Perceived Effectiveness of Added Sugar Labels

Objectives and hypotheses:

The goal of this study is to examine how added sugar restaurant menu labels influence U.S. adults' perceptions and reactions. Predictions:

In a between-subjects experiment with 3 arms (control label, icon-only added sugars label, and icon plus text added sugars label):

• Added sugar labels will be perceived as more effective than the control label.
• The text plus icon label will be perceived as more effective than the icon-only label.
• A larger proportion of participants who see the added sugar labels will report learning something new than those who see the control label.
• Added sugar labels will lead participants to more accurately identify restaurant menu items high in added sugars compared to the control label, and the text plus icon label will outperform the icon-only label on this outcome.

Additionally, using a within-subjects design:

The study will examine which label (control, icon, text plus icon) most discourages wanting to consume menu items high in added sugars.

Analyses will compare various icon and text options for the added sugars label to determine which icon and which text variations are perceived as most discouraging for wanting to consume items high in added sugars. The is no hypothesis about which will be perceived as more discouraging.

Planned analyses:

For predictions 1-2: linear regression model (OLS) regressing PME on indicator variables for experimental condition. The margins command in STATA will be used to conduct pairwise comparisons between each condition (i.e., icon-only vs. icon plus text label). Also, PME will be regressed on an indicator variable combining the added sugar label groups.

For predictions 3-4, Poisson regression with robust standard errors will be used to estimate relative probability, regressing each dichotomous outcome on indicator variables for experimental condition. The margins command in STATA will be used to conduct pairwise comparisons between each condition. The outcomes will be regressed on an indicator variable combining the added sugar label groups. If the Poisson regressions do not converge, logistic regression will be used.

For the within-subjects comparisons, mixed effects linear models will be used to assess the relationship between condition and rating of label discouragement for consuming items high in added sugars.

A critical alpha 0.05 will be used, and statistical tests will be two-tailed. Because this is an initial, exploratory study to help identify the best performing label to use in a larger trial, alpha level will not be adjusted to control for multiple comparisons.

If there is evidence of deviations from modeling assumptions required for the parametric tests above, non-parametric sensitivity analyses will be conducted.

Inclusion criteria

• Equal or greater than 18 years of age
• Less than 100 years of age
• English-speaking
• U.S. residents
• Had to have purchased food from a restaurant at least once per month prior to the COVID-19 pandemic
• Participants will reflect the U.S. Census Bureau's 2018 American Community Survey 5-year estimates for gender, race/ethnicity, educational attainment, and age

Exclusion criteria

• Failing the attention check question
• Completing the survey in less than 30% of the median completion time
• Straightlining
• Providing nonsensical responses to open-ended questions