The 2026 ACA Filings Split Into Two Morbidity Camps. One of Them Is Wrong

Every ACA rate increase that made headlines this year decomposes, inside the filing, into a handful of multiplicative factors: trend by service category, demographic shift, plan design change, and the morbidity adjustment, the factor that states how much sicker or healthier the carrier expects its single risk pool to be than the experience period it priced from. The first three factors are actuarial routine. The fourth is where 2026 happened. We have now parsed all 373 individual-market Unified Rate Review Templates in the CMS Rate Review public use file, weighted every factor by projected member months, and published the result as a standing reference: the ACA Filing Assumptions Benchmark. This piece is about what the morbidity column shows, because it shows something stranger than a market repricing a shock. It shows a market that could not agree on whether the shock existed.

An 8.6-Point Reversal in One Filing Cycle

The three-year trajectory of the enrollment-weighted morbidity assumption is the cleanest single measurement of what enhanced-subsidy expiry did to individual-market pricing. For PY2024, carriers filed a weighted average morbidity factor of 0.994, pricing the risk pool as marginally healthier than their experience period. For PY2025 the assumption was -0.1%, effectively neutral. These were the years of record subsidized enrollment, when the steady arrival of healthy marginal entrants let carriers price the pool as stable. For PY2026 the weighted average jumped to +8.5%. Nothing else in the filing moved comparably: Rx cost trend rose from 5.1% to 6.3%, inpatient cost trend held near 4.3%, the demographic shift factor stayed within two points of neutral. The rate action came from morbidity, and the morbidity came from the subsidy cliff.

The scale of the associated premium movement is worth stating precisely, because the enrollment-weighted figures differ from the simple averages that circulate. Across every renewing individual-market plan, the cumulative filed rate change for PY2026, weighted by current enrollment, is +26.1%, against +6.5% for PY2025 and +6.0% for PY2024. Terminated and new plans are excluded: a terminated plan carries its members at a 0% "change" that is a crosswalk, not a rate action. Weighting matters in the other direction too: the weighted median morbidity assumption is +5.2%, well below the +8.5% weighted mean. When a mean sits that far above its median, the mass is in the large books. The biggest carriers, the ones whose assumptions move the most member months, filed worse morbidity than the market's typical filing. Whatever the small carriers believed, the institutions with the largest claims datasets priced the pool as substantially sicker.

The small group market supplies the control. Small group plans share the carriers, the networks, the provider contracts, and the trend environment, but not the subsidy mechanism, and its filed morbidity moved from +0.4% to +1.5% while its weighted rate change went from +8.3% to +11.7%. A medical-inflation story would have moved both markets together. A subsidy-expiry story moves only the market where subsidies determined who enrolled. The URRT record is unambiguous about which story the filings tell.

The Zero Camp and the Fifteen-Plus Camp

Distributions earn their keep when the mean stops being representative, and the PY2026 morbidity distribution is genuinely bimodal. Sorting all 373 filings by their filed factor and weighting by projected enrollment: 15.6% of the market's projected member months sit in filings whose morbidity adjustment is exactly 1.000, no assumed change. Another 4.2% sit below 1.000, assuming improvement. At the other end, 23.6% of projected enrollment sits in filings assuming +15% or more. The four middle buckets between +2.5% and +15% hold roughly 42% combined, but no single bucket exceeds 14.3%. The market did not converge on a consensus deterioration estimate with dispersion around it. It split into a camp that priced the cliff and a camp that priced through it.

The camps are not a large-carrier-versus-small-carrier artifact. Among the twelve largest individual-market books in the country, Blue Cross and Blue Shield of Texas filed +16.9% morbidity on roughly 789,000 average projected members, and UnitedHealthcare of Texas filed +22.2%. Celtic Insurance Company, Centene's Texas Ambetter entity, filed +26.4%, the highest morbidity assumption among the largest carriers. Meanwhile Oscar's Florida HMO filed exactly 0.0% on roughly 692,000 members, and Centene Venture Company Florida, in the same corporate family as Celtic, filed 0.0% on 614,000. The same parent organization priced +26.4% morbidity in Texas and zero in Florida. Kaiser's California book filed +3.0%; Blue Shield of California filed +0.7%. State-level market structure, not carrier identity, is doing the sorting.

The state cut makes that mechanical. Arkansas filings carry the steepest weighted morbidity in the country at +29.9%, Texas +19.8%, Louisiana +18.0%, Mississippi +16.9%, Georgia +16.2%. These are precisely the states with the highest shares of heavily subsidized enrollment, where expiry removes the largest fraction of the healthy marginal buyer. California, New York, and other state-subsidy jurisdictions cluster near zero. And the morbidity factor does not travel one-for-one into the headline rate: Texas filings hold +19.8% morbidity inside a +35.6% weighted rate change, with the residual living in trend, demographics, and plan design, while Arkansas pairs the nation's worst morbidity assumption with a smaller +24.0% overall increase. For a pricing actuary benchmarking their own book, that decomposition is the usable part, which is why the benchmark page lets every factor be drilled state by state and carrier by carrier rather than stopping at the average.

What does a 1.000 factor on a 600,000-member book actually claim? Read literally, it claims the carrier's own experience-period population already resembles its projected population, either because its members skew unsubsidized and are staying regardless, or because the carrier expects to capture a healthier share of a shrinking market. Read skeptically, some zeros are placement choices: the URRT allows population change to be carried in the demographic shift or the catch-all other-adjustment fields, and a filing can hold morbidity at 1.000 while loading the same expectation elsewhere. The benchmark flags the filing as filed rather than reinterpreting it. But the two readings have very different consequences, and only one of them is benign. A genuine zero on a book exposed to subsidy-driven attrition is an underpriced renewal cycle: the mid-year deterioration arrives, the loss ratio overshoots, and the correction lands in the PY2027 filing with a year's compounding attached. Risk adjustment softens but does not neutralize the miss, because transfers reallocate relative morbidity within a state's pool while the zero-camp filing has mispriced the absolute level.

Where the Filed Record Sits Among the Season's Numbers

Three kinds of numbers circulate in every ACA rate season, and they answer different questions. Requested rates, tracked filing by filing through the summer, most comprehensively by Charles Gaba's ACA Signups project, answer what carriers asked for; they are timely but preliminary, and their weighted averages lean on enrollment estimates that redacted memoranda force the tracker to guess. Consultant morbidity guidance answers what carriers should assume: the effectuation-based factors in the 2.9% to 6.5% range that shaped early 2027 conversations, covered in our Wakely analysis, are of this kind. The URRT public use file answers a third question, what carriers actually filed, with the actual enrollment weights attached, and it is the only one of the three that is complete, official, and granular to the factor level. It arrives late, in the spring after open enrollment, which is why nobody builds news coverage on it. But the filed record is the reconciliation layer for the other two: PY2026 filings landed at +8.5% weighted morbidity against consultant guidance topping out near 6.5%, and the camp structure shows the guidance was not so much overridden as ignored from both directions at once.

Our own earlier reporting can be graded against it too. The 2026 benchmark-premium analysis, built from a sample of filings during the season, put carrier morbidity factors at 1.0025 to 1.044; the complete record shows the enrollment-weighted market at 1.085, with nearly a quarter of it above 1.15. Season-time samples systematically understated the shift because the largest books, which filed the biggest factors, publish last and redact most. That gap between what was visible in June and what the PUF shows in March is not a criticism of season coverage; it is the argument for a standing benchmark that rebuilds when the complete file lands. The first wave of 2027 requested rates is showing carrier-level morbidity spreads of 3% to 19.5% in state filings right now. The PY2027 PUF, due with open enrollment in late October, will show what the full market actually did, and the benchmark will rebuild on it.

What October Settles, and What It Cannot

The obvious question, which camp was right, gets a partial answer soon and a real answer slowly. The partial answer arrives with the PY2027 filings: carriers that filed zeros for 2026 and spent the year watching their loss ratios will encode the verdict in their next morbidity factor, and camp convergence, or its absence, will be visible in one chart. The real answer requires experience, and experience in this market is endogenous in a way that should discipline anyone's confidence. A +26.1% weighted premium increase is itself a selection event: it prices out another tranche of healthy unsubsidized buyers, which worsens the realized pool, which means even the +15% camp may discover it priced a moving target. Our adverse-selection coverage has followed this spiral mechanism through the season; the filed record now gives it a quantified baseline. The spiral question for PY2027 is not whether morbidity assumptions rise but whether they rise faster than the pool deteriorates.

Two features of the data limit what any grading exercise can claim, and they are worth naming rather than discovering later. First, filed is not final. Prior-approval states cut filed rates, sometimes drastically, as our New York DFS analysis documented when regulators halved carriers' filed 2026 increases; a filing's morbidity assumption can be actuarially vindicated even when its rate never survived review. Second, the zero-camp ambiguity cuts both ways at grading time: a carrier that embedded its morbidity expectation in another URRT field will look wrong in the morbidity column and right in the loss ratio, and only the filing-level detail distinguishes the cases. These are reasons to grade carefully, not reasons to skip grading. The URRT record is the only place where every carrier's expectation is written down in comparable units before the outcome arrives. As of this year, it is parsed, weighted, and public on the benchmark, and the October file starts the clock on the verdict.

Correction, August 28, 2026: an earlier version of this analysis contained two data-handling errors. First, blank URRT factor fields, which the public use file encodes as zero, were treated as filed values of -100%; excluding not-reported factors raises the enrollment-weighted morbidity series from -4.7%/-2.4%/+7.5% to -0.6%/-0.1%/+8.5%. Second, terminated plans, which the file carries at a 0% rate "change," were included in the weighted rate change; restricting to renewing plans raises it from +23.5% to +26.1% for PY2026, in line with KFF (+26%) and ACA Signups (+25.5% final approved). The camp structure and carrier-level figures are unaffected. The benchmark methodology now excludes both.