Medicare Advantage enrollment reached 51.6% of eligible Medicare beneficiaries in 2026, driven by 1.1 million net new enrollees concentrated in Special Needs Plans and non-SNP HMOs (Arnold Ventures, July 2026). KFF's tally of the same enrollment season puts the figure higher, at 55%, or 35.2 million of 64.2 million eligible beneficiaries (KFF, July 2026). Whichever number a plan actuary reaches for, the arithmetic underneath it has changed: more than half of Medicare is now priced and risk-adjusted through private plans, and the fee-for-service population left behind is not a random half of what it used to be.
That composition shift is not a future risk to flag for the 2030s. It is already showing up in MedPAC's payment estimates, in the CMS risk model, and in how plan leaders are describing their own 2027 pipelines. Three distinct mechanisms are converging on the same bid cycle: a shrinking favorable-selection subsidy, a second and separate risk-score compression layered on top of an already fully-weighted model, and enrollment growth that plan executives themselves now expect to flatten or reverse.
Two Measures of the Same Milestone
The gap between Arnold Ventures' 51.6% and KFF's 55% reflects two different denominators applied to a very similar enrollment count, not a rounding error, and it moves the answer by enough points to flip a county from apparent below-average penetration to above-average depending on which baseline a competitive analysis uses (KFF, July 2026; Arnold Ventures, July 2026). We walked through the denominator mechanics and the SNP-versus-non-SNP composition shift behind the 2026 growth figure in detail in an earlier analysis of the enrollment mix; the short version is that SNPs absorbed 85% of the entire net enrollment gain, expanding to 8.2 million enrollees, while non-SNP individual plans, the segment most actuaries think of as the core MA product, grew by only 0.2 million and group MA enrollment fell for the first time since 2010 (KFF, July 2026). The headline penetration number is rising on the back of a narrower, more clinically complex slice of the program than it was five years ago, which is the backdrop for the three implications below that a single enrollment-mix reading does not fully capture.
Why the Marginal Enrollee Looks Different Than the 2015 Enrollee
MedPAC's March 2026 status report to Congress puts a number on the mechanism that made two decades of MA growth profitable: MA payments in 2026 will run $76 billion, or 14%, above what the same beneficiaries would have cost in traditional Medicare, split into an 11-point favorable-selection component (healthier-than-average beneficiaries choosing MA) and a 4-point coding-intensity component (MA risk scores running higher than a comparable FFS beneficiary's score for the same health status) (MedPAC, March 2026). Coding intensity is a documentation and workflow variable a plan can influence directly through coding audits and provider engagement. Favorable selection is not; it is a function of who is left to enroll.
That distinction is the one plan actuaries should carry into 2028 bid modeling. The first beneficiaries to choose MA in the 2000s and 2010s were disproportionately younger-old, mobile, and healthier than the FFS population they left, because those are the beneficiaries most willing to trade provider choice for lower cost sharing and richer supplemental benefits. Every enrollment cycle since has drawn from a progressively less self-selected population. The CBO now projects national penetration climbing to 63% by 2034 and holding there through 2036 (CBO, cited in KFF, July 2026), leaving under 40 points of the Medicare population still to convert, split between beneficiaries who actively want an open network, rural beneficiaries in counties with thin MA plan availability, and higher-acuity beneficiaries for whom Traditional Medicare is a clinical necessity rather than a preference. As that pool of convertible, favorably-selected beneficiaries shrinks, the 11-point favorable-selection component of MedPAC's overpayment estimate should compress mechanically, independent of any regulatory action. A plan actuary who extrapolates 2020-2025 margin trends into a 2028 bid without adjusting for this compositional shift is implicitly assuming the favorable-selection subsidy holds constant even as its source population disappears.
2027 Layers a Second Compression Mechanism Onto an Already Full-Weight Model
It is common shorthand, including in some coverage of the 2026 enrollment data, to describe 2027 as the year the V28 risk model reaches full weight. That is not quite right, and the distinction matters because it changes which mechanism plans need to model. CMS completed the three-year phase-in of V28 with payment year 2026, moving from a 67%/33% V24/V28 blend in 2024 to 33%/67% in 2025 to 100% V28 in 2026, a transition CMS projected would compress average MA risk scores by 3.12%, worth roughly $11 billion in reduced payments (see our V28 full-weight analysis). That compression has already flowed through the current bid year.
What 2027 actually adds is a second, distinct compression layered on top of an already fully-weighted model. CMS finalized a rule excluding diagnoses drawn from unlinked chart review records, meaning diagnostic codes not tied to a specific clinical encounter, from risk score calculation starting in calendar year 2027, a change CMS estimates will cut MA payments by $7.12 billion (see our unlinked chart review analysis). Separately, the CY 2027 model recalibration updates the underlying FFS cost and diagnosis data the coefficients are built from, shifting from 2018 diagnoses and 2019 expenditures to 2023 diagnoses and 2024 expenditures, a five-year jump that reflects a different coding and utilization environment than V28's original calibration captured. CMS finalized a 5.33% effective growth rate for 2027, translating to a 2.48% average net payment increase, or roughly $13 billion in additional plan payments once all these components net out (CMS, April 2026). Plans that modeled 2027 as a continuation of the 2026 V28 transition are modeling the wrong mechanism. The two 2027 changes compound rather than repeat, and they land on the same book of business whose marginal enrollee is already trending toward higher acuity, leaving less margin to absorb estimation error in either direction.
The Enrollment Reversal Already Showing Up in Plan Survey Data
National growth in 2026 masks a reversal plan leaders are already pricing into 2027. In a January 2026 survey of leaders at more than 35 health plans, HealthScape Advisors found 23% expected overall MA enrollment to decline in plan year 2027, up sharply from 9% in the prior year's survey, while 69% expected their own plan's membership to hold flat or contract (HealthScape Advisors, 2026). Ninety-three percent said their MA line of business was not currently profitable, and 79% did not expect profitability for at least two more years. More than half had at least "somewhat seriously" considered exiting the MA market in the prior year (HealthScape Advisors, 2026).
That caution is consistent with what has already happened to disenrollment. Roughly 2.9 million MA enrollees were forced to switch plans for 2026 after their plan exited their county, a tenfold increase from the sub-2% annual rate that held from 2018 through 2024 (see our analysis of the 2026 exit wave). Most of those beneficiaries found another MA plan rather than reverting to Traditional Medicare, but the pattern of who stays and who leaves under benefit and network pressure is exactly the sorting mechanism that determines whether the remaining MA pool gets healthier or sicker relative to the FFS benchmark population it is measured against.
County-Level Penetration Already Exceeds the National Threshold
A national figure of 51.6% or 55% obscures wide dispersion. Some markets crossed the majority threshold years ago and are already operating in the low-conversion, high-acuity-marginal-enrollee environment the national data is only now approaching.
| Market | Approximate MA Penetration | Source |
|---|---|---|
| National (Arnold Ventures denominator) | 51.6% | Arnold Ventures, July 2026 |
| National (KFF denominator) | 55% | KFF, July 2026 |
| Monroe County, NY | 82% | Becker's Payer Issues, 2026 |
| Miami-Dade County, FL | 75%+ | Becker's Payer Issues, 2026 |
| Palm Beach County, FL | 75% | Becker's Payer Issues, 2026 |
Carrier concentration compounds the geographic dispersion. UnitedHealth Group holds 26% of national MA enrollment, or 9.3 million members, down 647,000 from 2025, while Humana holds 20%, or 7.0 million, up 1.3 million (KFF, July 2026). Combined, the two carriers hold 46% of national enrollment, unchanged from 2025, but in 889 counties, 28% of all U.S. counties, those two carriers alone control at least 75% of MA enrollment (KFF, July 2026). A plan actuary pricing a bid in one of those 889 counties is not modeling a market approaching the majority threshold. They are modeling one that reached it, and reached duopoly-level concentration, well before the national figures did. National trend data lags what actuaries in those counties have already had to price around for several bid cycles.
The growth mechanics in a saturated, concentrated county are also different in kind, not just degree. Below roughly 40% penetration, a plan can grow largely by converting FFS beneficiaries who have never had an MA option evaluated against their specific provider relationships, so a modest premium or benefit advantage is often enough to win volume. Above 75%, with two carriers already holding most of the book, incremental growth increasingly means taking share from the other incumbent rather than converting new FFS entrants, a dynamic that competes on network breadth and provider-contract terms more than on headline benefit value. That shifts the marginal cost of winning a member: in a low-penetration county, the marginal new enrollee costs roughly what the average enrollee costs, while in a duopoly county like Monroe or Miami-Dade, the marginal enrollee is disproportionately a switcher whose claims history and provider attachments the receiving plan does not yet have, which is exactly the kind of member for whom recent completion factors and IBNR development patterns carry the least predictive value. Pricing actuaries building county-level bids should flag saturated counties for a separate, more conservative credibility standard rather than applying the same statewide morbidity assumption used in counties still in the conversion phase.
Actuarial Implications for the 2028 Bid Cycle
The practical consequence for plan actuaries is that 2028 bids cannot simply extrapolate 2023-2026 trend, utilization, and risk score assumptions forward, because the population generating that trend data was itself in a favorable-selection phase that is now ending. Three specific modeling adjustments follow directly from the mechanisms above.
First, credibility weighting of experience periods should shift. The high-growth years from roughly 2010 through 2024 mixed a stable core population with a continuous inflow of newly-converted, favorably-selected enrollees, which understates the morbidity an actuary should expect once that inflow slows. HealthScape's finding that 69% of plan leaders expect flat or contracting membership in 2027 suggests the experience base is entering a steadier-state period that should carry more diagnostic weight for baseline morbidity than the high-growth years, not less.
Second, the denominator for fixed cost allocation needs an explicit enrollment-trend assumption rather than an implicit growth assumption. Administrative cost, network contracting overhead, and quality-improvement spending were priced against a book that was growing 3 to 4% a year through most of the last decade. A plan modeling flat or slightly declining 2027-2028 enrollment needs to reallocate those fixed costs across a smaller or flatter denominator, which raises the per-member cost baseline independent of any change in medical trend.
Third, the vintage effect needs to be disentangled from secular trend. A book's blended risk score reflects a mix of long-tenured members (who joined during the favorable-selection era) and recent entrants (who look more like the FFS population left behind). As the recent-entrant share grows and the long-tenured share ages further into higher morbidity, both forces push the blended risk score up simultaneously, and an actuary who reads that increase as pure trend will overstate how much of it reflects the underlying secular medical cost growth versus a one-time compositional shift that will not repeat once the mix stabilizes.
None of this argues that MA growth has ended. CBO's 63%-by-2034 projection still implies a decade of further conversion. It argues that the economics of each additional point of penetration are changing, and that the bid cycle building on 2026 and 2027 experience is the first one where plan actuaries need to price that change explicitly rather than assume it shows up gradually enough to ignore.