Why healthy people leave, sick people stay, and premiums spiral — individual mandates, subsidies, and risk adjustment as counterweights
Health insurance only works as a business because most enrollees in any given year cost far less than they pay in premiums, subsidizing the minority who file large claims. Under community rating — the rule underlying the ACA individual market — every enrollee in a rating area pays the same premium (adjusted only for age and tobacco use, not health status), which means healthy members are, by design, paying more than their expected claims to cross-subsidize sicker members.
Insurance is fundamentally a mechanism for pooling unpredictable individual risk into a predictable group average. If every enrollee paid a premium exactly equal to their own expected annual claims, there would be no pooling at all — just prepayment for expected costs, with no protection against the risk of unexpectedly becoming sick. The entire value proposition of insurance depends on ex-ante healthy people paying more than their expected costs in any given year, in exchange for the guarantee that if they become sick, the pool will absorb costs far above what they alone paid in.
Under guaranteed issue and community rating — the two pillars of ACA individual market regulation — insurers cannot deny coverage or charge higher premiums based on health status. This is precisely what makes the market accessible to people with pre-existing conditions, but it also removes the insurer's ability to price each enrollee according to their individual risk. The entire pool's viability then depends on a sufficient number of low-cost, healthy enrollees choosing to remain in the pool voluntarily — a participation decision no longer disciplined by individualized pricing.
The theoretical foundation for adverse selection was formalized by economist George Akerlof in his 1970 paper on used car markets, later extended to insurance by Rothschild and Stiglitz (1976). The core problem: enrollees typically know more about their own health status and expected future health spending than the insurer does. When insurers cannot price on health status (as under community rating), they must set a single premium based on the average expected cost across all enrollees.
But individuals do not enroll at random — they self-select based partly on private information about their own risk. If the community-rated premium is set at the average cost of a representative population, it will look expensive to healthy people (who know they are unlikely to use much care) and cheap to sick people (who know they will). This asymmetric response to a single price is what economists call adverse selection: the insured population is not a random sample of the population, but is skewed toward higher risk than the price was calculated to cover.
Adverse selection is not a market failure caused by bad actors — it is the predictable equilibrium outcome of voluntary enrollment under a single community-rated price when individuals have private information about their own risk. Any policy fix has to change either the price, the information asymmetry, or the voluntariness of the enrollment decision.
The ACA individual marketplace (HealthCare.gov plus 19 state-run exchanges) has functioned since 2014 as arguably the largest continuous real-world experiment in managed community rating anywhere in the world, covering over 21 million people in 2024 across a wide range of state-level policy choices — some states layered on their own mandates and reinsurance programs, others relied solely on the federal baseline rules.
This variation is analytically useful: because otherwise-similar states adopted different combinations of the four stabilizing tools (mandate, subsidy generosity, reinsurance, risk adjustment), health economists have been able to compare risk-pool outcomes across states as a natural experiment, rather than relying purely on theoretical modeling. The stage-by-stage narrative in this simulator — baseline pool, premium setting, exit decisions, deterioration, and intervention — mirrors the sequence regulators and insurers actually watch play out annually in each state's rate-filing cycle.
Premiums are not set arbitrarily — actuaries build next year's rate from this year's observed claims experience, projected forward for medical trend, and divided across the expected enrolled population. When the enrolled population turns out sicker than assumed, the resulting premium increase becomes the trigger for the next round of enrollment decisions — ratemaking and enrollment behavior are locked in a feedback loop.
A simplified actuarial premium formula: Premium ≈ (Total Projected Claims ÷ Projected Member-Months) × (1 + Administrative Load + Margin) ÷ (1 − Risk Adjustment Net Transfer Rate).
The critical input is "Total Projected Claims ÷ Member-Months" — average cost per enrollee. If this year's actual enrolled pool skewed sicker than last year's pricing assumed (because healthy members exited), next year's claims experience will run higher than budgeted, and the insurer must file a larger rate increase to remain solvent under the required 80% medical loss ratio floor. State insurance regulators review and can reject unjustified increases, but they cannot force an insurer to price below its actual claims experience without threatening market exit.
This creates a one-year-lagged feedback loop: this year's adverse selection becomes next year's premium increase, which becomes the trigger for the next round of healthy-enrollee exits. The lag is what allows the dynamic to compound across several years rather than resolving instantly.
A real-world illustration of how administrative decisions ripple through premium-setting: in 2017, the federal government stopped directly reimbursing insurers for Cost-Sharing Reduction (CSR) subsidies mandated by the ACA for low-income silver-plan enrollees, even though insurers remained legally obligated to provide the reduced cost-sharing. Insurers responded by loading the unfunded CSR cost entirely onto silver-tier premiums specifically ("silver-loading") rather than spreading it across all metal tiers.
Because ACA premium tax credits are calculated as a percentage of the benchmark silver premium, silver-loading paradoxically inflated tax credits, making bronze and even some gold plans nearly free after subsidy for many enrollees — an unintended stabilizing effect that partially offset the mandate repeal happening the same year. This episode illustrates how premium-setting mechanics interact with subsidy design in ways that can either amplify or dampen adverse-selection pressure depending on the specific rule structure.
Since individual health status cannot be used to price a policy under community rating, insurers instead price along a different axis: actuarial value, the average share of covered costs the plan pays versus the enrollee's out-of-pocket exposure. ACA marketplace plans are standardized into four metal tiers — Bronze (~60% actuarial value), Silver (~70%), Gold (~80%), and Platinum (~90%) — letting enrollees self-select their preferred cost-sharing structure while insurers still price each tier on the pool that actually selects into it.
This creates a secondary, tier-level adverse selection dynamic layered on top of the market-wide one: sicker enrollees who expect to hit their deductible anyway tend to prefer richer, higher-actuarial-value Gold and Platinum plans (lower marginal cost per visit), while healthy enrollees who rarely use care gravitate toward cheaper, high-deductible Bronze plans. Regulators partly address this second layer with the same risk-adjustment mechanism used market-wide, transferring funds between tiers, not just between insurers.
When a premium increase lands, not all enrollees respond the same way. Someone managing a chronic condition has few alternatives to coverage and will absorb a price increase rather than go uninsured. Someone who rarely uses care compares the premium to their perceived risk and often concludes the expected value no longer justifies the cost — so healthy enrollees exit first, precisely the group whose presence had been holding the average premium down.
Economic studies of ACA marketplace enrollment consistently find that price sensitivity is highly heterogeneous across the risk distribution. Low-risk enrollees — who by definition expect to file few or no claims — treat the premium largely as a pure cost with little offsetting expected benefit, making their enrollment decision quite sensitive to price (elastic demand). High-risk enrollees, especially those with ongoing chronic conditions or scheduled treatments, face a very different calculation: the expected benefit of coverage is large and immediate, making their demand comparatively price-insensitive (inelastic).
This asymmetry means a uniform percentage premium increase does not shed enrollees uniformly across the risk pool — it disproportionately sheds the lowest-risk members first. The mathematical consequence is that even a modest rate increase, applied to a community-rated pool, can measurably raise the pool's average risk score simply by changing its composition, independent of any change in the underlying health of any individual enrollee.
ACA premium tax credits are means-tested and pegged to the cost of the benchmark (second-lowest-cost silver) plan relative to a household's income, capping the enrollee's effective payment as a percentage of income for those who qualify. For subsidized enrollees, a sticker-price premium increase is largely absorbed by a larger tax credit, insulating their exit decision from the underlying rate increase almost entirely.
The exposure is concentrated among unsubsidized enrollees — those with income above the subsidy eligibility threshold (400% of the federal poverty line, prior to temporary American Rescue Plan enhancements that removed the cliff) — who pay the full sticker price and are the segment most likely to exit when premiums rise. Because unsubsidized enrollees also tend to skew healthier on average (subsidy-eligible lower-income populations have historically shown higher morbidity in some studies), this creates a specific channel through which rising premiums disproportionately erode exactly the subsidy-insensitive, price-sensitive, comparatively healthy segment of the pool.
This is why the subsidy slider in this simulator has an outsized stabilizing effect relative to its cost: subsidies do not need to reach every enrollee to blunt adverse selection — they only need to reach the price-sensitive healthy segment that would otherwise be first to exit.
A subtler version of the same exit dynamic operates through enrollment timing rather than a binary in/out decision. Outside the annual open enrollment window, ACA marketplaces allow Special Enrollment Periods (SEPs) triggered by qualifying life events — job loss, marriage, birth of a child, loss of other coverage — specifically so people are not locked out of coverage for a full year after a major life change.
Because a new diagnosis or planned medical procedure is not itself a qualifying event, SEPs are not supposed to let people enroll reactively purely because they got sick. In practice, however, studies of SEP claims experience have repeatedly found that SEP enrollees run meaningfully higher average claims costs than open-enrollment enrollees in their first year of coverage — consistent with at least some degree of adverse selection concentrating specifically in the SEP channel, prompting CMS to periodically tighten SEP eligibility verification requirements.
A premium death spiral is what happens when the exit-then-repricing feedback loop from Stage 3 runs for multiple cycles without a stabilizing intervention. Each round of healthy exits raises the pool's average risk, which forces a further premium increase, which drives out the next-healthiest tier — a compounding dynamic that, left unaddressed, can shrink an individual market to a rump of the highest-cost, least price-sensitive enrollees, sometimes prompting insurers to exit the market entirely rather than keep re-pricing a shrinking, sicker pool.
Washington State's 1993 health reform law introduced guaranteed issue and community rating to its individual insurance market — but, crucially, without an individual mandate to enroll. The predictable result unfolded over several years: healthy individuals, no longer facing medical underwriting but also under no obligation to buy coverage, increasingly waited to purchase insurance until they got sick, since insurers could no longer deny them for pre-existing conditions.
As the insured pool skewed progressively sicker, premiums rose sharply — reported increases of 40–70% for some carriers within a few years — and insurer after insurer exited the individual market rather than continue re-pricing into a shrinking, adversely-selected pool. By the late 1990s, the number of carriers offering individual coverage in Washington had fallen from around 19 to just a couple, and the legislature repealed guaranteed issue in 1995 specifically because the reform without a mandate had proven commercially unsustainable. This episode is frequently cited in the academic and policy literature as the clearest empirical case of a real-world adverse selection death spiral.
The Washington State case is often summarized as: guaranteed issue plus community rating minus an enrollment mandate equals eventual market collapse. It became a central piece of evidence cited when Congress designed the ACA's original individual mandate over a decade later.
In principle, an insurer could keep repricing indefinitely to match a deteriorating risk pool, in a pure actuarial sense. In practice, several forces make exit more likely than indefinite repricing:
• Regulatory friction: state regulators can slow-walk or reject large sequential rate increases, especially when the political optics of "another 40% increase" become untenable, leaving insurers pricing below true cost for a period. • Enrollment scale economics: as the pool shrinks, fixed administrative costs are spread across fewer members, adding cost pressure on top of the pure claims-driven increases. • Regulatory and reputational risk: a shrinking, loss-making line of business becomes an easy target for internal capital reallocation — insurers often find it simpler to exit a small, structurally unprofitable market segment than to continue in it. • Adverse-selection acceleration near the endgame: as remaining enrollees become an increasingly small, sick, price-inelastic group, further increases barely shed any additional members, meaning the marginal repricing does little to fix the loss ratio — reducing the incentive to keep trying.
The medical loss ratio (MLR) — claims paid divided by premiums collected — is the single number that most directly reveals where a pool sits on the spiral. A healthy, correctly-priced pool runs an MLR close to the 80% regulatory floor, leaving room for administrative costs and a small margin. As adverse selection progresses, realized claims outpace the premium that was set on last year's healthier pool, and the MLR climbs past 100% — meaning the insurer pays out more in claims than it collected in premium, before administrative costs are even counted.
Because the ACA's MLR rule works in only one direction — insurers owe rebates to enrollees if the ratio falls below 80%, but there is no equivalent cap forcing insurers to absorb losses above 100% — a sustained loss-ratio spike is a reliable leading indicator that a rate filing far larger than typical medical trend is coming, which is exactly the trigger that restarts the exit cycle for the next plan year.
| Product | Indication | Trial Design | Key Result |
|---|---|---|---|
| Washington State (1993–1995) | Guaranteed issue + community rating, no mandate | Healthy enrollees delayed purchase until sick; carriers repriced repeatedly then exited | Reform repealed 1995 — cited as the canonical spiral case |
| Kentucky (1994–1997) | Guaranteed issue + community rating, no mandate | Similar exit-and-reprice pattern; most individual-market carriers withdrew | Individual market effectively collapsed statewide |
| New York (1993–2013, pre-ACA) | Guaranteed issue + community rating, no mandate | Persistently small, high-cost individual market with few carriers for two decades | Chronically elevated premiums rather than outright collapse |
| Massachusetts (2006, Chapter 58) | Guaranteed issue + community rating + individual mandate | Mandate paired with subsidies from the outset — the direct template for the ACA | Stable, broad individual-market enrollment; near-universal coverage achieved |
The death spiral is not an inevitable feature of guaranteed-issue insurance markets — it is a specific consequence of leaving the enrollment decision fully voluntary and fully price-exposed under a single community rate. Four policy instruments, alone or combined, directly counteract the mechanism: individual mandates change the enrollment decision itself, subsidies change the effective price faced by price-sensitive enrollees, and risk adjustment plus reinsurance redistribute the financial consequences of risk without touching who enrolls.
An individual mandate attaches a financial penalty (or, in some designs, a tax benefit for compliance) to the decision not to enroll, directly counteracting the asymmetric exit problem from Stage 3 by making the "wait until sick" strategy more costly. Unlike subsidies, a mandate does not change the sticker price of insurance at all — it changes the relative cost of the alternative (staying uninsured).
The ACA's federal mandate penalty was reduced to $0 starting tax year 2019 (via the 2017 Tax Cuts and Jobs Act), effectively eliminating the federal mandate's financial teeth while leaving guaranteed issue and community rating fully in place — reproducing, at national scale, a milder version of the exact structural gap that caused the Washington State collapse. Several states (Massachusetts, New Jersey, California, Rhode Island, Vermont, and DC) responded by enacting their own state-level individual mandates to preserve the enrollment incentive locally.
Where mandates change the cost of exiting, subsidies change the cost of staying — directly targeting the price-elastic healthy segment identified in Stage 3. Because ACA premium tax credits scale with income and are pegged to the benchmark plan cost, they automatically grow larger when premiums rise, cushioning subsidized enrollees from exactly the rate increases that would otherwise push them out.
The American Rescue Plan Act (2021) and Inflation Reduction Act (2022) enhanced ACA subsidies by removing the 400%-of-poverty "subsidy cliff" and capping payments at 8.5% of household income at every income level, and lowering required contributions for lower-income brackets toward zero. Marketplace enrollment rose from roughly 11 million (2020) to over 21 million (2024) over this subsidy-enhanced period, alongside a documented improvement in the risk mix as previously price-sensitive healthy enrollees who had stayed out returned to the pool.
Risk adjustment and reinsurance solve a different, related problem: even with mandates and subsidies stabilizing overall enrollment, individual insurers still face the risk of attracting a disproportionately sicker sub-population relative to competitors (e.g., because their provider network is more attractive to people managing chronic conditions). Without a fix, this would punish insurers for attracting sick enrollees exactly as guaranteed issue is meant to protect.
ACA permanent risk adjustment transfers funds from insurers with lower-risk enrollee populations to insurers with higher-risk populations within the same state and market, based on a standardized risk score calculated from enrollee diagnoses — over $12 billion moved between insurers in a single recent year. Reinsurance, used federally from 2014–2016 and since revived through many state Section 1332 waiver programs, reimburses insurers directly for a share of very high-cost individual claims, reducing the amount that must be recovered through premiums and estimated to lower marketplace premiums by roughly 6–20% in states that have implemented it.
No single tool fully substitutes for the others: a mandate without subsidies can be politically unsustainable and regressive for low-income enrollees; subsidies without a mandate leave healthy-but-unsubsidized enrollees exposed to the original exit dynamic; and risk adjustment/reinsurance stabilize insurer finances but do nothing to change who enrolls in the first place. The four tools are complements, not substitutes — which is why this simulator models mandate strength and subsidy level as two independent sliders acting on the same pool.