The ACA Premium Spike Is Mostly Sticky Care Cost. Only a Thin Policy Slice Can Round-Trip.

On 28 October 2025, the health-policy group KFF put a number on what Americans who buy their own insurance were about to face: insurers were requesting an average premiumpremiumThe fixed monthly price you pay an insurer to keep a health plan, before you use any care. increase of roughly 26% for 2026 coverage, the steepest ask in more than five years. (A 'premium' is the fixed monthly price you pay to keep a plan, before you use any care.) Eight months later, in June 2026, early 2027 filings already point to a second straight double-digit year. The press treated it as one big number going up.

It is not one number. It is at least two, glued together, and they behave in opposite ways. One part is the plain, grinding growth of medical costs, which does not come back down. The other is a defensive markup insurers added because they expect a specific government subsidy to expire, and that part can be unwound, on a known deadline, if Congress acts. Telling them apart is the whole story.

We are not asking 'are premiums up?' They are. We are asking a decomposition question: of the gross increase insurers filed, how many percentage points are a reversible policy artifact, versus genuine, sticky medical cost growth? The gross premiumgross premiumThe full sticker price of a plan before any government subsidy is applied. is the full sticker price before any subsidy, the number insurers file and the press quotes, so it is the right place to look for a markup that could be taken back out.

The enhanced premium tax credits (ePTCs) are a temporary, more generous version of the ACA's subsidy, in place since 2021. They lowered what enrollees actually pay each month, and they are scheduled to lapse at the end of 2025 unless Congress extends them. Insurers price a full year ahead, so they had to guess, in their 2026 and 2027 filings, whether those credits would survive. Many assumed they would not.

That guess matters because of how insurance math works. When a subsidy shrinks, the healthiest customers, who use little care, are the first to drop coverage. The people who stay are sicker on average, so the cost per remaining member rises. Insurers call the cushion they add for this 'morbidity loadingmorbidity loadingAn extra cushion insurers add to a price because they expect the people who stay in the plan to be sicker (higher 'morbidity') than before. Here it is the markup insurers add because they expect healthy customers to drop out when subsidies shrink.', in plain terms, a markup for an expected sicker pool. On 15 May 2026, the federal government finalized its 2027 rulebook (the NBPP) and trimmed the fees insurers pay to operate on the marketplace to about 1.9% and 1.5%. That fee change nudges premiums slightly down rather than up, so it is close to premium-neutral and does not muddy the decomposition. (Other parts of the same rule are not neutral, CMS projects they will reduce enrollment by 1.2 to 2 million people, but those provisions sit outside the price breakdown we are tracing.)

Start with the headline insurers filed, then take it apart.

The 26% is KFF's estimate for the market as a whole, and it hides a lot. Dig into the individual filings and the picture is a little lower and far more revealing: across 312 insurers, the Peterson-KFF Health System Tracker found a median request of 18% and a mean near 20%, with a handful of outlier insurers and a lopsided geography pulling the headline average up. (Benchmark premiums are rising about 30% on Healthcare.gov states versus about 17% on state-run exchanges.) To find the reversible slice, we decompose the typical, median request.

Decompose that typical 18% request and the policy share separates cleanly from the cost share. Underlying medical trendmedical trendThe underlying year-over-year growth in health care costs from higher prices and more use of care, separate from any one-off policy change., the year-over-year growth in prices and use of care, is the largest block at roughly 8 points and is sticky. TariffsTariffsTaxes on imported goods; here, levies on imported drugs and medical supplies that insurers say add a few points to their costs. on imported drugs and supplies add around 3 points. And the subsidy-cliff morbidity load, the markup for the expected healthy exodus, is roughly 4 points. The policy-driven slice is therefore about 7 of the 18 points, but only the 4-point subsidy load is what a credit extension would directly claw back. The tariff points reverse only if trade policy does.

The cleanest evidence the load is real and separable comes from insurers' own filings. Peterson-KFF reviewed 23 early filings in Vermont, Oregon, Washington and Washington, DC, and found they added about 4% on average specifically for the expected subsidy expiration, a contingency written down on paper, not an atmospheric 'uncertainty' line. For 2027, the pattern repeats and is, if anything, more explicit: insurers cite the end of the credits as driving rates roughly 4% to 6% higher than they otherwise would be, and one Massachusetts insurer attributed 3.3 of its 25.7 percentage-point increase to morbidity changes from policy shifts including the expired credits. That is exactly the kind of itemized, conditional markup a policy reversal could claw back.

The load reads as reversible because insurers wrote it as a conditional bet. The question is whether the dollars come back out when the condition fails, and they may not, because by the time Congress acts, the healthy customers the markup was hedging against have already walked.

Keep the two numbers apart. The gross premium, the sticker price, carries the reversible 4-point subsidy load. What enrollees actually pay is a different quantity, and it moves the other way. KFF estimates that if the enhanced credits lapse, the average out-of-pocket (net) premium for subsidized enrollees more than doubles, from $888 to $1,904 a year, a 114% jump. That rise comes from losing the subsidy itself, not from the gross-premium markup, so extending the credits would undo most of it, but only by restoring the subsidy, not by lowering the underlying price of care.

So even in the best case for a reversal, the picture is asymmetric. A credit extension before 12 August could refile away the ~4-point morbidity load on the gross premium and spare enrollees the ~114% net jump. It would do nothing to the ~8 points of medical trend or the ~3 points of tariffs baked into the sticker price. The spike does not round-tripround-tripDesk shorthand for a number that spikes and then largely reverses back toward where it started, as opposed to a permanent step up. to where it started; it round-trips to a higher floor.

There is a clock on the reversible slice, and it is not just the legislative one. The morbidity load was a forecast: insurers priced for healthy people leaving. But the healthy people the markup was pricing against have, by mid-2026, largely already left, after a year of higher net costs and shrinking marketplace enrollment. Once the pool is actually sicker, the 'load' stops being a contingency and becomes the true cost of the members who remain.

That is the difference between this episode and the textbook reversible case, silver loadingsilver loadingA specific markup insurers added to mid-tier ('silver') plans starting in 2018 after the government cut certain payments; the classic example of a policy-driven price bump that, in theory, could be reversed.. In 2017 the government cut cost-sharing-reduction payments; in 2018 insurers marked up silver plans to cover the gap. That markup sat on the price mechanically and could in principle be unwound if the payments returned, because it was an accounting patch, not a change in who was insured. A morbidity load is stickier: it describes a real shift in the risk poolrisk poolThe whole group of people covered by a set of plans, whose collective health determines average costs; a 'sicker pool' means higher costs per person., and a pool, once it sorts sicker, does not un-sort just because a subsidy comes back. Each month the experiment runs, more of the 'reversible' 4 points calcifies into plain cost.

What makes this worth watching in near-real-time is that the policy that drove the markup is still in play at the deadline. Usually we only get to study a price shock after the fact. Here the trigger, whether Congress extends the credits, is unresolved with the 12 August refile windowrefile windowA period before plans are finalized when insurers can submit revised rates, for example to lower them if a feared policy change does not happen. open, so the reversal is a live test, not a postmortem.

The prediction is specific and falsifiable. If the credits are extended in time, watch the refiled 2027 rates fall by something on the order of the 4-point subsidy load insurers itemized, and no further. If they fall by far more, the markup was hiding ordinary trend. If they do not move at all, either the pool has already hardened past the point of reversal or the load was never really conditional. Either way, by mid-August the answer is locked for a year, and we will know which kind of number the spike was.