The best part is the table (page 5 in the pdf) showing per agency and program what percentage of payments were made incorrectly. The Small Business Administration's Paycheck Protection Program, for two different aspects of the program, managed rates of 49.1 percent and 40.5 percent respectively - that does seem pretty spectacularly awful. Though I could see how SMEs would have frequently shifting employee numbers and not the most optimal processes of submitting claims under such a program given their limited staff, so I'm sure it's very difficult to get such payments fully right and many of the wrong ones may not be because of fraudulent intent.

For what it's worth, the SBA PPP's 2024 numbers seem to be down to 'only' around 25 percent wrong payments - seems like very large parts of that huge total number of incorrect payments are related to various programs started because of Covid and slowly phased out since then.

One does also need to put the numbers in context - compared to the full federal spending, we're talking about less than 4 percent for 2023, still less for the full period since 2003. Still far more than it should be of course, but especially for the urgent programs like the ones during Covid, which unsurprisingly are also the ones with the worst accuracy rates, there is a trade off to be made between speed and accuracy. More rigorous processes might have meant a bit less money wasted, but people would've received the assistance later and that would've come with direct and indirect costs as well.

And not all of the amount was actually overpaid - in fact about 5 percent of it was underpaid, which seems quite confusing to add those things up like that, and then a good chunk that they can't conclusively show if it was right or not - as aero said, that part might often be a matter of minor bureaucratic issues.