High consumption of sugary drinks (or sugar-sweetened beverages (SSB)) is an important source of empty calories in people’s diets in the USA and globally( 1 ). Overconsumption of SSB has been linked to negative health outcomes, including obesity, chronic diseases like diabetes and CVD, and poor dental health( Reference Vartanian, Schwartz and Brownell 2 , Reference Malik, Popkin and Bray 3 ). As a result, policy makers and researchers alike have sought to reduce public consumption of SSB. Excise taxes on SSB have emerged as one potentially promising strategy( Reference Brownell and Frieden 4 ). An SSB excise tax was implemented in Mexico in early 2014, a ballot initiative to implement an SSB tax in Berkeley, California passed in 2014, and the Philadelphia City Council approved such a tax in 2016. Excise taxes on SSB have been introduced (without passage) in dozens of other US states and local jurisdictions and continue to appear on the public health policy agenda around the globe (with a planned tax in the UK, for instance, anticipated to go into effect in 2018). Traditional (i.e. neoclassical economic) theory suggests that raising the price of SSB will reduce their consumption, based on the price elasticity of demand( Reference Powell and Chaloupka 5 ). In fact, emerging evidence from an evaluation of the Mexico tax does suggest some declines in demand( Reference Colchero, Popkin and Rivera 6 ). While economic theory would suggest that the price increase is the primary mechanism through which consumption may be influenced by an SSB tax, literature in psychology and behavioural economics suggests that other factors also influence consumer decisions, such as how the price message is presented( Reference Rothman and Salovey 7 – Reference Tversky and Kahneman 9 ).
Public health researchers, advocates and policy makers have described several alternative justifications for SSB taxes, including their potential effect on improving diet quality or oral health, reducing obesity or health-care costs, protecting child health, and addressing state and municipal budgets( Reference Nixon, Mejia and Cheyne 10 , Reference Niederdeppe, Gollust and Jarlenski 11 ). For example, SSB tax advocates in California in 2012 emphasized the link between sugary drinks and negative health outcomes like obesity and type 2 diabetes( Reference Jou, Niederdeppe and Barry 12 ), advocates in Berkeley in 2014 emphasized negative industry actions( Reference Somji, Nixon and Mejia 13 ), while Mayor Jim Kenney in Philadelphia emphasized an SSB tax as an opportunity to fund social programmes( Reference Sanger-Katz 14 ). People who hear a health-related rationale may feel differently about the tax than those exposed to the argument that an SSB tax is necessary to reduce a city’s budget deficit. In fact, research shows that the public has varying levels of agreement with the major pro- and anti-SSB tax messages in public discourse( Reference Barry, Niederdeppe and Gollust 15 , Reference Donaldson, Cohen and Rutkow 16 ) and that SSB tax messages affect public beliefs about soda companies( Reference Niederdeppe, Gollust and Barry 17 ). These findings – combined with abundant research evidence across multiple disciplines that strategic communication can shape attitudes as well as behaviour( Reference Rothman and Salovey 7 – Reference Tversky and Kahneman 9 , Reference Bleich, Herring and Flagg 18 ) – suggest that messages about the rationale for an SSB tax could influence a variety of outcomes.
Messages about a price increase could influence consumption intentions directly or indirectly. For example, in addition to providing information about the price increase, messages could contribute to changes in public perceptions of beverage companies, the products or support for policies. These perceptions could shift norms surrounding sugary drinks and ultimately change the longer-term dynamics of SSB demand, a process that has been described for smoking and attitudes about cigarettes and tobacco companies( Reference Nathanson 19 ). However, effects are unlikely to be uniform. Individuals who regularly consume SSB may be more attentive to messages about a tax in the first place, or messages could affect consumers and non-consumers differently based on how relevant the message is to them or how much it resonates with their pre-existing knowledge, values and beliefs( Reference Haider-Markel and Joslyn 20 , Reference Chong and Druckman 21 ).
The objective of the present study was to examine whether rationales provided for a hypothetical SSB price increase influence young adults’ SSB purchasing intentions and their attitudes about SSB products, policies and the beverage industry compared with simply indicating a price increase. In addition, secondary objectives were to test the difference between messages naming a price increase as a ‘tax’ v. a ‘user fee’ and to compare message effects among individuals with different levels of SSB consumption.
Methods
Study participants
The study population was undergraduate students at the University of Minnesota, a large Midwestern US university, who were recruited to participate through a research programme managed by the psychology department. Participants were told that the study was about ‘Consumption of Snacks, Media and Beverages’ (questions about beverages were embedded among questions about their habits on purchasing snack foods and newspapers, magazines, movies and music, to de-emphasize the focus on beverages). A total of 543 students agreed to participate in the anonymous survey, exceeding our target (determined a priori based on power calculations) of at least 400. The final sample size was 494, after eliminating respondents who had missing data on more than half of the survey and anyone who did not complete the key independent variable (SSB beverage consumption). Participants received extra credit points for completing the survey, and they completed it between 25 April 2013 and 6 November 2013.
Study design and methods
A forty-six-item online survey was developed (after pilot testing for usability) in which participants were asked a series of questions about their beverage consumption (described below) and then were presented with a vignette about entering a convenience store on a hot day to purchase a beverage. Respondents were told that upon entering the store, they noticed a 15-cent price increase on their favourite 16-ounce SSB (based on respondents’ own report of their favourite such beverage; if they reported no favourite SSB, the vignette stated that the product subject to the price increase was ‘Coke’). This increase is about a 9 % increase based on the price (approximately $US 1·69) of 16-fluid-ounce (473 ml) beverages on and around this university’s campus. Following this vignette introduction, participants were randomly assigned to one of eight messages providing tax rationales (see Table 1 for the text of the messages): (i) no rationale was given; (ii) a new tax on the product; (iii) a new user fee on the product; (iv) a tax to raise revenue for state obesity prevention efforts; (v) a tax for state budget deficit improvement; (vi) a tax to offset health-care costs for oral health problems; (vii) a tax to offset the health-care costs for chronic health conditions; or (viii) a tax to protect children from harm. All participants were shown a consistent price increase of 15 cents to examine the effects of the rationale independent of the price increase. Messages (iv) through (viii) were based on past research exploring the arguments advocates have used to promote the tax( Reference Nixon, Mejia and Cheyne 10 , Reference Jou, Niederdeppe and Barry 12 ). In addition, given survey research showing low public support for SSB taxes( Reference Gollust, Barry and Niederdeppe 22 ), messages (ii) and (iii) were included to examine whether individuals respond differently if a price increase is described as a ‘user fee’ without the use of the word tax at all compared with ‘tax’( Reference Hardisty, Johnson and Weber 23 ).
Table 1 Study design and messages randomized to study participants
SSB, sugar-sweetened beverage.
Measures
The primary dependent variable was a measure of intention to purchase the SSB described in the vignette and measured immediately following the vignette: ‘Given this change in price, how likely are you to purchase [the beverage]?’ Response categories included: ‘much less likely to purchase’, ‘less likely to purchase’, ‘will not change purchase decision’, ‘more likely to purchase’ and ‘much more likely to purchase’.
Several secondary dependent variables that were also measured following the experimental vignette were also examined. Attitudes about sugary drinks were measured using a semantic differential scale (a measure where two words represent the opposite poles of a 7-point ordinal measure). Respondents were asked to rank sugary drinks on three dimensions. The item asked participants to ‘Think about the group of beverages with added sugar (e.g. non-diet soda, energy drinks, and other sugary drinks). In my opinion, these beverages are …’ Options included three scales: ‘very bad’ (= 1) to ‘very good’ (= 7); ‘very unappealing’ (= 1) to ‘very appealing’ (= 7); and ‘very unhealthy’ (= 1) to ‘very healthy’ (= 7). These three items formed an acceptably reliable scale (Cronbach’s α=0·68). Perception of SSB companies was measured using an item asking participants ‘How favourable is your impression about soft drink companies?’, with response options measured as an ordinal scale from 1 to 5 (‘very unfavourable’, ‘unfavourable’, ‘neither favourable nor unfavourable’, ‘favourable’ and ‘very favourable’)( Reference Gollust, Barry and Niederdeppe 22 ). Finally, participants were asked how much they would support or oppose a strategy that would ‘Require a penny-an-ounce tax on drinks with added sugar that would add 12 cents to the cost of a 12-ounce can of soda’, also measured as an ordinal scale from 1 to 5 (‘strongly oppose’, ‘oppose’, ‘neutral’, ‘support’ and ‘strongly support’)( Reference Gollust, Barry and Niederdeppe 22 ).
Participants’ reported beverage consumption was measured using the Beverage Intake Questionnaire (BEVQ-19), which was modified to focus on the drinks college students consume( Reference Hedrick, Comber and Estabrooks 24 ). Participants reported the number of times per week (or day, if they noted they drank that beverage daily) they drank thirteen different beverage types (see online supplementary material, Supplemental Table 1). A summary measure of consumption of SSB (defined as non-diet soft drinks, sweetened fruit drinks, sweetened tea, energy drinks and sports drinks/flavoured water( 25 )) was created by weighting the ordinal consumption scale to make all values scaled per week (e.g. 1 for once per week; 14 for twice per day, etc.), then summing these weighted consumption values across the five SSB categories, and then reconverting the aggregate scale back to an eight-category ordinal measure. For the subgroup analyses, SSB consumption was categorized into two groups: no SSB consumption or low (once weekly or less) compared with weekly or more.
Participants also self-reported demographic and social characteristics, including age, gender, work status, race, Hispanic ethnicity and parental household income.
Analysis
The initial step in the statistical analysis was to conduct one-way ANOVA to determine if purchasing intentions were different for any of the groups in the randomized conditions, the eight SSB price increase messages to which participants were exposed. Then, to examine whether there were any specific differences between the ‘no justification’ control group message and the specific SSB price increase messages, ordinary least-squares regression models were estimated of the primary outcome on the message variables with the ‘no justification’ message serving as the reference category. (Results were consistent when estimated with ordered logit regression, but ordinary least-squares regression is presented here for ease of interpretation of coefficients and the intercept.) Differences between the ‘SSB tax’ and ‘SSB user fee’ conditions were tested by comparing the regression coefficients using post-estimation Wald tests. Given random assignment to condition (and indeed, all observed characteristics reported in Table 2 were equally balanced across conditions), we did not include any covariates in the regression models, as per standard survey-based experimental methods( Reference Mutz 26 ).
Table 2 Descriptive characteristics of respondents (n 494) to an undergraduate student survey about a sugar-sweetened beverage (SSB) tax at a large US Midwestern university, April–November 2013
The same set of analyses was re-estimated for the secondary dependent variables (SSB product attitudes, company favourability perceptions and support for an SSB tax). To explore whether the messages affected participants’ outcomes differently based on their reported consumption of SSB, regression models with interaction terms by SSB consumption (once weekly or less compared with weekly or more) and experimental condition were estimated; where interactions were statistically significant, regression models were estimated within SSB consumption subgroups. All analyses were conducted using the statistical software package Stata version 13.
Results
Table 2 displays descriptive characteristics of the sample. The majority (88 %) were between the ages of 18 and 22 years (roughly equally divided across these age groups). The sample included more women than men, more students who were working compared with not working, and three-quarters of the sample was white. About 30 % reported that their parents earned less than $US 65 000 per annum. Study participants skewed more female than the university student body more generally, but the racial/ethnic breakdown was similar to the broader demographics of the university.
Just under 12 % of the sample reported consuming no SSB; 18·2 % consumed a little (once weekly or less); 33·2 % consumed a moderate amount (more than once weekly but less than daily); and 36·6 % consumed a high amount of SSB (daily or more).
Message effects on sugar-sweetened beverage purchase intentions
Message type was significantly associated with differences in participants’ purchasing intentions (F 7,485=2·53, P=0·014). To examine differences between the control group and the specific rationales to which participants were exposed, Table 3 displays estimates from regression models of purchase intention by condition for the full sample. These results indicate that within the full sample, a message describing a user fee, a message describing the tax goal to reduce obesity, a message describing the tax goal to offset chronic health-care costs (but not oral health-care costs) and a message to protect children were each associated with a reduction in participant intentions to purchase the SSB compared with a message indicating a price increase with no justification. The ‘user fee’ condition was associated with lower purchasing intentions than the ‘tax’ condition, but this difference was not statistically significant at conventional levels (F 1,485=3·18, P=0·08). There were also no statistically significant differences in message effects by level of SSB consumption (F 7,477=0·44, P=0·88).
Table 3 Effect of sugar-sweetened beverage (SSB) price increase rationales on SSB purchasing intentions and favourability towards soda companies among respondents (n 494) to an undergraduate student survey about an SSB tax at a large US Midwestern university, April–November 2013
Estimates are coefficients (β) and their standard errors from ordinary least-squares regression models. The reference category is the group that received a message describing a price increase with no justification.
*P≤0·05, **P≤0·01, ***P≤0·001.
Message effects on attitudes, perceptions and opinions
For the full sample, the mean level of favourable attitudes about sugary drinks (on a 7-point scale) was 3·2 (sd=1·1), mean level of favourability towards soda companies (on a 5-point scale) was 2·6 (sd=0·9) and mean level of SSB tax support (on a 5-point scale) was 3·0 (sd=1·2). No significant differences by message were observed for the sugary drink attitude measure (F 7,485=0·54, P=0·81) or the soda tax policy support measure (F 7,484=1·26, P=0·27). However, message type was associated with perceptions of soda companies (F 7,485=2·10, P=0·043). In particular, Table 3 shows that in the full sample, messages describing the price increase as a user fee or as a tax to reduce obesity led to less favourable attitudes about soda companies, compared with a message describing a price increase with no justification. While the ‘user fee’ condition was associated with more negative perceptions of soda companies than the ‘tax’ condition, again this was not statistically significant (F 1,485=3·39, P=0·07).
Message effects on soda company perceptions by level of sugar-sweetened beverage consumption
Regression models fitted with interaction terms by level of SSB consumption indicate that the message had different effects depending on whether the participant had lower or higher SSB consumption (F 7,477=2·26, P=0·03). Table 3 displays message effects among consumption subgroups. Within the sub-sample of respondents who reported low SSB consumption, a message describing a user fee applied to SSB or a tax devised to reduce obesity led to less favourable perceptions about soda companies compared with a message just describing a price increase. Within the low SSB-consuming group, the effect of the user fee message was significantly different from that of the SSB tax message (β=−0·69 v. β=0·03, F 1,140=6·30, P=0·01). None of the messages had a significant effect on perceptions of soda companies among those consuming SSB more than once per week.
Discussion
Economic and evaluation studies presume that the predominant or only effect of a tax increase on SSB consumption is through the change in the product’s price, shifting consumers’ willingness to purchase the product compared with alternatives( Reference Powell, Chriqui and Khan 27 ). The results of the present research suggest that there could be another mechanism through which a tax has effects – via how that tax is justified in the public discourse. The present results show that the message justifying a price increase can influence intentions to purchase a sugary beverage and perceptions of soda companies; and that messages have variable effects on soda company perceptions depending on the message recipient’s level of SSB consumption. These results add to previous research showing differences in public response to SSB tax rationales( Reference Niederdeppe, Gollust and Barry 17 ).
By demonstrating variation in the effectiveness of the many ways of presenting SSB taxes that advocates and policy makers use, our findings offer important implications for SSB tax policy making, especially for advocates considering the appropriate pro-tax messaging approach for a particular jurisdiction. Specifically, research in communication about the environment( Reference Hardisty, Johnson and Weber 23 ) has indicated the potential promise of changing how an SSB tax is presented to remove the controversial and politically charged term ‘tax’. In fact, the advocacy organization Coalition for a Healthy California supports a ‘health impact fee’ on sugary drinks (this is the term it is using to describe a price increase of 2 cents per fluid ounce, as introduced in a bill in California in March 2016)( 28 ). The present study’s findings offer weak support for the idea that people respond differently to these terms; a ‘user fee’ term in our study was associated with more negative perceptions of soda companies than the ‘tax’ term. However, at the time of the present study in 2013, the idea of a ‘fee’ applied to SSB was not currently debated, so the phrase may have struck study participants as odd.
The present study also provides evidence that messages that link the SSB tax to specific health outcomes may be more effective than those that describe the tax as a method to raise revenue. This finding is important in the current policy context, since the main pro-tax message debated in Philadelphia as of spring 2016 avoided connecting the tax to obesity prevention and framed the tax instead as a tax on the industry to generate money for social priorities (in this case, for funding early childhood education)( Reference Sanger-Katz 14 ). The message in the present study that connected the soda tax to obesity reduction also led to less favourable perceptions of soda companies among those who are low SSB consumers. Public health advocates planning future SSB tax campaigns should consider that SSB tax rationales (particularly those that invoke ‘Big Soda’, the term used by advocates to refer to the soda industry) may resonate differently with those who do or do not drink SSB.
Limitations
These results and their implications must be interpreted with some key limitations of our study design and sample in mind. First, participants were undergraduate students at a single university, so they are not representative of all young adults. Since young adults are the highest consumers of sugary drinks( Reference Han and Powell 29 ), this population is an appropriate one within which to examine potential interventions that could reduce consumption. In addition, with about 12 % of our study sample never consuming SSB, one-third consuming at least weekly but less than daily and almost 37 % consuming daily, our sample’s SSB consumption is on par with state and national estimates( Reference Park, Onufrak and Sherry 30 , Reference Park, Pan and Sherry 31 ). A national survey in 2010 found, for instance, that 29·4 % of 18- to 24-year-olds consumed no SSB, 32·8 % consumed SSB at least once weekly and 37·8 % consumed SSB at least daily( Reference Park, Onufrak and Sherry 30 ). However, our sample is not representative of young adults in other dimensions, such as education and racial/ethnic diversity. The study’s goal is not to make generalizable statistical estimates about a particular population; with an experimental study design, the study maximizes internal validity (the ability to make a causal inference that a given message shifts a given attitude or belief) at the expense of some external validity.
Second, the study leverages an online experiment, asking participants to imagine a hypothetical purchasing scenario. The key dependent variable is a self-reported intention, not an observed behaviour. A study design which exposes participants to an actual price change (such as studies measuring the effect of price changes on sugary drink consumption in real stores( Reference Block, Chandra and McManus 32 – Reference Jue, Press and McDonald 34 )) and observes actual purchasing behaviour will be required to confirm that these results hold up in naturalistic settings. However, given that there were eight different messages reflective of the actual policy discourse around SSB price increases that we wanted to test, the online survey mode is an appropriate first step to examine message effects.
Finally, in the present study, the price change was kept constant at 15 cents (roughly a penny per fluid ounce). Yet, the effects of a message might interact with the magnitude of the price increase, with, for instance, certain message–price combinations more effective in shifting behaviour. Future research might test the most promising messages revealed in the current study and simultaneously manipulate other price scenarios to further advance our understanding of the many price-related and non-price related influences on consumer behaviour.
Conclusion
Policies to reduce public consumption of SSB are urgently needed worldwide, especially given the emergent science surrounding the public health harms of sugar in excess( Reference Lustig, Schmidt and Brindis 35 ). It is important to consider the complex and intersecting mechanisms through which such policies might have an influence on the public. A growing body of research suggests that economic incentives may be as important for changing behaviour as shifts in norms and social beliefs( Reference Kraft-Todd, Yoeli and Bhanot 36 ). The findings from the present study suggest that SSB taxes may influence the public not only from the price increase( Reference Waterlander, Mhurchu and Steenhuis 37 ) but also from the language used to justify such a policy. Additional research – in experimental and observational settings in the USA and other international contexts – should continue to evaluate the many intersecting mechanisms through which a heightened nutrition policy focus on reducing SSB consumption will affect the public’s attitudes and behaviours.
Acknowledgements
Acknowledgements: The authors thank AnnaMarie Vu for her research assistance during survey development and implementation. Financial support: This research was supported by the University of Minnesota Obesity Prevention Center and the University of Minnesota Healthy Foods, Healthy Lives Institute. The funders had no role in the design, analysis or writing of this article. Conflict of interest: The authors have no conflicts of interest to disclose. Authorship: S.E.G., J.M.W., C.F.R. and A.J.R. conceived of the study, designed the survey and collected the data. X.T. conducted the data analysis along with S.E.G. All authors provided data interpretation, critical revisions of the manuscript and approved the final version to be submitted. Ethics of human subject participation: This study was determined to be exempt from human subjects review (exemption category 2).
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S1368980016001816
