Eating from out-of-home food outlets is common, is expected to increase globally over the next decade(1), and is associated with poorer dietary quality, increased energy intake and obesity(Reference Lachat, Nago and Verstraeten2–Reference Gesteiro, García-Carro and Aparicio-Ugarriza4) Eating food from out-of-home food outlets may lead to weight gain due to larger portion sizes and greater energy density of food from these outlets, which may cause consumers’ energy intake to exceed their energy requirements(Reference Lachat, Nago and Verstraeten2). Furthermore, previous work has found both experts and the general public tend to be poor estimators of their energy intake from restaurants(Reference Livingstone and Pourshahidi5,Reference Block, Condon and Kleinman6) .
Historically, there have been fewer regulations on labelling the nutritional content of foods purchased at out-of-home food outlets compared to those purchased in grocery stores. One policy response to help inform consumers about the nutrition content of out-of-home eating is to include energy labels on menus. Mandatory energy labelling policies may improve diets through various pathways including informing consumers about the energy content of food options to help them make a more informed selection, shifting food choices towards healthier options, and incentivising the food industry to offer reduced energy versions of their offerings, via reformulation and reduced serving sizes, or introducing new products(Reference VanEpps, Roberto and Park7,Reference Kaur, Briggs and Adams8) . Menu labelling policies are also thought to be cost effective population-level interventions to improve diets, reduce obesity and prevent associated chronic diseases(9). Several recent meta-analyses of evaluations of menu labelling interventions found that, although study quality tends to be mixed, energy labels may lead to small reductions in energy intake among adults at a population level, and energy labelling may reduce the amount of energy consumers purchase from restaurants(Reference Crockett, King and Marteau10–Reference Littlewood, Lourenço and Iversen12).
Although there is a vast literature on the effects of menu labelling on behaviour, to our knowledge no studies have examined the effects of implemented menu labelling policies in a multi-country context. Several reviews have been published examining the impact of menu labelling policies in ‘real world’ and laboratory settings(Reference VanEpps, Roberto and Park7,Reference Crockett, King and Marteau10,Reference Littlewood, Lourenço and Iversen12–Reference Cantu-Jungles, McCormack and Slaven18) . However, the majority of studies are small randomised controlled trials, include populations only from the USA or were implemented in small settings such as a university cafeteria. Evidence from the USA suggests energy labelling can lead to small improvements in fast food meal quality and small-to-moderate decreases in energy purchased from supermarkets and fast food restaurants, but reductions in purchases may diminish over time(Reference Petimar, Zhang and Rimm19–Reference Petimar, Grummon and Zhang21). Many studies also lack a comparison group to examine the effects of menu labelling policies(Reference Fernandes, Oliveira and Proenca22) and those that do are limited to analyses of individual food service chains or examine policy in individual cities, potentially limiting generalisability of findings(Reference Elbel, Mijanovich and Dixon23–Reference Downs, Wisdom and Wansink25). Although a randomized controlled trial (RCT) is typically assumed to have less risk of confounding than observational studies, it is essential to also understand the effects of policies implemented in the real world, particularly in a large state-wide or even national setting, where ‘real world’ effectiveness may differ from efficacy in a RCT. In the context of national-level diet surveys, many are conducted too infrequently to be compared with other countries during the same time period or they may capture different diet-related behaviours, which limits comparability(Reference Micha, Coates and Leclercq26,Reference Hammond, Vanderlee and White27) . Thus, a multi-country approach to evaluating the impacts of food policies addresses current gaps in national monitoring surveys(Reference Hammond, Vanderlee and White27).
Mandatory menu labelling has been implemented in national and subnational jurisdictions(Reference Sofía Rincón-Gallardo, Zhou and Gomes28), and this study presents a population-based evaluation to clarify the impacts of real world menu labelling policies. We utilised data from 2019 of the International Food Policy Study (IFPS), a multi-country repeated cross-sectional survey of five upper- and middle-income countries including Australia, Canada, Mexico, the United Kingdom, and the USA. The IFPS allows for comparisons of polices in countries or jurisdictions that have implemented compared to those that have not implemented(Reference Hammond, Vanderlee and White27). The five countries in the IFPS have varying mandatory menu labelling regulations, with some policies mandatory at the national levels, others at the state/province level and others with no mandatory menu labelling regulations. Thus, this multi-country survey includes large populations that were and were not exposed to mandatory energy labelling regulations at the time of data collection.
The purpose of this study was to examine the prevalence of noticing and using menu labels and the behaviours associated with menu labelling overall and by sociodemographic characteristics, comparing jurisdictions with and without mandatory menu labelling policies. The first research question was whether there were any significant differences in these behavioural outcomes according to policy status. We hypothesised that jurisdictions with mandatory menu labelling policies would have higher rates of noticing and use of menu labels compared to jurisdictions without. The second research question was whether differences by policy status varied for sociodemographic groups. Given the high agency requirement of menu labelling policies, we hypothesised that variations would exist across sociodemographic groups.
Methods
Dataset
Data are from the 2019 wave of the IFPS, a multi-country repeat cross-sectional study of eating patterns and policy-relevant behaviours, and include data from Australia, Canada, Mexico, the United Kingdom and the USA(Reference Hammond, Vanderlee and White27,Reference Hammond, White and Rynard29) . These countries were selected for the IFPS survey because of differences in food-related policies prior to the first wave and the potential for change in policies between subsequent waves (Table 1). The study sample was recruited from Nielsen Consumer Insights Global Panel, which provides standardised recruitment sampling across countries. A random sample of participants aged 18–99 from the Nielsen Consumer Insights Global Panel and their partners’ panels were invited by email to complete the IFPS survey(Reference Hammond, White and Rynard29). Online surveys were completed between November and December of 2019. Questionnaires and further details about recruitment are available from the IFPS: Technical Report 2019(Reference Hammond, White and Rynard29).
Table 1 Categorisation of jurisdictions according to presence or absence of mandatory menu labelling policies before 2019 data collection(Reference Sofía Rincón-Gallardo, Zhou and Gomes28)
Exposure
Policy status was treated as an indicator variable for either Policy Present or No Policy in the analysis, and Policy Present was defined as having a mandatory menu labelling policy in place during 2019. Table 1 shows the jurisdictions included in this study that did and did not have a mandatory menu labelling policy implemented during 2019. The Policy Present group includes the USA and jurisdictions of Australia and Canada with mandatory labelling regulations (Table 1). In April 2022, England introduced mandatory calorie menu labelling for large out-of-home food businesses, defined as those with more than 250 employees(30); however, at the time of the data collection for this study, England and Mexico had not implemented menu labelling requirements and served as ‘comparison’ conditions. The No Policy group includes Mexico, the United Kingdom, and segments of Australia and Canada without mandatory labelling regulations (Table 1). We separated regions with mandatory labelling policies from areas without in both Canada and Australia. Participants in Canada answered ‘What province or territory do you live in?’ and participants in Australia answered ‘What state or territory do you live in?’ For Canada, responses of Ontario were coded to Policy Present, and all provinces other than Ontario were coded to No Policy (Table 1). For Australia, responses of Australian Capital Territory, New South Wales, Queensland, South Australia and Victoria were coded to Policy Present, and responses of Western Australia, Tasmania, Northern Territory were coded to No Policy (Table 1).
Outcomes
There are myriad ways in which consumers make food-related decisions. For example, contemporary behaviour change theory conceptualises behaviour as a result of interacting capability, opportunity and motivation(Reference Willmott, Pang and Rundle-Thiele31). Price, taste and convenience are also key factors in making food decisions. Other potential psychological mechanisms are involved in eating behaviour status quo bias – people eat what is typical and available such as large restaurant portion sizes – simplicity and energy compensation(Reference Roberto and Kawachi32). The conceptual framework used in the present study assumes in order to make eating decisions, nutrition information must be noticed, then used, and finally used in a particular way. Previous work has examined the rates at which consumers notice and use nutrition information(Reference Vadiveloo, Dixon and Elbel24). This study examines several self-reported outcomes related to how mandatory menu labelling policies are theorised to affect behaviours associated with menu labelling. Outcomes measured were as follows: noticing nutrition labels, use of nutrition labels, ordered something different, ate less of the food they ordered, visited different restaurants or ate at restaurants less often. These measures, as well as sociodemographic characteristics, are defined in Table 2, including the survey questions and coding for the analysis. Responses to noticed nutrition information and used nutrition information questions refer to the last time the participant visited a restaurant. Reponses to the behavioural impact of labelling questions refer to behaviours that occurred within the last 6 months and were preceded by the question ‘In the past 6 months, have you done any of the following because of nutrition information in restaurants? (Select all that apply)’ (Table 2). These measures were adapted from previously validated measures and published research(Reference Goodman, Vanderlee and White33).
Table 2 IFPS 2019 survey questions and variable categorisation
Sociodemographic characteristics
Sociodemographic characteristics including age, sex, education, income adequacy and ethnicity were included as potential confounders in models. The wording, responses and categories used in analysis of covariates are described in Table 2. Age was categorised into 10-year age brackets, except for the youngest group which included participants aged 18–24 years (Table 2). Because this is a multi-country survey with diverse ethnicities, the most comparable ethnicity measure across all countries was comparing the majority ethnicity to combined minority ethnicities. For income, we used income adequacy as it is associated with economic resources and health and allows for comparability across the multiple countries of the IFPS(Reference Litwin and Sapir34).
Statistical analysis
All statistical analyses were conducted using Stata, Version 16. Data were weighted with post-stratification sample weights constructed using a raking algorithm with country-specific population estimates from census data based on age group, sex, region, ethnicity (except in Canada) and education (except in Mexico). A detailed explanation of survey weights can be found at http://foodpolicystudy.com/methods (IFPS: Technical Report 2019). Sample weights were used throughout the analysis to minimise the influence of differential non-response and selection bias on the representativeness of findings.
There were 20 968 observations in the dataset. Four of the six behavioural survey questions asked about consumer behaviour at restaurants within the previous six months. Therefore, we restricted our sample size to only those participants who visited a restaurant in the previous six months, reducing the sample size to 19 617. Ethnicity data was missing for 176 observations, and a further forty-eight observations were missing education data and were dropped from the analysis, leaving a complete case analysis sample size of 19 393.
Descriptive statistics were used to summarise the outcomes and sociodemographic characteristics of the sample by policy status. To assess whether there were any significant differences in the six binary outcome measures according to policy status, weighted estimates were calculated using a survey-adjusted logistic regression model for each outcome. Policy status was included as an indicator variable (0 = no policy, 1 = policy), and models were adjusted for covariates selected a priori: age, sex, education, income adequacy and race/ethnicity. The differences by policy status for behavioural outcomes were reported as OR with 95 % CI, and statistical differences between policy status groups were tested using Wald tests. Results are also presented as predicted probabilities for all behavioural responses calculated using the margins command in Stata(Reference Williams35), as marginal effects can aid interpretation of magnitude and are more comparable across populations than OR(Reference Mood36,Reference Norton, Dowd and Maciejewski37) . Predicted probabilities are the probability that the outcome will occur, estimated by the model. Differences in predicted probabilities were calculated using pairwise comparisons of margins(Reference Norton, Dowd and Maciejewski37,38) .
To assess whether differences in outcomes by policy status varied across sociodemographic groups, we added two-way interactions between policy and each sociodemographic variable of interest to logistic regression models. Predicted probabilities of all six outcomes were estimated for each level of demographic variable by policy status. Differences by policy status at all levels of each sociodemographic variable were tested using pairwise comparisons of margins(Reference Norton, Dowd and Maciejewski37,38) .
Results
Table 3 describes the sample characteristics, stratified by policy status for the 19 393 participants analysed. The majority of the sample reported high education, majority ethnicity and not easy income adequacy (i.e. not easy to make ends meet). The distribution of education varied by policy status, with more participants reporting High education in No Policy jurisdictions and more participants reporting Low and Medium education in Policy Present jurisdictions (Table 3).
Table 3 Sample demographic characteristics by policy status (unweighted n; weighted %)
Noticing and using nutrition information and changes in behaviours by policy status
Participants in jurisdictions with policies were more likely to notice nutrition information compared to jurisdictions without policies (OR = 1·67 (95 % CI 1·53, 1·83)). The predicted probability of noticing nutrition information was 21·2 % (20·2–22·1 %) in jurisdictions with mandatory policies compared to 13·9 % (13·1–14·7 %) in jurisdictions without mandatory menu labelling policies, a significant difference of 7·3 % (6·0–8·6 %) (P < 0·001). Participants in jurisdictions with policies were more likely to use nutrition information compared to jurisdictions without policies (OR = 1·56 (95 % CI 1·38, 1·76)). The predicted probability of used nutrition information was 10·6 % (9·9–11·3 %) in jurisdictions with mandatory policies compared to 7·1 % (6·5–7·7 %) in jurisdictions without mandatory menu labelling policies, a significant difference of 3·5 % (2·5–4·4 %) (P < 0·0001) (Table 4).
Table 4 Predicted probability weighted estimates for noticing, using and behaviour change from menu labels by policy status in 2019 (n 19 393)
All models adjusted for age, sex, education, ethnicity, income adequacy.
Bolded values are statistically significant at the 0·05 level.
Participants in jurisdictions with policies were more likely to order something different compared to jurisdictions without policies (OR = 1·17 (1·07, 1·28)). The predicted probability of ordering something different as a result of nutrition information was 20·2 % (19·3–21·2 %) in jurisdictions with mandatory policies compared to 17·9 % (17·0–18·7 %) in jurisdictions without mandatory menu labelling policies, a significant difference of 2·3 % (1·0–3·7 %) (P = 0·0005). Participants in jurisdictions with policies were more likely to eat less of their order compared to jurisdictions without policies (OR = 1·37 (1·23, 1·52)). The predicted probability of ate less of order was 13·2 % (12·4–14·0 %) in jurisdictions with mandatory policies compared to 10·1 % (9·4–10·8 %) in jurisdictions without mandatory menu labelling policies, a significant difference of 3·1 % (2·1–4·2 %) (P < 0·0001). Participants in jurisdictions with policies were more likely to change restaurant visited compared to jurisdictions without policies (OR = 1·18 (1·04, 1·35)). The predicted probability of changed restaurant visited was 8·1 % (7·5–8·7 %) in jurisdictions with mandatory policies compared to 7·0 % (6·4–7·6 %) in jurisdictions without mandatory menu labelling policies, a significant difference of 1·1 % (0·2–2·0 %) (P = 0·013) (Table 4). There was no significant difference in the odds of ate at restaurants less often between jurisdictions with and without mandatory menu labelling policies. We also examined the differences in outcomes by country and policy status descriptively, finding similar patterns of greater noticing, use and behavioural outcomes associated with menu labelling in Policy Present jurisdictions (online Supplementary Table 1).
Interaction results: differences in behaviour by sociodemographic characteristics
Next, we examined whether differences between policy groups varied by sociodemographic characteristics. Only those models with significant interactions are presented below in Tables 5–7. For noticed nutrition information, the greatest difference between policy groups was seen for 55–64 year olds (difference of 12·6 %, 95 % CI 9·9, 15·4; P < 0·001) and 65–74 year olds (difference of 9·6 %, 95 % CI 6·5, 12·8; P < 0·001) (Table 5). These differences were primarily due to lower rates of noticed nutrition information for those age groups in No Policy jurisdictions. There was a significantly greater difference between policy groups for high education (10·7 %, 95 % CI 8·9, 12·6) compared to low education (4·1 %, 95 % CI 1·8, 6·3; P < 0·001) participants (Table 5). For use nutrition information, the greatest differences between policy groups were again for the oldest groups: with a difference of 4·9 % (95 % CI 2·9, 7·0; P < 0·001) for 55–64 year olds and a difference of 4·8 % (95 % CI 2·7, 6·8; P < 0·001) for 65–74 year olds (Table 5). There was a significantly greater difference between policy groups for high education (4·9 %, 95 % CI 3·5, 6·4) compared to low education (1·8 %, 95 % CI 0·2, 3·5) participants (P = 0·006) (Table 5).
Table 5 Predicted probability weighted estimates – % (95 % CI) – for models tested with interaction between policy and sociodemographic variables
Table 6 Predicted probability weighted estimates – % (95 % CI) – for models tested with interaction between policy and sociodemographic variables
Table 7 Predicted probability weighted estimates – % (95 % CI) – for models tested with interaction between policy and sociodemographic variables
For ordered something different, the differences between policy groups were directionally different for the majority ethnicity group (difference of 4·7 %, 95 % CI 3·3, 6·1; P < 0·001) compared to the minority ethnicity group (difference of –4·8 %, 95 % CI –8·0 to, 1·6; P < 0·001). These differences were primarily due to high rates of ordering something different for minority groups in No Policy jurisdictions (Table 6).
For changed restaurant visited, the difference between policy groups was greater for medium (difference of 3·1 %, 95 % CI 1·4, 4·8, P < 0·001) and high education (difference of 2·3 %, 95 % CI 1·0, 3·6, P < 0·01) compared to low education (difference of –0·9 %, 95 % CI –2·4, 0·6) (Table 7). The difference between policy groups was greater for higher income adequacy (difference of 2·9 %, 95 % CI 1·4, 4·4) compared to lower income adequacy (difference of 0·3 %, 95 % CI –0·8, 1·4) (Table 7).
Discussion
To our knowledge, this is the first multi-country examination of national and state/province-level menu labelling policies and associated behaviours. In this online multi-country survey with 19 393 participants conducted in 2019, we find evidence that implementation of mandatory menu labelling in restaurants is associated with a range of behaviours that are on the pathway from exposure to mandatory menu labelling to change in individual purchasing and eating. When differences by sociodemographic factors were present, the greatest differences were seen in those of middle to older age and those with greater socio-economic affluence according to education or perceived income adequacy.
Interpretation and implications of findings
Our study builds on previous work measuring noticing and using nutrition information with a multi-country comparison of jurisdictions where energy labelling on menus was mandatory compared to where it was not. Jurisdictions with mandatory nutrition labelling policies had higher rates of noticing nutrition information, ordering something different, eating less of what was ordered and changing restaurants due to nutrition information availability. However, there were no differences between policy groups in frequency of eating out at restaurants. This suggests menu labels may affect behaviour within restaurants, but do not affect consumers’ decision of whether or not to eat at a restaurant.
Most importantly, these results suggest that mandatory menu labelling policies may be improving behaviours associated with menu labelling at restaurants when comparing jurisdictions with and without mandatory menu labelling policies. These changes are according to our proposed mechanism involving noticing, using and types of use all leading to changes in energy consumed. There is evidence for the link between noticing labels and behaviour change. For example, noticing other types of labels such as traffic light labels has been found to be associated with healthier items purchased(Reference Sonnenberg, Gelsomin and Levy39). Mandatory menu labelling policies have increased noticing and use of nutrition information in other contexts, but more evidence is needed to understand whether these findings are consistent for older age groups and in other countries and population subgroups(Reference Goodman, Vanderlee and White33). More recent evidence from the USA suggests the small-to-moderate reductions in energy purchased may diminish over time, potentially reducing the long term public health impact of energy labels. The present study found significant differences between mandatory and non-mandatory menu labelling jurisdictions for five out of six behavioural outcomes measured, which could potentially lead to improved diets across large populations.
Although noticing and use of nutrition information was greater in jurisdictions with mandatory labelling policies, estimates were relatively low (21·2 % noticing and 10·6 % using nutrition information), and the differences in behaviours associated with menu labelling are modest. However, these differences could still be meaningful for health when they include millions of people. To improve public health, there may be ways to augment the effects of menu labelling policies. First, menu labelling interventions may be optimised by further helping people notice nutrition information – for example by making the information more prominent via increased size or visual salience(Reference Peschel, Orquin and Mueller Loose40). Second, interventions could help people use nutrition information by including associated messaging such as choosing an option with lower energy content to benefit health. For example, evaluative labels may be easier to interpret than numerical labels(Reference Thorndike, Sonnenberg and Riis41), and adding a recommended daily energy intake alongside menu labels maybe increase their effects(Reference Roberto, Larsen and Agnew42). Previous work has also found that motivation to use nutrition information may be a more important barrier than mere nutrition knowledge(Reference Grunert, Fernández-Celemín and Wills43). Third, additional policies are needed to have a large impact on diets across populations. Mandatory menu labelling may be a component of an effective obesity reduction strategy, but it is unlikely to achieve government targets to reduce obesity without complementary policies as eating behaviours are influenced by numerous complex factors beyond individual decision-making processes.
Although our study suggests mandatory menu labelling policies may play a role in reducing energy consumption out of home, other mechanisms of action may have more important effects, and mandatory labelling policies alone may not be enough to greatly reduce energy intake out of home across large populations. Additional messaging about how to use energy information alongside mandatory menu labelling policies may augment consumers noticing and ability to use this information to make healthier food choices when eating outside the home. People may also lack guidance regarding how to understand and use nutrition information to eat healthier and ultimately improve their health. Some jurisdictions such as New York City have tried to supplement mandated energy information posted on chain restaurant menus by adding recommended energy intake per day or per meal, but with no effect(Reference Downs, Wisdom and Wansink25). Future work is needed to determine whether other policies that reduce energy consumed out of home are more effective than menu labelling, and more research is needed to understand whether the sociodemographic differences in self-reported behaviours found in this study also exist for dietary intake. Menu labelling may also spur reformulation of products to lower energy or healthier forms by reducing nutrients of concern. However, energy labelling in large chain restaurants was associated with minimal changes in energy content of menu items, primarily consisting of the introduction of new lower energy items(Reference Grummon, Petimar and Soto44).
Differences between policy groups by sociodemographic characteristics
In addition to estimating the differences between policy groups for each of the six behavioural outcomes, we found several differences between policy groups by demographic characteristics. Examining differences between policy groups by age, the youngest age group, aged 18–24 were the mostly likely to eat less of their order compared to other age groups. Among United Kingdom diners at catering establishments, younger groups were more interested in menu labelling than older groups(Reference Mackison, Wrieden and Anderson45), and this greater interest could translate into greater use. On the other hand, the greatest differences for noticing and using menu labels were found for the middle and upper-middle age groups. Noticing and using nutrition information were more common for younger age groups living in No Policy jurisdictions, and differences between policy groups were larger for older groups. This finding is supported by a systematic review of nutrition labels on pre-packaged foods that found older adults were less likely to use nutrition labels than middle-aged and young adults(Reference Campos, Doxey and Hammond46). Our study similarly found low noticing and use of nutrition information among older age groups, but living in a Policy Present area reduced some of the disparity.
Examining the differences between policy groups by education, rates of noticing nutrition information, using nutrition information and changing restaurants were roughly equal in No Policy jurisdictions, but the increases were greater for higher education and majority ethnicity groups in Policy Present jurisdictions (Tables 5 and 6). Differences between policy groups for changing restaurants were also greater in the higher income adequacy group compared to the lower income adequacy group, and the higher income adequacy group was more likely to change restaurants in the Policy Present group (Table 7). This suggests that higher education and higher income groups may be more sensitive to changing restaurants within Policy Present jurisdictions. Indeed, higher comprehension and use of nutritional labels has been found to be associated with higher income and higher education(Reference Campos, Doxey and Hammond46). Higher education and higher income levels were associated with a greater likelihood of ordering something different, changing restaurants and eating at restaurants less often. The highest income level group was also the most likely to eat less of what they ordered due to noticing nutrition information. This is concordant with the majority of previous evidence which also suggests higher education and income levels are associated with greater use of nutrition labels(Reference Campos, Doxey and Hammond46). Although some other research has not found any convincing evidence of sociodemographic disparities in responses to menu labelling, the present results suggest that menu labelling policies, specifically those containing only numeric information, could potentially widen inequalities in healthy eating and therefore health(Reference Robinson, Polden and Langfield47,Reference Robinson, Boyland and Christiansen48) . This insight suggests that other interventions may be needed to support lower education groups through mechanisms other than information-based policies that require significant cognitive demand on individuals. There is evidence that labels can be designed to avoid widening disparities between socio-economic groups. Other types of labels, such as warning labels, may be of greater use for people of low income or literacy, thereby reducing disparities(Reference Hall, Lazard and Grummon49). Therefore, our results suggest the association between label design and socio-economic disparities should be considered when designing labels for real world policies.
The difference between policy groups was greater in the majority ethnicity group compared to the minority ethnicity group for ordering something different, but these differences were primarily due to high rates of ordering something different for minority groups in No Policy jurisdictions (Table 5). The rates of ordering something different in No Policy jurisdictions for minority groups were closer to the rates found in Policy Present jurisdictions for both minority and majority groups. Previous work from the USA found Black or Hispanic participants were more likely to choose restaurants with menu labelling and to use caloric information compared to White participants(Reference Bleich and Pollack50). A similar pattern could explain our findings of greater ordering something different in No Policy jurisdictions if minority groups are more likely to seek restaurants that have menu labelling compared to majority groups. These findings suggest that there may be limited additional benefit of mandatory menu labelling policies for minorities if they already using nutrition information at a greater rate than in voluntary jurisdictions. However, further work is needed to determine whether this pattern exists across other populations. Overall, these differential effects of menu labelling in restaurants across groups suggest complementary policies may be needed to support healthy eating and reduce inequalities across more vulnerable socio-economic groups.
Strengths and limitations
To our knowledge, this is the first multi-country examination of national and state/province-level menu labelling policies and associated behaviours. Using the same survey questions across intervention and comparison policy jurisdictions allows for between-country comparisons that are otherwise more challenging to do between countries with limited capacity to conduct routine national diet surveys(Reference Hammond, Vanderlee and White27). Thus, these results may help provide more generalisable evidence for the effects of menu labelling on self-reported eating behaviours. Weighted IFPS estimates are close to the sociodemographic distributions in the countries studied, although there was a lower recruitment of low education participants from Mexico(Reference Hammond, Vanderlee and White27). Finally, the large study sample of nearly twenty thousand participants increases power to detect differences between policy groups.
Our study does have limitations as we cannot determine the degree to which self-reported behavioural changes translate to changes in dietary intake, obesity or other health outcomes. There are also some variations in menu labelling policies within our Policy Present and No Policy groups. Although we were able to categorise jurisdictions into either Policy Present or No Policy groups, the mandatory policies in Canada, Australia and the USA are not exactly the same – for example, businesses with twenty or more locations in Ontario, Canada, could refer to businesses with greater density of outlets compared to with twenty or more locations across the USA, given the different density of outlets required to meet the twenty outlet threshold in geographic areas of different size. Thus, our study has some challenges to consistency – variations of exposure (Policy Present in this study) do not differentially affect outcomes (behavioural measures in this study) – a core assumption of causal inference. The cross-sectional nature of a single data collection and a natural experimental design are more vulnerable to confounding bias than randomised controlled trials for demonstrating causal effects, and we cannot eliminate the possibility for residual confounding if some factor other than policy status is driving the differences between groups, such as country differences. Due to the single year of data used, we also cannot determine whether differences observed in 2019 are due to reverse causation: for example, if pre-policy rates of noticing and using menu labels were higher and thereby facilitated policy adoption. Finally, we did not examine all mechanisms through which menu labels could improve population health. Menu labels could improve health through pathways other than the behaviours examined here – for example through product reformulation – which could reduce energy intakes.
Conclusions
Participants living in jurisdictions with mandatory nutrition information were more likely to report noticing nutrition labels, ordering something different, eating less of what was ordered, and changing restaurants in jurisdictions where nutrition information in restaurants was mandatory. The magnitudes of differences between Policy and No Policy jurisdictions were relatively small. Mandatory menu labelling was associated with greater behavioural differences in more socio-economically affluent groups, which could potentially exacerbate existing inequalities in diet and health. Complementary interventions may be required to optimise mandatory menu labelling interventions by accounting for unequal effects across sociodemographic groups. Further research understanding whether menu labelling has similar inequitable effects on dietary intake will now be valuable.
Acknowledgements
The authors thank the participants from Wave Three of the IFPS.
Financial support
This independent research was commissioned and funded by the NIHR (NIHR200689, Policy Research Programme). The views expressed in this publication are those of the author(s) and not necessarily those of the NIHR or Department of Health and Social Care. This work was supported by the MRC Epidemiology Unit, University of Cambridge (grant number MC/UU/00006/7). Funding for this project was provided by a Canadian Institutes of Health Research (CIHR) Project Grant, with additional support from the Public Health Agency of Canada (PHAC) and a CIHR-PHAC Applied Public Health Chair (DH). Funders had no role in study design; collection, analysis and interpretation of data; writing the report; and the decision to submit the report for publication.
Conflict of interest
There are no conflicts of interest.
Authorship
Conceptualisation, M.E., J.A., M.W., E.R., A.J., R.S., L.V. and D.H.; Formal analysis, M.E., J.A., T.B., E.R. and A.J.; Funding acquisition for the International Food Policy Study, D.H.; Methodology, M.E., J.A., E.R., A.J., D.H. and L.V.; Project administration, C.W.; Supervision, J.A.; Validation, M.P.K., D.C., G.S., L.V. and D.H.; Writing – Original draft, M.E.; Writing – Review and editing, M.E., L.V., T.B., A.C., A.J., M.P.K., M.P., E.R., G.S., R.S., C.W., M.W., D.H. and J.A. All authors have read and agreed to the published version of the manuscript.
Ethics of human subject participation
This study was conducted according to the guidelines laid down in the Declaration of Helsinki. The study was reviewed by and received ethics clearance through a University of Waterloo Research Ethics Committee (ORE# 30829). A full description of the study methods can be found in the International Food Policy Study: Technical Report – Wave 3 (2019) at www.foodpolicystudy.com/ methods. Written informed consent was obtained from all participants.
For the purpose of open access, the author has applied a Creative Commons Attribution (CC BY) license to any Author Accepted Manuscript version arising from this submission.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S1368980023001775
Open access
