For the first time in the history of the Supplemental Nutrition Assistance Program (SNAP), starting in FY2028 states will be liable for a share of SNAP benefit costs. Each state’s cost share will range from 0% to 15% and will depend on its payment error rate. However, as I pointed out in a recent paper, a conceptual flaw in the formula treats overpayments and underpayments asymmetrically. Using FY2025 data, I show in this post that correcting this flaw would increase states’ payment error rates and have important implications for their shares of SNAP benefit costs. In the long run—though not in the short run—it would likely produce greater federal savings from the program.
The conceptual flaw in the payment error rate formula
The payment error rate formula is shown below. It is equal to the sum of all benefit payment errors E divided by total benefit payments B, where the errors and benefits are summed across all SNAP recipient households in the state. The error for each household is the absolute value of the difference between the monthly benefit it receives and the monthly benefit for which it is truly eligible.
This formula is prescribed by statute, and it is calculated by the U.S. Department of Agriculture (USDA) based on an audit of a sample of SNAP recipient households, as laid out in the accompanying regulation and guidance.
I showed in a recent working paper that the payment error rate formula suffers from a conceptual flaw, which penalizes overpayments less than underpayments. While underpayments and overpayments of the same magnitude have an identical effect on the numerator of the formula, overpayments increase the denominator more than underpayments and thus lead to a lower payment error rate. As a simple example, suppose there is a state with only one household receiving SNAP, and that household is truly eligible for a benefit amount of $300 per month. If the household receives $400 (an overpayment of $100), the payment error rate would be \(\frac{\$100}{\$400} = 0.25\). But if the household receives $200 (an underpayment of $100), the payment error rate would be \(\frac{\$100}{\$200} = 0.5\).
The simple fix to this conceptual flaw in the payment error rate formula is to divide the sum of all benefit payment errors by the total benefit amounts for which households were truly eligible B*.
In the previous example, the household truly eligible for a benefit amount of $300 would yield a corrected payment error rate of \(\frac{\$100}{\$300} = 0.33\) regardless of whether it received $200 or $400, thus treating the $100 underpayment and overpayment symmetrically.
Because the accurate benefit amounts are already determined via an audit, this correction is unlikely to be especially burdensome to implement.
Cost share implications of correcting the payment error rate formula
If Congress chose to fix this conceptual flaw, there would be important implications for states’ cost shares of SNAP benefit costs. In FY2025, the latest data available, overpayments exceeded underpayments in every state, and so correcting the formula would tend to increase states’ payment error rates. These changes could affect the share of benefit costs for which states are liable. For purposes of this illustrative exercise, I assume that Congress applies the revised formula to FY2025 payment error rates while leaving the cost-share thresholds and transition rules for implementing cost-sharing unchanged.
I approximate how the payment error rate would change in each state in FY2025 due to correcting the formula. To do so, note that the corrected payment error rate can be calculated by dividing the uncorrected payment error rate by one minus the overpayment rate \(\frac{E_\textit{over}}{B}\) (the sum of overpayments divided by total benefits paid) plus the underpayment rate \(\frac{E_\textit{under}}{B}\) (the sum of underpayments divided by total benefits paid), as demonstrated below.
USDA publishes payment error rates, \(\frac{E}{B}\), along with the overpayment rates, \(\frac{E_\textit{over}}{B}\), and underpayment rates, \(\frac{E_\textit{under}}{B}\), for all states. Because the published rates exclude discrepancies of $57 or less, these estimates do not fully adjust the denominator to reflect households’ correct benefit amounts. If excluded overpayments exceed excluded underpayments, the estimates understate the true corrected rate. If excluded underpayments exceed excluded overpayments, they overstate it.
Table 1 reports for each state the underpayment rate, overpayment rate, and the uncorrected and corrected payment error rates for FY2025. Because the overpayment rate is higher than the underpayment rate in each state, the corrected payment error rate is always higher than the uncorrected one. Across all states, the increase in the payment error rate ranges from 0.04 to 5.09 percentage points, and the average increase is 0.92 percentage points. The largest increases are in Alaska (5.09 percentage points), New Mexico (2.90 percentage points), the District of Columbia (2.54 percentage points), Delaware (2.41 percentage points), and Oregon (2.23 percentage points).
Table 1 also reports how correcting the payment error rate formula affects cost shares owed by states in FY2028. I report both the cost share percentage and the corresponding dollar amount on the basis of states’ FY2025 benefit costs. Cost shares depend directly on states’ payment error rates. Ordinarily, a payment error rate below 6% yields no cost share, a rate between 6% and 8% yields a 5% cost share, a rate between 8% and 10% yields a 10% cost share, and a rate of 10% or above yields a 15% cost share.
However, there are two special provisions for determining state cost shares in FY2028. First, states with payment error rates above 13.33% in FY2025 delay paying any cost share until at least FY2029. Thus, if correcting the payment error rate formula increases a state’s rate above 13.33%, then its cost share could fall from 15% to 0% as a result. The second special provision is that states may use either their FY2025 or FY2026 payment error rate to determine their FY2028 cost share, whichever produces a more favorable cost share. Thus, the FY2028 cost shares reported below could ultimately end up being lower for states that obtain a more favorable cost share based on their FY2026 payment error rates. These two special provisions do not apply for calculating cost shares for FY2030 or later.
As reported in Table 1, the FY2028 cost share would change for 16 states if the payment error rate formula were corrected. Among those, 9 states would owe a greater share, and in each of these states the share rises by 5 percentage points. The states facing the greatest dollar increase in state obligations are Michigan ($156 million increase), Alabama ($86 million increase), and Tennessee ($85 million increase). Together, the 9 states would owe an additional $548 million above the amount they would owe using the current payment error rate formula.
Meanwhile, 7 states would owe a smaller share. These 7 states’ cost shares fall from 15% to 0%, because correcting the formula would increase their payment error rate above 13.33%. The largest savings would accrue to New York ($1.15 billion), Florida ($1.03 billion), and Massachusetts ($392 million). Across all 7 states that would owe a smaller share, their savings total $3.24 billion.
Across all 16 states whose cost share would change, their net obligations for SNAP benefits would fall by $2.70 billion relative to what they would pay under the current payment error rate formula. Conversely, the federal obligation would rise by $2.70 billion. Correcting the payment error rate formula would increase federal spending for FY2028 only because of the temporary provision exempting states with especially high payment error rates from owing any cost share. Starting in FY2030, higher payment error rates can only increase state cost shares. Thus, the $548 million higher obligation among states that experience an increased cost share in FY2028, and the corresponding $548 million in additional federal savings, illustrates the potential federal savings once the transition provisions no longer apply, starting with FY2027 payment error rates that determine FY2030 cost shares.
As policymakers consider ways to improve the accuracy of SNAP, they should turn their attention to fixing the conceptual flaw in the payment error rate formula, which penalizes underpayments more than overpayments. Correcting the flaw would treat underpayments and overpayments symmetrically. It would also have implications for federal spending. Applying the correction to FY2025 rates would increase federal spending under this illustrative exercise, and applying it to FY2026 rates would likely have a similar effect. Beginning with FY2027 rates, however, the correction could only increase or leave unchanged state cost shares, assuming overpayments continue to exceed underpayments.
Source: USDA SNAP Payment Error Rates, Fiscal Year 2025; USDA SNAP State Level Benefit Data; Kevin Corinth, 2026, “Paying for Payment Errors: State Incentives and Caseload Responses to SNAP Cost Sharing,” AEI Center on Opportunity and Social Mobility Working Paper. Available at https://www.aei.org/research-products/working-paper/paying-for-payment-errors-state-incentives-and-caseload-responses-to-snap-cost-sharing/; Author’s calculations
Notes: Corrected payment error rate obtained by dividing uncorrected payment error rate by 1 minus the overpayment rate plus the underpayment rate. Correction is approximate because published overpayment and underpayment rates do not account for payment errors of $57 or less. FY2028 cost shares could ultimately be lower if FY2026 payment error rates correspond to lower cost shares. Change in FY2028 cost share amount calculated on the basis of FY2025 benefits.
