Dozens of states could face new costs because of high error rates in SNAP food aid

[BumRushDaShow]

Source: AP

Updated 11:22 PM EDT, June 24, 2026


Several dozen states could have to fork over millions of dollars to provide food aid to lower-income residents, if they don’t cut down on payment errors in the Supplemental Nutrition Assistance Program.

Nine states, meanwhile, won’t have to pay a penny toward SNAP benefits, because their error rates are so low that they won an exemption from a cost-sharing requirement included in a big tax-and-spending law signed by President Donald Trump.

Data released Wednesday by the U.S. Department of Agriculture provides a first look at the winners — and potential losers — under the new law.

The error rate refers to the percentage of SNAP benefits paid either above or below what people should have received, primarily because of mistakes. While low-error states are guaranteed to owe nothing when the annual cost-sharing requirement begins in October 2027, others will have another year to try to reduce their errors and decrease the hit to their budgets.

Read more: https://apnews.com/article/snap-food-aid-error-rates-4fa06549e3ec2bd1a1c665b1985cefea

---

Link to USDA PRESS RELEASE - USDA Announces FY 2025 State Payment Error Rates in SNAP

Link to USDA REPORT. (PDF) - https://www.fna.usda.gov/sites/default/files/resource-files/snap-qcfy25-per.pdf

[pat_k]

1. "Errors" (even underpayments) are ALL fraud in Trumpublican-speak.

Doesn't matter if the error is easily remedied, inadvertent, or if it was an error on the part of the state that resulted in underpayments, to the propagandists it is all "fraud."

And, perhaps unsurprisingly in Trump's 1984 executive branch, a wrongful denial of SNAP benefits is not counted as part of the standard SNAP Payment Error Rate (PER).

So, the best way to keep the "error rate" low is to deny, deny, deny, deny, and deny some more.

Furthermore, the systems used to detect the "errors" themselves have a completely unacceptable error rate. (i.e., flag problems that are not actually problems, particularly with the greater use of AI.)