Sixty-five billion dollars is not a rounding error. That is the kind of figure that makes you stop mid-scroll and ask a blunt question: how did a public health program quietly pay premiums for millions of people who were not supposed to be on the books? I kept coming back to that number because it is large enough to change how you look at every subsidy debate that follows. If even part of the estimate holds, the problem is not a handful of paperwork mistakes. It is a system that got sloppy at scale.
What The $65 Billion Estimate Actually Captures
The short version is uncomfortable. Researchers looking at federal survey data, program enrollment files, and spending totals concluded that taxpayers covered premiums in 2024 for people who either did not exist, did not qualify, or should have been sitting in a different program. The combined picture covers two Affordable Care Act pillars: the subsidized marketplace and Medicaid expansion.
Expanded Medicaid lets states enroll adults up to 138 percent of the federal poverty level. Traditional Medicaid generally stops at 100 percent. In 2024, that expansion cutoff sat near $35,600 for a family of three. Marketplace coverage, by contrast, stretched much higher. At the time, households earning up to 400 percent of poverty could still tap enhanced help. For that same family of three, the ceiling hovered around $103,000.
Both tracks run through commercial insurers. The federal government does not write a personal check to each household and call it a day. It pays carriers. That design is efficient when eligibility is clean. It is expensive when eligibility is not.
How Many People Were Flagged As Improper
The same analysis put combined improper enrollment at about 14.3 million people in 2024. That is not a small leak. On the marketplace side, researchers estimated that roughly 34 percent of enrollees that year were fraudulent, duplicated, or simply outside the benefit rules. The following year looked worse, not better. Improper exchange enrollment rose by more than 26 percent, reaching an estimated 6.5 million people.
I find that jump hard to ignore. Programs usually tighten after a scandal. This one appeared to loosen first and clean later. That sequence matters if you care about the public purse more than the press release.
Improper exchange enrollment increased by more than 26 percent from 2024 to 2025, up to an estimated 6.5 million enrollees.
On Medicaid expansion, the research team estimated that more than 9 million people on the rolls in 2024 probably did not belong there. Some likely earned too much. Others may have failed citizenship, immigration, or residency tests. A third group should have been in traditional Medicaid instead of the expansion category that draws a richer federal match.
Why Zero-Premium Plans Became A Magnet
After the pandemic years, many marketplace plans carried a $0 premium for the enrollee. That sounds generous. It also removed the last everyday check on a bad enrollment. If you never see a bill, you may never notice that someone signed you up. Brokers still get paid. Insurers still get a premium from Washington. The household may not know a policy exists until a tax form arrives or a clinic asks for an ID that does not match reality.
That is the part that feels almost too simple. Fraud does not always look like a Hollywood heist. Sometimes it looks like a commission check, a quiet enrollment file, and a person who never uses the plan. In congressional testimony, one policy researcher described people being enrolled without their knowledge. The government paid the broker. The government paid the carrier. The “member” never logged a claim.
Insurers later argued that a no-claims year can mean a healthy year, or a short spell of coverage. Fair point. A quiet file is not automatic proof of a fake person. Still, when quiet files stack up by the millions, and when some states show more marketplace enrollees than residents who even meet the income band, the healthy-year explanation starts to strain.
State Enrollment That Does Not Add Up
Twenty-eight states, according to the same body of work, had more people in subsidized marketplace coverage than there were residents who appeared to qualify on income. You can debate survey error. You can debate undercounting. You cannot wave away a pattern that wide without looking a little defensive.
Perhaps the most interesting aspect is not any single state. It is the clustering. When one jurisdiction overshoots, you check the survey. When more than half the map overshoots, you check the enrollment machinery.
- Income above the legal cutoff still showing as eligible
- Duplicate coverage across Medicaid and the marketplace
- Residency or citizenship gaps that never get resolved
- Accounts created by brokers the consumer never hired
- Identities that later match death records or unused Social Security numbers
None of those buckets is theoretical. Watchdogs later showed that fake identities can get through. In a controlled test, investigators enrolled 20 identities that did not belong to real people, using Social Security numbers that had never been issued and other easy-to-fabricate papers. Eighteen of those 20 accounts were still live many months later. The monthly taxpayer cost on that tiny sample ran above $10,000.
They also found tens of thousands of subsidy accounts tied to Social Security numbers that matched a death file. One year of those false enrollments cost more than $94 million. That is a sliver of the larger $65 billion claim, and it is already ugly.
The Pushback You Should Hear Before You Decide
Not everyone buys the 14.3 million figure. State marketplace officials have said they see no evidence of systemic fraud, waste, or abuse in their own portals. Insurance trade groups insist that a year without claims is not a confession. Budget analysts have argued that survey income and program income are not the same animal. A household can look “too rich” on a snapshot survey and still qualify under the rules that matter for enrollment.
That last objection is real. Annual income swings. Self-employment is messy. People lose hours. People pick up overtime. A survey answer in March is not a tax return in April. I have found that this is where honest readers split. One camp treats every mismatch as theft. The other camp treats every mismatch as noise. Both camps overreach.
A no-claims year can mean a healthy consumer or a short coverage spell. It does not automatically prove the policy was fake. It also does not prove the policy was clean.
So what do we actually know? We know eligibility systems were loose during and after the public-health emergency. We know enhanced subsidies made $0 premium plans common. We know brokers had a financial reason to maximize sign-ups. We know duplicate enrollment between Medicaid and the marketplace happened at scale. We know investigators could invent people and keep them enrolled. Those facts do not require you to accept every line of one think-tank spreadsheet. They do require you to stop pretending the integrity problem is imaginary.
Medicaid Expansion Versus Traditional Coverage
This distinction bores people until the money shows up. Expansion coverage draws a higher federal matching rate than many traditional Medicaid categories. If a state parks someone in expansion who belongs in a different bucket, Washington pays more than it should. If a person belongs in neither bucket, Washington pays for a benefit that should not exist.
Researchers said the likely ineligible expansion group included people over the income line, people who failed status rules, and people who should have been in traditional Medicaid. That mix is important. It is not all “ghosts.” Some of it is category gaming. Some of it is stale data. Some of it is outright fiction. Mixing those problems under one headline can inflate the shock value. Separating them still leaves a very large bill.
| Program track | 2024 income idea | Main integrity risk |
| Traditional Medicaid | Generally up to 100% of poverty | Stale eligibility and weak recertification |
| Medicaid expansion | Up to 138% of poverty | Wrong category and over-income adults |
| Marketplace subsidies | Up to 400% of poverty in that period | Phantom sign-ups and duplicate coverage |
Look at that table for a second. Three doors. One taxpayer. If the doors do not talk to each other, people walk through two of them. Sometimes they walk through a door that should have been locked.
Brokers, Commissions, And Consent
Let’s talk about the sales channel, because this is where the human mess lives. A broker who can enroll a household in a $0 premium plan does not need the household to feel the cost. The household may not even know the sale happened. Consent becomes a checkbox instead of a conversation. That is a terrible design if you care about accuracy.
Federal officials later suspended agents and brokers suspected of abuse. They canceled enrollments that looked like phantoms. They said hundreds of thousands of people had been put on coverage without consent. Another million-plus were stripped off after they showed up on two programs at once, or after they failed to file and reconcile old subsidy paperwork.
Those cleanups were billed as a $10 billion annual save. Fine. Celebrate the clawback. Then notice the implied admission. You do not recover $10 billion from a process that was humming along just fine.
- Spot the agent or broker pattern that looks industrial, not accidental.
- Match marketplace files against Medicaid and child coverage files.
- Kill the dual enrollments and the no-consent accounts.
- Force households to reconcile past subsidies or lose the new ones.
- Put a shorter clock on Medicaid recertification so dead files do not linger.
That sequence is basic program hygiene. It should not require a political fight to say out loud. In my experience, the fight starts when hygiene threatens enrollment totals that officials like to tout. Headcount became a victory lap. Integrity became a footnote. That trade is how you get a $65 billion conversation.
What “Fraud” Means Here, And What It Does Not
Words matter. Fraud implies intent. Improper payment can be sloppiness, lag, or a household that misunderstood the form. Bundling those ideas makes a sharper headline and a sloppier argument. I would rather keep the language tight.
Some cases are fraud in the ordinary sense. Invented identities. Dead Social Security numbers. Brokers stuffing files for commissions. Other cases are eligibility drift. A worker gets a raise in June and stays coded as poor in December. A student moves across state lines and the old plan never dies. A parent is on Medicaid while a marketplace plan is also paying a premium. That last one can be confusion. It can also be a farmed dual file.
Does the distinction reduce the taxpayer hit? Only at the edges. Money spent on the wrong person is still money spent on the wrong person. Intent changes who should be prosecuted. It does not change who paid.
Why The Timing Made Abuse Easier
Continuous coverage rules during the public-health emergency froze a lot of Medicaid reviews. States were told, in effect, not to kick people off. That made sense in a crisis. It also built a backlog of files that no longer matched life. When the freeze thawed, the recertification wave was late, uneven, and politically loud. Some eligible people lost coverage they should have kept. Some ineligible people stayed on longer than they should have. Both things can be true. Pretending only one happened is spin.
Enhanced marketplace subsidies layered on top of that thaw. Lower net premiums. More $0 plans. More broker advertising. More reason to treat enrollment like a volume business. If you were designing a stress test for integrity controls, you could hardly do better.
Pressure stack: Frozen Medicaid reviews + richer subsidies + zero-premium shopping + commission-driven brokers = a bigger error surface
That is not a conspiracy board. It is an incentive map. People follow incentives. Systems that ignore incentives get looted, slowly and then all at once.
The Taxpayer Angle People Keep Softening
Every improper premium is a transfer. It moves money from workers and firms who file taxes to carriers and intermediaries who bill the Treasury. Some of those transfers buy real care for people who qualify. That is the point of the law. Transfers that buy coverage for empty files, duplicate files, or ineligible files are just leakage.
$65 billion in one year is the estimate under discussion. Even if you cut it in half for caution, you are still staring at a sum that could fund a lot of actual care, actual deficit reduction, or actual premium relief for people who follow the rules. I get tired of the shrug that says public money has no opportunity cost. It does. Always.
There is also a fairness problem that does not poll as loudly as access. The household that reports income honestly and pays a partial premium is subsidizing the household that never should have been enrolled. That is not solidarity. That is a queue with a hole in it.
What Washington Already Tried To Fix
Over the last two years, federal managers tightened a few bolts. Suspensions for suspect brokers. Data matching across programs so the same person cannot collect two subsidies at once. Canceled phantom enrollments. A push toward more frequent Medicaid eligibility checks, including six-month recertification in some settings. Removal of more than a million people who were on marketplace coverage and Medicaid or child coverage at the same time, or who had not reconciled old credits. Another 250,000 removed after officials decided they had been enrolled without consent.
Those steps are not ideology. They are bookkeeping. If you run a private health plan and ignore dual coverage, you get fired. If you run a public subsidy and ignore dual coverage, you get a hearing and a slogan.
Cleanup that saves $10 billion a year is progress. It is also evidence that the earlier controls were not doing their job.
Will the next enrollment season look cleaner? Maybe. Enhanced subsidies, broker rules, and verification standards keep moving. The 26 percent rise in estimated improper exchange enrollment into 2025 is a warning that the cleanup and the leakage were happening at the same time. That is the awkward part. You can remove a million files and still watch the bad stock grow if the front door stays wide.
How To Read The Next Round Of Numbers
When the next report drops, ignore the first adjective and hunt for three measures. First, the share of marketplace enrollees with no claims over a full plan year, broken out by whether the premium was zero. Second, the overlap rate between Medicaid and marketplace files after matching. Third, the gap between survey-eligible populations and actual subsidized counts by state. Those three will tell you more than any press conference.
Also watch how agencies define “removed.” A person taken off a file in January can be back in March if the broker pipeline is still hot. Gross removals are not net integrity. I have seen too many dashboards that celebrate the outflow and stay quiet on the refill.
- Ask whether income verification is prospective or after-the-fact.
- Ask whether death-file matching is monthly or annual.
- Ask whether broker commissions are clawed back when a file is fake.
- Ask whether states lose federal match when expansion coding is wrong.
Those questions sound dry. They are the whole game. If commissions are not clawed back, the sales channel keeps selling ghosts. If match rates never fall, the dual-coverage problem is a feature. If expansion coding errors carry no fiscal penalty for the state, the higher federal match remains a temptation.
A Practical View For Readers Who Pay The Bill
If you buy coverage on the exchange, check your own account even if you think nothing changed. Confirm the plan, the premium, and the names on the file. If a broker you do not know appears in the record, treat that as a problem, not a courtesy. If you are on Medicaid, respond to renewal packets. Silence is how stale eligibility survives.
If you run a company that pays employer coverage, do not assume public-program leakage is someone else’s hobby. Subsidy design changes who shows up in the individual market, which changes risk pools, which changes premiums around the edges. The chain is longer than the slogan.
And if you just pay taxes and want the argument in one line, here it is. A health program can be popular and still be poorly policed. Those are not opposite ideas. They travel together more often than people admit.
The Part That Still Bothers Me
I can live with a contested estimate. Researchers will fight over methods until the next decade. What I cannot shrug off is the controlled test with unused Social Security numbers. If a small investigative team can mint coverage for people who do not exist, and if most of those files stay active, the front-end check is theater. Theater is expensive when the tickets are paid by wage earners.
I also cannot shrug off the death-file matches. Paying subsidies against numbers tied to dead residents is not a philosophical dispute about poverty measurement. It is a matching failure. Matching is a solved problem in plenty of private industries. Credit bureaus do it every hour. Banks do it before they wire a dollar. Public health finance can do it too. The question is whether anyone is rewarded for doing it well.
Maybe that sounds harsh. Good. Harsh is appropriate when the claimed one-year loss is $65 billion and the year after looks sloppier on the exchange side. Soft language is how leakage becomes tradition.
What A Serious Integrity Agenda Would Include
Start with identity. Real-time checks against issued Social Security numbers, death records, and existing public coverage. Not a yearly audit. A gate. Then fix the broker channel so a commission does not survive a canceled phantom file. Then align the definition of income across surveys, tax data, and enrollment engines as tightly as privacy law allows. Then publish state-level improper-payment rates with the same pride officials publish raw enrollment.
Notice what is missing from that list: a sermon about whether the Affordable Care Act should exist. You can support coverage expansion and still want the roster to be real. You can dislike the law and still want eligible people to keep the help they already have. The integrity fight is not a proxy war unless we let it become one.
In my view, the grown-up position is simple. Pay for eligible people. Stop paying for empty names. Measure both outcomes in public. Anything less is marketing.
Where The Story Goes From Here
The next enrollment cycle will tell us whether the removals were a one-time sweep or a new baseline. Watch the share of $0 premium plans. Watch broker discipline. Watch whether 28-state over-enrollment shrinks or just changes zip codes. And watch whether the official savings figure keeps pace with the estimated improper stock. If savings are $10 billion and the estimated waste stays several times larger, the broom is smaller than the mess.
Readers do not need another morality play about health reform. They need a ledger. Who was enrolled. Who qualified. Who paid. Who got a commission. Who never used the card because the card was never theirs. That ledger is how you keep a popular program from becoming a soft target.
Sixty-five billion dollars is the number that opened this piece. Treat it as a claim, not scripture. Then treat the supporting facts as what they are: evidence that eligibility got loose, sales incentives got sharp, and taxpayers funded the gap. If that does not bother you, the next estimate will not either. If it does, the work is not a think-tank report. The work is making the roster real.
