Blog
Is the insurance death spiral real? Adverse selection tested claim by claim
Six common claims about adverse selection and the death spiral, tested against how pools, subsidies, mandates and benchmarks work, with round-number math.
Whenever health insurance prices jump, someone says the market is in a "death spiral": healthy people quit, premiums rise, more healthy people quit, and the insurer eventually has nobody left but the very sick. The idea is simple enough to sound like a law of nature. It is also an idea that people use in opposite ways. Some use it to argue the individual market is doomed. Others use it to argue that the danger is overblown and the argument is a scare tactic.
The truth is in between, and you can see it with a small pool of people and a calculator. The numbers below are invented and round, an illustration and not a forecast. This post states six common claims, tests each one, and says what survives.
The pool every claim depends on
Adverse selection means that people who expect to use more care are more likely to buy coverage, or to buy richer coverage, than people who expect to use less. When an insurer must accept everyone at a price that does not depend on health, which is how the ACA's guaranteed issue rules and the ban on pre-existing-condition pricing work (see pre-existing conditions and health insurance: what the law says), the mix of people who sign up matters a great deal.
Here is the hypothetical pool used throughout.
- 800 healthy people, each expected to generate $2,000 in claims for the year.
- 200 sick people, each expected to generate $20,000 in claims for the year.
Total expected claims: 800 × $2,000 = $1,600,000, plus 200 × $20,000 = $4,000,000, for $5,600,000. Divided across 1,000 people, that is $5,600 each. For simplicity, say the insurer sets the premium at exactly the average claim cost, $5,600, and ignores administration and margin. Adding those would raise every figure, but it would not change the pattern.
A healthy person in this pool pays $5,600 and expects $2,000 in claims. A sick person pays $5,600 and expects $20,000. Insurance is a bargain for one and a poor deal, in dollars, for the other, which is the whole difficulty.
Claim 1: "If healthy people leave, the insurer collapses"
Test it. Suppose 100 of the 800 healthy people decide the price is too high and drop coverage.
- Remaining healthy: 700 × $2,000 = $1,400,000.
- Sick: 200 × $20,000 = $4,000,000.
- Total claims: $5,400,000 across 900 people.
- New break-even premium: $5,400,000 ÷ 900 = $6,000.
The premium rises from $5,600 to $6,000, which is an increase of $400, or about 7.1%. An insurer that sees this coming prices for it in the next filing, as the post on how state rate review decides whether my premium increase is allowed describes. Nothing has collapsed.
Verdict: partly true, mostly overstated. Losing healthy members raises the average cost, and the premium has to rise to match. That is an increase, not a collapse. A collapse needs the loop in the next claim.
Claim 2: "Premiums then spiral upward forever"
The spiral is a loop: a premium rise pushes out more healthy members, which causes another rise. Here is the loop run for several rounds, with 100 more healthy members leaving each time.
| Round | Healthy members | Sick members | Total members | Total claims | Break-even premium |
|---|---|---|---|---|---|
| Start | 800 | 200 | 1,000 | $5,600,000 | $5,600 |
| 1 | 700 | 200 | 900 | $5,400,000 | $6,000 |
| 2 | 600 | 200 | 800 | $5,200,000 | $6,500 |
| 3 | 500 | 200 | 700 | $5,000,000 | about $7,143 |
| 4 | 400 | 200 | 600 | $4,800,000 | $8,000 |
| 5 | 300 | 200 | 500 | $4,600,000 | $9,200 |
If every healthy person eventually left, the premium would end at $20,000 (200 sick people at $20,000 each, $4,000,000 ÷ 200). That is the death spiral in its pure form.
The table assumes that 100 healthy people leave in every round no matter how the price compares with their own situation. Real people do not behave that way. Some value protection against a large, unexpected bill even if they expect to use little care. Some need coverage for a job, a lender or peace of mind. Some are shielded from price changes entirely, which leads to the next claim.
Verdict: the arithmetic is right, the assumption is the weak spot. A spiral needs members to keep leaving as prices climb. Whether they do is an empirical question about each market and year.
Claim 3: "Subsidies change the math"
This claim is true, and the pool example shows why. Under the premium tax credit, a subsidized household pays a set share of its income for the benchmark plan, and the credit covers the rest. The mechanics are in How ACA premium tax credits work: income, benchmark plan, your share.
Say a hypothetical household's required contribution is $2,400 a year. This is an illustration, not an actual percentage of anyone's income. If the benchmark premium is $6,000, the credit is $6,000 − $2,400 = $3,600. If the benchmark premium rises to $7,000, the credit becomes $7,000 − $2,400 = $4,600. The household still pays $2,400. The rise lands on the government's budget, not on this household, and so it gives the household no reason to drop coverage.
Now rerun the pool with half of the healthy members subsidized in this way. Of the 800 healthy, 400 are subsidized and stay no matter what. The 400 unsubsidized healthy members can leave.
- If all 400 unsubsidized healthy members leave, the pool is 400 healthy and 200 sick, or 600 people.
- Claims: 400 × $2,000 = $800,000, plus $4,000,000, for $4,800,000.
- Break-even premium: $4,800,000 ÷ 600 = $8,000.
The premium rises to $8,000, which is an increase of $2,400 or about 42.9% over the starting $5,600, and then it stops. It cannot spiral further, because nobody is left who responds to the price. Subsidized members absorb the increase through larger credits.
This is a real stabilizer, and it is also a real cost. The guide on the enhanced subsidies ended: the 400% cliff and what changed explains why people above the income limit for credits are the ones most exposed. As of November 2026, credits end above 400% of the poverty level, so those buyers face the full premium increase. KFF reported that changes in the risk pool, meaning healthier people leaving after the enhanced credits expired, added about 4 percentage points to 2026 premiums, and that insurers expected a similar effect for 2027 (as of November 2026). That is a measurable adverse selection effect, and it is a four-point push on premiums, not a collapse.
Verdict: true, with a caveat. Subsidies stop a spiral among subsidized members. They shift the cost to the budget and leave unsubsidized members exposed.
Claim 4: "Only a mandate can keep healthy people in"
The individual mandate required most people to have coverage or pay a penalty. The federal penalty has been $0 since 2019, and a handful of states have their own mandate. Supporters of the mandate argued it pulled healthy people into the pool. Critics argued it was a blunt tool, or that the penalty was too small to matter. Both positions have been defended with data, and the evidence is contested.
What the pool example shows is that a mandate is one of several ways to keep the mix balanced. Others include:
- Enrollment windows. Outside the open enrollment period, a person generally needs a qualifying event to buy a plan. That prevents the classic strategy of waiting until you are sick to sign up.
- Subsidies, as shown above.
- Late-enrollment penalties in some programs, such as Medicare.
- Reinsurance and risk adjustment, which reduce the damage when one insurer ends up with sicker members.
Verdict: a mandate is one tool, not the only one. Whether the market is stable without it depends on the others working. The question of whether it should return is a political one, and reasonable people disagree.
Claim 5: "A risk pool is just healthy people paying for sick people"
There is a grain of truth. A pool works by spreading the cost of rare, expensive events across everyone. In the example, the healthy members pay $3,600 more than they expect to use each year, and the sick members receive $14,400 more than they pay. But two corrections matter.
First, "healthy" and "sick" are not fixed labels. A person who is in the low-cost group this year may be in the high-cost group next year, after a diagnosis or an accident. Coverage is a hedge against that shift, and many people value it for that reason.
Second, the pool is not a single national fund. Individual-market plans in a state are treated as one risk pool for rate-setting, employer plans are separate pools, and each insurer has its own mix of members. If one insurer ends up with a sicker mix than another through no fault of its own, risk adjustment moves money between them so that no insurer is punished just for attracting more costly members. The post on how risk adjustment moves money between insurers, and why members should care covers the mechanics. State reinsurance programs, covered in how state reinsurance programs lower marketplace premiums, do something related by covering a slice of very high claims.
Verdict: simplified, and misleading when treated as a moral story. A risk pool is a way to share uncertainty. It is not a fixed list of givers and receivers.
Claim 6: "The benchmark plan means premiums cannot spiral"
Credits are tied to the benchmark plan, the second-lowest-cost silver plan in the area (see benchmark plan). When benchmark premiums rise, credits rise too. Some readers conclude that spirals are therefore impossible for subsidized buyers, and that is largely right for them, as claim 3 showed.
Two cautions apply. First, the benchmark is one plan. If a buyer chooses a different plan, the credit does not move with that plan's price in the same way. Second, the benchmark is silver, and silver premiums have, in many areas, been set to include the cost of cost-sharing reductions, a practice often called silver loading. That makes the benchmark higher than it would otherwise be, which raises credits, which in turn lowers the net price of some bronze and gold plans for subsidized buyers. This is why a careful comparison, rather than a rule of thumb, is worth doing.
Verdict: a stabilizer for subsidized buyers, not a guarantee for the market. Unsubsidized buyers and the public budget absorb the increases.
How the real market differs from the example
The example has two kinds of people and one price. Real markets are messier in ways that cut in both directions.
- There are many levels of health, not two. Most people sit between $2,000 and $20,000 in a year, and most claims come from a small share of members. The pattern of the loop is the same, but the steps are smaller.
- Plans differ. A person who finds the premium too high can often switch to a lower-priced plan with a higher deductible instead of leaving. That keeps the person in the pool, though it shifts more of their own claims onto their cost sharing.
- Prices are set once a year. Insurers reset rates each plan year and can price in what they expect, so a spiral would play out in annual steps rather than in the rapid rounds of the table. Regulators and the filings behind each step are covered in how state rate review decides whether my premium increase is allowed.
- Insurers can leave. If a company judges that a market no longer pays, it may withdraw, which creates a different problem for members: fewer choices. That is a more common outcome than a literal spiral.
What the six tests add up to
Adverse selection is real: a pool that loses its healthiest members becomes more expensive per person, and the 4-point effect reported for 2026 premiums is an example. A full death spiral is a much stronger claim, and it needs members who respond to every price increase by leaving. Subsidies, enrollment windows, risk adjustment and reinsurance each reduce the loop's force. Their strength differs by state, by year and by income group, and policy changes can strengthen or weaken them.
For a person shopping now, the practical points are different from the policy debate:
- Your premium depends on your own age, location and plan, not only on pool dynamics.
- If you receive credits, check how a price change affects your net cost rather than the sticker price. A navigator, a licensed broker or a tax professional can run it for your income.
- If you do not receive credits, compare plans carefully. You are the one exposed to the increase.
- Do not drop coverage because of a headline. A year without coverage carries its own risks, and the post on The premium went up: should I drop coverage? The real cost of a year uninsured works through the numbers.
Key takeaways
- Adverse selection means sicker people are likelier to buy coverage. When healthy people leave, the break-even premium rises, as the example pool shows.
- A spiral needs members to keep leaving as prices rise. Subsidized members do not respond to price changes the same way, so the loop stops at a ceiling.
- In the example, subsidies cap the rise at $8,000 per person, about 42.9% above $5,600, instead of the $20,000 of a full spiral.
- A mandate is one stabilizer among several, with enrollment windows, subsidies, risk adjustment and reinsurance.
- Buyers above the credit limit absorb the increases directly, so they have the most reason to compare plans each year.