The Influence of Sugar-Sweetened Beverage Health Warning Labels on Parents’ Choices
- aDepartment of Medical Ethics and Health Policy, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania;
- bDepartment of Nutrition, Harvard T.H. Chan School of Public Health, Boston, Massachusetts; and
- cSchool of Public Health and Health Systems, University of Waterloo, Waterloo, Ontario, Canada
Dr Roberto conceptualized and designed the study, analyzed the data, and drafted the initial manuscript; Ms Wong and Ms Musicus helped design the data collection instruments, coordinated and supervised data collection, and reviewed the manuscript; Dr Hammond contributed to the study design, helped interpret the data, and critically reviewed the manuscript; and all authors approved the final manuscript as submitted.
BACKGROUND AND OBJECTIVES: US states have introduced bills requiring sugar-sweetened beverages (SSBs) to display health warning labels. This study examined how such labels may influence parents and which labels are most impactful.
METHODS: In this study, 2381 demographically and educationally diverse parents participated in an online survey. Parents were randomly assigned to 1 of 6 conditions: (1) no warning label (control); (2) calorie label; or (3–6) 1 of 4 text versions of a warning label (eg, Safety Warning: Drinking beverages with added sugar[s] contributes to obesity, diabetes, and tooth decay). Parents chose a beverage for their child in a vending machine choice task, rated perceptions of different beverages, and indicated interest in receiving beverage coupons.
RESULTS: Regression analyses controlling for frequency of beverage purchases were used to compare the no warning label group, calorie label group, and all warning label groups combined. Significantly fewer parents chose an SSB for their child in the warning label condition (40%) versus the no label (60%) and calorie label conditions (53%). Parents in the warning label condition also chose significantly fewer SSB coupons, believed that SSBs were less healthy for their child, and were less likely to intend to purchase SSBs. All P values <.05 after correcting for multiple comparisons. There were no consistent differences among different versions of the warning labels.
CONCLUSIONS: Health warning labels on SSBs improved parents’ understanding of health harms associated with overconsumption of such beverages and may reduce parents’ purchase of SSBs for their children.
- SSB —
- sugar-sweetened beverage
- SSI —
- Survey Sampling International
What’s Known on This Subject:
Research finds that large tobacco text warnings are associated with increased risk perceptions of the health harms of using tobacco products. Although this suggests sugar-sweetened beverage (SSB) warning labels will help educate consumers, few studies have investigated SSB warning labels.
What This Study Adds:
In an online study of 2381 demographically and educationally diverse parents, SSB health warning labels improved parents understanding of health harms associated with overconsumption of SSBs. The warning labels also lowered parents’ intentions to purchase SSBs for their children.
National US surveys have revealed that 66% of children 2 to 11 years old drink sugar-sweetened beverages (SSBs) daily.1 One study estimated that these beverages contribute 69 kcal (calories) and 118 calories daily to the diets of children 2 to 5 and 6 to 11 years old, respectively.2 Relative to white children ages 2 to 5, African American children consume almost twice as many calories from SSBs.2 Research has linked children’s consumption of SSBs with weight gain and risk of obesity in adulthood, as well as dental caries.3–5 Further, 60% of sodas are estimated to contain caffeine, a known addictive substance.6
Growing concern about the health harms associated with SSB intake has prompted policies and interventions to try to reduce their consumption. Most recently, San Francisco passed a law (although due to an ongoing lawsuit it has not yet been implemented) requiring SSB advertisements to include a conspicuous and legible warning label.7 The health warnings would be required to take up at least 20% of print advertisements on billboards, walls, taxis, and buses within city limits, but would not apply to advertisements in newspapers, on broadcast outlets, or on the Internet. Legislative bills have also been introduced in California and New York State, which would require SSBs to display health warning labels on product containers, much like tobacco warning labels.8,9
From a public health perspective, the first goal of a warning label is to educate consumers about the potential health harms of a product. In the case of SSBs, such education may be particularly needed for certain beverages. For example, 1 study of 982 parents revealed that many believed some SSBs, including flavored waters and fruit and sports drinks, were healthy options for children.10 The second goal of a warning is to reduce consumption by making salient the long-term consequences of drinking SSBs at moments of purchase and consumption.
Research has revealed that large tobacco text warnings are associated with greater knowledge and increased risk perceptions of the health harms of using tobacco products.11 Although this evidence suggests SSB warning labels will help educate consumers, there is a lack of data on the influence of SSB warning labels. Therefore, the current study addressed the following research questions:
Do warning labels educate parents’ about the health harms associated with SSB intake above and beyond current beverage industry standards of placing calorie information on beverages?
Do warning labels influence parents’ intentions to buy SSBs for their children and is this effect moderated by education level?
Do warning labels influence parents’ perceptions and intentions toward nonlabeled beverages?
Do the effects of warning labels differ across different label phrasings?
What are parents’ beliefs about proposals to put warning labels on SSBs?
Overall, we hypothesized that a warning label would be more likely to increase perceptions of the health harms of SSBs and reduce purchase intentions for SSBs relative to calorie labels or no labels. This research has the potential to inform regulatory efforts in states and municipalities considering SSB warning label policies.
We recruited 3136 primary caregivers of a child aged 6 to 11 years old through Survey Sampling International (SSI), a firm that recruits research participants through their online panels and other online communities, social networks, and Web sites by using banner advertisements, SMS and text messaging, and telephone alerts. SSI uses a 3-stage randomization process in matching participants with surveys they are likely to be eligible for and complete. First, randomly selected participants from SSI’s panels are combined with people entering the sample who have responded to other SSI recruitment advertisements. An invitation is sent asking them to “take a survey.” To reduce selection bias, no specific details of the survey are included in the invitation. After recruitment, potential participants complete proprietary quality control questions before inclusion in a study sample. Participants are then randomly assigned to surveys they are likely to be able to take. There are also quality controls to ensure participants do not take the same survey twice. Based on the target population being recruited, SSI offers a diversity of incentives, including cash, points, prizes, sweepstakes, or being able to donate to charity. For the current study, SSI determined whether a 6- to 11-year-old resided in the household and then randomly selected caregivers in those households to participate. The survey had to be completed on a computer size screen; smart phones or tablets were not permitted. Recruitment efforts were targeted so that the sample reflected the educational make-up of the United States on the basis of 2010 census data. We also oversampled Hispanics and African Americans because they have the highest obesity prevalence in the United States (Table 1).12 Of the 3136 participants who started the survey, 2492 completed it, and 2381 accurately answered the data integrity check question described below; these participants composed the final sample.
Label Development and Randomization
After agreeing to participate, child caregivers were randomly assigned to 1 of 6 label conditions. The conditions did not significantly differ on sociodemographic variables. The first condition (No Label Control) was a no health warning label control group in which participants viewed beverages without any label. The second condition (Calorie Label) displayed a “Calories per Bottle” label that appeared on all beverages, not just SSBs, which were identical to the American Beverage Association’s “Clear on Calories” labels.13 Conditions 3 to 6 displayed warning labels. The first warning label condition (California Label) included the text proposed in the California bill (Safety Warning: Drinking beverages with added sugar[s] contributes to obesity, diabetes, and tooth decay). We then tested 3 modifications of that text designed to make aspects of the warning more salient. A scientific advisory board and legal team reviewed all labels to ensure they accurately reflected current scientific evidence and would be legally viable.
The second warning label condition (Weight Gain Label) modified the California label by changing “obesity” to “weight gain.” The rationale was that obesity might seem like a distant problem affecting older adults and may therefore be less salient to parents making decisions for children. Because people tend to focus on the present rather than the future, we hypothesized that the phrase “weight gain” would be more influential than “obesity.” For the third warning label condition (Preventable Label), the words “preventable diseases like” were inserted before “obesity.” We hypothesized this would be more effective than the California label because it makes salient that these diseases could be prevented. Finally, to respond to concerns that the labels are misleading because the messaging does not apply to type 1 diabetes, we tested a fourth label condition (Type 2 Diabetes Label) that includes the words “type 2” before diabetes. See Fig 1 for label images. We hypothesized that the Type 2 Diabetes Label would not differ from the California Label.
Warning Label Criteria
We used the criteria in the proposed California legislation to determine which beverages qualified for a warning label. These criteria were as follows: any sweetened nonalcoholic beverages with added sweeteners that contain 75 or more calories per 12 fluid ounces qualified for a label.8 Beverages containing 100% natural fruit juice or natural vegetable juice with no added caloric sweeteners, liquid products used as “dietary aid,” products used for oral nutritional therapy and/or a source of necessary nutrition as a result of a medical condition, oral electrolyte solutions, infant formula, and milk were excluded.
This study was approved by the Harvard T.H. Chan School of Public Health Institutional Review Board. After providing informed consent, caregivers took a 20-minute survey (median completion time was 23 minutes). At the beginning of the survey, they were asked if they had more than 1 child between the ages of 6 and 11 years old, and if so, to answer the questions on the basis of their youngest child within that age range.
Vending Machine Choice
The first task required parents to imagine they were looking at a vending machine while shopping because they wanted to purchase a beverage for their child. They were asked to scroll down on 1 screen to view 20 popular, 20-ounce beverages (12 of which were SSBs) presented in 2 columns and then select 1. Beverages were presented in random order. Participants had to view all the beverages before they could advance to the next question. We included beverages with a wide range of added sugar content. Participants were told to select the beverage brand they wanted even if they typically buy a different flavor. Those in a warning label condition were also told that, “drinks with a lot of added sugar have a safety warning label on them.” When a calorie or warning label appeared on a product, it was enlarged and displayed above the beverage image (Fig 2). Because this is among the first studies on SSB warning labels, we tested labels under conditions when they were highly visible and salient. If no effects are detected, there is little reason to think warning labels would work better in the real world.
Beverage Perceptions and Intentions
After completing the vending machine task, participants answered questions about 14 of the twenty, 20-ounce beverages used in the vending machine task (9 of which were SSBs). After pilot testing the survey, we only included a subset of beverages from the vending machine task so the survey did not take too long. Beverages were shown in random order displaying labels on the basis of study condition (see Table 2 for survey questions).
After the perceptions and intentions task, participants scrolled down on 1 screen to view the same 20 beverages in random order that were presented in the vending machine task. They were asked to indicate all beverages they would buy for their child for which they would like to receive a coupon.
Health and Sociodemographic Information
After the coupon task, we asked several health-related questions, including a question asking participants to indicate their relationship with their weight. Participants also provided information about health conditions and whether their doctor ever told them their child is overweight or obese or has type 2 diabetes. Finally, participants indicated their age, gender, height, weight, number of children, ethnicity, race, educational level, marital status, household income, political party affiliation, and the US state or territory they reside in.
As a manipulation check, participants were asked whether they saw a warning label on any of the beverages (choosing among: yes, no, and I don’t know).
Support for Warning Label Policy
At the end of the survey, participants were presented with either the California warning label if they were in the control, calorie label, or California condition or a picture of the warning label they were assigned to for the other conditions (see Table 2 for survey questions).
Data Integrity Check
The last question asked people to indicate how many days are in a week (options ranged from 1–7). Those answering incorrectly were excluded from the analyses (N = 111).
First, we assessed differences across the 4 warning label conditions by using logistic regression for categorical outcomes and ordinary least squares regression for continuous outcomes. We regressed each dependent variable on label condition, controlling for self-reported frequency of purchasing SSBs and beverages that would not qualify for a warning label over the past month. Controlling for past behavior is important because people’s perceptions, beliefs, intentions, and behavior are partially a product of people’s past behavior. We ran all pairwise comparisons by varying which condition was the reference group.
We then collapsed across all warning label conditions and repeated these analyses, running all pairwise comparisons to assess differences among the No Label Control condition, The Calorie Label condition and the Warning Label condition. For this analysis, we examined label effects on both SSBs and beverages that would not qualify for a warning label to see if the labels had spillover effects on beverages that did not have a warning label.
To determine whether the effects of warning labels were similar across the education spectrum, we regressed each of the outcome variables on (1) a dummy variable for Warning Label condition, (2) a dummy variable for Calorie Label condition, (3) mean-centered level of education, (4) the interaction between the Warning Label condition and level of education, (5) the interaction between the Calorie Label condition and level of education, and (6) the self-reported average frequency of purchasing SSBs and beverages that did not qualify for a warning label. For each of these 3 sets of analyses, we used a P < .05 significance threshold and the Bonferroni-Holm procedure to correct for multiple comparisons.15
Do Different Warning Labels Exert Different Effects?
Our first set of analyses explored whether the 4 warning label conditions exerted different effects on the outcome measures. As shown in Table 3, the differences were minimal. Among the 16 measures analyzed, only estimated calories differed, whereby those who saw the California Label estimated that the SSBs contained fewer calories than those who saw the Preventable Label. However, the California Label group performed better, although not significantly so, on most other measures. These results suggest that the label modifications did not detectably affect our primary outcome measures, and thus are unlikely to be consequential. As a result, the warning label conditions were collapsed for the main analyses, described below.
Do Warning Labels Affect Choices, Perceptions, and Intentions?
As shown in Table 4, putting warning labels on the 9 qualifying SSBs exerted powerful effects relative to both the No Label and Calorie Label conditions. In the vending machine choice task, those in the Warning Label group were significantly less likely than The Calorie Label or No Label groups to choose an SSB for their child. Calorie labels did not have a significant effect compared with the control condition.
Second, warning labels led parents to believe that SSBs were significantly less healthy, less likely to make their child feel energized, less likely to help their child to focus, and more likely to increase their child’s risk of weight gain, heart disease, and diabetes relative to both the Calorie Label and control groups. Although calorie labels significantly increased parents’ estimates of the calorie content of SSBs, warning labels did so as well, but to a lesser extent. Participants in the warning label condition judged SSBs to have more added sugar, and they indicated they were less likely to purchase them for their child. Willingness to pay did not significantly differ across groups. Finally, participants in The Warning Label condition chose significantly fewer SSB coupons than did those in The Calorie Label and No Label conditions.
Analyses of beverages that did not qualify for a warning label (averaged across the 5 beverages without warning labels) revealed only a few significant effects of warning label on disease risk and calorie estimation, all of which were very small, and possibly caused by anchoring on the values participants had given for SSBs (see Supplemental Table 5). Judgments of healthfulness and purchase intentions were unaffected.
Do the Effects Vary Across Levels of Education?
The Education × Warning Label interaction was barely significant for only 2 of the 16 outcomes, indicating that the warning label increased calorie estimates for SSBs (P = .02) and perceived diabetes risk (P = .05) more for those who were less educated. Education level did not moderate the effect of warning labels on other outcomes, including vending machine beverage choice (P = .93), number of SSB coupons chosen (P = .34), SSB purchase intentions (P = .32), and perceptions of SSB healthfulness (P = .950). The Calorie Label × Education interaction was nonsignificant for all outcomes.
Do People Support Warning Label Policies?
Most participants reported that a warning label would change their beliefs about a beverage’s healthfulness (mean = 3.8 ± 1.2 on 5-point scale) and that a label would encourage them to purchase fewer of the beverages for their child (mean = 4.1 ± 1.1 on 5-point scale). Second, 73.3% of participants were in favor of an SSB warning label policy, with only 5.7% opposed (the average support was +1.1 ± 1.0 on a scale from −2 to +2). This did not differ across experimental conditions. Although Republicans (72.9%) and Independents (66.0%) favored the policy less than Democrats (79.2%), the policy had strong majority support among all 3 parties.
The first aim of this study was to test whether warning labels can effectively educate consumers about the health harms associated with SSB intake more so than current industry practices of printing calorie labels on beverages. The study also aimed to evaluate how warning labels influence parents’ intentions to buy SSBs, as well as beverages that would not qualify for a label, whether these effects were moderated by education level, and whether the different label phrasings would be more or less effective. Finally, we assessed support for SSB warning labels.
We found that SSB warning labels may be an important way to educate parents about the health harms of SSBs and encourage them to purchase fewer of these beverages. Warning labels reduced parents’ perceptions that SSBs are healthy beverages and that SSBs can increase their child’s energy or ability to focus. We also found that the labels increased parents’ perceptions of the child’s risk of weight gain, heart disease, and diabetes from consuming SSBs. As predicted, calorie labels increased parents’ ability to estimate the calories in SSBs as did warning labels, but to a lesser extent. However, warning labels led parents to judge SSBs to have more added sugar. Analyses of beverages that did not quality for a warning label suggested that SSB warning labels are unlikely to have spillover effects, either positively or negatively, on judgments of nonlabeled drinks.
Three outcomes measured in this study indicated that warning labels may influence behavior. When asked to make an in-the-moment hypothetical purchasing decision for their child, caregivers who saw SSBs with warning labels were significantly less likely to choose an SSB relative to those who saw calorie or no labels on beverages. When parents’ were asked to select beverages for which they would want to receive coupons, those who saw warning labels chose significantly fewer coupons for SSBs than the control and calorie label conditions. Finally, warning labels led parents to report being less likely to purchase SSBs for their child in the future. These results suggest that when noticed, warning labels may encourage parents’ to purchase healthier beverages for their children, while current efforts to place calories per bottle information may have little influence. However, research examining SSB purchases among low-income adolescents revealed that brightly colored signs displaying calorie-related information was associated with a decrease in purchases of these beverages over 6 months,16 but we do not know whether such decreases would be more dramatic if the signs had included warning labels.
The influence of warning labels on the vast majority of outcomes did not vary based on education level, suggesting they may be helpful for people across the education spectrum. However, the impact of more traditional nutrition labeling strategies tends to vary based on demographic subgroups. For example, research on posting calorie information on restaurant menus finds that those who are more educated or have higher incomes are more likely to use the information when make purchasing decisions.17,18 Although we are not seeing meaningful differences in this study on the basis of education level, our results might be unique to lower education populations in online samples.
Contrary to our hypotheses, we did not find that modified label phrasings differentially impacted the outcomes of interest. This provides support for keeping the proposed California text in future bills, but future research should continue to explore whether other types of label phrasings can increase the influence of text warnings and whether different phrasings resonate with different populations (eg, parents versus teenagers).
Finally, participants strongly believed labels would help change their beliefs about a beverages’ healthfulness and would encourage them to purchase fewer SSBs for their children. The majority of respondents favored a policy to place warning labels on SSBs and although there were differences across political parties, the policy had strong majority support among all 3 parties, suggesting that SSB warning label proposals are unlikely to be met with strong constituent opposition. Labeling strategies typically garner more public support than more controversial food policies such as taxing SSBs or limiting their portion size.19
This study has several limitations. First, we studied the influence of warning labels via an online survey, not in the real world. However, given that such SSB labels do not exist in the real world, this is 1 of the first studies to look at the potential effect of such a policy. In addition, tobacco research suggests that labels are most likely to influence consumers if they are visible and salient, whereas more obscure text warning are less likely to have an impact.11 Therefore, we wanted to study warning labels under conditions where they are highly visible and salient to understand how they may impact consumers who see them. This means the study may have overestimated the effect of the warning label, but if we had found no effect, it would suggest that such labels would not be influential in real-world settings. The survey is also limited because of potential desirability bias. Consumers may be inclined to indicate they would not want to purchase an SSB because it is the desirable answer. However, consumers were completing the survey online and anonymously, likely reducing the desire to please the researcher. In addition, if there was a strong social desirability bias, we would expect to also see strong effects from exposure to salient calorie labels, but this did not happen. Although we have a large racially and ethnically diverse sample and we recruited so that our sample reflects the educational make-up of the United States, this does not mean we have a nationally representative sample. Therefore, these results may not generalize to other populations. In addition, we do not know how our study sample differs from those who opted not to take the survey. This study is also limited to parents making purchasing decisions for their children. Future research should assess the influence of warning labels on adults and adolescents. Finally, we tested the warning label guidelines on the basis of the California law so that the study could inform current policy debates. However, other laws might seek to include 100% fruit juices in labeling requirements, making it important to understand how warning labels would affect perceptions of those beverages.
This study has a number of strengths, including a large sample size, a randomized-controlled design with both a no label control group and a calorie label group, and a sample that included a range of education levels, as well as a large proportion of racial and ethnic minority participants. This study is among the first to examine the potential influence of SSB warning labels and provides timely data on the potential for such labels to educate consumers and reduce SSB intake. Although this study provides preliminary support for placing warning labels on SSBs, more research is needed to understand how they would influence a range of consumers and whether they would impact overall dietary choices. For example, 1 concern is that warning labels on SSBs would be ineffective at reducing overconsumption of calories and sugar because people would simply compensate by buying other high sugar foods that are unlabeled.
These results suggest that SSB warning labels are likely to reduce parents’ perceptions of SSBs’ healthfulness, increase perceptions of the health risks posed by SSBs, and decrease parents’ likelihood of buying SSBs for their children.
This work was commissioned by the Healthy Eating Research Program of the Robert Wood Johnson Foundation.
- Accepted October 27, 2015.
- Address correspondence to Christina A. Roberto, PhD, Department of Medical Ethics and Health Policy, 1105b Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by RWJF Healthy Eating Research.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Copyright © 2016 by the American Academy of Pediatrics