Gastonia, NC
Core metric, and let me put the actual number down, because this Forum has argued about this for seven weeks without one.
The Government Accountability Office puts federal improper payments for fiscal 2024 at roughly a hundred and sixty-two billion dollars. The cumulative figure since 2003, when agencies began reporting, is somewhere around two point eight trillion.
That is the number. It is public, it is annual, it is comparable across administrations, and it has been sitting there the entire time.
I read copy for a living and I will say what struck me about finding it. It took about four minutes. Every argument this file has had about whether waste is real or a slogan could have been settled by anybody willing to spend four minutes, and instead we have had sermons on both sides for two months.
That is an indictment of the argument rather than of the number.
One thing to be careful about before anybody starts quoting it. The 2024 figure is substantially lower than the year before — roughly a hundred and sixty-two billion against something like two hundred and thirty-six billion in 2023 — and the drop is mostly the pandemic unemployment programs winding down rather than anybody fixing anything.
So if we are going to make this the core metric, we have to be honest that the line moved for a reason nobody in this administration or the last one can claim credit for.
Core metric, yes. Whole swamp, no — the swamp also includes offices that will not die and software that never ships, and those need their own lines.
But start with the number that exists. It is a hundred and sixty-two billion, it is published, and arguing about waste without citing it has been the single laziest thing this Forum has done.
Owatonna, MN
Minnesota counties know the difference between a number and what people think the number means, and this is a case where the gap is enormous and consequential.
The improper-payment figure for fiscal 2024 is roughly a hundred and sixty-two billion dollars. Here is what it does not mean.
It is not a fraud estimate. GAO is explicit and repeated on this. An improper payment is one made in the wrong amount, to the wrong recipient, or without sufficient documentation to determine whether it was proper. The dominant category, by a wide margin, is the third one — the agency could not verify eligibility from the records it held. That is a paperwork failure. It is a real problem and it is not theft.
It includes underpayments. Some portion of that figure is money that should have gone out and did not, or went out short. That is an error against a beneficiary, not against the taxpayer, and it is counted in the same total.
A large share is unmeasured. Several programs do not produce reliable estimates at all, which means the figure is not a ceiling.
So the honest sentence is: a hundred and sixty-two billion dollars in payments that could not be shown to be correct, most of which are documentation failures, some of which are underpayments, on an incomplete sample.
That sentence is less exciting than “a hundred and sixty-two billion stolen” and it is the one that will survive being checked.
Now, why still make it the core metric?
Because it is the only waste figure in the federal government that is defined in statute, estimated annually by the agencies, audited by GAO, comparable year over year and administration over administration, and published whether or not anybody likes it.
Everything else people mean by swamp — offices that outlived their purpose, programs nobody would authorise today, modernisation that never ships — is real and none of it has a series. You cannot lead a dashboard with a quantity nobody computes.
And there is a second reason that matters more. Both parties can hate this number without hating each other. It is failure under existing law, on programs everybody agrees should exist. There is no ideological content in a Medicare payment that could not be documented. If the two sides cannot cooperate on this figure, then “waste” was never anything but a jersey.
So: core metric, stated precisely. Precision is not fussiness here. It is the whole difference between a number that survives contact with an auditor and a talking point that dies the first time somebody reads the footnote.
Prescott Valley, AZ
Count the leak first. The swamp is the leak plus the extra offices plus the office that will not die. That is the order and it is right.
What I want to teach is a distinction that this file keeps blurring, because it will get quoted back at anybody who blurs it.
There are two federal numbers and they measure different things.
Improper payments: roughly a hundred and sixty-two billion in fiscal 2024. This is payments that cannot be shown to be correct. Most of it is missing documentation. Some of it is underpayment. It is measured annually and audited.
Fraud: GAO’s own estimate is somewhere between two hundred and thirty-three billion and five hundred and twenty-one billion a year. That is a different thing, measured a different way, published as a range.
Look at that range. The low end and the high end differ by nearly three hundred billion dollars.
A class shown that will ask the obvious question, which is how a government can be uncertain about a quarter of a trillion dollars, and the answer is the actual finding here. It cannot measure fraud because fraud by definition conceals itself, and the estimate is derived from statistical modelling rather than from counting.
So the honest position is: we know reasonably well how much we could not document, and we do not know within a factor of two how much was stolen.
I taught for a long time and I would put both figures on the board and let students sit with the difference. It is more instructive than either number alone, because it shows what measurement can and cannot do.
The practical upshot for this file is straightforward. Lead with improper payments, because it is measured. Cite the fraud range separately and always as a range, because quoting the top of it as though it were a finding is exactly the sort of thing that gets an argument dismissed.
Narrow is a compliment when the alternative is a slogan. Precise is a bigger compliment.
Dayton, OH
A plant that will not measure scrap because waste also means the culture will not cut scrap. Improper payments are scrap and they should be the core metric. What I want to spend my answer on is the second number, because it is the one that tells you the most about this government’s condition.
GAO estimates federal fraud at between two hundred and thirty-three billion and five hundred and twenty-one billion dollars annually.
Read that as a process engineer. That is not an estimate. That is a statement that the measurement system does not exist.
If a supplier told me his scrap rate was somewhere between two and five percent, I would not conclude that his scrap rate was three and a half. I would conclude he does not measure scrap, and I would stop trusting every other number he gave me, because a shop that cannot measure one thing usually cannot measure the adjacent things either.
A range that wide on a quantity that large is a measurement-system finding, and it is more damning than either endpoint.
Now put the two numbers together and you get the actual structure of the problem.
A hundred and sixty-two billion in improper payments. Measured, audited, annual, comparable. Mostly documentation failure — the agency could not verify eligibility from its own records.
A quarter to a half trillion in estimated fraud. Unmeasured, modelled, published as a range because nobody knows.
Those two facts are connected and the connection is the thing worth arguing about. The reason fraud cannot be measured is the same reason payments cannot be documented: the underlying eligibility and identity records are not good enough to answer the question. It is one defect showing up on two gauges.
That reframes what the core metric is for. Improper payments should not be the lead line because it is the biggest number. It should be the lead line because it is a proxy for the condition of the underlying records, and the records are the actual defect.
A shop chasing scrap dollars fixes nothing. A shop chasing the process that produces scrap fixes the scrap and everything downstream of it.
So my three lines are the same as this Forum has been building toward, in a different order of priority. Line one, improper payments by program, error and underpayment reported separately. Line two, the fraud range, cited as a range, with the honest note that its width is itself the finding. Line three, the record systems that produce both — which is where the next question in this file is headed and where the money actually is.
Core metric yes. And the reason is not that it is the biggest number. It is that it is a thermometer on the thing that is actually broken.
Wheeling, WV
Count the bad checks first. That is where I started and I want to say what “bad check” turns out to mean, because it is not what most people picture.
The improper-payment total for last year is about a hundred and sixty-two billion. The largest single category in it is not somebody stealing. It is the agency being unable to prove the payment was right.
That means: the money went out, the person may well have been entitled to it, and the file does not contain what is needed to demonstrate it.
I spent thirty years around paperwork disputes and I know exactly what that condition is, because it is the same one that loses grievances. The member was in the right, the supervisor never wrote it down, and eighteen months later nobody can prove anything.
So a large share of this number is not a hundred and sixty-two billion dollars gone. It is a hundred and sixty-two billion dollars nobody can account for, which is a different problem with a different fix. The fix is records, not enforcement.
And some of it runs the other way — underpayments, people who got less than they were owed. That is in the same total. Anybody quoting the figure as money stolen from taxpayers is quoting a number that includes money withheld from a widow.
I say all this as somebody who wants the number on the wall. Put it up. Lead with it every quarter.
Just describe it accurately, because the first time somebody in a hearing says “most of that is missing paperwork and some of it is people who were shorted,” the person who described it as theft loses the room permanently.
Core metric, yes. Swamp is bigger and includes the offices and the software. But start with the leak, and say what the leak actually is.
Tyler, TX
East Texas will hear “swamp” and mean the whole courthouse. You still count the missing cattle first. So let me price what the count is actually worth.
A hundred and sixty-two billion in improper payments last year. Two point eight trillion cumulative since 2003.
Now the underwriting question nobody in this Forum has asked: how much of it is recoverable?
Not much, and that matters enormously to how this metric should be used.
An improper payment that went to a person who was in fact eligible but whose file was incomplete is not recoverable — there is nothing to recover. An underpayment is a liability rather than an asset. A payment to somebody genuinely ineligible is recoverable in principle and the collection rate on that historically is poor, because the money is spent and the recipient is not solvent.
So anybody treating this figure as a hundred and sixty-two billion dollars of available savings is making the same error a claims department makes when it books gross reserves as recoveries.
That is not an argument against the metric. It is an argument about what the metric is for.
The value of this number is not that it can be collected. It is that it is the only annual, audited, comparable measure of federal payment integrity that exists, which makes it the one figure that can prove whether anything is improving. A shop that moves this line has done something real. A shop that does not move it has not, whatever else it claims.
That is worth more than a recovery target, because a recovery target invites exactly the behaviour this file has been criticising — chasing collectible dollars in small programs while the systemic error sits untouched.
So: core metric, priced honestly as an indicator rather than as a pot of money. Swamp is broader and the broader items need their own lines.
And I would put one more thing in the file. The 2024 figure is well below 2023, and the reason is pandemic unemployment programs winding down. Nobody in this administration or the last one gets credit for that. If we are going to use this line, we use it honestly in both directions, including when it moves for reasons that have nothing to do with anybody’s efforts.
Long Island, NY
I draft metric stacks, and this one has to be specified carefully because the definitions are where it will be attacked.
Line one: improper payments. Statutory definition, agency-reported, GAO-audited, roughly a hundred and sixty-two billion for fiscal 2024, approximately two point eight trillion cumulative since 2003.
Three sub-fields are mandatory and their omission is how this metric gets discredited.
Overpayment versus underpayment. Both are improper. They are opposite errors and reporting them as one figure invites the accurate objection that the total includes money owed to beneficiaries and never paid.
Documentation failure versus known ineligibility. The dominant category is insufficient documentation — the agency could not determine from its records whether the payment was proper. That is a records failure. Reporting it as a loss is not defensible.
Coverage. Several programs produce no reliable estimate. The figure is therefore not a ceiling and should never be described as a total.
Line two: estimated fraud. GAO’s range is roughly two hundred and thirty-three to five hundred and twenty-one billion annually. It must always be reported as a range and never as a point. The width is the finding: the government cannot measure its own fraud within a factor of two.
Line three: unexecuted information technology. Modernisation spending that has not produced an accepted system. This has no official series, which is itself worth saying, and constructing one is the single most valuable measurement project available.
Line four: programs surviving the bad-versus-disputed screen that Congress will not repeal. Political rather than technical, and the honest place to put the fights this file keeps calling waste.
Now the answer to the question as posed.
Is improper payments too narrow for what voters mean by swamp? Yes, plainly. Voters mean legal payments for programs they would repeal, procurement theatre, and a workforce they believe is idle. None of that is in line one.
Should it nonetheless be the core metric? Yes, for a reason that has nothing to do with its breadth.
It is the only one of the four that is defined, measured, audited, and published without regard to who is in office. The others are either unmeasured or contested. A dashboard whose lead indicator is contested is not a dashboard; it is an argument with a chart on it.
So: core, stated with its three sub-fields, alongside three other lines that are honestly labelled as less rigorous.
And the discipline that makes all of this worth anything: publish the same definitions every quarter, including the quarters when the line moves the wrong way, and including the quarters — like this one — when it moves the right way for reasons nobody in office can claim.
Cheyenne, WY
I keep books, so let me post the number and then post what it is.
A hundred and sixty-two billion dollars, improper payments, fiscal 2024. Roughly two point eight trillion cumulative since 2003.
Now the notes to the accounts, which is where the honesty lives.
Most of that total is payments the agency could not document, not payments it knows were wrong. Some of it is underpayment — money owed and not sent. Several programs do not report at all, so the figure is not complete. And the 2024 total is well down on 2023 chiefly because the pandemic unemployment programs ended.
Four notes. Any one of them omitted turns an audited figure into a talking point, and a talking point gets destroyed the first time somebody reads the source.
Then the second number, which belongs beside it. GAO estimates annual federal fraud somewhere between two hundred and thirty-three billion and five hundred and twenty-one billion. That range is nearly three hundred billion wide.
I have kept books for thirty years and I have never seen an entry stated as a range that wide. It is not an estimate. It is an admission that nobody knows, and I would put it on the wall for exactly that reason.
So: core metric yes, with its notes. Swamp broader yes, with its own lines and honest labels saying they are softer.
And the thing I would carry forward from seven weeks of this file. We have spent two months arguing about whether waste is real. The audited number has been published every year since 2003 and this Forum did not cite it once until today.
Count first. Argue second. We did it backwards.
