USDA lost in court over SNAP food bans. It's expanding them anyway.

South Carolina's food-stamp agency announced Thursday that it is postponing the state's new ban on using federal benefits to buy soda, candy and energy drinks, pushing the start date from Aug. 31 to Nov. 1 at the U.S. Department of Agriculture's own request. Framed by the state as a routine scheduling adjustment, the delay is better read as a symptom of something larger: a federal agency plowing ahead with a nutrition experiment on the roughly 37 million Americans still enrolled in SNAP even after a federal judge told it, in unambiguous terms, that it lacks the legal authority to do so.

That ruling came on June 22, when the U.S. District Court for the District of Columbia vacated USDA's approval of food-restriction waivers in Colorado, Iowa, Nebraska, Tennessee and West Virginia. The plaintiffs in Aragon v. Rollins — five SNAP recipients represented by the National Center for Law and Economic Justice — argued that the agency had no statutory basis to redefine what counts as "food" under the Supplemental Nutrition Assistance Program, and that it skipped the public notice-and-comment process required for changes with this kind of reach. Judge Amy Berman Jackson agreed on both counts, according to an analysis published by the Food Research and Action Center: USDA had invoked the wrong provision of the Food and Nutrition Act, one meant for administrative pilot projects, to pursue what was in substance a nutrition-and-health intervention that a different provision governs — and governs with a requirement for rigorous evaluation that the agency never conducted.

USDA's response was not to pause and rethink. Secretary Brooke Rollins called Judge Jackson, an Obama appointee, an "activist judge" and vowed the department would "keep fighting to Make America Healthy Again."

"SNAP is for food — not sugar bombs fueling obesity, diabetes, and skyrocketing healthcare costs for low-income families," Rollins wrote, as reported by the Washington Examiner.

In the two months since, the agency has not withdrawn a single one of the other waivers it approved on the same legal theory a court has already found defective. According to USDA's own tracking page, it has issued food-restriction waivers to more than twenty states, and states outside the five named in the lawsuit — including South Carolina, Ohio, Kansas and Missouri — remain on schedule to roll out their own versions in the coming months. The department is, in effect, betting that a ruling against five states does not bind its conduct toward the rest, an argument that may be technically defensible in the narrow sense that only those five approvals were formally vacated, but one that treats a court's reasoning about USDA's basic statutory authority as an inconvenience to route around rather than a signal to heed.

A policy without an evidence base, layered on procedure without a public record

Set aside the legal posture for a moment and consider what the waivers are actually trying to accomplish. The evidence that restricting SNAP purchases improves health outcomes is thin. A working paper by economists at Chicago, MIT and Stanford found that early restrictions reduced sugary-drink purchases by roughly 12 percent, a real behavioral effect, but one of the study's own authors, Stanford's Hunt Allcott, cautioned that partial bans invite substitution — recipients simply buy the products still on the eligible list, including energy drinks not covered by narrower rules — and that "excluding only some sugary drinks from SNAP is probably not a good idea" if the goal is actually improving diets rather than simply moving dollars around a grocery store. No state has been running restrictions long enough to show a change in obesity, diabetes or health spending, the outcomes Rollins invokes to justify the policy.

What the research does show, more consistently, is a stigma cost. A University of Michigan researcher, Kate Bauer, found in a July survey that the restrictions increased recipients' sense of shame while shopping, and warned that "stigma hurts health, so that may offset some of the calculated benefits," per the same CBS reporting. Anti-hunger advocates go further: Gina Plata-Nino of the Food Research and Action Center has argued that added friction and embarrassment at checkout discourage eligible families from applying for or using benefits at all, a dynamic that would worsen the food insecurity SNAP exists to prevent even if it never moves a single measure of diet quality.

This is the second problem with USDA's approach, and it is separate from the legal one. The Food and Nutrition Act provision that Judge Jackson found actually governs nutrition pilots — Section 2026(k), by the FRAC account of the ruling — requires the kind of rigorous outcome evaluation that would let policymakers know, before scaling a restriction to millions of households, whether it works and at what cost. USDA's waivers skipped that step. States are being permitted to bar candy and soda first and ask whether it helped anyone second, if at all, which is precisely backward for a program that low-income families rely on to eat.

The counterargument, and where it runs out

The administration's case is not frivolous. SNAP is a $100-billion-a-year program, and it is fair to ask whether taxpayer-subsidized benefits should function as an unrestricted grocery allowance, including for products with no nutritional value. States have asked USDA for permission to run these pilots going back decades — Minnesota and others were denied waivers under prior administrations, often on the grounds that retailers would find the rules unworkable to enforce at the register. There is a legitimate, bipartisan-adjacent policy conversation to be had about whether SNAP's design should more closely track its nutrition mission, and several states arrived at these waivers through their own elected governors, not federal imposition. South Carolina Gov. Henry McMaster, who requested his state's waiver against a backdrop of roughly 36 percent adult obesity, framed it as a way to "ensure federal taxpayer dollars are used to their maximum benefit" — a genuine public-health rationale, not a pretext.

But intent does not substitute for authority, and urgency does not substitute for evidence. If the policy is as sound as USDA insists, the agency had a lawful path available to it: propose the change through the pilot-project provision Congress actually wrote for this purpose, publish it for public comment as the Administrative Procedure Act requires, and build the evaluation data that would let the next round of waivers, or a permanent statutory change, rest on something sturdier than a press release. Instead the agency chose the faster route, got caught, and has continued down parallel tracks with everyone but the five states a court explicitly stopped.

What the delay actually signals

South Carolina's two-month postponement is not evidence of caution. Read alongside the June ruling, it reads as evidence that USDA knows its legal footing is unstable enough to warrant reshuffling implementation dates state by state while the broader theory remains unresolved on appeal, rather than pausing the program nationwide until a court — or Congress — settles the question the agency would rather avoid. Retailers in twenty-some states are being asked to build compliance systems, and SNAP recipients in those states are adjusting shopping habits, atop a legal foundation a federal judge has already called unauthorized in five states with functionally identical waivers.

The honest options were to litigate the authority question to a final resolution before expanding further, or to go to Congress and get explicit statutory permission to redefine what SNAP dollars can buy. USDA has done neither. It has instead kept approving waivers, kept delaying the ones drawing scrutiny, and kept daring the next set of plaintiffs to file the next lawsuit. That is not evidence-based policymaking. It is litigation as a compliance strategy, with low-income families and the retailers who serve them absorbing the uncertainty in the meantime.