SNAP Food Restrictions May Help or Hurt—But We Will Probably Never Know

The Supplemental Nutrition Assistance Program (SNAP) helps over 37 million Americans meet their food needs. Beginning in 2025, the US Department of Agriculture (USDA) granted states waivers allowing new restrictions on foods purchasable with SNAP. As of this writing, 23 states had been granted waivers. Implementation of restrictions began in some states in January 2026, although their legality has been hotly contested and stopped in some states. Waivers are approved for two years, after which USDA will decide if food restrictions should be extended, with or without evidence to guide that decision.  

Research shows that SNAP boosts food spending, improves food security, alleviates poverty (particularly for children), is associated with improvements in health outcomes, and reduces health care spending. Nonetheless, its most frequent criticism is the failure to improve participants’ diets. Although SNAP participants’ diets are below average, the difference from the overall population’s diet is small.

States hope the waivers will reduce SNAP participants’ purchases of restricted products and lead to improvements in diet, but there is little evidence to support this optimism. The range of restricted products varies considerably, but the most frequently restricted food category is sugar-sweetened beverages (SSBs). SSBs are significant contributors to the total caloric intake for the average American, while contributing little nutritional benefit. However, SSBs are a limited share of overall household grocery spending. Moreover, SNAP benefits are used for only a portion of SNAP participants’ spending on SSBs, suggesting that participants use other resources to buy them. The few studies testing impacts of restrictions on low-income consumers suggest that any effects will be small. Other restricted foods have even less evidence to support their inclusion.

Evaluations Should Include Assessment of Potential Negative Impacts

At the same time, the food restrictions could have unintended consequences that jeopardize SNAP’s role in reducing household food security and poverty and shaping the larger food retail environment (Figure 1). At the household-level, participants could face new stigmas or decide program participation is no longer worth the administrative burden, which could decrease their overall food spending and increase food insecurity.

Figure 1. Outcomes of Interest to Evaluate SNAP Food Restriction Waivers by Unit of Analysis

At the community level, SNAP-participating retailers will face increased costs to enforce food restrictions and potentially reduced SNAP due to participation declines that have already begun. They could also face expulsion from the program if they are found to accept SNAP benefits for restricted items more than twice, which might occur more than expected given the chaotic rollout of restrictions in some states. Smaller SNAP caseloads may make participating in the program less attractive to some retailers, especially small retailers who may find the also-being-implemented stricter requirements on foods they must stock particularly burdensome. These factors could impact retailers’ decisions to continue as authorized SNAP vendors and potentially their overall profitability. If retailers leave the program or go out of business, store access for participants and others in their community could be impacted.

Given these potential unintended consequences, it is important to rigorously assess whether restrictions yield meaningful benefits to diet and health. Yet evaluation faces numerous challenges. Prior to the waivers, the only limitations on SNAP purchases were that benefits must be used for food intended for at-home (FAH) consumption and could not be used on hot foods or alcohol.  Now there are 23 “waiver” states with different sets of restrictions, presenting a significant barrier to evaluation.

For example, several states restrict “soft drinks” or “soda” but use their own definition of these products. Indiana’s waiver does not restrict sugary canned coffee with milk, but does restrict those that do not contain milk. Louisiana restricts beverages sweetened with sugars other than cane sugar; Iowa restricts diet soda because it contains artificial sweetener; Nebraska restricts “energy drinks” with sweeteners and flavoring that are designed to increase energy and performance, but excludes “sports drinks” marketed to increase hydration; and West Virginia restricts any carbonated non-alcoholic beverage containing a sweetening agent and flavoring, even if that beverage is carbonated water with only minimal sweetener. Other states restrict candy or “junk food” that require definitions that lead to seemingly inconsistent rules, making flourless candy bars restricted, while those made with flour (like a KitKat) are not restricted.

Unfortunately, because state food item restrictions are adopted as a treatment bundle and vary significantly across each waiver state, it will be nearly impossible to distinguish results for any one restricted item from the suite of restricted items. For example, are bans that limit access to canned coffee drinks with milk more effective at reducing overall sugar consumption than those that exclude these products? Without evidence with high external validity, the foods that states choose to restrict will remain idiosyncratic, appearing more political than health-oriented.

Existing Data Sources Are Inadequate for Evaluating the Impacts of the Restrictions

Even though the Federal government spends nearly $100 billion annually on SNAP, there is no federally funded survey or data source that can evaluate this important change to the program. The only two surveys that potentially could be used, the Consumer Expenditures (CE) survey and the National Health and Nutrition Examination Survey (NHANES), are both too small to conduct cross-state analysis and lack the detailed information on which foods are purchased that are needed to assess spending changes due to restrictions (for the CE) or are released in 2-year waves (for the NHANES), thus limiting researchers’ ability to assess key timing issues for restriction rollouts. As a result, the research community will have to rely on private funders and proprietary vendors for data to evaluate the restrictions. 

But these proprietary data sources also have severe limitations. While they have large sample sizes and include a subset of households that provide data over multiple periods (“static” households that are a subset of the overall panels), making them potentially useful for conducting cross-state, difference-in-difference analyses, they may not represent SNAP and lower-income populations because they are not based on random samples of households. Instead, households are recruited with incentives to register to participate, and then selected to meet demographic and economic quotas and based on the quality and consistency of their food spending reports. Furthermore, while these surveys are large, it is unclear that they have adequate sample size to evaluate restrictions in all waiver states. They have poor measures of SNAP participation, often miss key sources of food spending (i.e., from smaller shopping trips and food away from home sources that, according to one estimate, is almost one-quarter of food purchases for SNAP households), and tend to underrepresent food spending relative to benchmarks from federal surveys.   

Some on-going evaluations are taking advantage of existing online general and public sentiment surveys that use address-based sampling (in the highest-quality cases) to recruit a large panel of respondents who fill out basic information about their household and are then invited to respond to surveys on different topics, often on a regular and recurring basis. These surveys have the advantage of already being in the field, with the ability to flexibly add content. The quality of these surveys for studying SNAP and food consumption behaviors is unknown, however. We have concerns about their ability to represent SNAP populations since these surveys are designed to capture broad US population information rather than SNAP participants, and their ability to collect adequate, reliable information on food consumption given their broad topical focus that is not designed to dig deeply into one topic such as food spending or dietary intake, which can be difficult to collect adequately.

Non-Experimental Study Designs Face Significant Challenges

None of the 23 waiver-implementing states have proposed using a randomized controlled trial (RCT) for evaluating new restrictions, although Congress and previous administrations have supported such high-quality studies for SNAP program changes in the past.  Instead, many states propose using a pre-post, time series design, in which participants’ spending and diets before restriction implementation are compared with those after implementation, or by using repeated cross-sections of participants before and after the restrictions. Other studies propose using cross-state, natural experiments, in which outcomes of participants in state(s) with restrictions are compared to those in states without restrictions, or using national-level data from the proprietary sources described above or from other national surveys. In addition to the data limitations discussed above, these studies will face significant challenges in drawing causal inferences.

First, the ability of these methods to estimate unbiased effects depends partially on the size of the effect. If effects of restrictions on food spending or diet are small, as we argue they are, these types of studies may not be able to detect causal effects, even with good data.

Second, there is no clarity on what success looks like. Is it enough to show that SNAP is no longer being used to purchase these food items if they are still being consumed using other resources? Or are household diets expected to change? If so, consumption of restricted items might be expected to fall, but is consumption of other healthy foods expected to increase? If overall food consumption falls as a result of these broad changes to SNAP, is that a problem?

Third, these restrictions come at a time of substantial change to the SNAP program because of the One Big Beautiful Bill Act of 2025 that includes stricter work requirements for some participants, eliminates eligibility for refugee and asylum seekers, tightens rules for deducting some expenses for eligibility determination, and shifts cost sharing for program benefits and administration to states. Some of these changes are already in place, while others do not fully take effect until 2028. As of April 2026, national SNAP caseloads had already dropped 11% since July 2025. Rapidly dropping caseloads mean that evaluations relying on a non-experimental pre-post design may not have comparable samples to compare diet or food spending outcomes before the restrictions are in effect and after because many of those in the “treatment” group (for which restrictions apply) or in the “control” group (for which there are no restrictions in effect) are not the same in the pre-policy period as in the post-policy period. For example, SNAP participation has declined by 50 percent in Arizona since July 2025, which is one of the comparison states in a new study meant to evaluate the restrictions in Texas. 

National Policy Decisions Should Be Evidence-Based

Current state waivers restricting some foods from SNAP purchases will impact millions of low-income households living in the 23 participating states. They would impact millions more if enacted permanently nationwide, with implications for health, food security, and overall wellbeing of millions of Americans receiving SNAP benefits. Such a decision should rest on strong evidence of the benefits of food restrictions versus any potential negative impacts on food security and food access.

Unfortunately, given the serious methodological problems identified, it is unlikely we will have the evidence needed to identify with confidence any positive or negative effects of restrictions sufficient to inform decisions about whether to make the waivers permanent and nationwide. Without better data and rigorous evaluation studies on how food spending, dietary intake, and other important outcomes are impacted, we will not know whether the restrictions improved nutrition nor whether there were any unintended outcomes.

About the Authors