The literature on policy interventions to address obesogenic dietary behaviors can be divided into 3 distinct categories. The first is the use of economic measures to alter food consumption, such as taxing unhealthy ingredients and subsidizing healthy foods, 1---4 in light of studies that have shown the price of a calorie obtained from unhealthful foods is lower than the price of a calorie from more healthful foods. 5,6 The second is targeting the food environment through zoning polices, including increasing the number of healthy food vendors in "food desert" communities 7 and restricting the opening of new fast-food restaurants.
8The third and final category is related to combating unhealthy eating norms, given research showing the power of food marketing to change dietary behaviors and proposing restrictions on the time, place, and manner in which obesogenic foods are marketed. 9,10 Conversely, pro-nutritional marketing focuses on education as a means of increasing consumers' awareness of dietary health (e.g., nutrition disclosure on menus and the issuing of dietary guidelines).
11, 12The research to date focusing on the effects of these various approaches has incorporated theoretical and empirical techniques that rely on the stable unit treatment value assumption, according to which there are no interactions among people who experience an intervention that would alter the effectiveness of the intervention. However, this assumption is known to be violated in the case of obesity-related behaviors.13, 14 What is therefore not clear from existing regression-based and experimental empirical work is the potential magnitude of the population-level impact of these policies if they were implemented in the real world. We performed simulations to contrast the potential of different approaches aiming at tackling unhealthy dietary behaviors in a population of urban US adults. Simulations involving systems dynamics or agent-based modeling (ABM) are increasingly being used in public health, 15,16 We developed a computer model explicitly representing how individuals make decisions to examine the impact of policies on their food consumption. Our model was based on a multilevel theory of population health that emphasizes the role of cognitive habits in human behaviors. 25 According to this theory, individual beliefs are influenced by incentivesas in rational-choice theory-but also shaped by cognitive habits that are reinforced within a population by social norms and culture. Thus, our model examined how individual beliefs are influenced by interventions either in the social network or in the food environment.
METHODSOur agent-based model contained 2 kinds of agents: individuals and food outlets. Individuals make dietary choices, and food outlets adapt to those choices. In our model, a single time period was defined as 1 day, and each simulation was run for more than 3 years (1100 days).In our simulations, individuals were assigned demographic characteristics (age, gender, Objectives. Unhealthy eating is a complex-system prob...