A working paper from economists at Stanford University, the Massachusetts Institute of Technology and the University of Chicago finds that state bans on buying sugary drinks with federal food assistance cut those purchases by 12.4 percent in the first half of 2026. Grocery Dive and Food Dive picked up the finding on September 2, moving a public health result into the trade press that grocers read.
The number matters to retailers because it is the first large sample measurement of what happens to a category when it loses eligibility under the Supplemental Nutrition Assistance Program. SNAP accounts for roughly 12 percent of overall US grocery spending, and SNAP households spend about 19 percent more per month on groceries than non-SNAP shoppers, according to figures cited in the trade coverage. A double digit drop in one aisle, concentrated in a customer group that over-indexes on grocery basket size, is a merchandising event as much as a policy one.
In short
- 12.4 percent decline: SNAP recipients in states with sugary drink restrictions cut those purchases by 12.4 percent between January and June 2026, per the National Bureau of Economic Research working paper.
- 34 fewer cans: the effect is equivalent to about 34 fewer 12-ounce cans of soda and other sugary drinks per person per year.
- No cash substitution: researchers found no evidence that recipients simply bought the same soda with their own money instead.
- But category leakage is real: where fruit juices and energy drinks stayed eligible, up to 39 percent of the redirected spending moved into those adjacent sugary products.
- 23 states, 10 live, 5 blocked: 23 states hold approved USDA waivers, the study covers the first 10 to implement, and five saw restrictions suspended by federal court order in June 2026.
What exactly did the study measure?
The paper was circulated through the National Bureau of Economic Research and has not been peer reviewed. STAT reported that the work was supported by a grant from Bloomberg Philanthropies. Those two facts sit at the front of any commercial read of the result, because a working paper is a first draft of evidence rather than a settled finding.
The headline estimate is a 12.4 percent reduction in sugary drink purchases among SNAP recipients living in states that removed those products from the eligible list. CBS News reported the window as January to June 2026, and the authors describe the effect as sustained across the full six months rather than a short adjustment that faded. That persistence is the part retailers should note, because a one-month shock and a durable shift call for different assortment responses.
The sample and the data behind it
According to STAT, the researchers worked with grocery purchase records covering roughly 15,000 households receiving SNAP benefits. Of those, 3,291 households were located in the 10 states that had implemented new restrictions. The remainder function as the comparison group, which is what allows the authors to separate a restriction effect from a general decline in soda consumption.
That design matters. US soda volumes have been drifting down for years across all income groups, so a simple before-and-after comparison in restricted states would overstate the policy effect. Using unrestricted states as a control is the standard way to strip that trend out. It also means the 12.4 percent figure should be read as a difference against a falling baseline, not against a flat one.
Why the size of the drop surprised the researchers
Hunt Allcott, the Stanford economist who co-authored the paper, told CBS News that “a 12.4 percent decrease is larger than most researchers had predicted, and indeed larger than I had predicted.” Matt Notowidigdo of the University of Chicago Booth School of Business is also named among the authors.
The prior expectation among economists was closer to zero. The standard argument is that money is fungible: if a household spends part of its own income on groceries anyway, removing soda from the benefit list should just shuffle which dollars pay for the soda. The finding that this did not happen at scale is the intellectually interesting result, and it is also the one that carries the most weight for category managers.
How did SNAP soda restrictions reach 23 states?
SNAP eligibility rules are federal, and states cannot narrow them unilaterally. The route open to states is a waiver from the US Department of Agriculture that redefines which foods a household may buy with benefits. Reporting from CBS News dates the start of the current wave to May 2025, when USDA began approving those requests.
The political framing came from the Make America Healthy Again agenda. CBS News reported that Robert F. Kennedy Jr., the Health and Human Services Secretary, announced in March that the federal government would move to end what he described as taxpayer-subsidized soda assistance. The waiver pipeline that followed is what produced the current patchwork.
From waiver approval to shelf-level enforcement
An approved waiver is not the same as a live restriction. States have to set an effective date, publish the excluded product list, and get that list into the eligibility file that point-of-sale systems check at the register. The gap between approval and enforcement explains why 23 states hold waivers while a smaller number were actually restricting purchases during the study window.
This is the same operational sequence grocers worked through when benefit rolls themselves moved. Our earlier reporting on how SNAP rolls fell 13 percent ahead of a November 4 stocking deadline covered the demand-side version of the problem, where the question was how many benefit dollars would arrive rather than what they could buy.
Where the courts intervened
Both STAT and CBS News reported that five states had restrictions suspended by federal court order in June 2026, following lawsuits brought by SNAP recipients. Grocery Dive described the outcome as court rulings overturning restrictions in five states.
The practical consequence for retailers operating across state lines is that the eligibility map is not stable. A chain with stores in a state that implemented, then had the rule enjoined, has to reverse a POS configuration and, in some cases, a planogram decision that was made months earlier. Compliance cost here is not the fine risk, it is the churn.
Where did the redirected spending actually go?
This is the finding with the clearest merchandising implication, and it cuts against the headline. Grocery Dive reported that when only soda was banned, researchers found no evidence of substitution into a household’s own cash. But in states where other sweetened beverages remained eligible, up to 39 percent of the redirected consumption moved into those products, specifically fruit juices and energy drinks.
Read those two results together and the picture is a partial leak rather than a clean reduction. A narrow rule that removes carbonated soft drinks while leaving juice and energy drinks on the list recovers a meaningful share of the lost volume inside the same aisle. A broad rule that covers all sugary drinks does not offer that escape valve, which is why the authors model the broader ban separately when they estimate benefits.
For a category manager, the operational question is therefore not “how much soda volume do I lose” but “how wide is the restriction line in each state I operate in.” Those are different numbers, and only the second one tells you whether the shelf space you free up gets refilled by an adjacent SNAP-eligible product or by a general merchandise item.
What does a 12.4 percent drop mean in category terms?
The per-person figure is the most usable one for planning. STAT and CBS News both report the effect as roughly 34 fewer 12-ounce cans of sugary drinks per person per year. That is a little under three cans a month, which is a modest individual change that becomes a visible number when multiplied across a benefit population.
The magnitude of the exposure depends on how concentrated SNAP redemption is in a given banner. Discount grocers, dollar chains and independent stores in low-income trade areas carry far more SNAP share than a premium format, so the same national percentage lands very differently store to store. That distribution question is the one that separates a rounding error from a category reset, and it is worth checking against the demographics behind how Kroger and Walmart split the US grocery market.
| Metric | Reported finding | Basis |
|---|---|---|
| Change in sugary drink purchases | Down 12.4 percent | SNAP recipients in 10 restricted states, January to June 2026 |
| Volume equivalent | About 34 fewer 12-ounce cans per person per year | Derived from the purchase-record estimate |
| Substitution using own cash | No evidence found | Soda-only restrictions |
| Substitution into adjacent drinks | Up to 39 percent of redirected spend | States where juice and energy drinks stayed eligible |
| Ten-year type 2 diabetes risk | Down 2.6 percent, about 34,000 fewer new US cases | Modelled from the purchase change |
| Estimated annual benefit of a full sugary drink ban | About USD 1.1 billion, roughly 70 percent from healthcare costs | Authors’ benefit model |
How does the restriction map break down right now?
The three-way split between approved, live and enjoined is the fastest way to size the operational problem. It is also the part most likely to change before the end of 2026, because waivers that were approved during 2025 and early 2026 continue to reach their effective dates.
| Status | States | What it means operationally |
|---|---|---|
| USDA waiver approved | At least 23 | Rule is coming; excluded product list needs mapping to internal item files |
| Restrictions implemented and measured | 10 | Live at the register; the basis for the 12.4 percent estimate |
| Restrictions suspended by federal court | 5 | Configuration must be reversed; litigation outcome still open |
| Broader unhealthy-food bans in force | Reported as 8 by CBS News | Extends beyond drinks into categories such as candy |
| Candy restrictions | About a dozen states | Estimated USD 300 million in industry sales impact |
The counts do not reconcile perfectly across outlets, which is normal when a policy map is moving weekly and different publishers snapshot it on different days. CBS News reported eight states with unhealthy-food bans and 10 actively implementing restrictions; STAT and Grocery Dive both anchor on 10 restricted states in the study sample. Treat the 23 waiver figure as the ceiling and the 10 as the measured base.
What have beverage and confectionery suppliers said?
Supplier commentary so far has been measured rather than alarmed. Grocery Dive reported that Hershey saw some impact in early-adopter states but that sales remained largely within the company’s expectations. The same coverage put an estimated USD 300 million industry sales loss against candy restrictions spanning roughly a dozen states.
That framing is consistent with how large CPG companies have handled category headwinds through 2026. A single-digit share of national volume moving in a subset of states is absorbable inside guidance, particularly for a portfolio that spans multiple categories and price tiers. It becomes a guidance event only if the waiver count keeps climbing and the restriction lines widen from soda to all sweetened drinks.
Beverage majors have been managing a softer US demand picture independently of SNAP policy. When PepsiCo beat on Q2 revenue while US snacking cooled, the pressure discussed on that call was consumer trade-down and category mix, not benefit eligibility. SNAP restrictions add a second, geographically concentrated drag on top of that.
What does the health and fiscal math look like?
The authors do not stop at purchase behavior. STAT reported a modelled 2.6 percent reduction in the risk of developing type 2 diabetes over 10 years, which the paper scales to roughly 34,000 fewer new US cases. Grocery Dive reported that extending the ban to all sugary drinks would produce about USD 1.1 billion in annual benefits, with around 70 percent of that coming from avoided healthcare costs.
These are model outputs rather than observed outcomes, and they should be labelled as such. The chain of inference runs from measured purchases to assumed consumption to projected disease incidence to costed healthcare savings, and each link carries its own error. The purchase measurement is the strongest part of the paper; the dollar benefit is the weakest.
The stigma finding that complicates the result
STAT reported survey evidence within the same work showing that recipients described feeling judged or disrespected under the restrictions. CBS News framed the study as finding that restrictions cut purchases but raised stigma concerns.
For grocers this is not an abstraction. Stigma at the point of sale shows up as declined transactions, longer checkout interactions, and in some cases a shift toward formats where the customer feels less observed, such as self-checkout or online pickup. A restriction that changes how a benefit customer experiences the register can move where they shop as well as what they buy.
What do grocers actually have to change?
The compliance burden sits in three places: the item file, the point-of-sale eligibility check, and the promotional calendar. None of them are difficult in isolation, and all three become expensive when the rule differs by state and changes mid-quarter.
Item files and point-of-sale configuration
Every excluded product has to be flagged against the state where the store operates, not against the chain as a whole. For a multi-state operator that means maintaining parallel eligibility maps and reconciling them every time a waiver reaches its effective date or a court suspends one. The failure mode is a legitimate purchase declined at the register, which is a customer service problem before it is a compliance problem.
Planograms, promotions and the freed shelf space
If a chunk of soda volume leaves a store permanently, the shelf allocation behind it is wrong. The substitution finding argues against a large reset in states with narrow rules, because juice and energy drinks recapture a meaningful share of the spend inside the same doors. In states with broad rules, the case for reallocating facings is stronger.
Promotional planning is the more immediate exposure. A multi-state circular that features a sugary drink promotion cannot run identically across a footprint where the product is ineligible in some states, and the endcap economics behind that promotion change when the funding retailer cannot redeem the volume it forecast.
Reading the effect in reported comps
The next real-world test comes through earnings. Grocers with heavy SNAP exposure report through September and October, and the question is whether any of them isolate a beverage category effect rather than folding it into general basket commentary. Kroger’s Q2 print on September 11 is the nearest large-format read, though its trade area mix means a national restriction effect would likely be small relative to the comp bar it faces.
Why the fungibility result is the most important number
Strip away the health projections and the political framing, and the finding that will be argued about longest is the absence of own-cash substitution. Economists expected it to be large. It was not detectable.
The reasoning behind the expectation is straightforward. A household on SNAP typically spends some of its own income on food as well as its benefit allocation. If soda leaves the eligible list, the household can pay for soda out of the cash side of the budget and shift benefit dollars onto something else it was already buying. Nothing about total consumption needs to change. That is the textbook prediction, and it is why several previous attempts to restrict benefit purchases were dismissed in advance as symbolic.
What the absence of substitution suggests about how benefits are used
The data say otherwise, which implies that SNAP dollars are not treated as fully interchangeable with cash inside the household. Behavioral explanations for this are well established in principle: mental accounting, where a benefit is treated as earmarked, or a simple friction where the household does not restructure its shopping to work around the rule.
The stigma finding points at a third possibility. If the restriction is experienced as a judgment rather than a neutral rule change, a recipient may drop the product rather than buy it in a way that draws attention at the register. That is a different mechanism from mental accounting and it would have different durability, because a norm-driven response can decay as the rule becomes routine.
The paper does not adjudicate between these explanations, and it does not need to for the purchase estimate to stand. But which mechanism dominates determines whether the effect persists past the six months observed, which is precisely what the ten-year health model assumes.
Why the 39 percent leak does not contradict it
The two substitution results describe different margins and are easy to conflate. The first says recipients did not fund the same soda from a different pocket. The second says that when the rule left neighboring sugary products eligible, a large share of the freed benefit spend went into those products instead.
One is substitution across funding sources, which the study did not find. The other is substitution across products inside the benefit, which it did find, at up to 39 percent of the redirected amount. A narrow rule therefore produces a real reduction in soda specifically and a much smaller reduction in sugar overall.
That distinction is what makes rule scope the single most consequential variable for both public health advocates and category managers, and it is why the authors model a full sugary drink ban separately when they put a dollar figure on benefits.
Who is most exposed by format and trade area?
National percentages hide the thing retailers need, which is where the volume actually sits. SNAP redemption is heavily concentrated by store format and by trade area income, so a 12.4 percent category change lands unevenly across a portfolio.
The framework below is a way of sorting that exposure rather than a set of measured results. The study did not publish format-level breakdowns, and the sample of 3,291 restricted-state households would not support them reliably. Treat it as a prioritization aid for internal analysis, not as reported data.
| Format | Typical SNAP redemption exposure | Likely response to a live restriction |
|---|---|---|
| Discount and limited assortment grocers | High | Largest absolute volume effect; strongest case for reviewing facings in broad-rule states |
| Dollar and small-format value chains | High | Beverage is a traffic driver; substitution into eligible drinks matters most here |
| Independent and neighborhood stores | High and concentrated | Least system capacity to manage state-specific eligibility files |
| Conventional national supermarkets | Moderate | Effect diluted across trade areas; visible in mix commentary rather than headline comps |
| Club and warehouse formats | Lower | Membership and pack size limit benefit share; smallest direct exposure |
| Premium and specialty grocers | Low | Compliance obligation without a material sales effect |
The multi-state operators face the harder version of the problem, and not because their exposure is largest. It is because the rule differs by state, changes by court order, and has to be reflected accurately at every register on the day it takes effect. A single-state independent has one rule to follow; a regional chain across six states may have three different rules and one under appeal.
Geography compounds it. The 10 states in the study are not evenly distributed across the country, so a chain concentrated in an early-adopter region carries a much higher effective exposure than its national SNAP share would suggest. That is the calculation worth running before any assortment decision, because the answer determines whether this is a shelf reset or a footnote.
What does the study not settle?
Three limits deserve stating plainly. First, it is a working paper that has not cleared peer review, and NBER circulation is not validation. Second, the six-month window captures the immediate response and cannot speak to whether the effect holds over multiple years, which is exactly what the health projections assume. Third, the sample of 3,291 restricted-state households is adequate for a headline estimate but thin for reliable state-by-state breakdowns.
There is also a selection question the design cannot fully close. The first 10 states to implement were not randomly chosen: they moved first because of their politics and their administrative capacity. Whether the same 12.4 percent shows up in a state that implements reluctantly in 2027 is an open question, not a settled one.
Finally, the paper measures purchases at participating retailers, not consumption. Volume bought outside the recorded channel, including at stores outside the panel, would not appear. The authors’ finding of no own-cash substitution is the strongest evidence against that concern, but it does not eliminate it.
What should retailers watch from here?
The waiver count is the leading indicator. If the 23 approved states convert to live restrictions through late 2026 and 2027, the exposed share of national SNAP redemption rises well above the base the study measured, and the supplier commentary that currently reads as manageable will get revisited.
The second thing to watch is the litigation. Five suspensions in June 2026 established that these rules can be challenged successfully, and the appellate outcome will determine whether the remaining waivers land on solid ground. A ruling against the waiver mechanism would unwind configuration work across every affected state at once.
The third is scope creep from drinks into candy and other categories, which is already visible in the roughly dozen states with candy restrictions and the estimated USD 300 million attached to them. That trajectory is what turns a beverage-aisle story into a center-store one.
Retailers are absorbing several state-level rule changes at once this autumn, and the eligibility patchwork is only part of it. The surveillance pricing bans that go live on October 1 land on the same store systems and the same compliance teams, which is why the practical constraint through Q4 is bandwidth rather than any single rule.
Frequently asked questions
What did the SNAP soda study actually find?
It found that SNAP recipients in states that removed sugary drinks from the eligible product list cut their purchases of those drinks by 12.4 percent between January and June 2026, roughly 34 fewer 12-ounce cans per person per year, compared with recipients in unrestricted states.
Who conducted the research and has it been peer reviewed?
Economists at Stanford University, MIT and the University of Chicago, including Hunt Allcott and Matt Notowidigdo, circulated it as a National Bureau of Economic Research working paper. STAT reported it was supported by a Bloomberg Philanthropies grant. It has not been peer reviewed.
How many states restrict sugary drink purchases with SNAP benefits?
At least 23 states hold approved USDA waivers. The study covers the first 10 to implement restrictions, and five states had restrictions suspended by federal court order in June 2026 following lawsuits from SNAP recipients.
Did recipients just buy soda with their own money instead?
No. Researchers found no evidence of that substitution where only soda was restricted, which is the result economists found most surprising because benefit dollars and cash are normally treated as interchangeable.
Where did the redirected spending go?
In states where fruit juices and energy drinks remained SNAP-eligible, up to 39 percent of the redirected consumption moved into those adjacent sugary products. Broader restrictions that cover all sweetened drinks close that route.
How much of grocery spending does SNAP represent?
Figures cited in the trade coverage put SNAP at roughly 12 percent of overall US grocery spending, with SNAP households spending about 19 percent more per month on groceries than non-SNAP shoppers.
What have suppliers said about the impact?
Hershey reported some impact in early-adopter states with sales largely within expectations, according to Grocery Dive. Candy restrictions across about a dozen states carry an estimated USD 300 million industry sales impact.
What are the projected health effects?
The authors model a 2.6 percent reduction in ten-year type 2 diabetes risk, about 34,000 fewer new US cases, and roughly USD 1.1 billion in annual benefits from a full sugary drink ban, with around 70 percent from avoided healthcare costs. These are modelled projections, not observed outcomes.
What should multi-state grocers do first?
Map excluded product lists to item files by state rather than by chain, build a process for reversing configurations when courts suspend a rule, and check that multi-state promotional plans do not feature products that are ineligible in part of the footprint.