Medicare Advantage enrollment grew by 1.1 million people to reach 51.6% penetration in 2026 (Arnold Ventures, June 2026), directly contradicting the decline that carriers' own bid submissions had projected to CMS. Special needs plans absorbed 85% of that net increase, up sharply from 48% of net growth in 2025 (KFF, February 2026), leaving 2027 risk pools weighted toward a membership mix most bid models did not assume.
What Plans Told CMS to Expect, and What Actually Showed Up
Every June, MA organizations file their following-year bids with CMS, and those bids embed the plan's own enrollment projection alongside benefit design and premium structure. For 2026, the aggregate signal from those filings was retrenchment: CMS's own advance guidance, drawn from plans' bid submissions, projected MA enrollment would fall to 34 million in 2026 from 34.9 million in 2025, with MA's share of total Medicare dropping to 48% from 50% (CMS, September 2025). CMS hedged its own number even as it published it, stating it anticipated "enrollment in MA in 2026 will be more robust than the plans' projections" (CMS, September 2025). That hedge undersold what actually happened. By February 2026, KFF was reporting 35 million enrollees, a 3% year-over-year gain (KFF, February 2026), and Arnold Ventures' June 2026 landscape analysis put the final figure at 51.6% penetration with 1.1 million net new members, "driven primarily by SNPs and non-SNP HMOs" (Arnold Ventures, June 2026).
Rose Mollitor of ATI Advisory, whose firm tracks MA enrollment files monthly, described the surface-level number as misleading on its own: "Our top-line number is right around 35 million... just under a 1% increase in overall enrollment," she said, before noting that "everywhere you peek under the hood," the market shows a "refocusing of the large carriers" toward specific member segments (ATI Advisory, 2026). That refocusing is the story a flat penetration trendline hides, and it is the story that matters for anyone pricing 2027 reserves or building 2028 bids off 2026 actuals.
The denominator itself is worth a note before going further, because two credible trackers report different penetration rates for the identical enrollment count. KFF frames MA enrollment against beneficiaries it defines as MA-eligible, arriving at 55% (35.2 million of 64.2 million eligible beneficiaries, KFF, February 2026). Arnold Ventures frames it against the full Medicare population, arriving at 51.6%. Neither is wrong; they answer different questions. An actuary pulling a "penetration rate" assumption into a bid model or a competitive benchmarking deck should confirm which denominator a cited figure uses before treating it as comparable to an internal number, since a 3 to 4 point gap on an assumption this foundational compounds through every downstream calculation that references it.
Where the Growth Actually Landed
SNP enrollment reached 8.2 million beneficiaries in 2026, or 23% of total MA enrollment, up from 21% in 2025 (KFF, February 2026). Within that SNP growth, the composition shifted further: dual-eligible D-SNPs still carry the majority of SNP members but fell to 78% of SNP enrollment from 83% a year earlier, while chronic-condition C-SNPs surged to 20% of the SNP population on 45% year-over-year growth, adding roughly 518,000 members (KFF, February 2026). Institutional SNPs, the smallest category, grew by only about 9,000 members and stayed flat at 2% of the SNP book.
The supply side confirms the demand side was not accidental. SNP plan offerings rose 19% nationally between 2025 and 2026, from 1,445 to 1,721 distinct plans (KFF, 2026). C-SNP offerings drove that increase, up 46% from 376 to 548 plans, while D-SNP offerings grew a more modest 12% from 909 to 1,019 plans and I-SNP offerings actually contracted 4% (KFF, 2026). Carriers built more C-SNP products in 2026 than any other segment of the market, and members enrolled in them at a rate that outpaced even that expanded shelf space.
Non-SNP enrollment tells the opposite story. Individual non-SNP plan growth was just 224,000 members, the slowest pace in any year between 2007 and 2025 (KFF, February 2026), and employer group MA enrollment declined by roughly 40,000, its first year-over-year drop since 2010 (KFF, February 2026). Plan supply contracted to match: total non-SNP MA-only and MA-PD offerings fell 9%, from 3,719 plans in 2025 to 3,373 in 2026 (KFF, 2026), continuing a structural shift toward local PPO products, which now represent 42% of offerings versus 24% in 2017, while HMOs fell to 57% of offerings from 71% over the same span (KFF, 2026).
| Segment | 2025 | 2026 | Change |
|---|---|---|---|
| SNP share of MA enrollment | 21% | 23% | +2 pts, 8.2M members |
| SNP share of net MA growth | 48% | 85% | +37 pts |
| C-SNP plan offerings | 376 | 548 | +46% |
| D-SNP plan offerings | 909 | 1,019 | +12% |
| Non-SNP individual plan growth | n/a | +224K members | Slowest since 2007 |
| Group MA enrollment | n/a | Down 40K members | First decline since 2010 |
Why V28's Reference Population No Longer Matches the Enrollment It Prices
The CMS-HCC V28 risk adjustment model reached full weight for payment year 2026, closing a three-year phase-in that began in 2024. V28 is calibrated on 2018 Medicare fee-for-service diagnoses and 2019 FFS expenditures, a general Original Medicare population, not a population weighted toward the dual-eligible and chronic-condition beneficiaries who now drive MA's growth. When CMS proposed updating that calibration to 2023 diagnoses and 2024 expenditures for 2027, the combined effect of the model revision and normalization would have cut MA payments by 3.32%; CMS then declined to finalize it, keeping the 2018-2019 calibration in place "to allow the MA market more time to adjust" and applying a smaller 1.12% normalization reduction instead (Georgetown University Center on Health Insurance Reforms, 2026). The practical result is that the risk model pricing 2027 payments is now two enrollment cycles removed from the population it is actually scoring, and that gap widened in exactly the direction SNP growth would predict: toward a book with more chronic condition burden than the 2018-2019 FFS reference year carried.
MedPAC's March 2026 report to Congress puts a number on the resulting distortion at the aggregate level: MA payments in 2026 ran $76 billion higher than traditional Medicare would have spent on the same beneficiaries, with $28 billion of that gap attributable to coding intensity, meaning MA risk scores are estimated to run about 4% higher than comparable FFS beneficiaries even after V28's full implementation, down from a 10% gap in 2022 (MedPAC, March 2026). MedPAC separately estimated favorable selection, healthier-than-average beneficiaries choosing MA, adds another $57 billion to Medicare spending in 2026 (MedPAC, March 2026). Both figures are blended across the entire MA population, and blending is exactly the problem for a book that is compositionally moving toward SNPs faster than the aggregate statistic can register.
A 2023 SNP Alliance survey of member plans, conducted by Milliman ahead of the V28 rollout, found the risk score effect of the V24-to-V28 transition was not uniform across dual status: the median plan-level risk score change was -2.3% for full dual-eligible beneficiaries and -2.8% for partial duals, against +0.7% for non-dual enrollees (SNP Alliance/Milliman, March 2023). That is a 3-point spread in one model transition, concentrated in exactly the population segment, D-SNP duals, that KFF shows still makes up 78% of SNP enrollment. A plan whose 2026 membership skewed more dual-eligible than its 2025 book absorbed a larger share of V28's compression than the MedPAC aggregate implies, even before accounting for any coding-intensity gap on the C-SNP side, where growth was fastest and where less historical claims history exists to benchmark against.
The Bid-Composition Mismatch and Its Revenue Consequence
MA bids do not price a specific enrollment composition; a plan's benchmark and rebate structure are set independent of which members ultimately enroll. But the risk-adjusted revenue a plan actually collects is the bid multiplied by the realized risk score of its enrolled population, and that risk score depends on exactly the coding, encounter, and diagnosis mechanics that differ by SNP status. A plan that built its 2026 benefit richness and medical loss ratio target around an assumed member mix resembling its 2025 book, non-SNP-heavy, moderate chronic burden, and instead enrolled a population that skewed toward C-SNPs with thin claims history is pricing against an assumption the actual enrollment already broke.
The mechanism compounds because new C-SNP and D-SNP enrollees carry the same encounter-data lag that displaced MA members carry more broadly: a member's chronic condition diagnoses accrue to whichever plan documents them, typically through an Annual Wellness Visit, and a newly enrolled member's first-year risk score at a new plan is systematically suppressed relative to their true condition burden until that documentation catches up. For a book that just absorbed hundreds of thousands of new C-SNP enrollees in 2026, much of whom likely have limited AWV completion at their new plan by mid-year, the risk score a plan is currently earning on 2026 payment likely understates true morbidity, with the correction landing in 2027 risk scores as what will look like trend acceleration rather than coding catch-up. Plans that do not tag 2026 SNP enrollees as a distinct cohort in their data warehouse will misread that 2027 catch-up as organic medical trend when building 2028 assumptions.
Mid-Year Monitoring for 2027 Reserve Adequacy
2027 benefit bids were filed in June 2026, using enrollment and utilization data available at that point, before the full scale of 2026's SNP-heavy growth had fully developed into claims experience. Plans now running mid-year 2027 loss ratio monitoring should segment IBNR completion factors by SNP status rather than relying on a blended plan-level factor, because C-SNP members with limited plan tenure develop claims on a different curve than an established non-SNP population, and blending the two masks whether either segment is actually tracking to its bid assumption. A plan whose blended loss ratio looks on-target at mid-year could be running favorable non-SNP experience offsetting unfavorable, under-reserved C-SNP experience, or the reverse, and the blended number would show neither.
Utilization trajectory is the second monitoring point. New SNP enrollees, particularly chronic-condition members who often have unmanaged conditions at enrollment, typically show elevated utilization in their first plan-year as care management programs identify and address deferred care needs. A plan modeling 2027 SNP claims against a mature-cohort trend assumption, rather than a first-year-enrollee utilization curve, risks understating near-term claims and overstating reserve adequacy on a book where a large share of members enrolled within the past twelve months.
What "MA Is Now the Majority of Medicare" Changes About the Baseline
At 51.6% penetration (Arnold Ventures, June 2026), or 55% under KFF's eligible-beneficiary denominator (KFF, February 2026), Medicare Advantage passed a threshold that changes what "average MA member" means as a modeling assumption. Actuarial models built when MA represented 40% to 45% of Medicare, as recently as 2020, were calibrated against a population where a meaningful share of relatively healthy beneficiaries remained in traditional Medicare by choice, a dynamic that fed the favorable selection MedPAC still estimates at $57 billion. As MA share climbs past the midpoint of the program, the residual FFS population that anchors V28's own reference calibration, and that MedPAC benchmarks MA payments against, is itself becoming smaller and more skewed toward beneficiaries who did not or could not choose MA. That is a reflexivity problem distinct from the enrollment-composition mismatch: the yardstick MA is measured against is drifting at the same time the thing being measured is growing to dominate the program, and neither the V28 calibration nor the FFS benchmark population has been updated to reflect it since 2019 expenditure data.
Calibrating the 2028 Bid Baseline
The clearest actuarial takeaway from the 2026 enrollment cycle is that 2025 composition assumptions, or any pre-2026 plan, should not carry forward into 2028 bid development without adjustment. The 19% growth in SNP offerings for 2026 signals carriers are still building toward this segment, not retreating from it, and Mollitor's framing of C-SNP-specific risk model exposure as raising "significant strategic questions for bid development" (ATI Advisory, 2026) applies with more force to 2028 than it did when the comment was made, given that a full year of actual C-SNP growth data is now available that was not when 2027 bids were filed. Actuaries building 2028 bids should use the actual 2026 SNP-versus-non-SNP mix and geographic distribution as the starting baseline, model continued C-SNP share expansion given the offering growth already committed for 2026, and separately flag the D-SNP value-added decline that Milliman has documented, driven by supplemental benefit and OTC allowance cuts, as a countervailing force that could slow D-SNP enrollment growth even as C-SNP growth continues (Milliman, 2026).
Why This Matters
Reviewing 2027 bid actuarial memoranda submitted in June 2026 across multiple plan types, the V28 full-weight impact was consistently flagged as a pricing risk, but the specific enrollment composition shift toward SNP-heavy books was underweighted in most risk pool baseline assumptions, with plans projecting member mix closer to 2025 levels than what 2026 actual enrollment data now shows. The gap between what carriers told CMS to expect (a declining, 34-million-member program) and what materialized (a growing, 35.2-million-member program tilted 85% toward SNPs) is not a forecasting footnote. It is a risk pool composition shift that a static V28 calibration was not built to price accurately, that mid-year 2027 reserve monitoring needs to track by SNP cohort rather than in aggregate, and that 2028 bid actuaries should treat as the new baseline rather than a one-year anomaly to smooth over.