The marketplace risk pool that issuers priced for 2027 is about 1% smaller than they thought. CMS said on October 14 that it has ended premium subsidies for 189,000 enrollees who also held Medicaid, Medicare or CHIP coverage, or who matched federal death records, with projected savings of about $1.5 billion over the next year (CMS, Oct. 14, 2026). The 2027 rate filings closed in August, built on a 2026 pool of 19.2 million people.
This is the second round. In 2025 the administration restarted the data matching the previous one had paused, and ended subsidies for about 551,000 Medicaid duplicates. That purge was large enough that issuers built its morbidity effect into 2027 pricing. The 2026 round, smaller but later, lands on a pool whose rates are already locked.
Key Takeaways
- 189,000 enrollees lost subsidies in the 2026 round: about 120,000 with Medicaid, 58,000 with Medicare and 11,000 deceased through July, with another 3,500 deceased expected by the end of October (Becker's, Oct. 14, 2026).
- $1.5 billion is CMS's projected saving over the next year; HHS Secretary Kennedy put this year's prevented payments at $738 million at the briefing.
- 551,000 to 120,000: the Medicaid duplicate count fell by nearly four fifths from 2025 to 2026, but Administrator Oz said the historical level is about 55,000, so the backlog is not cleared.
- 30 days is the dispute window. Enrollees who do not respond lose the subsidy and, for the deceased, the plan.
- At least twice a year is the data-matching frequency CMS now requires, restored from a four-year pause that Oz blamed for the backlog.
What CMS Did and Who Is Affected
Federal rules require exchanges to check periodically whether subsidized enrollees are also enrolled in Medicaid or CHIP. CMS calls this periodic data matching. It was expanded to twice a year in the first Trump administration and paused for Medicaid and CHIP during the Biden administration. The 2025 restart produced the 551,000 removals; the 2026 cycle produced 189,000.
The process is the same for each category. Enrollees whose records show duplicate coverage get 30 days to dispute the match or drop the duplicate. If they do nothing, CMS ends their advance premium tax credit. They can appeal afterwards, or show their Medicaid coverage is ending, and regain the subsidy. For people matched to death records, CMS contacts the individual or the estate, with the same 30 days.
Oz told the briefing that "no one is losing insurance coverage based on what's happening today," because non-responders default to Medicaid. That holds for the Medicaid group. A marketplace enrollee who was also on Medicare keeps Medicare. The marketplace plan, in either case, goes away with the subsidy, and so does its premium from the pool.
| Duplicate category | 2026 round | Comparison |
|---|---|---|
| Medicaid or CHIP | about 120,000 | 551,000 in 2025; about 55,000 historically |
| Medicare | about 58,000 | 93,000 in 2024 |
| Deceased | about 11,000 through July | 3,500 more expected by end of October |
| Total | 189,000 | 19.2 million effectuated enrollees in 2026 |
Sources: CMS release, Oct. 14, 2026 (total, 2025 figure); Becker's Payer, Oct. 14, 2026 (category figures from the CMS briefing); ASPE for the 2026 base.
What a One-Percent Removal Does to the Pool
The actuarial question is not the count but the claims behind it. Someone enrolled in both Medicaid and a subsidized marketplace plan is, in most cases, using one of them. If the marketplace plan was the idle one, removing it takes premium out of the pool with almost no claims attached, and the remaining pool gets sicker on average. If the marketplace plan was the active one, the removal takes claims out too.
The 2025 round was read the first way. HHS's own analysis attributed the whole 2.9 million drop in 2026 enrollment to improper-enrollee removal rather than affordability attrition, and the site covered how issuers converted that into a 2027 morbidity load. Wakely's data put the shift at 2.9% to 6.5%. Those loads were filed by August and approved by now.
The 2026 round changes the base those loads sit on. If 189,000 low-utilizing lives leave the 2026 pool in the second half of the year, 2026 experience comes in with a higher claims ratio than the first-half data showed, and 2027 rates were trended from that experience. The effect on any one issuer depends on how many of its members were among the 189,000, which CMS has not broken out by state or carrier.
Risk adjustment moves the other way. A member who leaves mid-year with little claims history carries a low risk score. Removing low-score members raises the issuer's average score relative to the state, which raises its transfer receipt or lowers its payment. For an issuer whose removals are concentrated, the transfer effect can offset part of the morbidity effect, but only after the 2026 transfers are calculated in mid-2027.
The Backlog Is Not Cleared
The numbers CMS presented as a success also say the matching is unfinished. The historical level of Medicaid duplicates, by Oz's figure, is about 55,000. The 2026 cycle found 120,000, more than twice that, a year after removing 551,000. Twice-yearly matching should drive the count toward the historical level over time, and each cycle until it gets there is another mid-year removal from a pool priced the previous August.
For 2028 pricing, which starts in the spring, the question is which base to trend from: the 2026 pool as experienced, or the pool after a third round of removals CMS has not yet run. CMS says it will keep matching at least twice a year. Issuers will file before the next count is known.
Further Reading
- ACA Enrollment Fell to 19.2M in 2026. HHS Says It Was a Fraud Purge, Not Attrition – the 2025 removals and how they reshaped 2027 morbidity assumptions.
- Wakely's Morbidity Data Reshapes 2027 ACA Rate Filing Assumptions – the 2.9% to 6.5% morbidity shift issuers filed.
- CMS Will Auto-Renew Lawfully Present Immigrants at Full Price – the other 2027 risk-pool shift arriving through non-payment.
- ACA Filing Assumptions Benchmark – the morbidity and trend assumptions filed by state and carrier.