Actuaries are not pundits. When an insurer files a rate increase with a state regulator, it has to justify the number, component by component, and an actuary has to sign it. The filings are boring by design and they are the most honest documents in American health policy, because lying in them is a professional and legal problem rather than a Tuesday.

So read them.

For 2027, across 77 insurers in 16 states and D.C. with public filings, the median proposed premium increase is 14%. The range runs from 1% to 52%. Twenty insurers want more than 20%.

That follows increases of roughly 20% for 2026. KFF's arithmetic on the two years together is the sentence worth pinning up: typical premiums for participating insurers will have jumped by more than a third over a two-year period.

And buried in the justifications is the number that ends the argument. Insurers attribute about four percentage points of their 2027 request not to hospital prices, not to drug costs, not to utilization, but to the changing composition of the risk pool. Healthy people left. The people who remain are sicker. That costs money, and they are charging for it.

That is a death spiral. Not a warning about one. A measured, priced, filed one.

The mechanism, stated plainly by the people running it

The sequence is not complicated and nobody involved disputes it.

Enhanced premium tax credits expired on December 31, 2025. Net premiums for subsidized enrollees roughly doubled. The people most likely to walk away from a doubled bill are the people who were not using their insurance much, which is to say the healthy ones. Their departure raises the average cost of everyone left. Insurers price for that next year. The higher price pushes out the next tranche of marginally healthy enrollees. Repeat.

The Congressional Budget Office called it in advance, projecting that without extension, healthier-than-average people would exit and insurers would raise premiums on those remaining, with gross benchmark premiums up 4.3% in 2026 and 7.7% in 2027 on that basis alone.

The exits have now happened. Enrollment peaked at 24.2 million in 2025 and stands at about 19.2 million. Five million people gone, a 13% year-over-year decline, ending five straight years of growth. Federal data showed the market shrank by more than 2.5 million over the past year, with some states losing close to a third of their enrollees.

The natural experiment nobody is talking about

Here is the finding I would lead any story with, and almost nobody has.

Internal CMS documents obtained by NOTUS show that 21% of HealthCare.gov enrollees were disenrolled in the early months of 2026 for failure to pay premiums.

On state-run marketplaces, the figure was about 8%.

The gap is attributed in significant part to supplemental subsidies that some states chose to fund themselves.

That is as close to a controlled experiment as health policy ever produces. Same country, same year, same insurers, same medical inflation, same expired federal credits. The variable is whether a state stepped in with money. Where it did, roughly a tenth of enrollees dropped out. Where it did not, more than a fifth.

The debate about whether affordability drives coverage loss, or whether the enrollment surge was fraudulent as the administration has suggested, does not survive that comparison. People did not stop being insured because their eligibility was suspect. They stopped because they could not pay, and where somebody helped them pay, they mostly stayed.

Insurers are voting with their capital

Carriers do not stay in markets out of loyalty.

At least six have announced ACA exits for 2027: Cigna, CareSource, PacificSource, Baylor Scott and White, Providence Health, and Mending, pushing roughly 650,000 people across a third of states into new plans. Cigna is leaving all eleven states where it participated, abandoning around 350,000 to 369,000 enrollees, after its marketplace book shrank 17%. Its incoming chief executive said there was no clear path to scale the business.

Read that as a solvency signal rather than a political one. Cigna is not making a statement about the ACA. It is making a statement about a shrinking, worsening pool it can no longer price with confidence, in a market where a competitor's exit dumps an unknown risk profile onto whoever stays. That last point is doing quiet damage: every exit makes the remaining insurers' pricing harder, because they inherit a population they have never underwritten.

Fewer carriers also means less competition, which means higher prices, which means more departures. The spiral has a second turbine.

The federal government is inside the machine, not watching it

The 2027 Notice of Benefit and Payment Parameters, finalized on May 20, 2026, adds income verification and paperwork requirements at enrollment. HHS's own estimate is that the rule will remove up to 2 million more people from the marketplace, at an implementation cost to taxpayers of roughly $1.34 billion a year. CMS has acknowledged that healthier enrollees may be the ones more likely to fall out under the new rules.

Sit with the shape of that. The government is spending over a billion dollars annually on paperwork designed to shrink the pool, has conceded the departures will skew healthy, and the resulting morbidity shift will raise premiums, which raises the cost of the premium tax credits still owed to everyone who remains.

It is a policy that pays money to make the risk pool worse and then pays more money because the risk pool got worse.

The timing compounded it. The rule was proposed in February and finalized on May 20, after 2027 rates were already due in some states. Vermont's Blue Cross filed while noting the rule was not final. A Washington insurer said it would seek approval to refile if material impacts emerged. Insurers were asked to price a market whose rules had not been written, and uncertainty in an actuarial filing does not come out as a shrug. It comes out as a margin.

What is actually being bought

The 2027 rule also opens the door to plans with lower premiums and much higher cost-sharing. Some maximum out-of-pocket exposures can exceed an enrollee's income.

About 40% of enrollees are already choosing bronze or other low-cost, high-deductible plans to survive the premium. Combine those two facts and the direction is clear. The market is not collapsing so much as converting: from insurance that pays for care into insurance that prevents bankruptcy, and then, at the margin, into insurance that does not quite do that either.

MVP Health Plan of Vermont has proposed a 109% increase on its catastrophic plan, because eligibility for catastrophic coverage is being broadened beyond the under-30s and the insurer expects the people who arrive to look like everyone else. Even the cheap door is being repriced as the people walking through it get sicker.

The forecast is not a forecast

A 40-year-old in Indiana earning $65,000 paid about $316 a month for a benchmark silver plan in 2025. In 2027, KFF puts the figure at about $546.

She earns too much for a subsidy now, because the 400% cliff is back. She will look at that number and make the same decision several million people already made.

Then the actuaries will price her absence into the 2028 filings, and sign them, and the filings will be correct.

Further reading