30% of 2023 FPL. PolicyEngine's computed values match to the dollar (e.g. 2-person: payment 575.17, benefit 493.00, net 628.03, gross 1,161.85). The manual is the operative, annually-refreshed expression of ARM 37.78.420 and supersedes the rule's stale 2011 fixed table; PE is correct. Dated the TANF 001 references in the two FPL-rate parameters to make the source hierarchy explicit.

The ARM post-employment payment-standard branch ($375/$275/$175 for months 1–3 after leaving TANF for work) is a narrow special-case program not modeled here; recommend closing #8842 as verified, with that branch tracked separately if desired.

Tests

All via policyengine_core ... test ... -c policyengine_us:

• Montana TANF: 158 passed
• Arkansas TEA: 59 passed (was 58; +1 Section 2362 Example case)
• Alabama TANF: 55 passed
• Combined: 272 passed

Touched Python files pass ruff format --check.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (f9e58e7) to head (80389be).
⚠️ Report is 127 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #8915   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            3         2    -1     
  Lines           55        31   -24     
=========================================
- Hits            55        31   -24     

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.