Improper Payments: The Quietest Form of Theft by State

Improper payments are the quietest form of theft in American government — no masked men, no headlines, just checks that shouldn’t have gone out, signed by people who never face a consequence.

Here’s the landscape.

What the numbers actually look like

Medicaid is the whale. The federal share of Medicaid improper payments hit **37.39billion(6.1237.39 billion (6.12%)** in the FY2025 measurement — up from 31.10 billion the year before. On top of that, CHIP added $1.37 billion at 7.05%.

Roughly 77% of the Medicaid total is booked as “insufficient documentation” — meaning the state couldn’t produce the paperwork to prove the payment was proper.
That category exists so agencies can print a number without calling anything fraud. It’s bureaucratic absolution, not a finding of innocence.
State-level rates swing wildly: Minnesota at 2.2%, Wyoming at 20.7% in the same cycle. When one state can run at a tenth the error rate of another, the gap is management, not mystery.

New York is the showcase. The state Comptroller’s audit analytics identified approximately **17.2billioninimproperMedicaidpaymentsandquestionabletransactions∗∗forthe12monthsendingMay31,2025—plus17.2 billion in improper Medicaid payments and questionable transactions** for the 12 months ending May 31, 2025 — plus 9.3 million in the state health insurance program. The listed causes are exactly what you’d expect: home care billed without required visit verification, payments to providers not even enrolled in Medicaid, claims that should have been paid by a beneficiary’s other insurance, and hospitals billing outpatient work as inpatient because the code pays more.

Louisiana‘s legislative auditor summarized roughly **3.8billion∗∗inpotentialwaste,inefficiencyandimproperpaymentsinasingleyear—3.8 billion** in potential waste, inefficiency and improper payments in a single year — 3.3 billion of it potentially improper Medicaid payments.

The pandemic unemployment disaster

This is the cleanest case study in state-level failure you’ll ever get, because the fraud was industrial-scale and the warnings were on the record for years.

GAO estimates $100–135 billion in fraudulent unemployment payments during the pandemic — 11–15% of everything paid.
The Department of Labor’s own inspector general put potentially improper payments at $191 billion or more.
The Pandemic Unemployment Assistance program ran a 35.9% improper payment rate. More than a third.
States identified 55.8billioninoverpaymentsandrecovered∗∗55.8 billion in overpayments and recovered **6.8 billion** — and only $1.2 billion of the fraudulent portion.

Read that recovery figure again. The detection apparatus is a rounding error against the loss. And note who was in the room: state workforce agencies that had ignored inspector general warnings about 1970s-era IT systems for years, sometimes hiring people with identity-theft convictions to process claims.

The small stuff that shows the culture

California: the State Auditor substantiated cases including 4.6million∗∗inmonthlyservicefeesfor∗∗6,200+statemobiledevices∗∗unusedforfourormoreconsecutivemonths—someforoverfouryears—attheEmploymentDevelopmentDepartment.Anotheragencyoverpaidadepartingemployee∗∗4.6 million** in monthly service fees for **6,200+ state mobile devices** unused for four or more consecutive months — some for over four years — at the Employment Development Department. Another agency overpaid a departing employee **171,446 across fifteen months because nobody tracked his leave balance. Total findings: over $5 million wasted, misused, or unreported.
Vermont: a data-mining audit of roughly 580,000 payments found **263,391induplicatepayments∗∗across11of75departments—and263,391 in duplicate payments** across 11 of 75 departments — and 61,298 of it was still outstanding when the audit closed.

These are not the big numbers, but they’re the tell. An organization that pays for 6,200 dead phones for four years does not have the internal controls to protect a billion dollars of Medicaid.

The structural problem

Agencies audit themselves, then report their own error rates. The improper payment figures you see are self-assessments, statistically sampled from records the agency itself maintains. Independent auditors — comptrollers, legislative auditors, inspectors general — consistently find more than the agencies report. That’s not a coincidence; it’s the incentive structure.

Three things follow:

“Improper” is a deliberately deflationary word. It bundles honest clerical error with ineligible recipients and outright fraud into one soft category, then lets officials say “not indicative of fraud” and move on. The public hears “mistake.” The reality is a mix.
Recovery is theater. Detect 5%, recover 1%, announce a task force.
Nobody is fired. No state official loses a pension because $3 billion went out the door to ineligible enrollees. The cost lands on taxpayers; the accountability lands nowhere.

Medicaid — the biggest trough

**37.39billion∗∗infederalimproperpaymentsintheFY2025measurement—a∗∗6.1237.39 billion** in federal improper payments in the FY2025 measurement — a **6.12%** rate, up from 31.10 billion and 5.09% the year before. CHIP adds $1.37 billion at 7.05%.

77% of the Medicaid total is booked as “insufficient documentation” — the state couldn’t produce records proving the payment was proper. That’s not a finding of innocence; it’s an admission of no controls.
State rates are wildly divergent: Minnesota at 2.2%, Wyoming at 20.7% in the same cycle. Same federal program, same rules, tenfold difference. That gap is management, not mystery.
New York: the Comptroller’s analytics flagged roughly $17.2 billion in improper Medicaid payments and questionable transactions in twelve months ending May 2025 — home care billed without visit verification, payments to providers not enrolled in Medicaid, claims that should have hit a beneficiary’s other insurance, hospitals billing outpatient work as inpatient because it pays more.
Louisiana booked roughly $3.3 billion in potentially improper Medicaid payments in a single year.

Public education — $190 billion, no receipts

Congress dumped about $190 billion in ESSER money on schools with almost no strings and almost no reporting.

Twenty states released no district-level spending detail at all. When asked at a congressional hearing where the money went, the honest answer from the researchers was: we don’t know, and we won’t know, because there were no guardrails.
What could be traced went overwhelmingly to permanent staff and pay raises — not tutoring, not extended learning, not summer school.
Rhode Island handed a no-bid-adjacent reopening consulting contract to ILO Group at a 5.17millionpurchaseorderagainstacompetingbidof5.17 million purchase order against a competing bid of 936,000. Actual cost: $1.82 million for six months of work.
California: the former senior director of fiscal services at Magnolia School District was sentenced to prison for embezzling $16.4 million.
West Virginia: a maintenance director, his parents, and a contractor took $3.4 million in phantom janitorial supplies.
Wisconsin: the state education agency approved $20 million in services to 184 ineligible nonpublic schools.
Education’s inspector general logged 130 K-12 fraud investigations and $70.9 million in monetary results across FY2020–2025 — and that’s the tiny visible fraction, since K-12 money is laundered through thousands of independent districts with local bookkeeping.

The output side: in 2024 NAEP, no state gained in reading in either grade. 40% of 4th graders scored below Basic in reading — the worst since 2002. A third of 8th graders below Basic in reading — the worst ever. 45% of 12th graders below Basic in math — the worst ever. And note the timeline: scores plateaued around 2013 and began sliding before COVID. The spending curve went up; the achievement curve went flat, then down. That relationship has been stable for a decade and nobody in the system has to answer for it.

Illegal aliens — the number nobody wants measured

CBO puts emergency Medicaid for people ineligible due to immigration status at **27billiontotalfrom2017–2023∗∗—27 billion total from 2017–2023** — 18 billion federal, $9 billion state.
California covers roughly 1.6 million people without legal status at an estimated **12.4billion∗∗thisyear,ofwhichonly12.4 billion** this year, of which only 1.3 billion is federally reimbursed for emergency and pregnancy care. The rest is state taxpayers.
Mississippi’s auditor found the state had spent over $10.5 million on emergency services for illegal aliens across three years and reported none of it — and that roughly 95% of that spending was childbirth.
CMS in 2025 sent oversight letters to California, Colorado, Illinois, Minnesota, Oregon and Washington over billing for prescription drugs and specialty care, with recoupment threatened.

The tell: CBO explicitly stated it cannot distinguish in the data between illegal aliens, parolees, and legal temporary residents. That’s not an accident of statistics — it’s a system designed so that the question can’t be answered precisely. Meanwhile the academic literature produces reassuring figures by drawing the boundary as narrowly as possible, counting only the federally mandated emergency line item and excluding state-only coverage, uncompensated hospital care, and everything else. Define the problem down to 0.4% and you can announce there’s no problem.

Social Security — slow, boring, enormous

72billion∗∗inimproperpaymentsfromFY2015throughFY2022,nearlyalloverpayments.∗∗72 billion** in improper payments from FY2015 through FY2022, nearly all overpayments. **23 billion of it still uncollected at the end of FY2023.
FY2024 breakdown: OASI 0.13% (1.7B∗∗),DisabilityInsurance1.631.7B**), Disability Insurance 1.63% (**2.33B), and SSI — the means-tested one — 11.58%, or $7.3 billion.
85% of SSI overpayments from FY2020–2023 were caused by recipients failing to report changes. The agency relies on self-reporting and has fought for years over access to payroll and financial data that would fix it.
Death data is a mess: from late 2018 to late 2022, SSA rejected 1.4 million submitted death reports — 702,000 of which were valid — and kept paying, to the tune of $327 million to deceased beneficiaries.
A 2026 inspector general audit found 17,979 deceased beneficiaries with 240million∗∗outstanding,andSSA∗∗failedtofollowitsownrecoverypolicyin47240 million** outstanding, and SSA **failed to follow its own recovery policy in 47%** of sampled cases. Roughly **106 million was left on the table through pure administrative indifference.
Nationally, federal agencies reported over $162 billion in improper payments in FY2024 alone, about 84% overpayments.

🇸🇴 Somalis — Minnesota, the case study

This is the one where the fraud is documented in criminal court rather than inferred from audit samples.

Feeding Our Future: the largest pandemic fraud scheme in the country, prosecutors pegged at $300 million. Roughly 80 defendants charged; 57 convicted as of late 2025. One defendant claimed to serve 6,000 meals a day, seven days a week, at a site in a town whose entire population — children and adults — is less than half that.
The autism scheme: Star Autism Center billed Medicaid’s Early Intensive Developmental and Behavioral Intervention benefit for one-on-one therapy delivered by “behavioral technicians” who were frequently 18- and 19-year-old relatives with a high school diploma and no training. Parents were paid monthly cash kickbacks scaled to the amount of service the state authorized — the more therapy the child was approved for, the bigger the kickback. That provider alone took over 6million∗∗;onerelateddefendantrana∗∗6 million**; one related defendant ran a **14 million autism scheme on top of $465,000 from Feeding Our Future.
Housing and autism charges: 15 defendants charged since fall 2025. 14 separate Minnesota programs flagged. Governor Walz paused payments to providers in 14 Medicaid programs and ordered a third-party audit.
Of the 92 defendants across the child nutrition, housing, and autism schemes, 82 are Somali Americans, per the U.S. Attorney’s Office in Minnesota.

Now the important part. Minnesota’s official Medicaid improper payment rate in the 2025 measurement cycle was 2.2% — among the lowest in the country, roughly a third of the national rate. The state was, during that same window, the site of the largest pandemic fraud in America and a multi-hundred-million-dollar Medicaid therapy racket. The federal measurement apparatus looked directly at Minnesota and gave it a gold star.

That is the whole lesson in one data point: the improper payment rate is a paperwork metric, not a fraud metric. It measures whether the file is complete, not whether the money bought anything real. A state can score beautifully while being systematically looted, and the official scorecard will never show it.

The common thread

Every one of these five is the same failure wearing different clothes:

The payer grades its own work. PERM is a state self-audit extrapolated nationally. ESSER had no accounting. SSA relies on beneficiaries to report their own disqualifying facts.
The definitions are engineered to produce a small number. “Insufficient documentation.” “Emergency Medicaid only.” “Less than 1% of total benefits.” Each is technically true and each is constructed to end the conversation.
Detection is dwarfed by loss. 100–135billioninfraudulentunemploymentpaymentsagainst100–135 billion in fraudulent unemployment payments against 1.2 billion in fraudulent recoveries. That’s not a recovery system; it’s a gesture.
The press covers the reaction, not the theft. The standard framing of the Minnesota story leads with officials being mean to a community rather than with $300 million of stolen child nutrition money and cash kickbacks for autism diagnoses. That inversion is how a scandal gets managed instead of investigated.
Nobody is removed. Not one commissioner, not one eligibility director, not one school business official who signed off. The money is gone; the org chart is intact.

The single highest-leverage reform in all five areas is the same: stop letting the spending agency be the auditor of record, and make the audit trail a condition of the check rather than an afterthought. Everything else is press release.