DOGE Watch · HHS · CMS · 11 Sources
$50.3B
FY2023 improper payments
8.6%
FY2023 error rate
$586B
Total Medicaid spending FY2023
§ DOGE Watch / Medicaid Improper Payments

$50 Billion. One Year. And CMS Knew Every Dollar Was Wrong.

§ 01 / The Number

One in Every Twelve Medicaid Dollars Goes Somewhere It Shouldn’t.

Medicaid is the largest single line item in the federal budget. In FY2023, it cost federal and state taxpayers a combined $586 billion. CMS’s own Payment Error Rate Measurement (PERM) program — the official federal methodology for measuring improper payments — found that $50.3 billion of that was improper. The error rate was 8.6%. One in every twelve dollars paid went to the wrong person, for the wrong amount, or for a service that was never properly documented.

That $50.3 billion is not an outside estimate or an advocacy calculation. It is the number CMS published in its own annual improper payments report, subject to OMB review, and submitted to Congress under the Payment Integrity Information Act. The federal government measured this, reported it, and moved on.

What “Improper Payment” Actually Means
An improper payment is any payment that should not have been made or was made in the wrong amount — including payments to ineligible recipients, payments for services not rendered or not sufficiently documented, and duplicate payments. It does not require criminal intent. Most Medicaid improper payments are eligibility errors: people who lost their eligibility — through income changes, state moves, incarceration, or death — but stayed enrolled because the eligibility system was not updated. Every month they stayed enrolled, a managed care organization received a capitation payment for covering them. The federal government paid that bill. No fraud required. Just no one checking.
§ 02 / What Went Wrong

The Pandemic Froze the System. Then No One Unfroze It.

Pre-pandemic, states were required to conduct annual eligibility redeterminations — verifying that enrolled beneficiaries still met income and residency requirements. The Families First Coronavirus Response Act of 2020 suspended this requirement during the public health emergency in exchange for enhanced federal matching funds. States could not disenroll anyone. Medicaid enrollment grew from 72 million pre-pandemic to over 90 million at peak.

The Consolidated Appropriations Act of 2023 ended continuous enrollment and required states to begin redeterminations. The “unwinding” process ran from April 2023 through December 2024. States disenrolled approximately 25 million people. A significant share of those disenrollments — perhaps the majority — were individuals who had lost eligibility months or years earlier but had never been removed from the rolls. For the duration of their improper enrollment, the federal government had been paying their managed care capitation every month.

Error CategoryShareNote
Eligibility errors — enrollees no longer qualifying~60%Largest share; includes deceased, incarcerated, income-ineligible
Managed care capitation — payment rate errors~25%Rates set on stale data; overpayments to MCOs
Fee-for-service claims — procedural & documentation~10%Missing prior authorizations, unsupported claims
Other (coordination of benefits, TPL failures)~5%Failure to bill third-party payers first
Source: CMS PERM FY2023 Report · Percentages are approximate, based on CMS published breakdowns
§ 03 / State-Level Variation

Some States Are Running 15%+ Error Rates. CMS Publishes It Anyway.

PERM measurements rotate through states on a three-year cycle — each state is measured roughly once every three years — meaning the national 8.6% figure is actually a blended estimate across different measurement years. States that were measured during the pandemic continuous enrollment freeze posted significantly higher error rates. CMS publishes state-level data; some large states have reported rates well above the national average in recent measurement years.

The structural problem is the same one that drives all Medicaid integrity failures: states administer the program, but the federal government funds the majority of it through the FMAP match. A state that improves its eligibility verification — reducing improper payments — bears the administrative cost of that improvement while most of the financial savings go to federal taxpayers. The incentive is inverted. The better a state gets at catching errors, the more it spends on administration relative to the savings it captures.

Medicaid improper payments have consistently exceeded $30 billion annually and represent one of the federal government's largest and most persistent payment integrity challenges.

GAO High-Risk Series — Medicaid Program Integrity (2023)
§ 04 / The Oversight Gap

GAO Has Flagged This for Twenty Years. CMS’s Response: More Planning Documents.

Medicaid has been on GAO’s High-Risk List since 2003 — twenty-plus years. In that period, GAO and HHS OIG have published dozens of reports documenting specific failure modes: eligibility systems that do not cross-check income data, managed care contracts that overpay for phantom services, states that fail to coordinate benefits with Medicare or private insurance before billing Medicaid. CMS has responded to nearly every recommendation with compliance plans, pilot programs, and guidance documents.

The error rate has moved in a narrow band — between 6% and 10% — for most of the past decade, with the pandemic spike pushing it higher. At no point in the program’s history has the improper payment rate hit the 3% target required by the Payment Integrity Information Act. The GAO has described CMS’s progress on Medicaid integrity as insufficient to meet program needs. Congress has held hearings. The dollar amount has continued to grow.

§ 05 / The Bottom Line
What This Means
$50.3 billion in a single fiscal year. An 8.6% error rate — nearly three times the statutory 3% target. The largest single driver of the federal government’s $186 billion total improper payment problem in FY2025. The primary cause is eligibility: millions of people stay enrolled in Medicaid after they stop being eligible because no one checks. The pandemic continuous enrollment requirement made the backlog worse for three years. The “unwinding” recovered some of it — but the structural problem that created the backlog in the first place remains: states have no financial incentive to invest in eligibility verification when the federal government is paying most of the bill.