CC BY-NC-ND
0american journal of
Preventive Medicine
|specialarticle|
Improving Nutrition by Limiting Choice in the ®
CrossMaik
Supplemental Nutrition Assistance Program
11 2 Jacob A. Klerman, MA,1 Ann M. Collins, MA,1 Lauren E.W. Olsho, PhD2
In contrast to the Special Supplemental Nutrition Program for Women, Infants, and Children, the Supplemental Nutrition Assistance Program (SNAP) currently allows the purchase of almost any food. This paper reconsiders the role of two forms of limiting choice in SNAP. Using economic theory, descriptive analysis of survey data, and discussion of random assignment evaluation evidence from the Summer Electronic Benefit Transfer for Children Demonstration, the paper argues that because households can substitute cash for SNAP, banning the use of SNAP for less nutritionally desirable foods (e.g., soda, candy) is unlikely to have a large impact. By contrast, because many households currently consume so little of more nutritionally desirable foods (e.g., whole grains, fruits, and vegetables), requiring that some portion of SNAP benefits be spent on those foods is likely to improve dietary intake. Summer Electronic Benefit Transfer for Children Demonstration impact estimates are consistent with this conjecture. Furthermore, these data and evidence from the Healthy Incentives Pilot implementation suggest that such a policy can be feasibly integrated into existing operational processes.
Am J Prev Med 2017;52(2S2):S171-S178. © 2016 American Journal of Preventive Medicine. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license (http://creativecommons.Org/licenses/by-nc-nd/4.0/).
INTRODUCTION
If improving the nutritional intake of Supplemental Nutrition Assistance Program (SNAP) recipients is a policy goal, two direct policy remedies are to: (1) require that at least some portion of SNAP benefits be used for the purchase of healthful foods; and (2) ban the use of SNAP benefits for the purchase of unhealthful foods. This article considers whether these two strategies—requirement and ban—are likely to improve nutritional intake.
Specifically, the next section briefly reviews the policy background. The third section presents the economic theory and some tabulations from survey data that inform the likely applicability of the economic theory. Building on the economic theory, the fourth section presents evidence from the Summer Electronic Benefit Transfer for Children Demonstration (SEBTC) on requirements. The fifth section draws on the Special Supplemental Nutrition Program for Women, Infants, and Children (WIC), SEBTC, and Healthy Incentive Pilot (HIP) experience to consider feasibility; that is, how hard would it be to implement a requirement in SNAP?
The final section summarizes the argument and its implications for policy. It then briefly considers the broader policy questions. Even if a requirement or a ban would improve nutritional intake, are they appropriate policies?
POLICY BACKGROUND
The U.S. Department of Agriculture (USDA) operates multiple food assistance programs, with varying goals, target populations, and program rules. These programs vary in how prescriptive they are about the foods that may be purchased. Traditionally, school meal programs have been prescriptive. Similarly, WIC prescribes specific and healthful foods for young children and their mothers, including juice, non-fat and low-fat milk, whole grains, eggs, fish, cheese, peanut butter, and (through a cash value voucher) fruits and vegetables.
However, unlike programs specifically targeted to children, SNAP benefits may be used to purchase nearly any food from participating retailers. The only
From the LAbt Associates, Social and Economic Policy Division, Cambridge, Massachusetts; and 2Abt Associates, U.S. Health Division, Cambridge, Massachusetts
Address correspondence to: Jacob A. Klerman, MA, Abt Associates, Social and Economic Policy Division, 55 Wheeler Street, Cambridge MA 02138. E-mail: jacob_klerman@abtassoc.com.
This article is part of a supplement issue titled The Supplemental Nutrition Assistance Program's Role in Addressing Nutrition-Related Health Issues.
0749-3797/$36.00
http://dx.doi.org/10.1016Aj.amepre.2016.07.018
© 2016 American Journal of Preventive Medicine. Published by Elsevier Inc. Am J Prev Med 2017;52(2S2):S171-S178 S171
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
exceptions are alcohol, hot foods, and foods that will be eaten in the store.
In principle, SNAP regulations could be changed to limit choice. USDA appears to have the authority at least to allow local pilots of limitations; new legislation might be required to make such a change nationally and permanently. It is useful to consider separately two different forms of limiting choice: requirements and bans.
Requirements
The USDA could require that some fraction of the SNAP benefit be spent on some basket of healthful foods. A proposal of this form—requiring that only some of the SNAP benefit be spent on designated foods—was passed by the Wisconsin Assembly in 2015, but did not become law.1 This paper takes no specific position on what foods might be designated as healthful. WIC's food list provides a natural starting point.
The USDA could ban the use of SNAP benefits to purchase unhealthful foods. The most prominent such proposal was New York City's attempt to ban the use of SNAP to purchase sugar-sweetened beverages.1 In 2011, USDA rejected that request, on implementation and evaluation grounds. Similar proposals have been considered in Minnesota (in 2004, proposed to USDA and denied),2 Mississippi (in 2012, proposed to USDA and then withdrawn), South Carolina (considered, no formal request to USDA),3 Maine (currently in discussion),4 Missouri (currently in discussion), and New York State (currently in discussion).
Again, this paper takes no specific position on what foods should be designated as unhealthful. Recent state proposals provide a natural starting point: soda, other sugar-sweetened beverages, and candy. Other possibilities include foods high in sugar, saturated fat, trans fat, and salt. Many states operationalize related concepts when they define which foods are excluded from the preferential sales tax treatment for most food.5
Anticipating some of the discussion, it is useful to consider USDA's stated reasons for rejecting or discouraging similar proposals6: (1) difficulty of defining unhealthful foods (not USDA's terminology); (2) difficulty of implementation; and (3) lack of evidence of effectiveness.
Tom Vilsack, USDA Secretary, made another argument, noting the "longstanding tradition of supporting and promoting incentive-based solutions that are better-suited for the working families, elderly and other low-income individuals" (as quoted by McGeehan1). Following the Vilsack line of argument, SNAP currently gives households unrestricted resources and then explores
voluntary approaches to improving dietary quality. Such approaches presumably include nutrition education (e.g., the SNAP-Ed program) and rebates for purchase of fruits and vegetables (e.g., HIP, a demonstration that provided rebates for fruits and vegetables purchased through SNAP7).
As long as SNAP benefits are required to be spent on food, why not require that at least some of these benefits be spent on healthful foods? And why not ban the use of SNAP to buy unhealthful foods? This paper returns to these questions in its final section.
ECONOMIC THEORY
The neoclassical economic theory of SNAP assumes that, up to program restrictions, households treat food assistance like cash.8,9 As long as the food assistance is less than what the household would have spent on food given its total resources (cash income plus food assistance), food assistance does not shift consumption choices, relative to what consumption would have been with cash assistance. Put differently, the assistance may have been intended for food and may only be used directly for food. Nevertheless, the household can substitute the food assistance for cash otherwise spent on food, freeing up that cash for non-food uses. Total food expenditures will increase, not by the amount of the additional food assistance, but instead as they would with an increase in unrestricted income (i.e., the Engel Curve10), perhaps 10% of the additional food assistance.9
This neoclassical argument generalizes directly to bans and requirements. Again, the key insight is that households can (often at least partially) offset any bans or requirements with non-SNAP funds.
For Bans
As long as total household expenditure on unhealthful foods was less than total household cash (i.e., non-food assistance restricted) expenditure on food, then a ban will have no effect. On imposition of a ban, the household can switch the healthful foods to SNAP and the unhealthful foods to cash—with no change in total purchase or intake of unhealthful foods.
This argument requires that the household's cash expenditures on foods be greater than the household's total expenditures on unhealthful foods. This is usually likely to be the case. By its very definition (and its name), SNAP is intended to be a supplemental program. Consistent with that intention, the SNAP benefit formula implicitly expects households to spend 30% of income on food. This theoretic argument suggests that bans are unlikely to have much effect, at least as long as the part of the food budget subject to the ban is smaller than household cash expenditure on food.
For Requirements
An analogous argument applies to requirements. If in total, a household was spending more on the healthful goods than was required, then a requirement would have no effect. On the imposition of a requirement, a household could again switch the healthful food purchases to SNAP and the unhealthful food purchases to cash, with no change in total purchase or intake of healthful foods. The crucial caveat is that the household be spending more than the requirement on the healthful foods.
Unlike the case for a ban, this underlying assumption will often not hold. Table 1 presents new tabulations of current intake for: (1) households in the National Health and Nutrition Examination Survey (NHANES) who report receiving SNAP; (2) households in NHANES who are income eligible but do not report receiving SNAP; and (3) SNAP households that participated in HIP. These are tabulations of 24-hour recall for a focal adult, which is the standard approach to measurement, but undoubtedly subject to imperfect recall and social desirability bias. The table presents both HIP and NHANES results because, although NHANES is a nationally representative sample, SNAP participation is known to be under-reported in survey data, including NHANES.11 By contrast, HIP data were collected from a random sample of known SNAP participants, though only in Hampden County, Massachusetts, where the pilot
was implemented. Thus, analyses of the two data sources are complementary.
Consistent with similar analyses,12-15 the striking finding is how far intake is from recommended levels. Results are similar across the three samples. Focusing on the HIP results, intake of total fruits and vegetables, excluding white potatoes, is barely greater than 2 cups per person per day (2.07 cups), as compared with Dietary Guidelines for Americans, 201016 recommendations of 4—5 cups; whole grains intake is 0.66 ounces per person per day, as compared with Dietary Guidelines for Americans, 2010 recommendations of 3— 4 ounces. It is easy to imagine a requirement for purchases larger than current intake levels.
Furthermore, note that these levels are at the mean. For many individuals, levels are much lower, or even zero. For example, about half of SNAP recipients eat no whole grains on any given day (lower panel of Table 1). For those individuals, a small requirement would lead to an increase in purchases—if the individual actually used the benefit to buy the required foods.
Finally, note that requirements and bans can have a direct effect that goes beyond pure budget considerations. Retailers would be expected to display point-of-sale reminders (e.g., "this item qualifies for SNAP-Prime"), as they currently do for WIC. Furthermore, the bans and requirements would themselves provide a form of implicit nutritional information or promotion to the participant
Table 1. Mean Daily Intake of Selected Foods
Food category6 NHANES (2011-2012)a HIP (March-June 2012)b
SNAP participants, M (SD) (n=1,374)c Income-eligible nonparticipants, M (SD) (n=1,152)d SNAP participants, M (SD) (n=2,001)
Mean daily intake (continuous measure)
Excluding white potatoes (cup-eq) 1.87 (1.99) 2.33 (1.95) 2.41 (2.28) 2.07 (2.11)
Dairy products (cup-eq) 1.72 (1.83) 1.72 (1.73) 1.61 (1.68)
Any daily intake (binary measure; yes/no) 16.19 (16.94)
Excluding white potatoes (cup-eq) 0.96 (0.21) 0.98 (0.14) 0.96 (0.20) 0.95 (0.22)
Dairy products (cup-eq) 0.93 (0.27) 0.92 (0.27) 0.93 (0.26)
Added sugars (tsp) 0.99 (0.10) 0.99 (0.10) 0.98 (0.12)
a24-hour recall data from the 2011-2012 National Health and Nutrition Examination Survey (National Center for Health Statistics, Centers for Disease Control and Prevention).
b24-hour recall data from USDA Healthy Incentives Pilot evaluation sampled respondent survey, control group, early-implementation sample, SNAP participants aged 16+ years.
cIncludes respondents aged 16+ years who report current receipt of SNAP benefits.
dIncludes respondents aged 16+ years with income r 130% of the federal poverty line who do not report current receipt of SNAP benefits. eUnits for outcomes defined as described in the Food Pattern Equivalents Database (USDA Agricultural Research Service). HIP, Healthy Incentives Pilot; NHANES, National Health and Nutrition Examination Survey; SNAP, Supplemental Nutrition Assistance Program; USDA, U.S. Department of Agriculture.
(i.e., a reminder about healthful and unhealthful foods). That reminder itself might directly affect intake. In the terms of the behavioral education literature, the implicit promotional message can be interpreted as a nudge.17-20
DEMONSTRATION EVIDENCE
The previous discussion is theoretic. There is also some empirical evidence from SEBTC. During the summer of 2012, at ten grantees, in 14 sites, SEBTC randomly assigned more than 37,000 low-income households (i.e., those with children who, in the prior school year, received National School Lunch Program meals) to $60 of additional food assistance per school-aged child per summer month (when few children were receiving school breakfast or lunch) or to a no-benefit control group. For households receiving SNAP, the SEBTC benefit was about a quarter of the SNAP benefit. Collins et al.21 provide more information on SEBTC.
Of relevance to the issues under consideration in this article, grantees proposed to implement SEBTC through either their state's SNAP or WIC EBT system (hereafter the SEBTC-SNAP or the SEBTC-WIC). The choice of EBT system brought along with it the required foods corresponding to SNAP (nearly any food) and at the wide range of SNAP retailers, or WIC (some foods in the WIC package appropriate for school-aged children) and only at the narrower set of WIC retailers. Thus, comparison across sites is insightful for the likely impacts of WIC-like restrictions in SNAP.
These results should be interpreted with care. SEBTC only collected data on school-aged children (using the NHANES dietary screener battery22,23), so inferences about adults are an extrapolation. Households were randomly assigned to the SEBTC benefit or not. Thus, within-model estimates have strong internal validity within the SEBTC-SNAP sites and within the SEBTC-WIC sites.
This paper attempts to use the SEBTC-SNAP/SEBTC-WIC comparison to make inferences about the likely impact of WIC-like restrictions in SNAP, but it is important to note that the choice of implementing model was not randomly assigned to a site. As a result, SEBTC-SNAP versus SEBTC-WIC comparisons should be viewed as observational; site characteristics of this relatively small number of sites could potentially confound differential impact estimates across SEBTC-WIC and SEBTC-SNAP sites. Systematic biases are also possible. Note, however, that there was random assignment within sites, so systematic bias would require that sites with larger impacts of SEBTC on healthful food intake systematically chose one of the models (not merely
that sites with larger levels of healthful food intake systematically chose one of the models).
With those crucial caveats, Table 2 presents new estimates from the SEBTC data on nutrition results separately for sites implementing SEBTC using SEBTC-SNAP and SEBTC-WIC, for households who self-reported SNAP receipt at the baseline survey in the spring (by contrast, Collins and colleagues21 analyze all SEBTC households). Unless explicitly noted, all findings —within the SEBTC-SNAP sites, within the SEBTC-WIC sites, and SEBTC-SNAP/SEBTC-WIC comparisons—are statistically significant at p<0.01.
More food assistance with minimal restrictions on purchases (i.e., SEBTC-SNAP) had statistically significant impacts on intake of fruits and vegetables, whole grains, and dairy products. These results are consistent with the conjecture that poor diets are caused in part by limited budgets, such that moderately more resources alone can lead to improved dietary choices. This result alone is crucial for the issues considered in this Special Issue. It is not, however, the crucial point for this article.
For a similarly valued benefit, the impacts of requirements (i.e., the restricted food package of SEBTC-WIC) are much larger than the impacts of a similarly valued benefit that could be used for any foods (i.e., SEBTC-SNAP). Impacts on foods in the SEBTC-WIC food package—fruits and vegetables, whole grains, and milk —are clearly statistically significant and several times larger for SEBTC-WIC than for SEBTC-SNAP. As noted earlier, the basic program eligibility criteria were common across the SEBTC-WIC and SEBTC-SNAP sites, but the sites themselves chose the model.
In addition, though SEBTC-WIC increased total household food resources and did not restrict how the basic SNAP benefit could be spent, intake of unhealthful foods decreased—sugar by 9% and sugar-sweetened beverages by 24%. By contrast, the unrestricted SEBTC-SNAP benefit had no statistically significant impact on these outcomes, and the SEBTC-WIC versus SEBTC-SNAP difference is clearly statistically significant.
From a naive perspective, the results on children eating more healthful foods are not surprising; require households to purchase healthful foods, and they will eat them. The economic theory discussion suggested (but discounted) the possibility that households could undo the requirements; these results suggest that households do not undo the requirements. The finding of less intake of unhealthful foods—in the case of SEBTC-WIC, sugar and sugar-sweetened beverages specifically—is striking, as is the fact that an increase in SEBTC-SNAP benefits did not generally result in an increase in consumption of unhealthful food. Neither model involved a ban.
Source: Authors tabulations from theSEBTC, Summer Survey, 2012, Sample limited to households that self-report SNAP receipt at baseline (i.e., in the spring) and who respond to all food frequency questions (n=11,938; in SNAP sites n=7,975; in WIC sites n=3,963).
Note: The p-values are based on a test of the difference between treatment group households and control group households. The null hypothesis being tested is that the treatment-control difference is zero (either the treatment-control difference in food consumption within a subgroup or a subgroup difference in the treatment-control difference in prevalence rates). Boldface indicates statistical significance (*p<0.05; **p<0.01). aDaily servings of fruits and vegetables and dairy are measured in cup-equivalents and in ounce-equivalents for whole grains, as defined by the Dietary Guidelines for Americans, 2010.16 One fruit and vegetable serving is 1 cup raw or cooked fruit or vegetables, vegetable juice, or fruit juice; 2 cups leafy green vegetables; or 1/2 cup dried fruit. One dairy serving is 1 cup milk, fortified soy beverage, or yogurt; 1Yi ounces natural cheese; or 2 ounces of processed cheese.
bWhole grain servings are measured in ounce-equivalents. One whole grain serving is one1-ounce slice bread,1 ounce uncooked pasta or rice, 1/2 cup cooked rice,pasta,or cereal; one 6-inch diameter tortilla; one 5-inch diameter pancake; or 1 ounce ready-to-eat cereal. "Respondents who reported that their child consumed more than one type of milk were included if any the milk types reported were nonfat orlow-fat. Those reporting only whole milk and/or 2% milk were not considered to usually consume nonfat or low-fat milk.
dTeaspoons of added sugars are derived from reported frequencies of consuming sugar-sweetened beverages (soda, fruit-flavored drinks, and sugar or honey added to coffee or tea),cookies/cakes/pies,doughnuts,ice cream,candy,and cereals.
SEBTC, Summer Electronic Benefit Transfer for Children; SNAP, Supplemental Nutrition Assistance Program; WIC, Special Supplemental Nutrition Program for Women, Infants, and Children.
Table 2. Impact of SEBTC-WIC and SEBTC-SNAPon Children's Food Consumption in Summer 2012, for Those Who Reported SNAP at Baseline
SNAP model WIC model SNAP/WIC difference
Food Control Impact Control Impact
Fruits and vegetables (servings per day)a 2.94 0.19** 2.94 0.59** 0.40**
Fruits and vegetables without fried potatoes (servings per day)a 2.83 0.19** 2.80 0.60*** 0.41**
Whole grains (servings per day)b 1.74 0.18* 1.90 0.82** 0.64*
Dairy products (servings per day)c 2.33 0.13** 2.35 0.46** 0.33**
Usually drank nonfat or low-fat milk (%) 17.19 -0.44 7.71 -0.29 0.15
Added sugars (teaspoons per day)d 17.74 0.17 19.07 -1.83** -2.00**
Added sugars excluding cereals (teaspoons per day)d 16.50 0.00 17.77 -2.35** -2.35**
Sugar-sweetened beverages (teaspoons per day)d 7.33 -0.19 8.86 -2.17** -1.98**
What mechanism might have caused this decrease? Perhaps parents receiving SEBTC through SEBTC-WIC fed children healthful foods to the extent that the children did not want or the parents did not feel the need to feed them the unhealthful foods (e.g., the children satisfied their interest in sweets from the fruit juice and fruit that was in the SEBTC-WIC package, or they were so full that they no longer had an interest in sweets). Perhaps it was an implicit nutrition education impact (i.e., a nudge).
Finally, note that the SEBTC results are only partially consistent with another argument of the opponents of bans and requirements: that they will lead to less food intake and worse food security.24,25 This might occur for several reasons, including: (1) not liking the required foods; (2) difficulty acquiring the required foods; and (3) additional stigma from using SNAP leading to lower participation (also discussed by Holden,2 writing for USDA). The SEBTC results do show substantially lower benefit redemption in the SEBTC-WIC sites compared with the SEBTC-SNAP sites (about 30 percentage points),21 yet other published analyses of the SEBTC data (e.g., Collins et al.21) find no differential impact on 30-day recall measures of Very Low Food Security for Children and Food Insecurity for Children (i.e., no
significant difference, even with the large household sample). Perhaps the inability to detect differential impact across the SEBTC-SNAP and SEBTC-WIC is because of insufficient power, given that there were only 14 sites, even though the household sample was very large. Alternatively, perhaps the households that were at risk of food insecurity were more likely to redeem benefits, even though they were not allowed to purchase their ideal foods. Finally, perhaps the interventions were too short to induce changes in entry and exit.
IMPLEMENTATION FEASIBILITY
Some have argued against requirements because of
concerns about feasibility.6,19,25,26 WIC, SEBTC-WIC,21
and HIP7 evidence suggests that feasibility issues— though real—can be overcome. This section briefly considers: (1) food lists; (2) retailer participation; and (3) software modifications.
The USDA (or a state or city) would need to maintain lists of required foods and retailers would need to incorporate those lists into their scanner systems. This is a nontrivial task. Food products change regularly. Nevertheless, the task appears feasible. Clearly, SNAP could simply adopt the existing WIC list. More broadly,
for WIC and for state sales taxes, such lists are already maintained (Pomeranz and Chriqui5 provide a complementary discussion). Similarly, lists of qualifying foods were created on short notice and maintained over the 15-month pilot period for both SEBTC-WIC and for HIP.
Imposing requirements in the SNAP program would require extra effort on the part of retailers. As a result, some retailers might stop accepting SNAP and some households (especially those with limited mobility) might have trouble accessing a retailer stocking the required foods. Consistent with this concern, there are many fewer WIC retailers than SNAP retailers.
Nevertheless, although real, issues of access are likely to be second order. SNAP represents a large share of the national food budget. It seems likely that the vast majority of current retailers would participate, even in a program with requirements.
Over a 15-month period, HIP offered rebates to some SNAP participants for the purchase of targeted fruits and vegetables. Implementing the program required identifying qualifying foods and processing them separately in the SNAP transaction. To do so, food retailers needed to modify their scanner systems. Despite the small (5,000 households) and short (15-month) nature of the pilot, all but one major food retailer and many smaller retailers did so. Even the one large retailer that did not implement the change indicated that it would have done so if the pilot had lasted longer—the timing was simply bad.
Even if some smaller retailers dropped SNAP, the broader literature27-29 suggests that, in net, the impact would not be large. However, there might be an impact for some subpopulations, in particular those without readily available transportation.30
Statewide EBT vendors and individual retailers would need to modify their computer systems. Many states already operate WIC via EBT and all are required to do so by 2020.31 Similar changes were required for HIP.
DISCUSSION
This discussion has two parts. First, it summarizes the theoretic, descriptive, and demonstration evidence on impact and feasibility of bans and requirements. Second, it considers whether—despite growing evidence that they can improve nutrition outcomes and are feasible— requirements are desirable.
Summarizing the Evidence
This paper has reviewed the policy background and theoretic basis for bans (i.e., prohibiting using SNAP for specific unhealthful foods) and requirements (i.e., requiring that SNAP be used to purchase specific healthful foods). The paper noted that the SNAP benefit formula
assumes that households with other income—and they are most households—already spend considerable cash on food. As a result, most households could theoretically circumvent any effect of a ban by moving unhealthful foods to cash and healthful foods to SNAP.
This paper also presented descriptive and evaluation evidence on the likely impact of requirements. The descriptive evidence on food intake of healthful foods among SNAP participants suggested that such intake was sufficiently low that many households would be unable to circumvent a requirement; that is, they are currently consuming at levels of healthful foods well below the requirement, such that their only option would be to consume the required foods or not to use part of the benefit.
As noted, the SEBTC evidence is only suggestive; SEBTC-SNAP versus SEBTC-WIC selection was not randomly assigned and the number of sites is small. With those crucial caveats, consistent with the conjecture that part of the cause of poor nutrition is lack of resources, the SEBTC-SNAP sites saw increases in intake of healthful foods (fruits and vegetables, whole grains, milk). More relevant to the focus of this paper, consistent with the conjecture from the descriptive analyses showing low intake of healthful foods, the evaluation evidence showed much larger impacts on nutritional quality from the SEBTC-WIC (i.e., distributing the benefit with restrictions) than from the SEBTC-SNAP (i.e., distributing the benefits without restrictions).
Although the paper began with a neoclassical economic theory, it also noted other perspectives, in particular implicit information or promotion effects. Such effects can be viewed as a nudge. The nudges perspective is another reason to favor requirements, that is, making SNAP more WIC-like. The nudges perspective would be a reason to favor bans, despite this paper's economic theory critique.
In addition, although the SEBTC results do show lower redemptions with restrictions, the SEBTC results do not show differential impact on food security with restrictions. Thus, granting the observational nature of the SEBTC-SNAP/SEBTC-WIC comparisons and the small number of sites, there is no evidence for one of the critiques of restrictions.
Finally, concerns about feasibility seem overstated. EBT technology grows more robust and scanner implantation more widespread. Experience with HIP and SEBTC-WIC has identified issues and shown that they can be overcome with moderate effort. Concerns about smaller retailers dropping out of the program seem overstated; even if this occurred, they represent only a very small share of SNAP expenditures (though they may be crucial for some subpopulations).
In sum, recent demonstration evidence about impacts (from SEBTC) strengthens the position that requirements would moderately improve nutritional intake (Dinour and colleagues32 and Richards and Sindelar19 provide theoretic, but not data-driven, arguments). Further reinforcing these arguments, recent evidence about feasibility (from SEBTC and HIP) strengthens the case for the feasibility of requirements. Other policies (e.g., taxing unhealthful foods) might also improve nutritional quality. Richards and Sindelar19 provide one catalog and the Government Accountability Office provides another list.33
Nevertheless, one attraction of requirements (and bans) is that they are (nearly) free. SNAP would simply require that some of the benefit be used for healthful foods. The benefit need not increase. There would be some one-time costs for reprogramming EBT systems and for some education and outreach efforts for participants, but those costs are second order relative to benefit costs. By contrast, Secretary Vilsack's "incentive-based solutions"—for example, HIP's rebates or more nutritional education—would cost more.
More direct tests, on different populations—not only children, not merely during the summer—seem worthy of consideration. A tentative movement toward requirements could set a low fraction of the SNAP benefit that must be spent on healthful foods, and allow considerable discretion as to the allocation within that fraction.
Are Requirements a Good Idea? This paper deliberately leaves open the question of whether requirements should be implemented. If the goal is to improve nutritional intake, the paper has presented some evidence that requirements are likely to help and that they are feasible. This paper has also presented some evidence against fears that requirements would worsen food security.
Nevertheless, requirements are coercive, a form of "hard paternalism."34 Rather than try to convince SNAP recipients that they want to eat healthful foods (e.g., nutrition education) and then trusting their judgment, the program would effectively force them to buy healthful foods or forego the benefit. Implicitly, such a policy is saying: "Government knows what is good for you better than you do." Although such advocacy of hard paternalism is standard in public health circles, in broader society it remains controversial (e.g., the repeal of Prohibition in 1933; the continued legal distribution of cigarettes, other tobacco products, and alcohol; and the spreading legalization of marijuana; Friedman34 provides a more detailed argument). Thus, the paternalistic nature of requirements (and bans)—and, in particular that they
would apply only to the poor—is an argument against them.
At least in the absence of clear evidence of large externalities,35-38 some would view such hard paternalism as odious, especially as requirements would apply only to those so poor that they must rely on SNAP to help cover their food budgets.39 That SNAP's requirement that the benefit be used only on food already incorporates some degree of hard paternalism does not imply that additional amounts of hard paternalism in the SNAP program would be better.
ACKNOWLEDGMENTS
Publication of this article was supported by the Physicians Committee for Responsible Medicine. The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Physicians Committee for Responsible Medicine.
The discussion reflects the position of the authors only. It does not represent the position of USDA, the Commonwealth of Massachusetts, or Abt Associates.
This paper draws extensively on the SEBTC and the Healthy Incentives Pilot. We acknowledge funding for those efforts from the USDA/Food and Nutrition Service. We also acknowledge the assistance of USDA/Food and Nutrition Service, the Massachusetts Department of Transitional Assistance, and the broader project teams in implementing the interventions and interpreting the results. Anne Wolf generated the SEBTC estimates. Suzanne Efrurth edited the document.
Partial funding for preparation of this article was provided by the Physicians Committee for Responsible Medicine and by Abt Associates internal funds.
No financial disclosures were reported by the authors of this paper.
REFERENCES
1. McGeehan P. U.S. rejects mayor's plan to ban use of food stamps to buy soda. New York Times. www.nytimes.com/2011/08/20/nyregion/ ban-on-using-food-stamps-to-buy-soda-rejected-by-usda.html. Published August 19, 2011. Accessed April 14, 2016.
2. Holden O. Letter to M. Gomez in response to request for waiver from State of Minnesota, May 4, 2004. web.archive.org/web/*/http:/heart land.org/sites/all/modules/custom/heartland_migration/files/pdfs/15364. pdf. Accessed April 14, 2016.
3. Holleman J. SC food stamp restrictions face tall hurdles. The State. March 2, 2013.
4. Kamp J, Newman J. Maine wants candy, soda excluded from food stamps. Wall Street Journal. www.wsj.com/articles/maine-wants-can dy-soda-excluded-from-food-stamps-1448447401. Published November 25, 2015. Accessed April 14, 2016.
5. Pomeranz JL, Chriqui JF. The Supplemental Nutrition Assistance Program: analysis of program administration and food law definitions. Am J Prev Med. 2015;49(3):428-436. http://dx.doi.org/10.1016Ai. amepre.2015.02.027.
6. U.S. Department of Agriculture. Implications of restricting the use of food stamp benefits. www.fns.usda.gov/sites/default/files/arra/FSPFoo dRestrictions.pdf. Published 2007. Accessed April 14, 2016.
7. Bartlett S, Klerman, J, Olsho L, et al. Evaluation of the Healthy Incentives Pilot (HIP): final report. www.fns.usda.gov/sites/default/ files/HIP-Final.pdf. Published 2014. Accessed April 14, 2016.
8. Southworth HM. The economics of public measures to subsidize food consumption. J Farm Econ. 1945;27:38-66. http://dx.doi.org/10.2307/ 1232262.
9. Hoynes HW, Schanzenbach DW. Consumption responses to in-kind transfers: evidence from the introduction of the Food Stamp Program. Am Econ J Appl Econ. 2009;1(4):109-139. http://dx.doi.org/10.1257/ app.1.4.109.
10. Chai A, Moneta A. Retrospectives: Engel curves. J Econ Perspect. 2010;24(1):225-240. http://dx.doi.org/10.1257/jep.24.L225.
11. Meyer BD, Mok WKC, Sullivan JX. Household surveys in crisis. J Econ Perspect. 2015;29(4):199-226. http://dx.doi.org/10.1257/jep.29.4.199.
12. Gregory C, Ver Ploeg M, Andrews M, Coleman-Jensen A. Supplemental Nutrition Assistance Program (SNAP) participation leads to modest changes in diet quality. Washington, DC: U.S. Department of Agriculture, Economic Research Service. www.ers.usda.gov/media/ 1083334/err147.pdf. Published 2013. Accessed April 14, 2016.
13. Bhattacharya J, Currie J, Haider S. Poverty, food insecurity, and nutritional outcomes in children and adults. J Health Econ. 2004;23 (4):839-862. http://dx.doi.org/10.1016/jjhealeco.2003.12.008.
14. Han E, Powell LM, Isgor Z. Supplemental nutrition assistance program and body weight outcomes: the role of economic contextual factors. Soc Sci Med. 2012;74(12): 1874-1881. http://dx.doi.org/10.1016/j.socscimed.2012.02.032.
15. Ludwig DS, Blumenthal SJ, Willett WC. Opportunities to reduce childhood hunger and obesity: restructuring the Supplemental Nutrition Assistance Program (the Food Stamp Program). JAMA. 2012;308 (24):2567-2568. http://dx.doi.org/10.1001/jama.2012.45420.
16. U.S. Department of Agriculture and U.S. Department of Health and Human Services. Dietary Guidelines for Americans, 2010. 7th ed, Washington, DC: U.S. Government Printing Office; December 2010.
17. Downs JS, Loewenstein G, Wisdom J. Strategies for promoting healthier food choices. Am Econ Rev. 2009;99(2):159-164. http://dx. doi.org/10.1257/aer.99.2.159.
18. Thaler R, Sunstein C. Nudge: The Gentle Power of Choice Architecture. New Haven, CT: Yale University Press; 2008.
19. Richards MR, Sindelar JL. Rewarding healthy food choices in SNAP: behavioral economic applications. Milbank Q. 2013;91(2):395-412. http://dx.doi.org/10.1111/milq.12017.
20. Wilde P, Klerman J, Olsho L, Bartlett S. Explaining the impact of USDA's Healthy Incentives Pilot on different spending outcomes. Appl Econ Perspect Pol. In press. Online November 30, 2015. http://dx.doi. org/10.1093/aepp/ppv028.
21. Collins AM, Briefel R, Klerman JA, et al. Summer Electronic Benefits Transfer for Children (SEBTC) Demonstration: summary report. Alexandria, VA: U.S. Department of Agriculture, Food and Nutrition Service. www.fns.usda.gov/sites/default/files/ops/sebtcfinalreport.pdf. Published 2016. Accessed July 4, 2016.
22. National Cancer Institute. Dietary screener questionnaire in the NHANES 2009-10: background. http://epi.grants.cancer.gov/nhanes/ dietscreen/. Published 2015. Accessed April 14, 2016.
23. National Cancer Institute. Dietary Screener Questionnaire (DSQ) in the NHANES 2009-10: data processing & scoring procedures. http://epi.grants.cancer.gov/nhanes/dietscreen/scoring/. Published 2015. Accessed April 14, 2016.
24. Schanzenbach DW. Proposals to ban purchase of sugary drink with food stamps won't work. Christian Science Monitor. www.csmonitor.
com/Commentary/0pinion/2013/0313/Proposals-to-ban-purchase-of-sugary-drink-with-food-stamps-won-t-work. Published March 13, 2013.Accessed April 14, 2016.
25. Gundersen C. SNAP and obesity. In: Bartfeld J, Gunderson G, Smeeding TM, Ziliak JP, eds. SNAP Matters: How Food Stamps Affect Health and Well-Being. Palo Alto, CA: Stanford University Press; 2015: pp 161-185.
26. You W, Mitchell PD, Nayga RM. Improving food choices among Supplemental Nutrition Assistance Program recipients. Health Econ. 2012;21(7):852-864. http://dx.doi.org/10.1002/hec.1758.
27. Taylor R, Villas-Boas SB. Store choice among low income households. Am J Agr Econ. 2016;98(2):513-532. http://dx.doi.org/10.1093/ajae/ aaw009.
28. Kyureghian G, Nayga RM. Food store access, availability, and choice when purchasing fruits and vegetables. Am J Agr Econ. 2013;95 (5):1280-1286. http://dx.doi.org/10.1093/ajae/aat043.
29. Cummins S, Flint E, Matthews SA. New neighborhood grocery store increased awareness of food access but did not alter dietary habits or obesity. Health Aff. 2014;33(2):283-291. http://dx.doi.org/10.1377/ hlthaff.2013.0512.
30. Fitzpatrick K, Greenhalgh-Stanley N, Ver Ploeg M. The impact of food deserts on food insufficiency and SNAP participation among the elderly. Am J Agr Econ. 2015;98(1):19-40. http://dx.doi.org/10.1093/ ajae/aav044.
31. U.S. Department of Agriculture. WIC policy memorandum 2011-3. Implementation of WIC-related electronic benefit transfer (EBT), provisions of P.L. 111-296. http://www.fns.usda.gov/sites/default/ files/2011-3-ImplementationofWICRelatedEBTProvisions-PL111-296. pdf. Published March 22, 2011. Accessed April 14, 2016.
32. Dinour LM, Bergen D, Yeh M-C. The food insecurity—obesity paradox: a review of the literature and the role food stamps may play. J Am Diet Assoc. 2007;107(11):1952-1961. http://dx.doi.org/10.1016/j. jada.2007.08.006.
33. Government Accountability Office. Food Stamp Program: options for delivering financial incentives to participants for purchasing targeted foods. http://www.gao.gov/new.items/d08415.pdf. Published 2008. Accessed April 14, 2016.
34. Friedman DA. Public health regulation and the limits of paternalism. Conn L Rev. 2014;46(5):1689-1770. http://dx.doi.org/10.2139/ssrn. 2332988.
35. Bhattacharya J, Sood N. Health insurance and the obesity externality. Adv Health Econ Health Serv Res. 2007;17:279-318. http://dx.doi.org/ 10.1016/S0731-2199(06)17011-9.
36. Bhattacharya J, Sood N. Who pays for obesity? J Econ Perspect. 2011; 25(1):139-158. http://dx.doi.org/10.1257/jep.25.L139.
37. Basu S, Seligman HK, Gardener C, Bhattacharya J. Ending SNAP subsidies for sugar-sweetened beverages could reduce obesity and type 2 diabetes. Health Aff. 2014;33(6):1032-1039. http://dx.doi.org/ 10.1377/hlthaff.2013.1246.
38. Philipson TJ, Posner RA. Is the obesity epidemic a public health problem? A review of Zoltan J. Acs and Alan Lyles's "Obesity, Business and Public Policy." J Econ Lit. 2008;46(4):974-982. http://dx.doi.org/ 10.1257/jel.46.4.974.
39. Ferdman R. Missouri Republicans are trying to ban food stamp recipients from buying steak and seafood. Washington Post. www. washingtonpost.com/news/wonk/wp/2015/04/03/missouri-republican s-are-trying-to-ban-food-stamp-recipients-from-buying-steak-and-sea food/. April 3, 2015. Accessed April 14, 2016.
