Discussion

Results presented in this study show the extent to which the French tax on sweetened beverages is known, perceived and supported by the French population in 2021, 3 years after its revision and 9 years after its initial implementation. Overall, a majority of respondents knew about the existence of the tax, although few of them had specific knowledge regarding the 2018 revision. The overall perception of the tax on sweetened beverages was favourable, particularly in regard to its ability to improve the health of the population. Support for the measure was high with almost two-thirds of respondents being in favour of it. Finally, multivariable analysis showed that support for the tax varied among subgroups of the population.

Overall, awareness of the tax on sweetened beverages among the population was high, with 63·4 % of respondents identifying the tax as existing, in line with the international literature on the subject. Several studies have focused on the awareness of similar taxes – one in the municipality of Berkeley (USA)^{(Reference Falbe, Thompson and Becker23)}, one in the municipality of Oakland and Berkeley (USA)^{(Reference Altman, Madsen and Schmidt24)} and one in Mexico^{(Reference Álvarez-Sánchez, Contento and Jiménez-Aguilar25)} – and have found mixed results (varying from 39 % in Oakland (immediately after its implementation), 65·2 % in Mexico (3 years after) to 68·2 % (1 year after), and 70 % (3 years after) in Berkeley). Local context plays an important role in making people aware of taxation schemes^{(Reference Claudy, Doyle and Marriott26)}, especially due to varying media coverage and communication campaign from public authorities or private actors. Additionally, one study^{(Reference Falbe, Thompson and Becker23)} found that tax awareness increased over time, which may further explain the variability in results. Very high levels of awareness obtained for other taxes, such as taxes on petrol or tobacco, tend to be in line with these elements as these taxes have been enforced for a longer time and have been covered extensively by the media. In contrast, non-existing taxes were incorrectly cited by a substantial and steady proportion of the respondents (about 30 %). Nevertheless, achieving higher rates of awareness is of importance as tax policies have a ‘signalling effect’, which may contribute to raising consumers’ awareness about the health consequences of consuming the taxed products^{(Reference Barigozzi and Villeneuve27)}. In Mexico, this effect led to a greater reduction in sweetened beverage consumption in groups aware of the tax^{(Reference Álvarez-Sánchez, Contento and Jiménez-Aguilar25)}. Additionally, it appeared that in our study, respondents had poor knowledge of the modifications made in 2018 and also of the calculation method. However, given that the questionnaire was administered 3 years after the revision of the tax and that specific knowledge questions required to be aware of technical details, it is not surprising that very few participants were able to answer correctly to these items. Furthermore, the French government, during the 2018 parliamentary debates, made clear that the main purpose of the revision of the tax was to incentivise food manufacturers to reformulate their recipes and thus reduce the sugar content in sodas. In this regard, the use of a progressive tax rate was indeed adequate as the lower the sugar content is, the lower the tax rate will be. Since consumer behaviour was not directly targeted with this revision, the lack of specific knowledge should not be regarded as a limit.

The perceived scope of the tax provided interesting insights into the different perceptions of beverage categories. For drinks containing added sugar (with the exception of fruit nectars), we observed a gradient according to the added sugar content with products containing the highest added sugar content (sodas and to a lesser extent still drinks) being the most reported as covered by the tax. These perceptions suggest that products with the greatest added sugar content were more frequently identified as sweetened beverages and thus are covered by the tax, whereas products such as flavoured waters or teas which contain in comparison less sugar were less often considered to be part of the scope of the tax. Interestingly, artificially sweetened beverages were identified as covered by the tax in somewhat similar proportions than sugary beverages. This result suggests that artificial sweeteners are perceived by a majority of the public in beverages as having somewhat similar effects as sugar, even though international recommendations mainly focus on the reduction of sugar-sweetened beverage consumption according to nutritional criteria⁽⁸⁾. Finally, fruit-based beverages, and especially fresh fruit juices, were among the less identified as being targeted by the tax as they tend to be considered as healthy^{(Reference Coelho28)}. Remarkably, fresh fruit juices and fruit juices from concentrates appeared with differing perceptions, while neither are targeted by the taxation scheme.

Public support for the taxation scheme in this study was relatively high when compared with previously reported data. A French study performed in 2015 in the NutriNet-Santé cohort found that 48·5 % of the population was in favour of the tax implementation following the launch of the tax in 2012^{(Reference Julia, Méjean and Vicari19)}. More globally, a recent systematic review^{(Reference Eykelenboom, van Stralen and Olthof14)} found an overall 42 % (CI 95 % = 39 %, 47 %) of the public supporting the implementation of a tax on sugar-sweetened beverages with great disparities between studies/countries (ranging from 27 % to 55 % of public acceptance). However, as pointed out by the authors, most studies were conducted in the USA, Australia and in the UK, which may make difficult to extrapolate these results to others countries as local context is of importance. In addition, very few studies in the systematic review were conducted after the implementation of a tax which may further make our results difficult to compare. Consistently with the body of literature, our findings also indicate that allocating tax revenues to specific health-promoting initiatives impacted positively the acceptance of the taxation scheme^{(Reference Eykelenboom, van Stralen and Olthof14)}. Unlike other jurisdictions which already implemented a tax on sweetened beverages^{(Reference Hagenaars, Jeurissen and Klazinga29)}, tax revenues of the French version of the tax are not dedicated to specific actions but rather to the overall healthcare system. In this regard, the French public health agency – HCSP – recommended using tax revenues to create targeted measures to incentivise consumers (e.g. through coupons or direct help) and food manufacturers (e.g. through subsidies) to adopt more sustainable practices⁽³⁰⁾. Additionally, our results indicate a clear rise in acceptance over time which has been supported by changes in perception. When comparing with the 2015 French study, we observed a slight increase in the proportion of the population who perceived the measure as useful (57·7 % in 2015 v. 60·2 % in 2021) and an important decrease in the proportion of the population who qualified the tax as unfair (49·3 % in 2015 v. 34·4 % in 2021). Perceived effectiveness and perceived fairness have been found to be positively associated with public acceptance^{(Reference Gunarathne, Spiller and Risius31)}. The 2018 revision of the tax appeared to be an opportunity for the government to reframe the policy as a public health measure, and these changes may have further led to an increase in acceptance^{(Reference Le Bodo, Etile and Julia32)}.

The policy development and media coverage of the tax on sweetened beverages may have also played a role in its perception. Indeed, the literature suggests that the social acceptance of taxes, such as soda taxes, is influenced by public debate and the political context. For example, in the UK, the soft drink industry levy – among other measures – gained traction as the ‘Responsibility Deal’ was found inadequate to solve nutritional issues^{(Reference Elliott-Green, Hyseni and Lloyd-Williams33)}. In France, although the 2012 version of the tax was adopted after fierce parliamentary debates^{(Reference Le Bodo, Etilé and Gagnon1)} and other nutritional policies are continuously considered⁽³⁴⁾, the implementation of food taxes continued to arouse the interest of stakeholders as shown by several parliamentary reports^{(Reference Le Bodo, Etile and Julia32)}. It should also be noted that when the bill was voted in 2017, the vote of the members of parliament was rather consensual and several factors may have played a role. First, this revision was an opportunity to clearly reframe the policy as a public health measure^{(Reference Le Bodo, Etile and Julia32)}. Then, in 2016, in the framework of the revision of national food-based dietary guidelines, the French National Agency for Food Safety published reports underlining the association between sugar-sweetened beverages consumption with weight gain and comorbidities⁽³⁵⁾. Finally, analyses showed that in 2017 media coverage was less intense and more favourable towards the measure when compared with the extensive coverage in 2011^{(Reference Le Bodo36)}. Private stakeholders, including the beverage industry, appeared to not be as opposed to a revision of the tax^{(Reference Le Bodo, Etile and Julia32)}, which may have contributed to a less heated debate. These elements may further contribute to explain the rise in acceptance and the overall better perception of the tax observed in the study.

The associations between tax acceptance and demographic characteristics highlighted disparities among subgroups of the population. We found that being a man increased the likelihood of supporting the taxation scheme. This result is surprising as numerous previous research suggested that there was no association with sex^{(Reference Eykelenboom, van Stralen and Olthof37–Reference Petrescu, Hollands and Couturier40)}. However, despite the lack of association reported in the past study conducted in the NutriNet-Santé cohort study in 2015^{(Reference Julia, Méjean and Vicari19)}, in univariate analysis, men were found to be significantly more prone to be in favour of the taxation scheme. Thus, the observed effect may be caused by residual confounding and it needs to be further explored in other studies on the French population. Then, the odds of supporting the tax were greater for respondents older than 45–55 years of age and lower for respondents younger than 45–55 years of age when compared with respondents aged 45–55 y/o. This could be interpreted by the fact that older individuals tend to be more health-conscious^{(Reference Kearney, Gibney and Livingstone41)} and thus would be more supportive of health prevention initiatives or by the fact that they tend to consume less sweetened beverages⁽⁴²⁾ or have greater financial resources⁽⁴³⁾ and thus would feel less affected by this kind of measure than younger individuals. Lastly, having at least one or more young children at home was associated with less acceptance of the tax, which was already highlighted in the previous study in the NutriNet-Santé cohort^{(Reference Julia, Méjean and Vicari19)}. However, in the Netherlands, among a representative sample of the Dutch population, no significant association between tax acceptance and household composition was found^{(Reference Eykelenboom, van Stralen and Olthof37)}. An explanation of these results could be that parents of young children engage less in health-promoting behaviours. Indeed, studies found that young parents (between 25 and 44 years old) spent less time doing physical activity^{(Reference Corder, Winpenny and Foubister44,Reference Pot and Keizer45)} . For diet, results are mixed with studies finding higher energy and saturated fat intake as well as higher BMI^{(Reference Corder, Winpenny and Foubister44,Reference Rattay and von der Lippe46)} in young parents v. non-parents and other studies finding no significant differences^{(Reference Laroche, Wallace and Snetselaar47,Reference Smith, McNaughton and Gall48)} . Finally, older parents (over 45 years old) appeared to engage more in behaviours favourable to health^{(Reference Rattay and von der Lippe46)}, which could explain why parents of adolescents in the study, which were on average older, were as likely as individuals without children to support the tax scheme, unlike younger parents. However, this hypothesis needs to be further explored as intended health behaviours may differ widely from actual behaviour, and both may interact in the acceptance of public health policies such as taxation.

Socio-economic variables also appeared to be important factors associated with tax support. Adults with a higher level of education were found to be more likely to be in favour of the tax scheme. It has been reported consistently in past research that taxes on sweetened beverages are more heavily supported by individuals with higher educational level^{(Reference Julia, Méjean and Vicari19,Reference Eykelenboom, van Stralen and Olthof37,Reference Miller, Dono and Wakefield38,Reference Gollust, Barry and Niederdeppe49)} . An explanation may be that education is associated with a higher level of nutrition knowledge^{(Reference De Vriendt, Matthys and Verbeke50)} and thus awareness of nutritional-related health issues. Additionally, greater income was associated with a higher level of support. Individuals with greater disposable resources were found to be less sensitive to price increases^{(Reference Muhammad, Meade and Marquardt51)}, whereas others may fear a reduction in their purchasing power with the implementation of new taxes. In turn, people who were overweight or obese tended to be less supportive of the tax than those with a BMI between 18 and 25 kg/m². Current body of literature has found inconsistent results with studies reporting no associations between weight status and tax acceptance^{(Reference Petrescu, Hollands and Couturier40,Reference Gollust, Barry and Niederdeppe49,Reference Sainsbury, Hendy and Magnusson52)} and other studies showing the decreased likelihood of support for overweight or obese people^{(Reference Eykelenboom, van Stralen and Olthof37,Reference Rivard, Smith and McCann53)} . Lower support for individuals with higher BMI observed in the study could be partly explained by the fact that the consumption was only assessed through a qualitative aspect. In other words, we did not make a distinction between low and high consumers, and BMI could have been a proxy for this dimension. A systemic assessment of beverage consumption in the world found that greater BMI was associated with greater sugar-sweetened beverage consumption^{(Reference Singh, Micha and Khatibzadeh54)}. All other things being equal, individuals with greater BMI would be more financially affected by the tax and thus potentially less supportive of the measure.

Consumption of sugar-sweetened sodas was found to be associated with less support for the tax. On the other hand, no significant effects were detected for artificially sweetened beverages or fruit juice consumption. Regarding the association between sugar-sweetened soda consumption and tax acceptance, several studies have identified a similar effect^{(Reference Eykelenboom, van Stralen and Olthof37,Reference Donaldson, Cohen and Rutkow55,Reference Richardson, Yanada and Watters56)} even though in the 2015 study in France, only sociodemographic factors played a role in tax acceptance. Artificially sweetened beverage and fruit juice consumption was rarely explored in the past as potential factors influencing tax support. As expected, individuals with a negative attitude towards sweetened beverages whether in regard to health or the environment were more favourable to a taxation scheme. The importance of personal opinion and perceived detrimental effects especially on health have been consistently shown in previous studies^{(Reference Miller, Dono and Wakefield38,Reference Pell, Penney and Hammond39)} . This finding reinforces the importance of public campaigns to raise awareness regarding the impact of consuming sugar-sweetened beverages, especially in regard to health, as these factors were found to have a higher OR than demographic or socio-economic determinants.

To our knowledge, this study is the first to explore public acceptance and perception since the 2015 study in the NutriNet-Santé cohort in a country that has already been enforcing a tax on sweetened beverages. In addition, we explored multiple dimensions (i.e. awareness, knowledge, perception and acceptance) which allowed us to identify important stakes when advocating in favour of a food tax. Several limitations should also be acknowledged. Given that the questionnaire was available for a limited amount of time (4 months) and that answering the questionnaire was optional, we aimed to collect at least 30 000 questionnaires before closing access. This procedure selected the more active members of the cohort, considering that the cohort is already voluntary. Thus, respondents were more likely to have a higher socio-economic status and educational level and to engage in health-conscious behaviours than the general population. In addition, to assess the representativeness of the final weighted sample, we compared sodas and juices consumption in the study with the national representative consumption study INCA 3⁽⁴²⁾. According to INCA 3, 41·1 % of the population consumed non-alcoholic soft drinks in 2017 which includes any sugar-sweetened beverages except fruit juices (v. 31·2 % for equivalent products in our study (weighted data)), and 50·3 % consumed fruit juices (v. 52·2 % in our study (weighted data)). Furthermore, the questionnaire was administered 3 years after the tax revision. Knowledge and perception may have been altered over time. Finally, as respondents were more likely to be aware of nutritional issues, answers may have been subject to social desirability bias, and we may have overestimated tax acceptance. Another concern might be that a self-reported online questionnaire with knowledge items could incentivise respondents to search for the correct answer on the Internet. However, we observed very high percentages of non-correct answers in questions regarding knowledge, which suggest that respondents answered spontaneously, as expressly asked by the instructions. We further limited the possibility of correction of wrong answers by preventing returns to previous questions.