Washington Admits It Sends Billions to the Wrong People, by Law Every Year

Most scandals have to be uncovered. This one is published. Every year, by order of Congress, the federal government counts the money it paid out that it should not have, and posts the total where anyone can read it. No whistleblower, no leaked memo, no investigative team. The government confesses on schedule, and then the country goes on as before.

That is the topic. I want to make three points about it, each from the government's own documents: the number is large, the number is misunderstood, and the right response is structural reform, not another round of promises to try harder.

Define the term, carefully

An improper payment is not the same thing as fraud. This is the single most important sentence in the debate, and it is the one most often botched, by critics who say "fraud" and by defenders who say "it's only paperwork."

Under federal law, an improper payment is any payment that should not have been made or that was made in an incorrect amount. The definition covers overpayments and underpayments. It covers payments to an ineligible recipient, payments for an ineligible good or service, duplicate payments, and payments for which the agency cannot tell, from the documentation, whether the payment was proper. That last category matters: a payment can be counted as improper because the file is missing the paper that would prove it was correct, even if the money went to the right person.

Fraud is a narrower thing. It is a deliberate deception to obtain money one is not entitled to. Some improper payments are fraud. Many are error: a worker enters the wrong figure, a recipient's income changed, a provider billed under the wrong code, a state agency did not verify a record. Some are neither, in the sense that nobody can say whether they were right.

I mention this up front because the argument in this post does not need the number to be fraud. It needs only to be what it is: money the government itself says it should not have sent, or cannot prove it should have.

The statutes, in order

Congress has been at this for more than two decades, and the record is worth knowing in sequence.

The Improper Payments Information Act of 2002. Signed in November 2002 and made Public Law 107-300, it required agencies to review their programs, identify those susceptible to significant improper payments, estimate the amounts and report them. It is the founding statute.

The Improper Payments Elimination and Recovery Act of 2010. Public Law 111-204 expanded the requirements, added recovery audits for certain programs and tightened the reporting.

The Improper Payments Elimination and Recovery Improvement Act of 2012. Public Law 112-248 pushed further, and among other things it wrote the "Do Not Pay" initiative into law, a set of databases that agencies are supposed to check before they release payments.

The Payment Integrity Information Act of 2019. Public Law 116-117, signed in March 2020, repealed and consolidated the earlier laws into a single statute and kept the core: agencies must assess risk, estimate improper payments in programs that are susceptible, report them, and say what they are doing about them. Under it, a program is generally treated as susceptible to significant improper payments when its estimate crosses a threshold, stated as a share of the money paid and a dollar amount; check the statute for the exact figures in force.

The same instinct, a promise to try harder, runs through The Federal Workforce Is Smaller Than You Think, and That Is the Problem, where the real lever turns out to be the program rather than the payroll. Four statutes in eighteen years, each with more detail than the last. One would think that by the fourth, the problem would be shrinking. The government's own scorekeeping does not say so.

What the government's own books show

The Office of Management and Budget runs a website, paymentaccuracy.gov, where agencies post their estimates by program and year. The Government Accountability Office, Congress's auditor, reviews them and reports on them. Two features of GAO's reports are worth knowing.

First, the cumulative figure. GAO has summed agencies' estimates from fiscal year 2003, when reporting began under the 2002 act, forward. In its report on fiscal year 2023 GAO put that cumulative total at about $2.7 trillion, including $236 billion for that year alone, and a later report put it near $2.8 trillion. I will not give you a figure to the billion, because it grows with each annual report and because the exact number depends on the year you read. Go to GAO's most recent report on improper payments and read the cumulative line. Then compare it with the national debt, which you can see on the national debt clock.

Second, GAO's recurring caveat: the estimates are incomplete. In fiscal year 2023, for example, GAO reported that agencies had not developed and reported required estimates for nine risk-susceptible programs. When the government says it cannot estimate how much it is wasting in a program, it is making an admission that a private firm would never be allowed to make to its auditors.

Third, and this is the part that surprises people, the biggest programs are not obscure. The same few appear year after year.

Which programs recur

Three programs are worth knowing, for different reasons.

Medicare. Including both the traditional fee-for-service program and Medicare Advantage, Medicare is among the largest programs in the federal budget, so even a modest rate produces a very large dollar figure. Check paymentaccuracy.gov for its current estimate. A common cause of improper payments in health programs is insufficient documentation: a provider billed for a service and the medical record does not support it.

Medicaid. Medicaid is a joint program in which states run the day-to-day work and the federal government pays a share. It is likewise very large, and its estimate depends on state-level reviews of eligibility and of the program's own rules. The federal government writes the check, the states determine who gets it, and each can point at the other.

The Earned Income Tax Credit. The EITC is a refundable credit for low-income working families, administered by the IRS through the tax system. Its estimated improper payment rate has been among the highest of any large program: GAO's table of programs with rates of 10 percent or higher for fiscal year 2023 lists it at 33.5 percent, about $21.9 billion, and an older Treasury inspector general figure put it at 21 to 25 percent of the credit in fiscal 2012. The credit is complicated, and many errors involve claimants who misstate a child's residency, income or filing status.

Other programs appear too, including unemployment insurance, the Supplemental Nutrition Assistance Program and Pell Grants, but the three above are the ones to remember. They are the ones with large dollar amounts or high rates, and they are the ones worth watching from year to year.

A worked example, labeled as a hypothetical

Let me show you how to read a rate. This is hypothetical, to illustrate the arithmetic.

Suppose a program pays out $800 billion in a year and its reported improper payment rate is 5 percent. Five percent of $800 billion is $40 billion. That is the estimate of dollars paid improperly. Now suppose that of that $40 billion, a quarter is overpayments the agency could recover and the rest is documentation failures. A quarter of $40 billion is $10 billion, which would be recoverable in theory. The remaining $30 billion represents payments that may have been correct, or not, and nobody can say.

Two lessons follow. The headline number is not all lost money. And a low-looking percentage on a huge program is still a very large number of dollars. A rate of 1 percent on that $800 billion program would still be $8 billion. Defenders who stress that "the rate is only a few percent" are right about the rate and quiet about the base.

The strongest case against my conclusion

Let me do what any careful argument should do, and state the other side as its best defenders would.

They say: the improper payment number is routinely misread. Much of it is documentation error rather than theft. Some of it is underpayment: people who were owed more and got less. Aggressive anti-fraud controls impose costs on legitimate recipients, who face more paperwork and more delay, and in programs for the poor and the sick the cost of a wrongly denied benefit is human. A program with a double-digit improper payment rate may also be a program that delivers benefits to millions of eligible people who would otherwise go without. And the numbers are, by design, estimates produced by statistical sampling, with wide error bars.

All fair. Here is why they do not change the conclusion.

1. The government itself reports the figure under a statute it passed. I am not importing a critic's number. The law requires agencies to produce the estimate, and the figure comes from agencies and is reviewed by GAO.
2. Documentation failure is a failure. A program that cannot show that a payment was proper has no basis to tell taxpayers that it was. In the private sector, an expense that cannot be documented is disallowed. A rule that applies to a family's tax return should apply to the Treasury.
3. Underpayments are in the number too, and they count against the program. Whether a payment is too high or too low, it signals that the program's rules cannot be applied correctly. A system that makes errors in both directions is not working for either recipient or taxpayer.
4. Controls and compassion are not opposites. A well-designed program verifies eligibility at the front, with data matching, and spends less on chasing people afterward. The sloppy approach, pay first and look later, hurts the honest recipient too, because recovery from a family that received too much and spent it is cruel, and the cost lands on the same people the defenders worry about.
5. The problem has lasted two decades. Reporting began in fiscal 2003, and GAO's cumulative total is now in the trillions. If this were a matter of one bad quarter, I would agree that patience is the answer. When the problem has lasted this long, the design deserves a look.

Why oversight alone has not fixed it

Here is the part that conservatives should say plainly. Congress has tried the reporting route four times. It has asked agencies to estimate, to report, to recover and to check databases. These are good ideas and I would not repeal them. But look at what they assume: that the program is basically sound and that the failure is one of attention. If a program is built so that the payment is made first and eligibility is checked afterward, or if a state determines eligibility with federal money and bears little of the cost of a mistake, then attention is not the missing ingredient. Incentives are.

That is where I want the reader to go from here. The size of Medicare and Medicaid and the high rate of the EITC are not a reason to exhort the agencies to try harder. It is a reason to ask whether the structure of those programs gives anyone an incentive to get the payment right. A state that pays only part of a Medicaid claim and gets the rest from Washington has a different calculus than one that pays all of it. A tax credit that is paid out of a refund system designed to deliver money quickly, based on a form the filer completes himself, is a poor fit for a task that requires verification. Reform means changing those structures: matching federal and state responsibility, verifying eligibility on the front end, simplifying rules so that honest people can follow them, and giving agencies a reason to care about the result.

GAO makes a similar observation about programs that have remained on its high-risk list for a long time. For the longer story of that list, see GAO's High-Risk List Has Named Some of the Same Programs Since 1990. For the practical business of finding and reading these reports, see How to Read an Inspector General Report and Use It to Hold an Agency to Account.

Measured against the Founding

The Constitution's treatment of public money is short and strict. Article I, Section 9, Clause 7: "No Money shall be drawn from the Treasury, but in Consequence of Appropriations made by Law; and a regular Statement and Account of the Receipts and Expenditures of all public Money shall be published from time to time." The framers believed that those who spend other people's money owe an account of it. Here the account exists, and the account says the government cannot be sure that it has spent the money as the law required.

The Clause was meant to make public accounts a rule, not a courtesy. The Founders did not imagine a budget measured in trillions. They imagined a government in which the people could read the books. We have the books. What we lack is the political will to act on what is in them.

There is also a moral point, which belongs to our tradition, and I will state it as a value rather than a fact. A government that takes money from a family by force of law holds that money as a trust. A trustee who cannot say where the money went has failed in the first duty of a trustee. The people who pay the bills are not abstractions. They are the families who file returns and who understand perfectly well that an expense without a receipt does not survive an audit.

What reform looks like

A short list, and each item is a place where the government's own reports point.

• Pay it right the first time. Use the Do Not Pay data and other matching tools before payment, not after, and give agencies the authority to use them.
• Align the cost with the control. Where states run a program with federal money, give them a stake in accuracy by tying part of their funding to payment accuracy.
• Simplify the rules. A program whose eligibility rules most honest filers cannot follow will produce errors no matter how hard anyone tries.
• Require every agency to produce an estimate. A program that cannot say how much it pays improperly should say so, in public, with a date by which it will.
• Sunset what does not improve. A program that stays at the top of the list for a decade should face a vote on whether it continues in its present form.

The case in brief

1. An improper payment is money paid in the wrong amount or to the wrong party, or without proof that it was right. It is not always fraud, and I do not need it to be.
2. Congress has legislated on the subject four times, in 2002, 2010, 2012 and 2019, and the government's own reports still show large annual totals, with GAO's cumulative count at about $2.7 trillion since fiscal 2003, as of its fiscal 2023 report.
3. Medicare, Medicaid and the Earned Income Tax Credit are the programs to watch, and the Credit's high rate points to structure and incentive, not to one bad year.
4. The fair objections, that the number includes documentation errors and underpayments, do not change the conclusion, because a program that cannot prove its payments are proper is not entitled to the taxpayer's trust.
5. The remedy is reform: verify first, align responsibility, simplify the rules, and sunset programs that do not improve.

What you can do: Open paymentaccuracy.gov, find one program you or your family help pay for through taxes, read its reported improper payment rate and year, and send the figure to your member of Congress through My Representatives with one question: what is being done to change the program's design?