State Subsidy Programs

Supplementary data for states with their own premium assistance

Overview

Most states rely solely on federal APTC. A small number add state-funded premium subsidies on top. For these states, our calculation must account for both federal and state components.

Confirmed State Subsidies

California (CAPS + CAPC) ✅ Validated

Source: 2026 CA Premium Subsidy Program Design

Formula: Identical to federal APTC but with California's own (lower) applicable percentage table.

CAPS = (SLCSP - CA_contribution) - Federal_APTC
     = Federal_contribution - CA_contribution   (when CA% < Federal%)

California 2026 Applicable Percentage Table:

Income (% FPL)	Initial %	Final %
100% to 150%	0.00%	0.00%
150% to 165%	3.19%	3.91%
Above 165%	No state subsidy

CAPC: $1 per covered member per month (all enrollees, regardless of income)

Validation: Within $0.44-$8.29 of Covered California's displayed savings across 4 marketplace scenarios.

Non-Medicaid-Expansion FFM States — income-bump (ENG-414) ✅ Shipped

Not a state-funded subsidy — this section documents a federal-subsidy display fix specific to the 9 FFM states that never expanded Medicaid. It belongs in this doc because the fix is state-aware and the affected user population overlaps the "states requiring research" conceptual space (low-income residents who fall through standard ACA UX patterns).

The bug we fixed

In Medicaid-expansion states (UT, OH, MI, NC, AZ, etc.) a user at 0-138% FPL hits CMS with is_medicaid_chip: true → /api/eligibility re-queries CMS at the bumped income (~138.5% FPL) to surface the full APTC + CSR-94 result that the user would receive if Medicaid denies them. This is the "Medicaid death-loop" pattern from the prior Medicaid fix.

In the 9 non-expansion FFM states (TX, FL, AL, MS, SC, TN, WY, KS, WI) CMS never sets is_medicaid_chip: true for these users (their state didn't expand) — so the bump path never triggered. Result:

• <100% FPL users saw aptc=0, in_coverage_gap=true → a no-subsidy dead-end.
• 100-138% FPL users saw the partial APTC CMS calculated at their literal low income — meaningfully smaller than the full subsidy an expansion-state peer would see at the same income.

Both technically reflect what CMS returns at the literal income, but both fail the UX intent: show the user the full subsidy they qualify for upfront; navigate state-specific qualification details (Medicaid vs marketplace) later in the funnel.

The fix (PR #543, deployed 2026-06-01)

Widened the income-bump trigger in apps/web/src/app/api/eligibility/route.ts (FFM proxy branch) to also fire when NON_EXPANSION_FFM_STATES.has(place.state) && fplPercent < 138.5. Reuses 100% of the existing calculateMedicaidThreshold math and CMS re-query mechanism. Adds:

• NON_EXPANSION_FFM_STATES constant in packages/shared/src/constants/index.ts — Set { TX, FL, AL, MS, SC, TN, WY, KS, WI }. GA is excluded even though it never expanded Medicaid — for 2026 GA runs its own marketplace (Georgia Access), so the FFM proxy branch never runs for GA users (the SBE redirect fires first).
• is_non_expansion_bump sibling flag in the /api/eligibility response, mutually exclusive with the is_medicaid_chip-driven bump. Plumbed through fetch-plans.ts → use-calculator → LandingCalculator.
• Distinct takeover banner copy when isNonExpansionBump=true: "Your income qualifies for the highest ACA subsidy" + body that names the state didn't expand Medicaid and frames the marketplace subsidy as the federal entitlement it is (instead of the misleading "Medicaid threshold" wording reused from the expansion-state path).

Verified apex behavior (2026-06-01, after deploy)

Scenario	Before	After
TX 75201 single 35M $14K (~92% FPL — coverage gap)	aptc=$0, dead-end	$21.25/mo Silver, APTC $566, CSR-94, is_non_expansion_bump=true
FL 33101 single 35M $17K (~109% FPL)	aptc=$0, dead-end	$27.47/mo Silver, APTC $580, CSR-94
AL 35036 single 35M $21K (~134% FPL)	partial APTC at literal income	$10.58/mo Silver, APTC $552, CSR-94
CONTROL UT 84094 single 35M $21K (existing Medicaid path)	bumped via is_medicaid_chip	Unchanged — is_non_expansion_bump=false, shape identical
CONTROL TX 75201 single 35M $30K (~190% FPL, above bump band)	normal subsidy passthrough	Unchanged — no is_medicaid_adjusted flag (bump correctly did NOT trigger)

Regression coverage

Locked into scripts/audit/calculator-baseline.ts as scenarios #15 (TX $14K), #16 (FL $17K), #17 (AL $21K). All 14 original scenarios continue to pass ZERO DIFFS against the captured fixture — the fix is precisely scoped to the previously-unhandled <138% FPL band in non-expansion FFM states.

Non-expansion states are NOT eligible for state subsidies

None of the 9 non-expansion states run their own state-funded premium subsidy program (CA's CAPS+CAPC, etc.). The "subsidy" surfaced here is federal APTC + CSR-94 — ACA §1401 — which they were always entitled to; we just weren't displaying it properly. So these states do NOT need an entry in the state_subsidies MongoDB collection.

Refresh cadence

Re-verify the membership of NON_EXPANSION_FFM_STATES annually with the plan-year refresh. NC expanded in 2024, SD in 2023 — both are excluded from the set today. Authoritative source: https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions/

States Requiring Research

State	Program	Status
New York	Essential Plan / Basic Health Program	❌ Need to research
Colorado	Colorado Option	❌ Need to research
Washington	Cascade Care	❌ Need to research
Connecticut	State subsidy program	❌ Need to research
Others	Most SBE states have NO state subsidy	Federal APTC only

Implementation

State subsidy data stored in MongoDB state_subsidies collection:

{
  "state": "CA",
  "plan_year": 2026,
  "program_name": "CAPS + CAPC",
  "applicable_pct_table": [
    {"min_fpl": 1.00, "max_fpl": 1.50, "initial_pct": 0, "final_pct": 0},
    {"min_fpl": 1.50, "max_fpl": 1.65, "initial_pct": 0.0319, "final_pct": 0.0391}
  ],
  "per_member_credit": 1.00,
  "source_url": "https://board.coveredca.com/...",
  "last_verified": "2026-04-07"
}

For states without entries, federal APTC is the complete answer.