Skip to content

Zero Medicaid federal share for state-funded immigrant coverage - #10950

Merged
MaxGhenis merged 5 commits into
mainfrom
fix-medicaid-state-funded-federal-share
Nov 21, 2026
Merged

MaxGhenis merged 5 commits into
mainfrom
fix-medicaid-state-funded-federal-share

Conversation

@MaxGhenis

@MaxGhenis MaxGhenis commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

Summary

medicaid_federal_share gave a Title XIX match to enrollees who reach Medicaid only through state-funded coverage. An undocumented adult in California, an adult Illinois Health Benefits for Immigrants (HBI) enrollee or an undocumented adult in DC got the 90% expansion FMAP. Children in those routes and in New Jersey's Cover All Kids got the regular FMAP. Their federal share now falls to 0, which generalizes the existing Rhode Island postpartum case.

What went wrong

The formula gave the category-based FMAP to every enrollee in a Medicaid category, whatever their immigration status, except Rhode Island postpartum enrollees who failed is_medicaid_immigration_status_eligible. That immigration variable also counts state-funded coverage: the undocumented_immigrant parameter (DC, MA, NY, and OR from 2023-07), is_ca_medicaid_immigration_status_eligible (California's undocumented, DACA and TPS residents) and regardless_of_immigration_status.child_age_limit. That last parameter holds the age limit for New Jersey's Cover All Kids, which has covered children under 19 of any status since January 1, 2023. Separately, is_medicaid_eligible ORs in ca_ffyp_eligible and il_hbi_eligible with no immigration test, and medicaid_category ignores immigration status.

Fix

The formula now uses is_federal_medicaid_immigration_status_eligible. Every enrollee outside a federally matched route (category NONE, or failing the federal immigration test) gets 0. Everyone else keeps the existing category-based FMAP. That condition replaces the Rhode Island term, and its four existing cases in medicaid_federal_share.yaml keep their expected values.

Several provisions together limit Title XIX payment for noncitizens:

  • 42 U.S.C. 1396b(v)(1) bars payment for an alien "who is not lawfully admitted for permanent residence or otherwise permanently residing in the United States under color of law", except under paragraphs (2), (4) and (5).
  • 8 U.S.C. 1611 limits federal public benefits to qualified noncitizens, and 8 U.S.C. 1613(a) bars most qualified noncitizens for five years after entry. A permanent resident inside the bar meets this limit.
  • From October 1, 2026, § 1396b(v)(5) limits payment to citizens and nationals, permanent residents, Cuban and Haitian entrants and Compact of Free Association migrants.
  • The (v)(4) lawfully residing option for children and pregnant people applies "notwithstanding sections 1611(a), 1612(b), 1613, and 1631 of title 8", and the federal test already includes it.

Military exemption added to the federal test

Once the federal test decides the match on its own, a gap in it matters, and the round-1 review found one. 8 U.S.C. 1613(b)(2) exempts from the five-year bar a veteran or active duty member who holds a qualified status, and their spouse or unmarried dependent child, and § 1612(b)(2)(C) protects their Medicaid eligibility. The helper left them out, so an Illinois permanent resident on active duty, three years after entry, would have enrolled through HBI and gotten share 0 instead of 0.9.

is_federal_medicaid_immigration_status_eligible now counts that exemption through is_snap_alien_military_connected, which already implements 1613(b)(2) for SNAP from is_veteran, is_military, the marital unit and tax-unit dependency. Its documentation lists those proxies and how they over-grant. Neither 1613(b)(2) nor 1612(b)(2)(C) sets ages for a dependent child, so the model borrows SNAP's (7 CFR 273.4(a)(6)(ii)(G)(3): under 18, a full-time student under 22, or disabled), without testing when a disability began. Medicaid, the Medicare Savings Programs and Nevada LIHEAP now share that variable and its parameters, and their documentation says so: an override or reform aimed at SNAP changes those programs too. The exemption still requires a qualified status, so from October 2026 it applies only to the (v)(5) statuses.

State-funded routes into medicaid_enrolled

Route How it enrolls Funding source Share before After
California state-funded Medi-Cal for undocumented, DACA, TPS is_ca_medicaid_immigration_status_eligible inside the immigration test Cal. Welf. & Inst. Code § 14007.8(f): DHCS claims FFP where available and otherwise uses "state funds appropriated for this purpose" 0.9 adult, regular FMAP for others 0
California Former Foster Youth Program, for people who fail the federal test ca_ffyp_eligible OR'd into eligibility § 14007.8(b)(3), (j)(4), (k)(2): former foster youth "who but for their immigration status are eligible" under § 14005.28 keep full scope to 26, funded under (f)(2) 0.9 0
Illinois HBI (All Kids for immigrant children, HBIA 42–64 until 2025-07-01, HBIS 65+) il_hbi_eligible OR'd into eligibility 305 ILCS 5/12-4.35 covers noncitizens not eligible under Article V because of § 1-11; HFS calls HBIS a "state-funded program"; HFS's HBIA FAQ gives a program end date of June 30, 2025 0.9 adult, 0.5138 child 0
States in the undocumented_immigrant parameter: DC, MA, NY, OR inside the immigration test 1396b(v)(1) bars a Title XIX match. The programs: the DC Health Care Alliance, for residents who can't get Medicaid, New York's Child Health Plus for undocumented children, and ORS 414.231 for adults who "would be eligible for medical assistance but for the adult's immigration status" 0.9 adult, regular FMAP for others 0
New Jersey Cover All Kids, children under 19 regardless of immigration status regardless_of_immigration_status.child_age_limit inside the immigration test New Jersey's FY2023 to FY2027 appropriations acts let the Department of Human Services enroll children who meet every rule but immigration status, and the Department began enrolling them on January 1, 2023, with state funds only (sources in child_age_limit.yaml) 0.5 0
Rhode Island postpartum continuation outside the pregnancy group ri_medicaid_state_funded_postpartum_eligible already 0 0 0

Federally matched routes for people who fail the immigration test

  • Lawfully residing children and pregnant people (1396b(v)(4)) pass is_federal_medicaid_immigration_status_eligible, so they keep the regular FMAP. The new California test checks a TPS child at 0.5. The model records New Jersey as electing the lawfully residing option for children under 21. Its permanent resident, TPS and other lawfully residing children pass the federal test and keep the regular FMAP, while undocumented and DACA children who reach Medicaid only through Cover All Kids get 0.
  • Emergency Medicaid (1396b(v)(2)) gets a match, but it isn't a route into medicaid_enrolled. is_emergency_medicaid_eligible has no consumer in the model, and it requires failing the combined immigration test. California directs DHCS to claim federal participation where available (§ 14007.8(f)(1)), and Illinois HFS calls emergency services for these residents "federally matchable". medicaid_cost holds one cost per enrollee with no split by service, so this PR attributes none of it to the federal government. The variable's documentation says so. Since 1396d(kk) (Pub. L. 119-21 § 71110), that match can't exceed the regular FMAP from October 1, 2026.
  • California FFYP for people who pass the federal test falls in the Title XIX former foster care group (42 CFR 435.150). This PR doesn't zero it: an FFYP citizen in the ADULT category keeps 0.9, and the new test checks that.
  • The enacted Pub. L. 119-21 amends 1396d only in §§ 71110, 71112 and 71114. It has no expansion-FMAP cut for states that cover undocumented immigrants.

I didn't add the work-requirement result to the federally matched test. The only people the model can enroll after a work-requirement failure come through FFYP or Rhode Island postpartum. 42 U.S.C. 1396a(xx)(9)(A)(ii) exempts former foster youth (subclause (I)) and pregnant or postpartum people (subclause (IX)), so requiring compliance would wrongly zero matched enrollees.

Tests

  • Four new YAML cases in medicaid_federal_share.yaml:

    • California: undocumented adult, five-year-bar LPR reached only through FFYP, DACA child → 0; citizen adult and citizen FFYP adult → 0.9; TPS child → 0.5. The case also checks federal and state costs for two of them.
    • Illinois HBI: undocumented adult and child → 0; citizen and LPR past the bar → 0.9.
    • DC: undocumented adult → 0; citizen → 0.9.
    • Illinois military exemption (2024): a permanent resident on active duty and one married to a veteran, both three years after entry → federal test true, not HBI, 0.9; the same resident without the connection → HBI, 0.

    The CA, IL and DC cases fail on main's formula, and the 17 existing cases pass both before and after the change.

  • Six cases in is_medicaid_immigration_status_eligible.yaml, numbered 23 to 28 after Cover New Jersey children under 19 in Medicaid and CHIP regardless of immigration status (Cover All Kids) #10987's four New Jersey cases:

    • Texas 2025: a recent permanent resident on active duty and a recent permanent resident veteran pass the federal test; a TPS adult on active duty doesn't, since the exemption needs a qualified status.
    • Texas 2027, after the (v)(5) narrowing: a permanent resident on active duty still passes; a refugee on active duty fails.
    • Alabama 2025, which doesn't elect the lawfully residing option: the 10-year-old permanent-resident dependent of an active-duty parent passes through the exemption; a 19-year-old dependent who isn't a student or disabled doesn't.
  • test_medicaid_federal_share_invariants.py (Hypothesis, 8 derandomized examples of up to 40 people across 10 states, 11 statuses, 13 categories and 2024–2027) checks every person. The share lies in [0, 1], and it equals 0 for category NONE or a failed federal test. Categories other than ADULT and NONE get the regular FMAP, an ADULT gets the regular or the expansion FMAP, and federal plus state cost equals medicaid_cost. It fails on main's formula. Locally it takes 16–36 s with 2.3 GB peak RSS. As a root-level test file it falls into the remaining Rest group.

  • Local runs at 6f21b53 used policyengine-core 3.34.2, which uv.lock pinned before the merge with main, and at most four processes. uv.lock now pins 3.36.0.

    • tests/policy/baseline/gov/hhs/medicaid: 1,277 passed.
    • Consumers of the immigration tests (hhs/medicare/savings_programs, IL HBI, CA FFYP, is_aca_ptc_eligible, Healthier Oregon, Mississippi working disabled): 311 passed.
    • Every other test file that sets is_military or is_veteran with an immigration status (SNAP immigration and unit size, MO SSP, Utah HEAT, and the NH, ND, NV, SC, WA and WY energy programs): 926 passed.
    • The property test: passed.
    • At 27f6c36, after the round-2 tests and notes: is_medicaid_immigration_status_eligible.yaml, hhs/medicaid/costs and hhs/medicare/savings_programs, 209 passed; SNAP immigration and the Nevada and Oklahoma LIHEAP tests, 326 passed.
  • No partner test names these variables, and this PR edits none.

Pull request run 44908242039 at 11336aa108 passed all 38 jobs on GitHub's merge with main a5f0f07c98.

Interaction with #10891

#10891 edits the same docstring, reference list and formula, so whichever merges second has a small conflict in both files. To resolve it, keep #10891's docstring up to its Rhode Island sentence (through "so the expansion FMAP stops at 133 percent of the poverty line there."). Then append this PR's text from "Title XIX pays only…", which restates the Rhode Island rule in general form. Keep both reference sets. Keep #10891's expansion_adult block, then no_federal_medicaid_match = ~federally_matched_route. For the YAML, take #10891's file plus this PR's appended cases. The new tests set medicaid_income_level: 1 and household size 1, so they hold under #10891's income test. Before the merge with main, on a tree merged with #10891 at edc1f51, the resolution matched the one at 42ec388 except for line offsets. The Medicaid costs tests and is_medicaid_immigration_status_eligible.yaml passed 91/91, and the property test and #10891's test_medicaid_expansion_fmap_income_limit.py (17) passed.

Impact

For enrollees in state-funded coverage, federal cost falls to 0 and state cost rises by the same amount, and total medicaid_cost stays the same. The military exemption changes eligibility. Veterans, active duty members and their spouses and dependent children who hold a qualified status inside the five-year bar now pass the federal test in every state. That holds for Medicaid and for the Medicare Savings Programs, which use the same test. In Illinois such a person moves from HBI to regular Medicaid.

Known biases in the federal test

These come from is_federal_medicaid_immigration_status_eligible as it stands on main, and this PR now leans on it as the funding gate.

  • It takes a YEAR period and reads the status list at January 1, so the § 1396b(v)(5) narrowing that starts October 1, 2026 applies only from 2027. For October–December 2026, the model overstates federal cost for people in statuses that (v)(5) drops, such as refugees and asylees. The model's 2026-10-01 list keeps only citizens, permanent residents and Cuban and Haitian entrants.
  • It assumes every state covers every optional qualified noncitizen, those outside the mandatory groups of 8 U.S.C. 1612(b)(2), though § 1612(b)(1) lets each state decide. In a state that opts out, the model overstates enrollment and federal cost.
  • The military proxies mostly over-grant the exemption (see is_snap_alien_military_connected), which overstates Medicaid and MSP eligibility and federal cost. They also miss unremarried surviving spouses (1613(b)(2)(C)) and the children of deceased veterans that SNAP's rule names.

Invariants

  • medicaid_federal_cost + medicaid_state_cost == medicaid_cost for every enrollee.
  • 0 <= medicaid_federal_share <= 1.
  • medicaid_federal_share == 0 whenever not is_federal_medicaid_immigration_status_eligible or medicaid_category == NONE.

Follow-ups (not in this PR)

Microsimulation impact

We estimate that this change moves $4.85bn of Medicaid spending from the federal government to the states in 2025, and $1.82bn in 2026, on the default dataset. Total Medicaid spending, enrollment and household net income stay the same. We compared this branch at 11336aa108 with main at 2574529521, its merge base.

Output 2025 main 2025 change 2026 main 2026 change
medicaid_federal_cost $619.2bn -$4.85bn $617.5bn -$1.82bn
medicaid_state_cost $299.5bn +$4.85bn $301.2bn +$1.82bn
medicaid_cost $918.7bn $0 $918.7bn $0
medicaid_enrolled 71,259,156 people 0 70,367,964 people 0
is_medicaid_eligible 76,172,926 people 0 75,203,072 people 0
household_net_income $16,440.1bn $0 $17,246.0bn $0

The shift comes from 291 sampled people in 2025 (466,836 weighted) and 200 in 2026 (147,199 weighted), whose medicaid_federal_share falls to 0. No record's is_medicaid_eligible or is_medicaid_immigration_status_eligible changes in either year.

axiom: us/statute/42/1396b/v encoded-correct (alien_described_in_payment_bar; companion test auto_positive_alien_described_in_payment_bar in us/statutes/42/1396b/v.test.yaml); TheAxiomFoundation/rulespec-us#2331 queued (8 U.S.C. 1613(b)(2) and 1612(b)(2)(C), the military exemption from the five-year bar and its Medicaid protection). The FMAP cost split itself stays microsim-only.

🤖 Generated with Claude Code

MaxGhenis and others added 2 commits November 13, 2026 19:07
medicaid_federal_share treated anyone who passed
is_medicaid_immigration_status_eligible as federally matched. That
variable also counts state-funded coverage (California's state-funded
Medi-Cal, states flagged in the undocumented immigrant parameter), and
is_medicaid_eligible adds California FFYP and Illinois HBI without any
immigration test. So undocumented, DACA and five-year-bar enrollees got
the 90% expansion FMAP or the regular FMAP.

42 U.S.C. 1396b(v)(1) bars Title XIX payment for anyone who fails the
federal immigration test, apart from emergency services ((v)(2)) and the
lawfully residing option ((v)(4), already in the federal test). The
formula now uses is_federal_medicaid_immigration_status_eligible and
gives no federal share to any enrollee without a federally matched
route, which generalizes the existing Rhode Island postpartum case.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
For generated enrollees across states, immigration statuses, categories and
years, the federal share is 0 whenever the federal immigration test fails
or the category is NONE, is the regular FMAP for other categories, and
federal plus state cost equals medicaid_cost. The test fails on the
previous formula, which matched state-funded enrollees.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
MaxGhenis and others added 2 commits November 15, 2026 09:49
The federal Medicaid immigration test now gates the federal share, so it
needs the 8 U.S.C. 1613(b)(2) exemption for qualified noncitizens who are
veterans, on active duty, or their spouses or dependent children. It reuses
is_snap_alien_military_connected, which implements the same exemption.

Also cite 42 U.S.C. 1396b(v)(1), (4) and (5) and 8 U.S.C. 1611 and 1613
for their separate limits instead of attributing all of them to (v)(1).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…children

Adds 2027 cases (a permanent resident on active duty passes, a refugee on
active duty fails) and an Alabama dependent-child case. Notes in the SNAP
military variable and parameters that Medicaid, the Medicare Savings
Programs and Nevada LIHEAP share them, and says the dependent-child ages
are borrowed from SNAP since federal Medicaid law sets none.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@MaxGhenis

Copy link
Copy Markdown
Contributor Author

Hand-off to the US + core hub.

  • Head: 27f6c36. Not a draft, mergeable.
  • Review (Subfleet, review/hard):
    • Round 1, at d10e541: REQUEST_CHANGES. One P1: the federal immigration test lacked the 8 U.S.C. 1613(b)(2) military exemption. Two P2s: citations, and the NY follow-up. All fixed in 6f21b53.
    • Round 2, at 6f21b53: APPROVE, no P0 or P1, so the review has finished. Its P2s were missing 2027 and dependent-child tests, and the shared SNAP variable needed notes. Both P3s were wording. All four are fixed in 27f6c36. The out-of-scope CHIP gap is a follow-up in the PR body.
  • Passed locally (policyengine-core 3.34.2, matching uv.lock):
    • At 6f21b53: gov/hhs/medicaid, 1,277; immigration-test consumers (MSP, IL HBI, CA FFYP, ACA PTC, Healthier Oregon, MS WD), 311; other files that set military flags with an immigration status, 926; property test.
    • At 27f6c36: 209 and 326 on the touched areas.
  • Pending: CI at 27f6c36 is queued. At d10e541 it passed: 37 pass, 3 skipping.
  • Depends on: merge order with Give partial-expansion Medicaid adults the regular FMAP and add FY2027 FMAPs #10891. Each conflicts with the other in medicaid_federal_share.py and its YAML. The PR body gives the resolution. On Give partial-expansion Medicaid adults the regular FMAP and add FY2027 FMAPs #10891 at 42ec388 it passes 88/88 plus the property test.
  • Decisions and holds: none. Not merged.
  • Follow-ups owed (all listed in the PR body):
    • FFYP precedence and the regular FMAP;
    • MA/NY undocumented_immigrant over-enrollment, keeping NY pregnancy and postpartum coverage;
    • a possible Oregon double count;
    • the CHIP five-year-bar military exemption;
    • a program-neutral name for the military-connection variable.

Keep main's New Jersey Cover All Kids cases 19 to 22 in
is_medicaid_immigration_status_eligible.yaml and renumber this
branch's six military-exemption cases to 23 to 28.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@MaxGhenis MaxGhenis closed this Nov 21, 2026
@MaxGhenis MaxGhenis reopened this Nov 21, 2026
@MaxGhenis
MaxGhenis merged commit fccedbc into main Nov 21, 2026
77 checks passed
@MaxGhenis

Copy link
Copy Markdown
Contributor Author

US + core hub merge audit, us#10950 at 11336aa108713b90b309e3f0cb6df845d0c4b700 (squash): Zero Medicaid federal share for state-funded immigrant coverage.

  • Reviews:
    • The PR has no GitHub reviews. Two independent review rounds ran outside GitHub, with a fix commit after each. Neither round read the second fix commit, 27f6c36ea2, or the merge of main that followed. A separate check read that merge.
    • The first round, at d10e54134a, read the change against 42 U.S.C. 1396b(v), 8 U.S.C. 1612 and 1613, Public Law 119-21 and the California, Illinois and Oregon sources. A formula-only reproduction ran. The round stopped the Medicaid costs suite after 586.9 seconds, still in parameter loading, before any assertion ran.
    • It asked for changes. Its blocking finding: the federal test lacked the military exemption of 8 U.S.C. 1613(b)(2). An Illinois permanent resident on active duty, three years after entry, would enroll through Health Benefits for Immigrants (HBI) with share 0 instead of 0.9.
    • It also found that the comments credited the whole limit to 1396b(v)(1), and that the New York follow-up left out pregnant and postpartum residents. Outside the PR's scope, it asked the description to state the federal test's two upward biases.
    • 6f21b53488 answered each finding. It added is_snap_alien_military_connected to the federal test, inside the qualified-status term, with an Illinois federal share case and three Texas cases. It rewrote the comments to cite (v)(1) and (5) and 8 U.S.C. 1611 and 1613. The description gained the New York pregnancy coverage and a section on the federal test's known biases.
    • The second round raised no blocking finding at 6f21b53488, and the review ended there. It had no shell and ran no test. It read the code, the statutes and CMS letter SHO make a test for max snap benefit, add to existing tests #26-001, and it found that the exemption lifts only the five-year bar, holds after the October 2026 narrowing and adds no dependency cycle.
    • It listed two P2 findings: no 2027 or dependent-child case, and no note that Medicaid now shares SNAP's variable and age parameters. It listed two P3 wordings, in the description and the federal test's documentation. Outside the PR's scope, it named the CHIP immigration test's missing exemption and the lack of a COFA status.
    • 27f6c36ea2 answered all four. It added the 2027 permanent-resident and refugee cases and the Alabama dependent-child case, and notes in is_snap_alien_military_connected.py and both SNAP age parameters. It changed tests, documentation and parameter comments in five files, and no formula. The description took both wordings and listed the CHIP gap as a follow-up.
    • 11336aa108 merges main at 2574529521. Cover New Jersey children under 19 in Medicaid and CHIP regardless of immigration status (Cover All Kids) #10987 had appended four New Jersey Cover All Kids cases to is_medicaid_immigration_status_eligible.yaml, where this PR appends six. The merge keeps main's file and renumbers this PR's cases 23 to 28, with no other change, and the merged tree differs from main in this PR's nine files alone. With main, uv.lock moves policyengine-core from 3.34.2 to 3.36.0.
    • The check of that merge found nothing to change. It rebuilt the resolution two ways and matched the commit's blob byte for byte. It recomputed by hand this PR's six cases, the four federal share cases and every New Jersey noncitizen-child case in the tree under the combined code. It also traced Require a Medicaid adult group parent's children to have coverage (42 CFR 435.119(c)), with New Jersey's under-21 age #11012's child-coverage path and Build omitted household marital units from explicit couples and the tax-role fallback #10000's marital-unit hook through the PR's cases. Its five findings corrected notes, not code.
    • Before merge, the description names the New Jersey route and corrects the Rhode Island count to four cases. It also updates what the merge made stale: the case numbers, the uv.lock pin and the CI history at 27f6c36ea2.
  • Methodology choices for maintainers:
    • The federal share follows is_federal_medicaid_immigration_status_eligible alone. An enrollee who fails it gets 0 in every category, and the PR attributes no emergency Medicaid match to the federal government, because medicaid_cost holds one cost per enrollee with no split by service.
    • The military exemption reuses SNAP's is_snap_alien_military_connected and its age parameters. Neither 1613(b)(2) nor 1612(b)(2)(C) sets ages for a dependent child, so Medicaid takes SNAP's: under 18, a full-time student under 22, or disabled, without a test of when a disability began. A change aimed at SNAP's variable or parameters now changes Medicaid, the Medicare Savings Programs and Nevada LIHEAP too.
    • The proxies is_veteran and is_military mostly over-grant the exemption, and the model has no unremarried surviving spouse. Both points sit in the variable's documentation and the description.
    • A New Jersey child who reaches Medicaid only through Cover All Kids now gets share 0, where main gives New Jersey's FMAP of 0.5. The PR reads the program as state-funded from the sources in child_age_limit.yaml. Lawfully residing New Jersey children under 21 pass the federal test and keep 0.5.
    • The work-requirement result stays out of the federally matched test, since 42 U.S.C. 1396a(xx)(9)(A)(ii) exempts former foster youth and pregnant or postpartum people.
    • California former foster youth who pass the federal test keep the category-based FMAP, 0.9 in the adult group, though 42 CFR 435.119(b)(4) points to the regular FMAP. A follow-up handles it.
    • The federal test reads its status list at January 1, so the (v)(5) narrowing applies from 2027, and it assumes every state covers optional qualified noncitizens. The PR keeps both and states the bias in its description.
  • Gates checked live at merge:
    • gh pr checks exits 0;
    • the latest Pull request run at the head, 44908242039, passed and tested this head merged into main at a5f0f07c98;
    • MERGEABLE;
    • not a draft;
    • no reviewer's latest review requests changes;
    • the semantic-overlap guard: since a5f0f07c98, main has changed no area this PR touches. This head merged main at 2574529521. After that, main took Date California Medi-Cal expansions for undocumented residents at their effective dates #11125, which adds two California cases to is_medicaid_immigration_status_eligible.yaml, the only file both sides change, and Add Arkansas SNAP broad-based categorical eligibility from April 4, 2025 #11124, which adds Arkansas cases to is_snap_eligible.yaml in an area this PR shares. The run above tested this head merged into a5f0f07c98, which holds both. git merge-tree --write-tree of the head into main at a5f0f07c98 returns tree 08e849c295 with no conflict. That tree differs from main in this PR's nine files alone: eight equal the head's blobs, and the YAML keeps main's 24 cases unchanged and in order and appends this PR's six as cases 23 to 28;
    • the merge pins the head.
  • Impact: We estimate the change moves $4.85bn of Medicaid spending from the federal government to the states in 2025 and $1.82bn in 2026 on the default dataset, against main at 2574529521. medicaid_federal_share falls to 0 for 291 sampled people in 2025 (466,836 weighted) and 200 in 2026 (147,199 weighted). Total Medicaid spending, enrollment, Medicaid eligibility, the immigration test and household net income don't change in either year.
  • Next:
  • Axiom: us/statute/42/1396b/v encoded-correct, unchanged at rulespec-us main 25ddba37d0. Encode the veteran and active duty exception to the five-year bar and its Medicaid protection (8 U.S.C. 1613(b)(2) and 1612(b)(2)(C)) (PE-US #10950) TheAxiomFoundation/rulespec-us#2331 queued, a new pe-parity issue for 8 U.S.C. 1613(b)(2) and 1612(b)(2)(C), with companion rows from this PR's YAML. The signed encoder can't run yet. The corpus release that rulespec-us pins has 23 US statute scopes, and the 22 with inventories in git hold no row for the title 38 sections that define a qualifying veteran. 1613.yaml defers its subsection (a) output on those definitions. The runtime also supplies no spouse or dependent-child relation. Encode Medicare and Medicare Savings Program immigration tests (42 USC 1395o(a), 42 USC 1395mmm, 8 USC 1611, 42 USC 1396b(v)) (PE-US PR 10632) TheAxiomFoundation/rulespec-us#2009 queued as it stands, since its rows set no military input.

🤖 Generated with Claude Code

@MaxGhenis
MaxGhenis deleted the fix-medicaid-state-funded-federal-share branch November 21, 2026 14:32
MaxGhenis added a commit that referenced this pull request Nov 21, 2026
Main now holds #10950 as one squashed commit; this branch carried it as
commits through 95b4e29. Git merges every path cleanly. The result is
main's tree plus this branch's six New York and Medicaid immigration
files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant