A Behavioral Economics Experiment on the Effect of Incentives on Food Choice in the National School Lunch Program

Year: 2014

Research Center: Institute for Research on Poverty, University of Wisconsin-Madison

Investigator: Samek, Anya

Institution: University of Wisconsin, Madison

Project Contact:
Anya Samek
University of Wisconsin-Madison
School of Human Ecology
4204 Nancy Nicholas Hall
1300 Linden Drive
Madison, WI 53706
Phone: 608-262-5498


Obesity and its associated disease burdens (for example, coronary diseases and diabetes) disproportionately affect the low-income urban population in the United States. A major cause of this problem is the food environment: low-income communities tend to have fewer grocery providers, creating what has been termed a “food desert.” U.S. Department of Agriculture estimates suggest that approximately 11.5 million low-income Americans live more than a mile from a supermarket. Indeed, the lack of availability of nutrient-rich foods in poor urban areas is associated with higher obesity rates. Unfortunately, interventions to improve the urban landscape have shown limited success, as setting up supermarkets in food deserts has not been profitable for firms, and nonprofit endeavors have not resulted in the health impacts that were anticipated.

The problem also extends to children, as evidenced by the rise in childhood overweight rates in the United States. U.S. children consume less than 20 percent of the recommended amount of whole grains and just 10 percent of the recommended amount of vegetables/legumes. The tendency to consume an unhealthful diet is learned at an early age and persists throughout adulthood, as individuals are more likely to eat familiar foods. Children’s health and weight are also closely linked to dietary behavior. Moreover, students from lower income families are at greatest risk.

This study proposes that one reason efforts to improve health through changes to the food environment have been unsuccessful is that they have largely ignored the behavioral side of the problem. This study designed interventions that harness behavioral economics principles to improve food choice among adults and children. It focused on theories of temptation and self-control. Models incorporating temptation impulses and self-control are among the most prominent in behavioral economics. The dynamic inconsistencies predicted by these models provide a reason for the observed difficulty of people to save more for the future, exercise more, eat healthier, and quit smoking, despite stated desires to make these changes.

The study evaluated the success of the interventions through two experimental studies. While it is inherently difficult to intervene in someone’s home, the studies use the setting of Federal food assistance programs in the United States as “teachable moments” for the interventions. Study 1 focused on the food choices of children who participate in the National School Lunch Program (NSLP). It conducted a field experiment in which over 1,500 children in grades K–12 in a Chicago-area school district were randomized to a control group or to one of two treatment groups. In the first treatment group, children were able to set a goal to choose the healthier white milk rather than less healthful chocolate milk prior to heading to the lunchroom. In the second treatment group, children were offered an unconditional gift of a sticker to choose the white milk. The choices of children in all groups were then recorded during lunch. Study 2 focused on the food purchase decisions of participants in the Supplemental Nutrition Assistance Program (SNAP) at a grocery store in Chicago. The investigator designed an intervention in which shoppers at a grocery store selected items for a food delivery program to be received in 1 week or immediately. Over 200 low-income shoppers participated in the experiment, which lasted 3 weeks. In addition to comparing the selection of foods made for the future versus the present, the study also asked participants to decide whether to “commit” to future healthful choices.

In the NSLP study, data from the Chicago Heights, IL, school district prior to the intervention showed that children choose white milk about 12 percent of the time, opting for chocolate milk the rest of the time. Having the teacher read a short educational message provides evidence for improved choice—47 percent of children start choosing white milk when such a message is read. The intervention treatments had significant effects. First, the study found that small, unconditional gifts significantly increase the likelihood of choosing the more healthful white milk by 50 percent relative to a control group (68 percent choose white milk). Allowing children to set a goal to choose healthier also improves choices, but only by 28 percent (60 percent choose white milk). These findings suggest a role for nonbinding gifts and plans in health-related decisions among children.

In the SNAP grocery store experiment, the study found that 46 (21 percent) of individuals exhibited dynamic inconsistencies, meaning they chose different foods for the future delivery as compared to in the present. Of these 46, 44 (96 percent) selected more healthful combinations of foods in their future deliveries as compared to their present deliveries. Consistent with models of temptation and self-control, participants’ immediate decisions were tilted toward less healthful alternatives than their advance decisions. Participants also had the option to “commit” to a healthful future delivery, without the option to change their mind later. Seventy-three (33 percent) chose to commit, and interestingly those who commit are more likely to be dynamically consistent. Hence, it seems committing subjects are systematically those that overcome temptation rather than those who succumb to it.

The study targeted the interventions at the time of food choice, and also reached a critically important group, low-income individuals who participate in Federal food assistance programs. The findings provide evidence of relevance to policymakers, namely that providing opportunities to make food choices in advance leads to more healthful choices. Moreover, Study 1 showed that the advance choice need not be binding, and Study 2 showed that at least some individuals are willing to commit to advance choices. The results also have implications for theory, since much of the empirical evidence for dynamic inconsistency to date has come from laboratory experiments, rather than from real-world decisions.