▣ DATA BRIEF · US FOOD STAMPS
☀ Key Stats
◆ Restrictions on buying sugary drinks with US food stamp benefits reduced participating households’ retail purchases of excluded drinks by an estimated 12.4% in the first 10 states studied.
◆ The researchers translate the decline into about 404 fewer fluid ounces per covered person per year, equivalent to roughly 34 12-ounce cans.
◆ In states with narrower restrictions, households shifted toward sugary drinks and fruit juices that remained eligible, offsetting about 39% of the excluded-drink calorie reduction.
◆ When fruit drinks remained eligible, purchases increased by an estimated 5.9%, replacing about 36.3% of the calories cut from excluded drinks in that comparison.
◆ The primary analysis followed 15,261 households receiving Supplemental Nutrition Assistance Program benefits from January 2025 through June 2026.
◆ Survey evidence indicates the restrictions increased three measures of perceived stigma by about 0.10 to 0.17 standard deviation.
◆ A separate model estimates that excluding all sugary drinks nationwide could generate about $1.1 billion a year in benefits, with roughly 70% coming from projected reductions in public health-care spending.
◆ That health model projects about 33,000 fewer new cases of type 2 diabetes over 10 years, but this is not an observed health outcome from the study.
Continue reading ↓
Restrictions on buying sugary drinks through the Supplemental Nutrition Assistance Program, commonly known as food stamps, reduced retail purchases of excluded drinks by an estimated 12.4%, according to a new NBER working paper.
The researchers estimate that the decline corresponds to about 404 fewer fluid ounces per covered person per year, or roughly 34 12-ounce cans.
The study examined food stamp restrictions implemented in 10 states during the first half of 2026 using nationwide grocery-purchase data from Numerator.
The authors are Hunt Allcott of Stanford University, Amy Finkelstein of MIT, Anna Grummon of Stanford and Matthew J. Notowidigdo of the University of Chicago.
◎ HEADLINE NUMBER
12.4% lower
Estimated reduction in retail purchases of drinks excluded from food stamp eligibility.
Households did not simply replace food stamp dollars with cash
One reason the policy’s effect was uncertain is that food stamp benefits usually pay for only part of a household’s grocery spending.
Households could therefore have continued buying the same drinks by paying with their own money instead.
The researchers instead estimate that the restrictions reduced retail purchases of the drinks covered by each state’s rules by 12.4%.
The decline appeared soon after implementation and remained similar in size over the study’s first six months.
−12.4%
Excluded-drink purchases
−404 fl oz
Estimated annual change per covered person
The authors say the pattern is consistent with what economists call mental accounting: households may treat food stamp benefits as a separate grocery budget rather than as perfectly interchangeable with cash.
What “food stamps” means: The official program is the Supplemental Nutrition Assistance Program, or SNAP. “Food stamps” is the widely understood informal name used here for accessibility.
↳ POLICY DESIGN · SUBSTITUTION
Narrow restrictions shifted some calories to other drinks
States did not all restrict the same products, allowing the researchers to examine what happened when some sweetened drinks remained eligible.
In those states, households increased purchases of some drinks that could still be bought with benefits.
When fruit drinks remained eligible, purchases increased by an estimated 5.9%.
The researchers calculate that the increase replaced about 36.3% of the calorie reduction from excluded drinks in that comparison.
↔ NARROWER RESTRICTIONS
~39% offset
About 39% of the calorie reduction from excluded drinks was replaced by calories from other sugary drinks and fruit juices that remained eligible.
This does not mean households cut their total calorie consumption by 39%.
It means substitution toward other eligible beverages offset part of the calorie reduction generated by the restricted drinks.
The researchers found no substantial evidence of substitution toward sugary foods or soda purchased at the limited-service restaurants captured in their data.
The restaurant data do not cover all drinks consumed away from home, however.
The analysis followed more than 15,000 food stamp households
The primary analysis followed 15,261 households identified as receiving SNAP benefits.
The purchase panel covered January 2025 through June 2026, giving researchers data from before and after the restrictions took effect.
The study compared changes in 10 states that implemented restrictions during the first half of 2026 with changes in states where restrictions were not yet operating.
The researchers used a staggered difference-in-differences design intended to isolate changes associated with the restrictions from broader changes occurring over the same period.
They also registered their primary research design and planned analyses before receiving the post-policy data.
↳ SURVEY FINDING · STIGMA
Recipients reported feeling more judged
The researchers also surveyed participants before and after the restrictions were introduced.
The policies increased SNAP recipients’ reports of feeling judged while paying for groceries by an estimated 0.10 standard deviation.
They also increased agreement that state SNAP policies were disrespectful by 0.12 standard deviation and that the policies took away personal freedom by 0.17 standard deviation.
The researchers found no significant effect on recipients’ perceptions of the health risks of soft drinks.
Key tradeoff: The paper’s later economic model does not put a dollar value on the increased stigma detected in the survey. The authors identify that omission as an important limitation.
Restrictions spread to more states in 2026
By July 2026, the paper reports that 23 states had received federal waivers allowing them to exclude at least some products from SNAP eligibility.
The rules varied by state, so a restriction in one state did not necessarily cover the same drinks or foods as a restriction elsewhere.
Readers can check the USDA’s current SNAP food-restriction waiver tracker for the latest status of state approvals.
↳ MODELED SCENARIO · NOT AN OBSERVED SAVING
Nationwide model estimates $1.1 billion in annual benefits
The paper separately models what could happen if all sugary drinks were excluded from SNAP nationwide.
Under the authors’ assumptions, the policy would generate about $1.1 billion per year in total benefits.
~$800M
Modeled annual reduction in public health-care costs after 10 years
$336M
Modeled annual consumer-welfare gain
The projected health-care savings account for about 70% of the modeled overall benefit.
The health model estimates that sustained nationwide restrictions would reduce the average adult SNAP recipient’s weight by about 0.27 pounds.
It also projects approximately 33,000 fewer new cases of type 2 diabetes over 10 years.
Neither figure represents a health outcome directly observed during the first months of the state restrictions.
The projections combine the study’s purchasing estimate with results and parameters from outside research on calorie intake, weight, diabetes and health-care spending.
✦ Why it matters ✦
The findings address a basic question about food stamp restrictions: does changing what benefits can buy actually change what households purchase?
The estimated 12.4% decline suggests households did not fully compensate by simply paying for the restricted drinks with other money.
But the study also shows that the design of the restriction matters.
When other sugary drinks remained eligible, households shifted some purchases toward them, offsetting about 39% of the beverage-calorie reduction in the paper’s narrower-restriction comparison.
The survey findings introduce a separate cost that cannot be seen in supermarket receipts: recipients reported greater feelings of judgment, disrespect and lost personal freedom.
The study therefore provides evidence of both a measurable change in purchases and possible social costs associated with the policy.
Its strongest evidence concerns what households bought during the first months of implementation.
The longer-term health, health-care and economic effects remain projections that depend on additional assumptions.
ⓘ How to read the findings
This is an NBER working paper circulated for discussion and comment. It has not been peer reviewed or reviewed by the NBER Board of Directors in the manner that accompanies official NBER publications.
The headline 12.4% estimate refers to retail purchases of the particular drinks excluded by each state’s rules. It does not mean every SNAP recipient reduced all sugary drink consumption by exactly 12.4%.
The primary analysis contains 15,261 SNAP households from an opt-in Numerator purchase panel. The researchers use weighting intended to make the sample more closely resemble the US SNAP population on selected characteristics.
The study uses purchases as a proxy for consumption. Food or beverages purchased by a household are not necessarily consumed in full or divided equally among its members.
Numerator receipts do not include package sizes or nutrition facts, so the researchers estimated drink volumes using additional data and a statistical model based on factors including price, category and brand.
The researchers use a staggered difference-in-differences design. Treating the 12.4% decline as a causal policy effect depends on assumptions including that treatment and comparison states would otherwise have followed sufficiently similar purchasing trends.
The authors registered their primary research design before receiving post-policy data, reducing the scope for choosing the main statistical approach after seeing the results.
The approximately 39% offset applies to narrower restrictions where other sugary drinks and fruit juices remained eligible. It is not a 39% decline in total calories and should not be interpreted as a universal substitution rate for every state.
The $1.1 billion annual benefit, approximately $800 million in health-care savings, 0.27-pound weight change and 33,000 diabetes cases are model-based projections rather than outcomes observed during the study.
Those health projections combine the purchase results with parameters from outside studies. The authors describe these predictions as more speculative than their empirical purchasing estimates.
The modeled welfare estimate does not assign a monetary cost to the increased stigma detected in the survey. The authors identify that as a potentially important limitation.
The study was funded through the Food Policy Research Program at the University of Illinois Chicago, which is supported by Bloomberg Philanthropies’ Food Policy Program. The paper states that its findings do not necessarily reflect the funder’s views or policies.
