End OR ERDC and VT CCFAP child immigration tests per 2024 rule changes

[hua7450]

Summary

Oregon and Vermont both ended their child care programs' child citizenship/qualified-immigrant requirement in 2024, serving federally ineligible children with state funds. The model applied is_ccdf_immigration_eligible_child unconditionally in both states, so it wrongly returned $0 for households with non-qualified children in periods on or after mid-2024.

Fixes #9358

Rule changes encoded

State	Program	Requirement ends	Source
OR	ERDC	2024-05-01	OAR 414-175-0021: "Until May 1, 2024, a child in a filing group must meet one of the following citizenship and noncitizen status requirements"; DELC: "regardless of their immigration status"
VT	CCFAP	2024-06-30 (gone from 2024-07-01)	DCF Act 76 updates: "Citizenship status of participating children will no longer be a factor for CCFAP eligibility"; DCF eligibility overview: "This requirement ends June 30"

Implementation

Follows the existing nm_ccap_eligible_child pattern for New Mexico's 2025-11-01 Universal Child Care change: a dated immigration_test_in_effect boolean parameter per state gates the is_ccdf_immigration_eligible_child call, so pre-2024 periods keep the test and historical results are unchanged. Monthly parameter resolution means the flip lands exactly on the May 2024 (OR) / July 2024 (VT) benefit months.

Test plan

• New boundary tests: an undocumented 5-year-old flips ineligible→eligible across 2024-04→2024-05 (OR) and 2024-06→2024-07 (VT), plus current-law eligible cases in both states
• Full OR ERDC + VT CCFAP suites pass locally (218 tests)
• CI passes

🤖 Generated with Claude Code

https://claude.ai/code/session_011UW7tQydLfV769EbeAd6f7

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (3540ad3) to head (4d5f65e).
⚠️ Report is 9 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main     #9359   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files            1         2    +1     
  Lines           17        33   +16     
=========================================
+ Hits            17        33   +16     

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.

[DTrim99]

PR #9359 Review — OR ERDC + VT CCFAP: end the child immigration/citizenship eligibility test

Program review

This PR sunsets the child immigration/citizenship eligibility test for two child-care subsidy
programs, per 2024 rule changes:

• Oregon ERDC (Employment Related Day Care) — test ends 2024-05-01.
• Vermont CCFAP (Child Care Financial Assistance Program) — test ends 2024-07-01.

Each program gains a boolean parameter immigration_test_in_effect (unit: bool,
period: month) that is true while the citizenship/qualified-noncitizen test applies and
false from the first month it no longer applies. The consuming *_eligible_child formula
reads the parameter and drops only the immigration conjunct after the sunset:

• OR or_erdc_eligible_child.py: age_eligible & immigration_eligible → age_eligible.
• VT vt_ccfap_eligible_child.py: age_eligible & is_dependent & immigration_eligible →
  age_eligible & is_dependent.

Both formulas continue to call the existing shared federal CCDF variable
is_ccdf_immigration_eligible_child (reused, not reinvented). The VT formula also refactors its
parameter handle from ...ccfap.age_threshold up to the ...ccfap root so it can read the
sibling immigration_test_in_effect — verified value-neutral (the age_threshold subtree is
byte-identical and no other consumer used the old handle).

Source documents

Source	Jurisdiction	Supports
OAR 414-175-0021 "Citizenship and Noncitizen Status"	OR	"Until May 1, 2024, a child … must meet one of the following citizenship and noncitizen status requirements" — anchors both 2023-07-01 (on) and 2024-05-01 (off)
DELC ERDC program page	OR	Secondary corroboration of post-2024 policy (eligible regardless of immigration status)
VT DCF, Act 76 and CCFAP updates (June 2024) blog	VT	Citizenship "will no longer be a factor," effective June 30, 2024
VT DCF, "More Vermont families qualify…" news	VT	"This requirement ends June 30"
33 V.S.A. § 3512(a)(6) (via Act 76, 2023)	VT	Durable statutory hook for the repeal — present in the sibling .py, absent from the parameter YAML
CCFAP Regulations PDF (Feb 2009)	VT	Durable source for the 2009-02-09 "on" start — present in the .py, absent from the parameter YAML

No PDF value mismatches — this is a logic/date PR, not a rate-table PR.

Branch status

The PR branch is 6 commits behind main. This is informational only and does not affect the
review; a merge/rebase with main before merge is advisable but is not a review finding.

Critical

None.

Should address

1. VT parameter YAML relies only on secondary DCF blog/news citations. The
   vt/dcf/ccfap/immigration_test_in_effect.yaml reference: list cites only a DCF blog post
   and a DCF news page. These corroborate the change and the July 1, 2024 effective date but are
   not durable primary sources, and neither establishes the 2009 start date. The durable primary
   cites already exist in the sibling vt_ccfap_eligible_child.py: 33 V.S.A. § 3512(a)(6)
   (the repeal, enacted via Act 76 2023) and the Feb 2009 CCFAP Regulations PDF (which anchors
   2009-02-09: true). Recommend adding both to the parameter YAML's reference: list (retaining
   the DCF pages as plain-language corroboration). Reference-durability, non-blocking. (OR is clean
   — OAR 414-175-0021 is durable, primary, correctly cited, and quotes the exact sunset date.)

2. Prefer a vectorized where() over the scalar if on the toggle parameter. Both formulas
   use if p.immigration_test_in_effect: to select the branch
   (or_erdc_eligible_child.py:35, vt_ccfap_eligible_child.py:24). This is correct — the
   parameter resolves to a scalar within a single computed period, both variables are
   definition_period = MONTH so each month picks the right branch, and this matches an
   established repo idiom (e.g. il_tanf_payment_level_for_grant_calculation.py). It is not a
   correctness bug. A fully vectorized form is the model's preferred style and removes any
   dependence on the parameter being a Python scalar:
   immigration_eligible = where(p.immigration_test_in_effect, person("is_ccdf_immigration_eligible_child", period.this_year), True)
   then return age_eligible & immigration_eligible (VT: & is_dependent). Style, non-blocking.

Suggestions

1. VT date encoding — mirror the statute exactly. The statute/DCF phrase the change as
   effective June 30, 2024 (last day the test applied), so a day-exact encoding would be
   false: 2024-06-30. The PR uses false: 2024-07-01. Because these are period: month
   booleans evaluated at the month start, 2024-06-30 and 2024-07-01 produce identical
   monthly results (June 2024 reads true, July 2024 reads false under either). The current
   encoding is operationally correct; switching to 2024-06-30 would only make the YAML mirror
   the statutory date more literally, with no computed change. Cosmetic. (OR's 2024-05-01 is
   exactly right — "Until May 1" is exclusive and May 1 is a month boundary.)

2. Parameter descriptions could use the standard naming pattern. Labels correctly spell out
   the state and abbreviate the program (Oregon ERDC, Vermont CCFAP) with no trailing period.
   The description sentences use a free-form verb ("applies … if this is true") rather than the
   standard closing … under the [Full Program Name] program. clause. Minor.

3. Two optional regression cases would fully lock the post-sunset truth table (see Validation
   summary): a VT case proving a non-dependent undocumented child stays excluded after the
   toggle, and an OR case proving an over-age undocumented child stays excluded after the
   toggle. Both surviving gates are already exercised in isolation by pre-existing cases at other
   periods, so these are additive, not blocking.

Validation summary

Check	Result	Notes
Regulatory correctness	PASS	Both formulas drop only the immigration conjunct; OR keeps age eligibility, VT keeps age + is_dependent
OR effective date (2024-05-01)	PASS	"Until May 1" is exclusive; May 1 is the first untested day and a month boundary — exactly right
VT effective date (2024-07-01)	PASS (operationally)	Statute says June 30; month-start evaluation makes 2024-07-01 and 2024-06-30 identical — suggestion to mirror statute
Shared variable reuse	PASS	is_ccdf_immigration_eligible_child reused, not reinvented
VT .age_threshold→.ccfap refactor	PASS	Value-neutral; age_threshold subtree byte-identical; no other consumers of old handle
Code pattern (Category 9 toggle)	PASS	Scalar if on scalar param is the established idiom; both vars MONTH-defined; dates aligned; no hard-coded dates
References — OR	PASS	OAR 414-175-0021 durable, primary, correctly cited
References — VT	PASS w/ warning	Corroborated; durable primary cites exist in .py but not the parameter YAML
Tests	PASS	Both toggles exercised on both sides at the exact boundary month (OR 2024-04/05, VT 2024-06/07)
Test coverage gaps	2 non-blocking	No VT post-toggle non-dependent case; no OR post-toggle over-age case
Changelog	PASS	changelog.d/or-vt-ccdf-immigration.fixed.md (.fixed) present
CI	PASS	33/33 green

Review severity: APPROVE

No critical issues. Both effective dates and the formula branching correctly reflect the 2024
OR/VT rule changes, no other eligibility condition is disturbed, the refactor is value-neutral,
and CI is green. The VT items are a reference-durability improvement and style/cosmetic
suggestions — none blocking.

Next steps

Run /fix-pr 9359 to apply the should-address items (VT durable citations; optional vectorized
where()) and, if desired, the suggestions (VT 2024-06-30 encoding, description naming, two
regression cases). Rebase on main (branch is 6 behind) before merge.

🤖 Generated with Claude Code /review-program

[DTrim99]

Fix minor nits, then ready