Limitations of Improper Payment Data

Why the government's payment error estimates deserve closer scrutiny

I recently wrote about why improper payments are a better target for policymakers than “waste, fraud, and abuse,” emphasizing the lack of consistent and reliable estimates of fraud and the subjectivity of waste and abuse.

That said, improper payment data have its own flaws. Some high-risk programs do not report estimates at all, other reported estimates have been deemed unreliable by auditors, and the government’s own categories often tell policymakers little about where the errors actually occur.

My goal is not to dismiss improper payment data, but to interpret them correctly. Doing so requires looking beyond the headline estimates to how they are produced, what they exclude, and how they classify payment errors.

Self-Reporting

Each year, the Government Accountability Office (GAO) publishes a report on improper payments, noting the total improper payments made by the government in the previous year, the sources of improper payments, and various other aspects. While helpful, this seems to have created a misconception among some that GAO is the originator of these estimates.

It is not.

Improper payments are estimated by agencies and reported in their annual financial reports. These estimates are then published on a Treasury-run website, PaymentAccuracy.gov, prior to GAO’s report.

Statutes, such as the Payment Integrity Information Act of 2019 (PIIA), set payment integrity requirements, including for the estimation of improper payments. But the self-reported nature of improper payment data creates an inherent risk of underreporting. Agencies estimate their own error rates, often using sampling methodologies that vary across programs.

That alone increases the risk of measurement error. Self-reporting also creates an incentive concern. Agencies that report high improper payment rates may face reputational or political costs, while reductions in improper payments can be publicly presented as evidence of improved management. For example, in November 2024 Biden’s Office of Management and Budget (OMB) highlighted that the FY 2024 government-wide improper payment rate was the lowest in more than a decade, while a separate press release from Treasury touted an increase in prevented and recovered fraud and improper payments.

This does not prove that agencies intentionally understate improper payments. But it does mean that agency-produced estimates should be interpreted with appropriate caution. This limitation is partly mitigated by independent inspectors general (IGs), who report on agencies’ compliance with PIIA and on the reliability of agencies’ improper payment estimates.

But compliance with the PIIA and its legislative predecessors has been inconsistent.

Compliance Failures and Data Reliability

Major agencies frequently fail to meet statutory standards. Since 2011, the 24 agencies covered under the Chief Financial Officers (CFO) Act have averaged compliance rates of just 48 percent. The table below shows this compliance by agency since 2011.

Notably, four of those agencies—the Departments of Health and Human Services, Treasury, Agriculture, and Veterans Affairs—have not been compliant a single time. Six other agencies, including the Department of Defense, have been compliant just once or twice over the same period.

One particularly important aspect of noncompliance is the reliability and completeness of improper payment estimates. In FY 2024, eight agencies were flagged during their annual PIIA compliance audits for providing unreliable improper and unknown payment estimates.

The next table lists 26 programs with missing or unreliable improper and unknown payment estimates. Ten of these programs were deemed in their audits to be at risk of significant improper payments and therefore required to report improper payment estimates but did not. The remaining 16 reported estimates that auditors found unreliable for a variety of reasons. In total, these programs comprised more than $478 billion in federal spending, or more than 7 percent of spending in FY 2024.

That is not just a paperwork problem. Agencies can fail compliance for many reasons, including failure to publish corrective action plans, failure to meet reduction targets, or failure to report required payment integrity information. But missing or unreliable estimates go directly to the usefulness of the data itself.

Incompleteness

Beyond reporting and compliance issues, improper payment estimates are incomplete by design. Not all programs are required to report, and some high-risk programs fail to produce estimates, including those mentioned in the preceeding table.

Improper payments are required to be estimated only if a program has more than $10 million in outlays and that program has been determined to be at risk of having significant improper payments. This is determined through a risk assessment, conducted at least once every three years, to check if a program likely has either:

1. both 1.5% of its outlays are improper or unknown and it has at least $10 million in total improper payments, or

2. more than $100 million in improper payments, regardless of the combined improper-unknown payment rate.

Consequently, programs with outlays of $10 million or less are not required to conduct risk assessments. And programs that are expected to have up to, but not exceeding, $100 million in improper and unknown payments are not required to conduct a formal estimate so long as their combined improper-unknown payment rate is below 1.5%.

In some cases, agencies report that they are unable to estimate improper payments at all. The most notable example is Temporary Assistance for Needy Families (TANF)—a $16.5 billion welfare block grant to states created in President Clinton’s signature welfare reform package in 1996. The Department of Health and Human Services (HHS) provides the following reason for this problem:

Statutory limits prevent HHS from collecting the necessary data, as section 411 of the Social Security Act only permits specified data elements that exclude case and payment accuracy. In addition, section 417 restricts HHS from regulating states without express congressional authority. Until the law is amended to allow a national error rate, HHS cannot project when TANF will be able to provide such estimates.

Official Classifications Tell Us Too Little

We can draw several conclusions from improper payments data:

1. they are large and rising

2. they are mostly overpayments

3. they are primarily due to a failure to access existing and available data

4. they are almost universally classified by agencies as being outside of each agency’s power to control

5. they are concentrated in a handful of programs

But beyond whether agencies classify improper payments as within or outside their control, the data provide limited information about where the error occurred.

The “outside agency control” label collapses very different problems into a single category. Some improper payments occur because beneficiaries or third parties fail to provide accurate or timely information. Others arise because states fail to verify eligibility information. Still others occur because the relevant data needed to verify eligibility do not exist or cannot be accessed. Treating all of these as simply outside agency control tells policymakers too little about the nature of the problem or the types of reforms that might address it.

An Alternative Way to Classify Improper Payments

The categories below are my attempt to reclassify the causes of improper payments by who or what is responsible, using the agencies’ own descriptions as the underlying basis:

1. agency noncompliance

2. when the required data or information to verify eligibility do not exist

3. state noncompliance

4. noncompliance by nongovernment entity or beneficiary

Examples of these alternative primary causes of improper payments are shown in the next table. The subsequent table lists each reporting program, along with its alternative primary cause and total programmatic overpayments.

$67.2 billion, or 44%, of all overpayments in FY 2025 were associated with programs whose primary reported cause of improper payments was noncompliance by a nongovernment entity or beneficiary. Examples of this include individuals failing to update income information for eligibility of federal programs and insufficient documentation from healthcare or insurance providers.

$47.3 billion, or 31%, of all overpayments were associated with programs in which state noncompliance was the primary reported cause. Examples include states failing to obtain, document, or verify eligibility information.

$28.2 billion, or 18%, of overpayments were associated with programs in which the inexistence of data was the primary reported cause of improper payments. This predominantly occurred with a handful of tax credits.

And $10.5 billion, or 7%, of overpayments were associated with programs in which agency noncompliance was the primary reported cause. These include administrative or procedural errors, such as failing to update information, verify eligibility, or acceptance of unacceptable documentation.

Conclusion

Improper payment data remain a better starting point than broad claims about “waste, fraud, and abuse.” But they are not perfect. They are agency-produced estimates, subject to reporting thresholds, audit findings, missing data, and classification choices that can obscure where payment errors originate.

The data are most useful when they are used to ask better questions rather than provide definitive answers. Why are estimates missing for some programs? Which reported estimates are reliable? Who actually failed to comply, and what information was unavailable? Those questions are ultimately more useful for designing reforms than the government-wide improper payment total alone.