On "improper" enrollment

 Now cometh Brian Blase of Paragon with his third iteration of the so-called “Great Obamacare Enrollment Fraud” (rechristened “The Persistent Obamacare Fraud”) — this time claiming 6.2 million “improper” enrollments .

I wanted to get “fraud” and “improper enrollment” in one sentence, because Blase conceptually segues from outright fraud— people enrolled without their consent, or under false pretenses — to “improper” enrollment, which he defines as enrollments in which the applicant’s the future-looking income estimate may have been optimized to the her advantage — usually by a broker. In the original report, he deemed all enrollments with allegedly massaged income “fraud.”

Blase’s core claim, which has not changed substantially since his first iteration of this report in 2024, is that the number of enrollees who estimate 2026 income in the 100-150% FPL range — 10.7 million, or 46% of all enrollees — far exceeds the Census Bureau’s survey-based estimates of the eligible population with income in that range. It’s doubtless true that many enrollees optimize their income estimates, usually at a broker’s direction - -e.g., to get over the 100% FPL eligibility thresholds in the 10 states that have not expanded Medicaid (available to adults with income up to 138% FPL in expansion states), or to get below 150% FPL, the eligibility threshold for the highest level of Cost Sharing Reduction (CSR).

Such income massaging of course happens, as it must — and to some extent should — in an application that requires an estimate of next-year income (do you maximize your taxable income?). But Blase exaggerates both the extent and the nefariousness though some broker-induced estimates are genuinely fraudulent and therefore nefarious). I engaged with Blase’s analysis in some detail in 2024, and I don’t want to go over all that ground, the I will borrow a point from a prior TLDR of mine:

ACA subsidies are based on an estimate of future income, which is inherently uncertain, especially for people at low incomes, who often work uncertain hours, change jobs, are self-employed, or depend on tips. Mismatches between income reported to the IRS and income projected in ACA applications probably have as much to do with inaccuracies in tax reporting as with inaccurate income projections in the ACA application. As for mismatches between income data based on ACA enrollment and data from the Census Bureau’s consumer surveys, those, like mismatches between IRS data and survey data, are perpetual.

In a followup post, I want to engage with Blase’s newer types of alleged evidence of “improper” enrollment, including: 1) the percentage of enrollments that end with no filed claims (EWOC); 2) the percentage of enrollees who do not identify their ethnicity (optional questions on the application; 3) the percentage of enrollees who actively vs. passively re-enroll; and 4) the percentage of low-income enrollees who forgo strong Cost Sharing Reduction subsidies by choosing bronze or gold plans.

First, a couple of big-picture points.

1) Among the alleged 6.2 million “improper” enrollments, Blase estimates that 3.42 million have income below 100% FPL. Almost all of these are in the ten states that have refused to enact the ACA Medicaid expansion; in nine of those states, most adults with income below 100% FPL get no government help obtaining health coverage (Wisconsin extends Medicaid eligibility to adults with income up to 100% FPL and thus has no ‘coverage gap’). In OEP 2026, 6.5 million marketplace enrollees in the nine “coverage gap” states reported income in the 100-138% FPL range. If the Medicaid expansion were nationwide, almost all of those enrollees would presumably be in Medicaid, except those who are underestimating their income (who, thanks to Republican rule changes, will repay all excess advanced tax credits at tax time, as the Republican megabill removed the former caps on such repayment). The irrational and self-harming refusal of red-state governments to take the 90% federal match for the Medicaid expansion population and cost-effectively provide for their low-income residents is the chief driver of whatever “improper” enrollment takes place.

2) As Blase notes, almost all fraud and income massaging (appropriate or inappropriate) in the marketplace is executed by brokers, and almost all broker fraud is in the 30 states currently using HealthCare.gov (with, he alleges, some apparent residual effects in states that have transitioned from HealthCare.gov to state-based exchanges (SBEs) in recent years). As he further emphasizes, the SBEs have effectively deterred the broker fraud that took off in HealthCare.gov states in 2023 and 2024 (with some curbs thereafter). That fraud outbreak was spurred by the combination of zero-premium silver plans at low incomes, year-round Special Enrollment Periods at low income (not just year-round enrollment, but, more damagingly, monthly SEPs enabling plan changes) — and, mostly damagingly, too-easy broker access to all existing accounts via commercial enhanced direct enrollment (EDE) platforms. SBEs deter broker fraud mainly by requiring two-factor client authorization of the broker as broker of record before the broker can act on the account.* CMS, for all its imposition of new administrative burdens on enrollees, has unaccountably failed so far to replicate the comparatively simple client sign-offs required in the state exchanges.

All of which is to say, fraud would not be a problem in the marketplace without a) a coverage gap in nine states, including population behemoths Florida and Texas, and b) simple controls that would keep brokers’ fingers off the accounts of people who have not authorized them to act.

As to a), the outbreak of fraud concentrated at the lowest incomes underscores that Medicaid is a more appropriate form of coverage for low-income people.

As to b), CMS is reportedly on the brink now of requiring 2-factor authorization and i.d-proofing before a broker can act on an account, and the 2027 NBPP standardizes client consent forms and restricts misleading marketing incentives. Once fraudster brokers have glommed onto a particular market, they are hard to root out. But perhaps the new controls will do the job.

As for the possibility that large numbers of people in nonexpansion states are claiming an income above 100% FPL when their actual income is lower, my response is and always will be: good. The coverage gap is an abomination, violating the ACA’s original intent (enactment was mandatory for states until the Supreme Court made it optional in 2012) and every tenet of public health, justice, and economics. If someone who waits tables and babysits projects an income of $16,000 and only ends up with $13,000 taxable — and gets coverage — that’s a good thing.

That said, according to CMS’s own estimates, the numbers are small. The 2027 Notice of Benefit and Payment Parameters reasserts a rule (so far stayed by the courts) requiring the exchanges to demand income verification from an applicant claiming an income above 100% FPL when trusted data sources indicate an income below. CMS estimates (p. 29836) that enactment would reduce enrollment by just 81,000 — 50,000 in the federal exchange, and 31,000 in the SBEs. Either that is a gross underestimate, or Blase’s claim that 3.4 million people with income under 100% FPL are obtaining subsidized marketplace coverage is a gross overestimate. Or both.

One fact Blase’s report does highlight is that the ACA marketplace cannot be understood without taking into account the extent to which enrollment depends on brokerage — good brokerage, sloppy brokerage, “gray” brokerage feeding off misleading ads, and outright fraudulent brokerage. In HealthCare.gov states, about 75% of active enrollments are broker-assisted; in SBE states, the percentage is somewhat lower but exceeds 50% (in aggregate, not in every state; a handful don’t even allow brokerage). Ignorance of ACA offerings among the general public is endemic and persistent; the market depends on brokers to find people who need insurance.

I personally think it’s a peculiarly American ideological sickness that generates a public health coverage program in which people need brokers to navigate a complex application, in which small differences in projected income can put thousands or tens of thousands of dollars at stake, and a dizzying array of commercial insurance products that differ chiefly in their offsetting defects, from narrow networks to high out-of-pocket exposure sliced in dozens of complex ways. Given that complexity, a good broker is priceless; a bad broker will not waste effort ensuring that a client doesn’t face unnecessarily high out-of-pocket costs or enroll in a plan that lacks providers or disease management programs that person needs. A fraudulent broker may expose an enrollee to huge costs at tax time, or switch an enrollee out of a plan that meets her needs into one that doesn’t. 

It is difficult to determine the relative prevalence of good brokerage, minimally decent brokerage, negligent brokerage, borderline fraudulent brokerage, and outright fraudulent brokerage. Some of the enrollment patterns that Blase cites reflect broker influence, and some, broker malpractice; with some it’s difficult to tell. In Part 2, we’ll take them one by one.