Model historical Arkansas Medicaid work requirements

[daphnehanse11]

What changed

• adds an Arkansas-specific historical Medicaid work-requirement overlay instead of changing the Arkansas adult Medicaid income limit
• introduces explicit Arkansas parameters for when the waiver approximation applies, the affected age range, the annualized income cutoff, the dependent-child exemption age, and the monthly hours threshold
• wires the Arkansas overlay into is_medicaid_eligible so it only affects the adult expansion category in Arkansas
• adds Arkansas-specific tests plus Medicaid eligibility regression coverage

Why

Issue #8272 tracks the need to represent the Arkansas Works work-requirement period more faithfully without pretending it was an income-threshold change.

Modeling approach

This PR uses an annual approximation because PolicyEngine's default microdata are primarily annual even though the underlying waiver operated monthly.

The approximation is:

• 2018: adults ages 30-49 at or below 100% FPL are subject
• 2019: adults ages 30-49 at or below 138% FPL are subject
• 2020+: the Arkansas historical overlay turns off

Observable exemptions included in the approximation:

• pregnancy
• full-time student status
• disability / blindness / incapacity proxies
• caring for a dependent child under 18
• caring for a disabled dependent
• unemployment compensation receipt

Explicitly not modeled here:

• monthly reporting compliance
• good-cause months
• treatment-program participation
• three-month lockout administration
• the brief 2019 phase-in for ages 19-29

That last omission is intentional: modeling ages 19-29 in annual data would likely overstate exposure because the phase-in was only partially rolled out before enforcement stopped.

Confidence

I'm reasonably confident in this as an annual-data approximation. I'm much less confident that annual data can support a literally correct monthly Arkansas waiver model, which is why the assumptions are documented in both parameters and code comments.

Validation

• uv run python policyengine_us/tests/test_batched.py policyengine_us/tests/policy/baseline/gov/states/ar/dhs/medicaid/eligibility --batches 1
• uv run python policyengine_us/tests/test_batched.py policyengine_us/tests/policy/baseline/gov/hhs/medicaid/eligibility --batches 1
• uv run python policyengine_us/tests/test_batched.py policyengine_us/tests/policy/baseline/gov/hhs/medicaid --batches 1

Closes #8272

[codecov]

Codecov Report

❌ Patch coverage is 98.03922% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 95.58%. Comparing base (f9a4a0f) to head (80007a2).
⚠️ Report is 1320 commits behind head on main.

Files with missing lines	Patch %	Lines
...v/hhs/medicaid/eligibility/is_medicaid_eligible.py	87.50%	1 Missing ⚠️

Additional details and impacted files

@@             Coverage Diff             @@
##              main    #8274      +/-   ##
===========================================
- Coverage   100.00%   95.58%   -4.42%     
===========================================
  Files            7        3       -4     
  Lines          215       68     -147     
  Branches         6        2       -4     
===========================================
- Hits           215       65     -150     
- Misses           0        2       +2     
- Partials         0        1       +1     

Flag	Coverage Δ
unittests	95.58% <98.03%> (-4.42%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.

[daphnehanse11]

PR review

I checked the PR out in a fresh worktree, validated the implementation against the cited regulations, and pushed CI/quality fixes.

Fixes pushed (c2cf764)

• Lint: ran ruff format on ar_medicaid_work_requirement_eligible.py and ar_medicaid_work_requirement_subject.py (CI Lint was failing).
• Changelog: added changelog.d/codex-issue-8272-ar-medicaid-work-requirement.added.md (CI Check changelog fragment was failing).
• Parameter description, monthly_hours_threshold.yaml: changed "more than this number" → "at least this number". The formula uses monthly_hours_worked >= p.monthly_hours_threshold, and the cited rule (016.20.18 Ark. Code R. 003) reads "80 hours or more per month", so 80 hours qualifies.
• Parameter description, applies.yaml: appended the standard ", if this is true" suffix used by the federal work_requirements/applies.yaml to make the boolean semantics explicit.

Regulatory accuracy

Cross-checked against the cited KFF brief and Arkansas Code Rule 016.20.18; the annual approximation matches the documented rollout:

• 2018: ages 30-49 at ≤100% FPL, 80 monthly hours, exemptions for pregnancy/full-time student/disability/dependent minor/unemployment receipt.
• 2019: same age band, income limit expands to 138% FPL.
• 2020+: applies turns off (court set aside the rule in March 2019).

The intentional omissions (monthly reporting, three-month lockout, 19-29 phase-in, TEA/SNAP-derived exemptions, AI/AN, medical frailty, treatment programs) are clearly documented in both the PR description and inline parameter comments — fair tradeoff for annual microdata.

Verified locally

• policyengine-core test policyengine_us/tests/policy/baseline/gov/states/ar/dhs/medicaid/eligibility -c policyengine_us — 10/10 pass
• policyengine-core test policyengine_us/tests/policy/baseline/gov/hhs/medicaid/eligibility -c policyengine_us — 80/80 pass (no regressions in the existing Medicaid integration test)
• ruff format --check . — clean

Validation summary

Check	Result
Regulatory accuracy	Matches cited sources
Reference quality	Each parameter has linked authoritative sources
Code patterns	Followed (fixed formatting)
Test coverage	10 new YAML tests + 3 integration tests; codecov flagged 1 untested line on the federal-branch where(...) in is_medicaid_eligible.py because p.work_requirements.applies is false in tested years — pre-existing federal pathway, not introduced by this PR
CI status	Lint & changelog fixed; other failing checks should re-run