Tests: H3.1 (direct-care turnover/vacancy constrains HCBS capacity more tightly than waiver funding does — the childcare-H4.2/housing-H4.1 rhyme, and the protocol's §12 kill condition for this workstream) and H3.2 (immigration restriction as a first-order supply constraint).
Process note carried into full execution: neither hypothesis was pre-registered with M3's refuted if / supported if / indeterminate if criteria before this pass — the seed protocol states the hypotheses but not their adjudication conditions, unlike housing's research/ws03-adjudication-criteria.md. This pass reports what the evidence actually shows and labels the verdict conservatively rather than retrofitting criteria to match the result; a pass-2 job is to write H3.1/H3.2 adjudication criteria before any further evidence collection, per M3.
H3.1: workforce vs. funding as the binding constraint — INDETERMINATE
The best available natural experiment is ARPA §9817 (2021): a temporary 10-point FMAP increase earmarked for HCBS, spent almost entirely on the workforce side.
- 37.1Bplannedacrossallstates + DC * *(MACPAC, Jan2026issuebrief), * * 7126.3B) to workforce recruitment/retention — provider rate increases, wage add-ons, hiring/retention bonuses. Only 5% ($1.7B) targeted waiting-list reduction directly, and only three states (CA, NM, TX — which together hold over half of all waitlisted individuals) specifically targeted that category.
- Wages moved substantially where states used the money that way: Colorado's direct-care hourly wage rose from 12.41to 18; national median home care worker wage rose $13.07 (2014, inflation-adj.) → $16.77 (2024), with commentary attributing part of that rise to §9817.
- Vacancy/turnover did not improve commensurately. KFF's 2023 50-state survey found every responding state still reported workforce shortages despite near-universal rate increases; Pennsylvania's direct-care turnover held at 44–65% with a projected 37,000-worker shortfall by 2026.
- Waiting lists did not shrink at the national level. KFF's tracker shows ~692,679 people waitlisted in 2021 growing to ~710,000+ by 2024 (+2.6% 2023→2024) — the opposite direction from what new funding aimed at the problem should produce if funding, not workforce, were binding. (Note: this is higher than the protocol's anchor-8 seed prior of "roughly 500,000+" — anchor 8 should be corrected to this range when it's filled in, not left at the lower seed figure.)
- No source found tracks hours of care delivered per enrollee in a way comparable across states — the single most direct measure of whether money bought more actual care is not being collected anywhere.
Verdict: the pattern — money went overwhelmingly to wages, wages rose, and neither vacancy/turnover nor waiting lists improved — is compatible with H3.1, but it is not a test of H3.1. Economy-wide post-pandemic wage inflation can produce the same wages-up/vacancies-flat pattern without HCBS-specific content. Per MACPAC's January 2026 assessment, §9817 carried no evaluation requirement, states kept no mandated common baseline, and states running multiple concurrent workforce initiatives "encountered difficulty isolating the impact of each initiative." A federally commissioned national evaluation (GSA/American Institutes for Research) was still pending as of that brief. Call this INDETERMINATE, not SUPPORTED or directionally suggestive. Revisit when a comparison sector or the federal evaluation supplies a macro-confound control.
H3.2: immigration restriction as a first-order supply constraint — SUPPORTED ON EXPOSURE, NOT YET ON MEASURED EFFECT
- Foreign-born share: 28% of the direct-care workforce nationally (PHI, ACS-based, up from 21% in 2011); home health aides specifically at 36.5% (Migration Policy Institute/American Immigration Council, 2019 ACS) — more than double the foreign-born share of the overall US workforce as of that year (17%), with 6.9% undocumented in that occupation. Dual as-of note added on Phase 1 verification (2026-08-10, V6): the 36.5% is 2019 data. Against the current base (BLS puts the foreign-born share of the US labor force at 19.2% in 2024) the ratio is ~1.9×, not "more than double." Cite the figure with its year, and say "roughly twice" rather than "over twice" for any present-tense framing.
- Regional concentration is extreme relative to the national figure: ~69–74% of NYC's home health/personal care aide workforce is immigrant; California's health workforce ~35% immigrant; a Chicago-metro estimate puts "more than half" of home care workers as non-citizens. A national exposure number understates the constraint badly in the highest-need metros.
- Documented 2025–2026 policy changes: DHS terminated the "sensitive locations" enforcement-exemption policy for healthcare facilities (Jan 21, 2025); an executive order (Feb 19, 2025) threatens federal payments to entities seen as facilitating sanctuary protections; TPS and humanitarian parole terminations for several nationalities with documented direct-care employment followed through 2025–2026.
- Actual labor-supply effect: anecdotal only, not yet quantified nationally. Chicago local reporting (WTTW/WBEZ, Aug 2025) documents named clients losing home-care coverage as workers stop showing up out of fear of enforcement, independent of any actual enforcement action against them. Justice in Aging's own Feb 2026 fact sheet states explicitly that it found no documented data on actual staffing disruptions or enforcement actions — it's a preparedness guide, not an effects study. EPI's "400,000 jobs threatened" figure is a modeled projection, not an observed outcome, and should be cited as such, not as a measured loss.
Verdict: the exposure half of H3.2 is strongly supported — this is a workforce with a foreign-born share more than double the general labor force, concentrated even higher in the metros that need the most HCBS capacity. The effect half is currently anecdotal-only: real policy changes, real fear-driven absenteeism reported locally, but no hard national participation-rate data yet shows a measured decline. Report both halves separately in the whitepaper — collapsing "exposure is high" and "harm is already measured" into one sentence would overstate what's known.
Sources
MACPAC, "Implementation of ARPA Section 9817" (Jan 2026); KFF, "Payment Rates for Medicaid HCBS: States' Responses to Workforce Challenges"; KFF, "A Look at Waiting Lists for Medicaid HCBS from 2016 to 2024/2025"; ADvancing States, "Efforts to Evaluate the Impact of ARPA HCBS Investments" (Feb 2024); PHI, "Immigration and the Direct Care Workforce" / "Direct Care Workers in the United States: Key Facts 2025"; Migration Policy Institute / American Immigration Council (2019 ACS occupation data); Justice in Aging fact sheet (Feb 2026); WTTW/WBEZ Chicago reporting (Aug 2025); EPI blog (labeled as a projection, not a finding).
Confidence tiers
H3.1: Low (spending/wage data is high-confidence; the capacity-outcome linkage is correlational and has an unresolved macro-wage confound). H3.2 exposure: medium-high. H3.2 effect: low-medium (anecdotal, no national participation-rate data yet).