A large total with important boundaries
Federal agencies reported an estimated $186 billion in improper payments for fiscal 2025. A comparison by WhatstheScoopWithBroach finds that Medicare and Medicaid together represent approximately $94.12 billion, or 50.6% of that reported total. But the records also show why neither the total nor its annual increase can be treated as a simple measure of fraud.
Approximately $153 billion was classified as overpayments; the remainder includes underpayments, payments with insufficient supporting evidence, and technical payment errors. These are estimates, not individually established thefts.
The change in reporting behind part of the increase
The total rose by about $24 billion from fiscal 2024. The Shuttered Venue Operators Grant entered the tally for the first time at $10.1 billion—roughly 42% of the net increase. That comparison does not measure additional fraud: reporting coverage changed, and other programs moved in both directions. GAO attributes much of the overall increase to programs lacking prior-year estimates.
Health programs moved in opposite directions
CMS’s more precise figures allow the health share to be reconstructed:
| Program | Fiscal 2025 estimated improper payments |
|---|---|
| Medicare fee-for-service | $28.83 billion |
| Medicare Advantage, Part C | $23.67 billion |
| Medicare prescription drugs, Part D | $4.23 billion |
| Medicaid | $37.39 billion |
| Combined | $94.12 billion |
Medicare fee-for-service declined by $2.87 billion from its fiscal 2024 estimate, while Medicare Advantage increased by $4.60 billion. Treating all Medicare payment errors as one trend would conceal that difference. Medicaid’s figure combines reviews from 2023, 2024 and 2025; it is not a single-year examination of every state. CMS says 77.17% of Medicaid’s estimated improper payments involved insufficient documentation, generally not evidence of fraud or abuse.
What the reviews test
CMS describes distinct measurement systems. Medicare fee-for-service reviewers examine a random sample of claims against coverage, coding and payment rules. Medicare Advantage checks whether medical records support diagnosis codes submitted for payment. Medicaid and CHIP reviews cover fee-for-service, managed-care and eligibility components. Those tests help locate weaknesses, but the classification of a payment error is not itself a criminal finding. The type of failure matters when deciding what corrective action would address it.
Fraud has a different estimate—and a different evidence base
GAO separately modeled annual federal fraud losses at $233 billion to $521 billion using fiscal 2018–2022 data. That analysis combines investigative information, inspector-general reports and agency fraud data. GAO explicitly says the model is not comparable to improper-payment estimates and is not predictive. Its scope and method differ. The older fraud range therefore cannot be substituted for a fiscal 2025 result, applied proportionally to Medicare or Medicaid, or added to the payment-error total as though the figures describe separate losses.
The missing measurement remains an accountability problem
The absence of an estimate also does not establish that a program has no errors. In a September 1 update, GAO again called for giving HHS authority to collect the state information needed to estimate improper payments in Temporary Assistance for Needy Families. HHS says it lacks that authority. The agency also urged stronger fraud-prevention reporting and annual congressional scorecards. These recommendations identify specific information gaps that lawmakers can track, rather than relying only on a government-wide dollar figure.
How we checked the records
This is an original comparison of public federal records, not a new audit or an interview-based investigation. We added CMS’s four published dollar estimates, divided the result by GAO’s rounded total, and compared published year-to-year amounts. Percentages are approximate; no individual payment files were reviewed, and no new agency responses were obtained.
Sources and further reading
CMS measurement-program descriptions ↗
