The ACA coverage loss stops being a projection: 2.6M fewer people actually paid for marketplace coverage in 2026 — and New Mexico, the one state that replaced the federal subsidies itself, grew 14%

KFF published state-level ACA marketplace enrollment comparisons on July 28, 2026 showing effectuated enrollment fell from 21.8 million in 2025 to 19.2 million in 2026 — about 2.6 million fewer people, a 12% decline. This matters because of WHAT is being measured.

KFF defines it precisely: 'Effectuated enrollment in this current analysis refers to enrollees with any coverage in February. For 2025, effectuated enrollment is as of March 15, 2026; for 2026, as of May 5, 2026.'

These are people who actually paid a first premium — an OBSERVED COUNT, not a model. Nearly every projection this platform has tracked about the expiring enhanced premium tax credits has now been replaced by an outcome.

KFF reports Ohio and Oklahoma with the steepest declines at about 32% each; Ohio Capital Journal (July 9) puts Ohio's loss at 161,385 people. Note carefully that KFF's published analysis gives percentages by state and does NOT publish state-level absolute numbers, so any state headcount must be sourced to the outlet that actually reports it, not to KFF.

The most consequential single finding is the exception: 'New Mexico was the only state to see an increase in effectuated enrollment, growing 14% between 2025 and 2026, coinciding with the state's premium assistance program that fully replaced the expiring federal enhanced tax credits with state-funded subsidies.' One state chose to backfill, and it is the only state whose coverage went up.

For FQHCs the read is direct: in the steep-decline states these are not people who vanished, they are people moving onto sliding-fee panels and self-pay, which raises uncompensated care exactly as the December 31, 2026 CHC Fund cliff approaches. Do NOT add this 2.6M to the separately tracked 5M+ combined Medicaid-plus-marketplace decline — those are different measures over different denominators and summing them double-counts.