Understanding Food Choices

The important role that diet plays in health and disease is well established. Excessive intake of energy, saturated fat and sodium increase the risk of heart disease, diabetes and certain cancers. As a result, interventions aimed at encouraging healthier food consumption have been pursued by many countries. These can be broadly grouped into the following categories: price manipulations, food labeling, and behavioral nudges.

No study has previously assessed the potentially interactive effects of a multi-component intervention that incorporates the strongest features of each intervention component while discarding those that do not meaningfully contribute to healthier consumption. That is the goal of this effort.

To this end, the investigators chose a full-factorial design because this experimental design allows us to estimate not only the independent (main) effects of the interventions but also their interaction effects. The full-factorial design includes all possible combinations of the interventions' status. Because the investigators have four interventions, each of which has two levels (intervention On or Off), there are 16 (i.e. 2^4) experimental conditions/arms in total. The four interventions for this study are outlined below:

• Explicit Tax: To impose an explicit tax on food and beverage items that are eligible for red stop-sign (explained in b.) as less healthy foods.

Foods and Beverages: The investigators will impose a 20% tax on sales price of the food items.

• Food Labels (with the summary of healthiness of shopping baskets \& targets): To provide the green circle 'healthy choice' food label to items with Nutri-Score "A" and "B", the amber circle 'in between healthy and unhealthy choice' food label to items with Nutri-Score "C", and the red stop-sign 'Unhealthy choice' food label to items with Nutri-Score "D" and "E". In the presence of food labels, before shopping, subjects will watch a video briefly explaining the food labels on the store and be provided a live visual indicator of the healthiness of shoppers' current basket (called "My Cart Summary") with the recommended healthy baskets goals (i.e., Green logo products ≥ 60% and Red logo products ≤ 15% of the weighted number of servings of products.). My Cart Summary will be displayed as a pie chart with the proportion of servings in their current shopping baskets, according to the three food labels. This way, as shoppers add to their basket while browsing the store, they will have visual feedback on how their latest addition contributes to their total basket healthiness.
• Ordering: To order food items by Nutri-Score such that healthier items are shown upfront on NUSMart.
• Within Group Substitution: At the checkout, to provide consumers with an opportunity to replace their original items with an item chosen from two or four recommended substitutes that are healthier within their corresponding food categories based on the similarities of price, ingredients, flavor, and other characteristics.

Participants will be randomly assigned to one of the 16 arms and instructed to perform a one-time hypothetical grocery shopping on NUSMart.

The investigators will collect participants' demographic and health characteristics as well as hunger at the time of the survey and their self-control in the baseline survey. The collected data will be used to precisely estimate the causal effect of the interventions and address their underlying mechanism to change consumers' food choices.

Our hypotheses about the effects of the interventions on diet quality, measured by the weighted average Nutri-Score (primary) of finalized shopping baskets, are as follows:

1. There will be significant and positive main effects of each of the interventions.
2. There will be significant and positive interaction effects on diet quality of: Explicit tax \& food labels, Food labels \& ordering.

We will also run models both with and without including covariates that include demographic variables (e.g., age, minority status, income, BMI, sex, and household size) and measurements of self-control, hunger, and health-status. To test the moderating effects of hunger, self-control, health-status, and education level, we will include interaction terms between the intervention arms and these variables.