Healthcare — total premium
Family health insurance premium including employer share: KFF EHBS national; MEPS-IC state with cross-source calibration.
Sources
Precedence ladder for healthcare — total premium. Lower precedence wins where multiple sources cover the same period. The engine walks the ladder per period with chain-link rescale at source transitions.
| Precedence | Source | Series ID | Cadence | Geography |
|---|---|---|---|---|
| 0 | KFF Employer Health Benefits Survey KFF EHBS family total premium (employer + worker share). Continuous annual 1999 → 2024. Post-publication-tail months (typically Oct of year Y → KFF release of year Y+1 the following Oct) are uncovered under the no-fallback doctrine — engine truncates rolling 12-month at last KFF-covered month rather than bridging through CPI SEME (which back-revises retrospectively as MEPS-IC arrives and produces spurious 6-7% MoM drops). |
kff_ehbs_premium_family_total | annual | national |
Weight across archetypes
How healthcare — total premium is weighted in each of the 10 published archetypes. Sorted by weight, highest first. National Average is the headline citation surface.
| Archetype | Weight | Action |
|---|---|---|
| Income — Lowest Quintile | 22.5795% | View archetype → |
| Income — Second Quintile | 18.4347% | View archetype → |
| Income — Middle Quintile | 15.5775% | View archetype → |
| Renter | 15.4041% | View archetype → |
| National Average | 14.9495% | View archetype → |
| Homeowner | 14.7628% | View archetype → |
| Income — Fourth Quintile | 13.4530% | View archetype → |
| Working Parent | 13.2223% | View archetype → |
| Income — Highest Quintile | 9.9857% | View archetype → |
| Retiree | 8.9527% | View archetype → |
Substitution history
Substitution log entries that touched this stratum. Each entry
bumps the basket version and is reproducible from observations
tagged with their basket_version.
2026-05-29 — Zero-CPI recomposition, chain-linked headline, CPI ingester removed
Version: v1.1.4 → v2.0.0
What changed. Retired BLS CPI from the index entirely and removed its last infrastructure. (1) Five CPI/proxy-dependent strata EXITED and weights renormalized over the surviving 15: household_goods, personal_care, alcohol (pure CPI primaries), food_away (single-chain Big Mac proxy, rejected as non-representative), and tobacco (CPI Tobacco was its only 2020+ path). (2) The last CPI bridge — CUUR0000SEMF01 (Rx CPI) — removed from healthcare_oop, now NADAC (5 generic drugs) + CMS PFS (2 CPT codes) only. (3) shelter_rent moved to Zillow ZORI (2015+) spliced to Census HVS median asking rent (≤2014), and the state healthcare_premium chain-link to BLS CPI SEME was dropped — these SUPERSEDE the v1.1.2 'chain-link to BLS CPI SEME for 2025+' and v1.1.3 'leaving BLS CPI Rent of Primary Residence as sole source' entries below, which describe the now-replaced v1.1.x state. (4) Headline switched to a drift-free fixed-base-within-spans construction, chain-linked only at composition boundaries, so late-starters splice in at entry with no level jump and the index spans the full window. (5) The dormant bls_cpi ingester, its CLI/freshness/reingest wiring, its fixture, and 7,174 inert CUUR rows were deleted from price_observations (local + prod).
Why. The load-bearing commitment is that the index contains NO BLS CPI anywhere — not a primary, not a bridge, not a fallback — because claiming to improve on CPI while still consuming CPI is incoherent. These changes realize it fully: non-CPI share = 100%.
Effect on headline. Published window 2000-01 → 2025-12 (312 months, up from a truncated 2010-2023); long-run decay 3.440%/yr; Dollar Half-Life 20.5 yr; CPI series in index 0; strata count 20 → 15. Price rows remain tagged basket_version=1.0.10 (a weighting + engine-construction change, not a re-ingest); exited strata's rows stay in the store at weight 0 so prior versions remain reconstructable.
2026-05-27 — Full annual KFF state series + post-2024 trim
Version: v1.1.1 → v1.1.2
What changed. State-level healthcare_premium rebuilt from the complete 2013-2024 annual MEPS-IC publication (12 years × 51 geographies = 612 measurements), replacing the 3-anchor (2014/2019/2024) v1.1.0 simplification; post-2024 hold-forward removed; state series chain-link to BLS CPI SEME for 2025+.
Why. The 3-anchor approach forced stepped hold-forward interpolation between anchors and a post-2024 flat tail that depressed long-run decay rates by ~20-40 bp vs national; KFF publishes annually so ingesting all 12 years removes both artifacts.
Effect on headline. National unchanged. State-vs-region gaps moved from -240/-380 bp to -3/-125 bp.
2026-05-27 — Per-period geography splice + unified series_ids
Version: v1.1.0 → v1.1.1
What changed. EIA and BLS-CPI now write per-state and per-region rows under the same internal series_id as their national equivalent with the geography column differentiating; the §5.4 fallback ladder rewritten from per-ref to per-period with chain-link rescale at geography transitions.
Why. The prior per-ref geography routing silently dropped KFF national 1999-2013 rows for state queries (because the engine picked a single best geography for the whole window); the audit revealed pop-weighted state-vs-national headline gaps of -240/-380 bp driven by a -1860/-2120 bp gap in healthcare_premium that traced to this routing.
Effect on headline. National unchanged. State headlines moved 25-40 bp toward their region's value; per-stratum decomposition now surfaces real state divergence in 4 of 20 strata.
2026-05-27 — Cross-source calibration for state healthcare_premium
Version: v1.0.2 → v1.0.3
What changed. State-level healthcare_premium rows are now emitted with the calibration calibrated_state(s, y) = MEPS_IC_state(s, y) × (KFF_EHBS_US(y) / MEPS_IC_US(y)), with the calibration ratio, raw MEPS-IC value, and adjustment basis surfaced in each observation's raw_payload.
Why. MEPS-IC (the state-level source) and KFF EHBS Figure 1.12 (the national source) measure the same construct through different operational definitions and persistently disagree at the US level by 3-5% in levels; without calibration, state-vs-national comparisons mixed two surveys and state pop-weighted aggregates did not match the national headline.
Effect on headline. None at national level. State family premium values run ~1-4% higher than raw KFF State Health Facts; per-state ranking unchanged.
Methodology
See the basket page for all 20 strata and the per-stratum source map. The methodology page covers the load-bearing pillars including the rules around source-precedence ladders, chain-link splicing at source transitions, and the no-fabricated-data invariant.