Oscar Health's Loss Ratio Fell to 79.2 Percent. It Booked Only a Sliver of the Risk-Adjustment Upside Behind It

Oscar Health's medical loss ratio fell to 79.2% in the second quarter of 2026 from 91.1% a year earlier, effectuated membership rose 46% to 2.96 million, and net income swung to $362 million from a $228.4 million loss (Oscar Health, August 2026). Underneath that print sits an accrual judgment few outside the reserving team will ever see directly: Oscar's actuaries received a first look at 2026 market morbidity built on claims through April, found it favorable to pricing, and booked only a small piece of it. The rest waits on data that has not finished arriving.

CFO Scott Blackley drew the line explicitly on the earnings call. Oscar had also just received the final 2025 CMS risk-adjustment reconciliation, a number with an actual settlement behind it, and treated it very differently from the preliminary 2026 read. "In the second quarter, we received the final 2025 CMS risk adjustment report, which was approximately $160 million favorable to our first quarter accruals and fully recognized in the quarter," Blackley said, before turning to the newer data: "We also received the first risk adjustment report for 2026, covering claims through April, which showed market morbidity tracking quite favorable to both our pricing and first quarter accruals. With only 4 months of claims in the data, we recognized only a small portion of that favorability, which we believe is appropriate at this stage in the year" (Scott Blackley, Oscar Health Q2 2026 earnings call, August 2026, via Motley Fool transcript). One number, final and reconciled, went into results in full. The other, preliminary and thin, went in at a discount management chose not to quantify publicly.

How an Interim Morbidity Read Becomes an Accrual Decision

The distinction Blackley drew is not a communications choice. It reflects how the HHS risk-adjustment program actually produces information over the course of a benefit year. Under the HHS-HCC methodology, CMS scores each ACA-compliant plan's enrolled population using diagnosis codes mapped to Hierarchical Condition Categories, compares that score to the statewide average by metal level and rating area, and transfers money from lower-risk plans to higher-risk plans, a zero-sum mechanic funded within the market rather than by the federal government (CMS, Risk Transfer Formula for Individual and Small Group Markets). The catch is timing. CMS does not publish even an interim summary report on a given benefit year until the following spring; the benefit year 2024 interim summary risk-adjustment report, for instance, came out in March 2025, and the final reconciliation lands later still (CMS, BY2024 Interim Risk Adjustment Report). An insurer setting quarterly reserves in the middle of the coverage year has no official CMS number to work from at all.

What Oscar and most large ACA carriers use instead is a modeled proxy: actuarial consultancy Wakely benchmarks each issuer's own EDGE server submissions, the claims and enrollment data plans upload to CMS continuously through the year, against a market-wide sample to approximate where the eventual transfer will land. That is the same process Centene uses to track its own Marketplace book, and it is the process that produced Oscar's April-claims read this quarter (see actuary.info's coverage of Centene's Wakely-informed risk-adjustment swing). Four months of paid claims is a thin base to build a full-year morbidity assumption on for a book with meaningful claims lag and a utilization pattern that shifts hard as the year progresses. Booking most of the signal and holding the rest back is not conservatism for its own sake; it is treating a low-credibility estimate as a low-credibility estimate, which is a more defensible reserving posture than pricing 79.2% into a number that a fuller data set could still revise.

A Net Payer Managing a Payable, Not a Receivable

Oscar's structural position inside the risk-adjustment program cuts the other way from Centene's, and that difference is what makes "favorable" mean something specific here. Oscar's population skews younger and more digitally engaged than the ACA market average, which historically leaves it with a lower risk score than the statewide benchmark, meaning it owes money into the transfer rather than collecting it. As of June 30, 2026, Oscar carried a net risk-adjustment payable of $4.85 billion, split between $2.37 billion accrued for the current 2026 benefit year and $2.47 billion for prior years still working through final reconciliation (Oscar Health, Form 8-K, August 2026). "Favorable" morbidity data for a net payer does not mean a check arrives; it means the bill Oscar expects to owe the rest of the market shrinks relative to what it had reserved. That is precisely the mechanic behind the $164 million of favorable prior-period reserve development in the quarter, most of it traceable to the fully recognized $160 million final 2025 settlement landing better than Oscar's first-quarter accrual assumed.

The same mechanic ran in reverse just two quarters earlier. In its fourth-quarter 2025 results, Oscar increased its risk-adjustment payable accrual by roughly $275 million after concluding its members looked healthier than the broader market, the opposite of what a struggling risk pool would produce but still a charge against earnings because a healthier-than-average population owes more into the transfer, not less (Baird analyst note via Yahoo Finance, July 2026). That charge helped drive Oscar to an $87.4 full-year 2025 medical loss ratio increase, worse than the 81.7% posted in 2024, and a full-year net loss of $443 million (FierceHealthcare, February 2026). The same accrual discipline that inflated the payable in the fourth quarter of 2025 is now shrinking it in the second quarter of 2026. Oscar's actuaries are applying one consistent standard to both directions of the same estimate, which is the detail that should reassure a reader more than either individual swing does on its own.

Decomposing the Move to 79.2 Percent

Three separate items sit inside the 11.9-point year-over-year improvement in the reported ratio, and only one of them is the risk-adjustment story. Pricing discipline accounts for a meaningful share: Oscar filed a weighted-average rate increase near 28% for 2026 plans, built on an assumption that market morbidity would deteriorate further as the enhanced premium tax credits lapsed (Seeking Alpha, 2025). Pricing that conservatively against a morbidity assumption that has come in better than feared is, by itself, enough to move a loss ratio several points without any risk-adjustment item touching the number at all. The $164 million of favorable prior-period development is the second, smaller piece, and it is a one-time item tied to a reconciliation of business already earned, not a recurring feature of the current accident-year loss ratio.

The third item is the one worth watching most closely, because it is the one management explicitly declined to fully recognize: the partial credit taken on the initial 2026 risk-adjustment signal. A regulator reading Oscar's reported 79.2% ratio in isolation should not confuse it with the ACA's statutory medical loss ratio calculation, which sits well above the reported figure. The Affordable Care Act requires individual-market issuers to spend at least 80% of premium on claims and quality-improvement activities or rebate the shortfall to policyholders, but that calculation runs on a three-year rolling average, nets out taxes and fees differently than a GAAP income statement does, and is computed separately from the quarterly ratio a company reports to investors (CMS, Medical Loss Ratio Fact Sheet). Oscar's 79.2% GAAP-basis quarterly ratio sitting below the 80% statutory rebate floor is a coincidence of the current quarter's math, not evidence that a rebate is imminent or that one is off the table for the smoothed three-year regulatory calculation.

The guidance table makes the conservatism visible in a different way. Oscar tightened its full-year MLR range down by roughly a point and roughly doubled its earnings-from-operations midpoint, but it did not lift revenue guidance and it left the top of the operating-earnings range well short of what a fuller recognition of the favorable 2026 risk-adjustment signal could support if the April trend holds through the fall. That gap between what the interim data suggests and what guidance assumes is the same conservatism showing up twice, once in the accrual and once in the outlook built on top of it.

Why a Record First Half Still Points to a Second-Half Loss

Oscar's own guidance implies the company loses money in the back half of 2026 even after the strongest first half in its history. First-half 2026 earnings from operations reached $1.1 billion, a company record, while the raised full-year guidance tops out at $700 million, meaning the third and fourth quarters combined are expected to run at a loss on an operating basis (Oscar Health, August 2026). That is not a warning sign so much as how ACA individual-market economics work every year: plan-year deductibles reset each January, member cost-sharing absorbs a larger share of claims early in the year, and deductible leverage erodes as members work through their out-of-pocket maximums through the fall, pushing more of each dollar of claims onto the insurer as the year progresses. The full-year 81.5% to 82.5% MLR guidance, well above the 79.2% first-half print, is management's way of saying it expects that seasonal pattern to reassert itself on schedule, not a hedge against the risk-adjustment estimate proving wrong. The arithmetic behind that expectation is straightforward: if the full year averages roughly 82% against a first half that ran 79.2%, per-member claims in the second half need to run meaningfully higher than the first half's pace even before membership growth is layered on top, purely because the deductible dollars that insulated Oscar's loss ratio in January through June stop insulating it once enrollees have satisfied their annual out-of-pocket exposure.

Growth of 46 Percent Against a Market That Shrank

The more striking number sits in the contrast between Oscar's trajectory and the market it competes in. Oscar's 46% membership growth to 2.96 million ran directly against the national tide: total ACA Marketplace enrollment fell by roughly 3 million people between 2025 and 2026, the first nationwide decline in seven years, as the enhanced premium tax credits lapsed and roughly 27% of the drop in sign-ups concentrated among enrollees just above the 400% federal poverty line subsidy cliff (KFF, 2026). A carrier gaining share that fast inside a contracting market is either winning healthier shoppers on price, absorbing the sicker members other issuers are shedding as they exit unprofitable rating areas, or some mix of both, and the 2026 risk-adjustment data is the mechanism that will eventually reveal which. Rapid growth complicates the morbidity assumption an actuary is pricing against in a way that is easy to understate. Oscar's growth concentrated in individual and small-group offerings including ICHRA-funded plans, a channel with limited multi-year claims history at Oscar's scale (Oscar Health, Form 8-K, August 2026). A book expanding that fast is, by definition, weighted more heavily toward new enrollees whose diagnosis coding and claims experience are still developing, which is exactly the population an interim, partial-year risk-adjustment read is least equipped to characterize with confidence.

Management also flagged that churn will run hotter than previously modeled: Blackley told analysts, "I would expect that churn, we previously thought it was 1% to 2%. It's probably going to be closer to twice that amount," attributing the increase mainly to CMS eligibility-verification activity rather than to underlying dissatisfaction with coverage (Scott Blackley, Oscar Health Q2 2026 earnings call, August 2026). A book that is both growing 46% against a shrinking national market and about to churn at double the previously modeled rate is not a stable population to extrapolate four months of claims across, which is the practical case for management's caution even before the accrual policy is considered on its own terms.

Pricing Into a Subsidy Cliff That Has Already Landed

The backdrop Oscar priced its 28% 2026 rate increase against was not hypothetical by the time enrollment closed. The enhanced premium tax credits that had subsidized ACA Marketplace coverage since 2021 expired on schedule at the start of 2026, and the effect showed up immediately in enrollment data: national Marketplace enrollment fell by roughly 3 million people between 2025 and 2026, the first decline in seven years, while average annual premium payments for subsidized enrollees rose 114%, from $888 to $1,904 (KFF, 2026). The subsidy cliff at 400% of the federal poverty level, which had been softened by the enhanced credits since 2021, returned as a hard cutoff, and KFF's tracking found the drop in enrollment concentrated disproportionately among people just above that threshold. Insurers that priced for a sicker retained population after subsidy-driven attrition, as Oscar did with its 28% increase, were betting that the healthiest marginal enrollees would leave first. The April 2026 claims data Oscar is now sitting on says that bet is coming in better than the pricing assumed, at least so far.

That is the frame worth carrying into the next rate-filing cycle. If Oscar's actuaries fully validate the favorable 2026 signal once more claims mature, and open enrollment for 2027 confirms the retained population is not deteriorating the way 2026 pricing assumed, next year's rate increases have room to moderate from 2026's levels. If instead the four-month read reverses as later-developing claims and the accelerated churn Blackley flagged work through the book, Oscar's conservative accrual posture means this quarter's 79.2% MLR will not have overstated the underlying trend, because most of the favorable signal was never booked in the first place. Either way, the discipline on display this quarter is a bet that under-recognizing a thin signal costs less than over-recognizing one that does not hold up, a trade every ACA-exposed reserving actuary is making right now on some version of the same four months of data.