Builds on Phase 0's headline: the AARP/CBO/RAND valuation figures diverge for three named methodological reasons (scope, wage basis, and hours source) and cannot be collapsed to a single point estimate — see phase0-findings.md §1. This pass covers the rest of §8's scope: paying family caregivers directly, the Credit for Caring Act, and caregiver health/labor-market effects.
Cash & Counseling / self-direction: the genuine RCT, and what it actually found
The 1998–2003 Cash & Counseling Demonstration (AR/FL/NJ, RWJF/ASPE-funded, Mathematica-evaluated) is one of the only true randomized controlled trials in this entire policy domain — worth flagging precisely because it's so rare here. Findings, treatment (cash allowance, hire own workers including family) vs. control (traditional agency care):
- Satisfaction: consistently and substantially higher under treatment (e.g., AR 60.8% vs. 42.7% "very satisfied").
- Health/strain: treatment caregivers reported less physical strain and better self-rated health across all three states.
- The actual mechanism was unmet need, not efficiency. Control-group members frequently didn't receive services they were authorized for; the RCT's core finding is that self-direction closed that gap — it didn't do the same job cheaper, it did more of the job.
- Cost: higher under treatment in year 1 in most states (AR elderly +17%, FL nonelderly +14%, both significant), narrowing to non-significant in Arkansas by year 2 but persisting in FL/NJ.
Today: self-direction is available in all 50 states + DC via 1915(c)/(i)/(j)/(k) and 1115 authorities; ~1.5M people self-directed Medicaid HCBS in 2023, with usage among 65+ dual-eligibles ranging from 0.1% to 99.6% by state (a spread wide enough to be its own finding about measurement/reporting inconsistency, not just genuine policy variation).
Open question, explicitly flagged as untested by the researchers who'd know: does self-direction substitute for or complement the paid agency workforce §3 is about? The one study that comes closest (PMC12605748) explicitly calls this a hypothesis, not a finding — high self-direction uptake may partly reflect limited agency access rather than genuine preference, but nobody has tested it causally. This is a real gap for full execution, not a corner to round off: if self-direction's growth is partly agency-workforce-shortage-driven, that's a second, independent piece of evidence for §3's H3.1 (workforce capacity as the binding constraint) — but it needs its own dedicated pass, not an inference from one authors' aside.
Credit for Caring Act: no cost estimate exists, and that absence is a finding
Reintroduced in the current Congress as S.925 / H.R.2036 (March 2025, bipartisan) — the same bill, same fate as every prior version since 2016: referred to committee, no markup or floor action in any Congress it's been introduced in. Current parameters: nonrefundable credit, 30% of qualified caregiving expenses above a $2,000 floor, capped at $5,000; caregiver earned-income floor $7,500; phases out above 75Ksingle/150K joint MAGI.
No CBO/JCT score exists, which tracks — JCT typically only scores bills that reach markup, and this one never has. No independent think-tank static estimate was found either. Advocacy materials cite the size of the caregiving population and its imputed economic value, not a fiscal cost to Treasury. Per M1, "the data does not exist" is itself the finding here — any full-execution cost comparison between this bill and OAA/NFCSP expansion or Medicaid HCBS de-capping will have to build its own static estimate rather than cite one, and should say so explicitly rather than quietly filling the gap with an unsourced number.
For scale, using confirmed figures: NFCSP's entire FY2026 appropriation is ~$209M; Medicaid HCBS spending is on the order of $116–162B. A national caregiver credit at even modest uptake would very plausibly land between those two — bigger than NFCSP, smaller than HCBS — but that is an extrapolation stated as such, not a sourced number.
Caregiver health and labor-market effects: federal-sourced where it matters; trend needs confirmation
- Labor-force exit (AARP/NAC 2025): >25% of the 63M caregivers reduced work hours, 16% turned down a promotion, >15% left paid work temporarily — specifically due to caregiving.
- Lifetime cost (MetLife 2011, still the standard cited figure — dated, no clear successor study found): $324,044 for a female working caregiver (lost wages, Social Security, pension), $283,716 for men.
- Health effects (CDC BRFSS Caregiver Module, 2021 — a federal source, not survey-advocacy): caregivers show significantly higher poor-mental-health prevalence (19.6% vs. 12.9% non-caregivers) plus elevated depression, obesity, and smoking. CDC reported a widening gap from 2015–16 to 2021–22, but the endpoint falls inside the pandemic period; persistence after recovery has not been checked. This is suggestive federal evidence for the caregiver-relief case, independent of the contested dollar-valuation question.
Effect on the protocol
H8.1 (imputed value exceeds paid LTSS spending) is superseded by Phase 0's finding that the value isn't a single derivable number — report the caregiver-health and labor-force-exit evidence as the load-bearing case for caregiver relief instead of the dollar figure, since the CDC/AARP hours-and-exit data is far more solidly sourced than any valuation total. H8.2 (OAA/NFCSP scale) is confirmed and sharpened: three orders of magnitude below HCBS, not two. Add a new candidate finding for §11: Cash & Counseling is the one architecture in this entire space with actual RCT evidence behind it, which should weigh heavily in scoring relative to architectures resting on projections or single-state natural experiments.
Sources
ASPE, "How Cash and Counseling Affects Informal Caregivers"; Mathematica/PMC1955330 (cost effects); PMC12605748 (self-direction current scale); MACPAC June 2025 Ch. 5; congress.gov S.925/H.R.2036; AARP/Alzheimer's Association Credit for Caring fact sheets; AARP/NAC Caregiving in the US 2025; AARP Valuing the Invaluable (2026); MetLife Mature Market Institute (2011); CDC Healthy Aging Data, BRFSS Caregiver Module (2021); ACL/NFCSP budget data; KFF (Medicaid HCBS spending).
Confidence tiers
Cash & Counseling RCT findings: high (primary ASPE/Mathematica reports). Current self-direction scale: medium (figures vary by dataset/year). Workforce-interaction question: low — explicitly an open research gap, not an answer. Credit for Caring status/parameters: medium-high. Cost comparison: low — no scored estimate exists, extrapolation stated as such. Caregiver health/labor effects: medium-high for the current level (CDC federal source; AARP figures corroborated across years), but the 2015–2022 trend is suggestive and pandemic-confounded.