ACA Enrollment Fell to 19.2M in 2026. HHS Says It Was a Fraud Purge, Not Attrition

19.2 million people carried effectuated ACA Exchange coverage as of February 2026, down from 22.1 million at the close of 2025, and the Department of Health and Human Services says every one of those 2.9 million lost enrollees was improper or phantom, not price-driven attrition (HHS ASPE, June 2026). Improper enrollment itself fell from 5.6 million in 2025 to 2.6 million in 2026. If the composition claim holds, the residual pool got smaller and sicker in a very different way than a subsidy-driven exodus would produce.

The ASPE issue brief, published June 26, 2026, is doing something more specific than reporting a headcount. It is making a claim about who left, and that claim is the actuarially load-bearing part of the report. A 13% enrollment decline driven by young, healthy people walking away from a suddenly unaffordable premium worsens the risk pool. A 13% decline driven by removing people who were never real claimants, ghost enrollments generated by broker fraud, duplicate records, or misreported income, does something close to the opposite: it removes low-or-zero-cost membership from both the numerator and the denominator of the morbidity calculation without materially changing the claims side of the ledger. Those two stories produce very different 2027 rate filings, and the dispute over which one is true is not settled. KFF's own tracking survey found that among returning 2026 marketplace enrollees, 73% worried about affording emergency care or hospitalization and 44% said higher premiums made it harder to afford groceries, utilities, or rent, evidence that at least some of the exit was priced-out attrition, not fraud cleanup.

What the ASPE Brief Actually Measures

ASPE's core data source is CMS's monthly effectuated enrollment count, the same series that CMS's effectuated enrollment reports have published since the exchanges opened, cross-referenced against subsidy verification and program-integrity screening data. The February 2026 figure of 19.2 million is 85% above February 2019's baseline, a comparison ASPE uses to argue that even after the 2026 contraction, the exchanges remain a structurally larger market than in the pre-pandemic era (HHS ASPE, June 2026). CMS's own historical effectuated-enrollment series put the 2019 full-year average near 9.8 million, which is the pre-enhanced-subsidy floor the market has been building from for six years. The improper-enrollment estimate is where the report's argument concentrates. ASPE traces improper and phantom enrollment growing from roughly 4.4 million in 2024 to a peak of 5.6 million in 2025, then falling to 2.6 million in 2026 as verification tightened, including more than 1 million enrollments tied to no valid Social Security number (HHS ASPE, June 2026). The brief credits two separate program-integrity tracks for the 2.9 million removed: about 1.5 million enrollees blocked from receiving subsidies they did not qualify for, and roughly 1.4 million more ended or blocked through the elimination of year-round special enrollment pathways that had allowed off-cycle sign-ups. Add those two tracks and the total matches the headline decline almost exactly, which is the numerical basis for ASPE's claim that the net change is a program-integrity story, not a demand story.

The Utilization Evidence Behind the Phantom-Enrollee Claim

The stronger evidence for ASPE's framing does not come from the enrollment count itself. It comes from claims data showing what the removed population was actually costing the system. Paragon Health Institute's analysis, built on CMS's Enrollee-Level External Data Gathering Environment claims files from 2019 through 2024, found the 100% to 150% of federal poverty level income band, the fully subsidized tier where net premiums run at or near zero, carried a dramatically elevated zero-utilization rate relative to every other income band. Paragon separately estimated that roughly 6.4 million people were placed in that fully subsidized 100-150% FPL category after the 2025 open enrollment period despite not actually having that income, a misreporting pattern that produces enrollees with a policy and a subsidy but no plausible reason to file a claim. That distinction, zero utilization rather than low utilization, is the actuarial fact that separates this cleanup from ordinary adverse selection. A healthy 28-year-old who lapses because the premium tripled was, while enrolled, a legitimate low-cost risk that priced favorably against the pool average. A phantom enrollee generated by a broker double-enrolling a household member, or by income misreporting that qualifies someone for a zero-premium plan they never intended to use, was never a real risk in the pool at all; the person behind the enrollment record frequently was not accessing the coverage, filing claims, or in some documented cases aware the policy existed. Removing the first type shifts the morbidity distribution upward, because the departing cohort's below-average cost is gone from the denominator. Removing the second type does comparatively little to the morbidity distribution, because a policy that generated no claims contributed almost nothing to the claims numerator in the first place and only diluted the per-member average through its presence in the denominator.

Translating a Composition Shift Into a 2027 Morbidity Load

Single risk pool rate development starts from a base-period claims experience and projects it forward with a trend and a morbidity adjustment for any expected change in the population's risk profile, a mechanism the site's 2027 risk pool morbidity load methodology explainer lays out in full. The standard adjustment answers one question: is the 2027 target population, on average, sicker or healthier than the 2026 base-period population that generated the claims data actuaries are trending forward? A pure headcount decline answers that question badly, because a shrinking pool could be getting sicker (healthy attrition), staying flat (proportional exit across risk levels), or getting marginally healthier at the margin (zero-cost phantom removal diluting a claims base that never included their cost anyway). The ASPE decomposition matters here because it tells the actuary where on that spectrum the 2026 exit actually falls, and the honest answer from the available evidence is: partway between the two competing narratives, not fully at either end. If the full 2.9 million departure were phantom enrollees with functionally zero claims experience, their removal would be close to morbidity-neutral for the residual pool's average cost per member, because a zero-cost enrollee contributes nothing to the claims numerator and its removal from the denominator raises the per-member average only mechanically, in exact proportion to how far below the mean its true cost sat. That mechanical effect is real but bounded: a policy with genuinely zero claims removed from a 22.1-million-member pool with some non-zero average cost raises the survivors' average cost by simple arithmetic, roughly the amount that phantom enrollee was diluting the denominator by. But KFF's affordability survey data indicates the exit also includes real, previously-claims-generating members who left because the enhanced subsidy expiration made the plan unaffordable, which is a genuine adverse-selection removal of a below-average-but-nonzero-cost enrollee. A carrier's 2027 filing that assumes the entire decline was morbidity-neutral phantom-enrollee cleanup will understate the residual pool's true claims cost; one that assumes the entire decline was healthy-lapse adverse selection, on the Wakely-style model documented in the site's enrollment cliff and subsidy expiry analysis, will overstate it. The actuarially defensible position sits between the two, weighted by how much of the 2.9 million each mechanism plausibly explains, and that weighting is precisely the judgment ASPE's brief does not resolve on its own; it asserts one end of the range as the whole story.

Reading the Retention Data as a Morbidity Signal

One data point in the reporting around the ASPE brief helps narrow that weighting: 27.8% of zero-premium 2026 enrollees had their coverage cancelled by May, according to the Paragon analysis of the same underlying data ASPE drew on. A near-30% cancellation rate within the zero-premium tier, concentrated in the same income band Paragon flags for elevated misreporting, is consistent with a meaningful share of the 2026 base-period enrollment already being unstable, low-engagement membership before the year is even over, which supports weighting some of the 2026-to-2027 transition toward the morbidity-neutral end of the range rather than the full adverse-selection end. It does not resolve the dispute; it shifts the actuarial prior.

The Separate, Adverse Mechanism Still Coming in 2027

The ASPE brief describes 2026. It does not describe 2027, and the mechanism that will dominate the next filing cycle is a different one entirely: the enhanced premium tax credits that kept net premiums low through 2025 have expired, and the resulting affordability shock is a textbook adverse-selection driver, not a program-integrity story. The site's coverage of that separate mechanism, most recently in the 2027 rate filing morbidity load and subsidy expiration analysis, documents carrier-level morbidity adjustments already running well above the levels a pure enrollment cleanup would justify, because subsidy-driven exit selects specifically against the healthiest members who have the most price elasticity. That is the mechanism actuaries should treat as adverse and load conservatively for; the phantom-enrollee removal ASPE documents for 2026 is a different, comparatively benign mechanism operating in the same period. Conflating the two, treating the full 2.9 million decline as either entirely benign cleanup or entirely adverse selection, is the modeling error the ASPE brief's framing invites if read uncritically in either direction. The two mechanisms will also show up differently in a carrier's own book. A carrier whose 2026 membership loss was concentrated among duplicate or unverifiable enrollments, the population program-integrity screening targets first, should see a smaller morbidity deterioration in its 2027 base period than a carrier whose losses concentrated among price-sensitive, previously subsidy-dependent members. Carrier-level effectuation and lapse data, not the market-wide ASPE aggregate, is what actually separates those two populations for a given filer, and the effectuation-based morbidity methodology work Wakely published earlier in 2026 is built on exactly that carrier-level distinction rather than a national program-integrity aggregate.

Risk Adjustment Recalibration and a Cleaner Base Period

There is a second-order consequence to a genuinely cleaner enrollment base that has gotten less attention than the headcount: risk adjustment transfer calculations run on the same enrollee-level claims and demographic data that program-integrity screening is now scrubbing. Phantom enrollees with fabricated or misreported demographic and income data do not just dilute morbidity averages; they introduce noise into the risk score distribution that risk adjustment transfers are calibrated against. A pool with 2.6 million fewer improperly documented records, per ASPE's 2026 estimate against 2025's 5.6 million, is a pool where the HHS-HCC risk scores CMS uses to calculate transfers are being computed from more reliable inputs. The site's 2027 risk adjustment recalibration coverage has tracked how enrollment contraction shifts transfer magnitudes; a contraction driven partly by data-quality cleanup, rather than purely by population change, should produce somewhat more stable transfer estimates for 2027 than a contraction driven by pure adverse selection would, because the remaining risk scores better reflect real utilization rather than a mix of real claims and unverifiable phantom records. That stability is a modest positive for actuaries building 2027 risk adjustment transfer projections, but it should not be overstated. Program integrity verification targets income and identity documentation, not the accuracy of diagnosis coding or the completeness of HCC-relevant claims submission for enrollees who remain in the pool. A cleaner enrollee roster is not the same thing as cleaner risk scores across the board, and the credibility gain from removing 2.6 million improperly documented records is partial, concentrated in the demographic and eligibility layer of the data rather than the clinical layer.

Index Rate Development in a Disputed Data Environment

The practical problem for a 2027 rate filer is that the ASPE brief and the KFF survey data are not reconcilable into a single clean composition estimate; they are two credible sources reading the same 2.9 million decline through different lenses, and a state regulator reviewing an actuarial memorandum will expect the filer to have grappled with both rather than adopting whichever framing produces a more convenient index rate. The defensible approach is not to pick a side in the fraud-versus-affordability dispute. It is to build the 2027 base period from carrier-specific effectuation, lapse, and claims data, the same data ASPE and Wakely both draw on at the market level, and let that carrier-level evidence determine how much of the observed 2026 membership change was morbidity-neutral cleanup versus adverse-selection attrition, rather than importing either narrative's implicit assumption wholesale. A morbidity load calibrated to the market-wide ASPE framing alone risks understating claims cost for any carrier whose actual losses skewed toward price-driven exit; a load calibrated to the Wakely-style full-adverse-selection framing alone risks overstating it for any carrier whose losses skewed toward the improperly documented population. Given that the true 2026 exit was almost certainly a blend of both mechanisms, and that carrier composition of that blend will vary by market and by state, the safer actuarial default heading into 2027 filings is to treat the ASPE brief as evidence that some, not all, of the decline is morbidity-neutral, and to document that partial weighting explicitly in the actuarial memorandum rather than letting either advocacy framing set the assumption by default.