Two insurers filing 2027 individual-market rates loaded morbidity adjustments of 6.0% and 4.7% onto the same underlying trend, both citing the identical mechanism: enrollees who lapse after losing enhanced premium tax credits skew healthier than those who stay (Peterson-KFF Health System Tracker, July 2026). Getting that single factor right takes more than a trend study.

The Multiplicative Stack Behind the 14% Median

KFF's July 8, 2026 review of preliminary 2027 filings from 77 insurers across 16 states and DC put the median proposed increase at 14%, down from an 18% median proposed and 20% median finalized a year earlier, but still a second consecutive double-digit cycle (KFF, July 2026). Insurers attribute the bulk of it to two components: a median underlying medical and prescription trend of 10%, itself above the roughly 8% average of prior years and driven partly by GLP-1 utilization that one filer reported rising from 1.6% to 5.4% of members between 2025 and 2027, and a second consecutive subsidy-expiration morbidity load of roughly four percentage points, mirroring the load insurers applied a year earlier (Peterson-KFF Health System Tracker, July 2026).

The correct index-rate build stacks these multiplicatively, not additively: Projected Index Rate PMPM equals Base Period Allowed Claims PMPM times a trend factor, times a morbidity or population adjustment, times a benefit and induced-utilization adjustment, times a load for administrative expense and margin, before the plan-level adjustment for risk adjustment transfer and reinsurance. A 10% trend factor and a 4% morbidity factor compound to 1.10 times 1.04, or 14.4%, which happens to land close to the reported 14% median once benefit and admin residuals are netted out. That near-coincidence is deceptive at the plan level. The insurer applying a 6.0% morbidity factor against its own trend assumption, or the one applying 4.7%, is not adding a fixed four points to a common base; each is compounding a carrier-specific morbidity estimate against a carrier-specific trend, and the two must be derived independently before they are ever multiplied together.

Deriving the Induced-Lapse Morbidity Adjustment

The morbidity factor is not a market benchmark to be copied from a competitor's filing. It is derived from a lapse-rate assumption and a relative-morbidity assumption for the members who lapse. Start from the net-of-subsidy premium increase a cohort faces after enhanced credits expire, apply a price elasticity to size the resulting lapse rate, then weight the retained pool by the claims each departing member would otherwise have generated. If L is the lapse rate and r is the relative morbidity of lapsers to the pool average, expressed as a factor below 1.0 because departing members skew healthier, the retained pool's morbidity index relative to the base period is (1 minus r times L), divided by (1 minus L).

Plugged into 2026's realized, not projected, national experience, the formula is a useful sanity check before it becomes a 2027 assumption. National marketplace plan selections fell from 24.2 million to 23.1 million, an aggregate decline near 4.5%, against a realized average net premium increase of 58%, from $113 to $178 per month, well below the 114% increase KFF had projected before enrollees shifted plan mix and the most price-sensitive members exited (KFF, 2026). Setting L at 4.5% and r at 0.6, a plausible relative-morbidity factor for members healthy enough to be highly price-sensitive, returns a retained-pool morbidity index of roughly 1.019, an aggregate shift of under two points. That is well short of the four-to-six-point loads insurers actually filed, and the gap is the article's second-order point: an aggregate, market-wide elasticity applied to an aggregate lapse rate systematically understates the true induced-lapse morbidity shift, because lapse and morbidity are not distributed evenly across the pool.

Why Cohort Segmentation Closes the Gap

KFF's 2026 enrollment data shows exactly where the concentration sits. Enrollees between 400% and 500% of the federal poverty level, the cohort that lost subsidy eligibility entirely at the enhanced-credit cliff, made up just 3% of 2025 plan selections but accounted for 27% of the total 2025-to-2026 decline, a 44% drop within that cohort alone, or roughly 321,000 people (KFF, 2026). Running the same retained-morbidity formula on that segment in isolation, with L at 0.44 and a lower r of about 0.5 to reflect its concentration of price-sensitive, low-utilization members, returns a segment-level morbidity index near 1.39, a swing of nearly 40 points within that narrow slice of the pool. Weighted by its 3% pool share, that single cohort alone contributes more than a full point to the aggregate morbidity load, and adults 18 to 34, who accounted for 542,000 of the total enrollment decline, or 46% of it, contribute a second concentrated slice (KFF, 2026).

This is why the actuarially defensible version of the calculation segments the pool by income band and, within each band, by health status, before applying any elasticity, rather than running one blended L and r against the whole pool. Silver-loading and cost-sharing reduction interactions add a further wrinkle: CSR-eligible enrollees near the subsidy cliff face a different net-premium schedule than unsubsidized enrollees in the same metal tier, so the differential elasticity is not just an income effect but an actuarial-value effect, and a load built on a single blended elasticity will misallocate the morbidity shift across metal tiers even if the pool-wide total happens to land near the right number.

The Double-Counting Trap Between 2026 and 2027

The four-point load insurers are citing for 2027 is described in filings as mirroring the four-point load applied for 2026, and that phrasing is the double-counting risk. If a carrier's 2027 rate development uses a 2025 base period, re-deriving the full four-point induced-lapse shift as if none of it had yet occurred, it re-bills deterioration that already ran through 2026 experience and 2026 premium. The defensible construction uses the 2026 base period, which already embeds the first wave of subsidy-driven lapses, and loads only the incremental, still-to-emerge second wave: the further lapses expected between the 2026 and 2027 plan years as remaining price-sensitive members continue exiting and as OBBBA-driven administrative enrollment barriers compound the effect. Partitioning total anticipated deterioration into an already-realized component, visible in 2025 run-out and early 2026 paid claims, and a still-anticipated component is the step that separates a defensible incremental load from a re-application of last year's number under a new label.

Credibility-Weighting Thin 2026 Experience

The complication is that the "already-realized" component is not yet fully observable. Mid-2026 claims run six to seven months deep, immature relative to a typical 12-month completion pattern for individual-market business, and the 2026 open enrollment cohort's morbidity signal is diluted by incurred-but-not-reported reserves and seasonal utilization patterns that have not yet played out. A limited-fluctuation or Buhlmann credibility weight, Z equal to n divided by the sum of n and a stabilizing constant k, applied to the emerging 2026 morbidity indication against the elasticity-derived manual estimate, keeps a thin six-month experience period from swamping a structurally sound assumption. A carrier with a small book and six months of paid claims might credibly weight the emerging signal at 20% to 30%, leaning on the manual elasticity estimate for the remainder; a large statewide carrier with a full calendar year of a comparable prior cohort might weight emerging experience considerably higher. Skipping this step and pricing directly off six months of immature claims is as much an error as ignoring emerging experience entirely.

Netting the Load Against the Risk Adjustment Transfer

A market-wide morbidity shift does not flow through to premium dollar for dollar, because the HHS-HCC risk adjustment transfer formula compares each plan's risk score, adjusted for actuarial value and induced demand, against the statewide average of the same measures (CMS, Risk Adjustment Methodology Overview). When the entire state's risk pool gets sicker, the statewide average Plan Liability Risk Score rises along with any individual carrier's own score. A carrier only retains the morbidity load's premium impact to the extent its own risk score diverges from that rising state average, not to the extent the state average itself moved. A common ratemaking error is loading the full gross morbidity shift into premium as though risk adjustment were absent, when in fact a carrier whose enrollee mix tracks the state average closely should expect most of a market-wide shift to be neutralized through transfers, while a carrier whose book skews sicker than the state average, because its price point or provider network attracts higher-utilization members, retains a disproportionate share of the load even before its own experience deteriorates.

Pricing the Subsidy-Restoration Contingency

Every 2027 filing built on a subsidy-expiration morbidity load carries an unstated policy assumption: that Congress does not restore enhanced credits mid-cycle. A restoration would reverse the induced-lapse dynamic, pull healthier members back into the pool, and leave a filed rate that priced for a sicker pool sitting above the emerging reality. That is a materially different risk than ordinary trend uncertainty, and it argues against burying it inside a single point-estimate morbidity factor. ASOP No. 12 calls for actuaries to identify and document the basis for risk classification and rating variable assumptions, and ASOP No. 8 calls for disclosure of significant assumptions and their sensitivity in a regulatory rate filing (Actuarial Standards Board). A documented sensitivity, showing the filed rate alongside a restoration scenario and the resulting premium deficiency or margin, satisfies that disclosure obligation more directly than folding a policy contingency into base-rate margin where a regulator or auditor cannot isolate it. The spread between the roughly four-to-six-point loads insurers self-applied and the 7.7% subsidy-expiration estimate CBO produced, cited in at least one insurer's own filing, is itself evidence of how wide the plausible range runs (Peterson-KFF Health System Tracker, July 2026); a single point estimate with no disclosed range understates that uncertainty.

Carriers that decompose the load into its trend, elasticity, cohort, and risk-adjustment-net components, and that document the restoration contingency as a disclosed sensitivity rather than an unstated assumption, will be positioned to explain a 2027 rate to a regulator in a way that a single blended morbidity percentage borrowed from last year's filing cannot support.

Further Reading

Sources