When US states barred people from using food-assistance benefits to buy soda, soda purchases fell — but much of the money simply moved to other sugary drinks.

Researchers at the National Bureau of Economic Research, led by the University of Chicago’s Matt Notowidigdo, analysed grocery records from about 15,000 households using SNAP. Among 3,291 households in 10 states that added new restrictions, soda purchases dropped 12% — roughly 34 fewer 12-ounce cans per person per year.

The policy context

Twenty-three states have received USDA waivers to restrict soda, candy and other junk food from SNAP purchases — a substantial shift in a programme that has historically placed few restrictions on what benefits buy.

The argument for restriction is straightforward: public money should not subsidise products contributing to diet-related disease, particularly in a population with elevated rates of obesity and type 2 diabetes.

The argument against is also straightforward: restrictions single out low-income people for dietary supervision that wealthier shoppers do not face, and add administrative complexity at the checkout for benefits that are already inadequate.

What the data show

A 12% reduction is real. Roughly 34 fewer cans per person per year is a measurable behaviour change, and it demonstrates that purchase restrictions do influence what people buy.

The number is also more modest than the policy rhetoric implies. It is not that people stopped buying soda — they bought somewhat less of it with benefits.

The substitution problem

SNAP recipients redirected up to 39% of the money they would have spent on soda to other sugary drinks and fruit juices not covered by the restrictions.

That figure is the study’s central finding, and it substantially undercuts the intervention. If roughly two-fifths of avoided soda spending goes to other sweetened beverages, the reduction in actual sugar intake is far smaller than the reduction in soda purchases.

Fruit juice is the awkward case. It carries a health halo and is frequently exempt from restrictions, while containing sugar in quantities comparable to soda. A policy pushing people from soda to juice may achieve little nutritionally while appearing successful on the metric being measured.

“If the goal is to reduce sugar consumption, you want the ban to be more comprehensive,” Notowidigdo said.

Why comprehensive is harder than it sounds

The obvious response — restrict all sugary drinks — runs into definitional problems that have defeated similar efforts elsewhere.

Where does the line fall? Sweetened coffee drinks, flavoured milk, sports drinks, sweetened teas, juice blends, 100% juice, smoothies? Each has advocates for exemption, and each exemption creates the substitution route the next study will document.

Retailers must also implement whatever definition emerges at the point of sale, across thousands of products with changing formulations — which is where administrative burden becomes real rather than theoretical.

The stigma finding

The study also found increased stigma reported by some SNAP recipients.

This is easy to dismiss as soft and worth taking seriously. Being told at a checkout, in front of other shoppers, that an item cannot be purchased with your benefits is a visible marker of receiving assistance.

Stigma has documented effects on programme participation, and SNAP already has substantial non-take-up among eligible people. A restriction reducing sugar purchases modestly while deterring some eligible households from enrolling could produce a net harm on the nutrition outcomes it targets — a trade the sugar figures alone do not capture.

The caveats

This analysis is a preprint that has not yet been peer-reviewed, which matters for a finding with immediate policy relevance.

The sample may not represent all SNAP users. Grocery records capture households that participate in tracking panels, who differ systematically from those who do not, and the analysis covers purchases at participating retailers rather than all food acquisition.

Purchases are also not consumption. Records show what was bought with benefits, not what was drunk, by whom, or what was bought with cash instead — a substitution the data cannot see at all.

The useful conclusion

What the alternative approaches look like

Purchase restriction is one of several levers, and the comparison is instructive given the substitution problem.

Incentive approaches work in the opposite direction — increasing the value of benefits spent on fruit and vegetables rather than restricting what can be bought. Pilot programmes of this kind have shown increased produce purchasing, and they avoid the stigma problem entirely because nothing is refused at the checkout.

Taxation operates on everyone rather than on benefit recipients specifically, which removes the fairness objection, and evidence from jurisdictions that have implemented sugary drink taxes generally shows reduced purchasing — though substitution appears there too.

The honest summary across all three is that dietary behaviour responds to price and availability in modest, partially offsetting ways. No single mechanism produces a large change, and the policies most likely to work combine approaches while defining categories broadly enough that the obvious substitutes are covered — which is exactly what this study concludes.

Narrow bans move consumption more than they reduce it. That is a finding about policy design rather than an argument against restriction in principle — and it suggests the current wave of state waivers, most targeting soda specifically, may deliver considerably less than expected. This summarises an early-stage study, not settled policy evidence.