Bulletin of the World Health Organization

The influence of market deregulation on fast food consumption and body mass index: a cross-national time series analysis

Roberto De Vogli a, Anne Kouvonen b & David Gimeno c

a. Department of Public Health Sciences, School of Medicine, University of California Davis, One Shields Avenue, Med Sci 1-C, Davis, CA 95616, United States of America (USA).
b. School of Sociology, Social Policy and Social Work, Queen’s University Belfast, Belfast, Northern Ireland.
c. Division of Epidemiology, Human Genetics and Environmental Health Sciences, The University of Texas Health Science Center at Houston, Houston, USA.

Correspondence to Roberto De Vogli (e-mail: rdevogli@ucdavis.edu).

(Submitted: 28 February 2013 – Revised version received: 23 September 2013 – Accepted: 24 September 2013.)

Bulletin of the World Health Organization 2014;92:99-107A. doi: http://dx.doi.org/10.2471/BLT.13.120287

Introduction

In the last decades, there have been substantial increases in mean body weight in wealthy countries.1,2 Such changes accompanied dramatic transformations in people’s dietary patterns, most notably an increase in the consumption of ultra-processed foods, including fast food,3 herein defined as “food that can be prepared quickly and easily and is sold in restaurants and snack bars as a quick meal or to be taken out”.4

Although some authors argue that fast food consumption has played a negligible role in the obesity epidemic,5,6 numerous studies have shown the opposite to be true.7,8 A cohort study by Pereira et al. showed that participants who visited fast food restaurants more than twice a week at baseline and were still doing so at a follow-up 15 years later had gained an average of 4.5 kg.9 Significant associations between the density of fast food restaurants and obesity have also been shown by neighbourhood-1012 and state-level analyses.1315 So far, little cross-national research has been conducted to investigate whether the spread of fast food has led to an increase in population-wide obesity rates over time.16,17 However, in a recent ecological analysis, the density of Subway outlets, used as a marker of fast food penetration, was positively associated with the prevalence of obesity across 26 advanced economies.18 Another cross-national ecological analysis revealed an association between increases in soft drink consumption and higher rates of overweight and obesity.19 The research conducted to date has revealed little about the factors that drive or contain the spread of fast food and obesity.16 Some authors argue that the rising consumption of unhealthy foods seen worldwide has been facilitated by trade liberalization20 and foreign investment in the food and beverage industries,8,2123 which have resulted in the proliferation of large transnational food companies.20,24,25 Offer et al. have found that high-income countries with market-liberal welfare regimes – most of which are also English-speaking – have a higher prevalence of obesity and easier access to fast food.17 A study by Cutler et al. has shown that regulations in the agricultural sector are negatively correlated with obesity.26

In this article we use a novel measure – the number of per capita fast food transactions (local and transnational) – to test the hypothesis that rising fast food consumption has been a major determinant of population increases in body mass index (BMI) among high-income countries belonging to the Organisation for Economic Co-operation and Development (OECD). We also examine whether market deregulation may have contributed to higher BMI by facilitating the spread of fast food.

Methods

We conducted multivariate panel data analyses of 25 high-income OECD countries over the period from 1999 to 2008. Data on fast food consumption and age-standardized mean BMI were available for only 27 of the 31 high-income OECD members. Such data were missing for Estonia, Iceland, Luxembourg and Slovenia. To limit biases in international comparisons between Asians and Caucasians due to different interpretations of BMI in Asian populations,27 we excluded Japan and the Republic of Korea. However, we ran additional analyses including these countries as robustness checks. We also developed separate models excluding Anglo-Saxon economies (Australia, Canada, Ireland, New Zealand, the United Kingdom of Great Britain and Northern Ireland and the United States of America) that, as previous studies showed, have a higher prevalence of obesity and easier access to fast food.17

Data sources

Fast food consumption

Data on per capita fast food transactions were taken from Euromonitor’s Passport Global Market Information Database (GMID), 2012 edition. The data comprise industry records of annual sales of meals and refreshments delivered in local and transnational fast food outlets,28 including chain restaurants, independent eateries and convenience stores (Appendix A, available at: http://goo.gl/36c7ai). This measure is the most comprehensive indicator of fast food consumption for comparisons across nations. Appendix B (available at: http://goo.gl/gThiG5) shows the scatterplot and strong correlation coefficient (r = 0.8501; P < 0.001) for the association between fast food transactions per capita, as obtained from the GMID, and Subway restaurants per 100 000 population, an indicator used in a previous paper as a proxy measure of the density of fast food restaurants at the country level.18

Age-standardized mean body mass index

Our main dependent variable, age-standardized mean BMI (in kg/m2), was obtained from the Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group, which produced comparative estimates of cross-country differences and changes over time in BMI for adults aged 20 years or older.1 Although data on BMI are reported separately for men and women, we developed an overall indicator by estimating the female to male ratio using the proportion of female population from the World Development Indicators from 1999 to 2008.29 We also ran sex-specific analyses as robustness assessments.

Market deregulation

Market deregulation is the degree to which market forces are allowed to operate without interference from outside intervention, especially in the form of government ownership, regulations and taxes.30 We used the index of economic freedom (IEF) created by the Heritage Foundation and the Wall Street Journal, which is based on a scale from 1 to 100. The score indicates the extent to which a country has adopted market deregulation policies. The index is calculated as the mean of 10 subcomponents measuring different aspects of economic freedom, as determined from national laws and regulations as well as written questionnaires completed by experts and investors (Appendix C, available at: http://goo.gl/M76H7I).31

Covariates

We included in our analyses several potential confounders of the association between fast food and BMI: gross domestic product (GDP) per capita (expressed logarithmically in constant 2005 United States dollars, adjusted for purchasing power parity for comparability between countries); the proportion of the population living in urban areas; national population size; openness to trade (imports and exports as a percentage of GDP); foreign direct investment (FDI, or net inflows as a percentage of GDP); and a time-invariant (2008) measure of motor vehicles per 1000 people. All these measures were taken from the World Bank’s World Development Indicators database.29 We also included as confounders time-invariant measures (2008) of the percentage of the population doing insufficient physical activity (i.e. less than 30 minutes of moderate activity five times per week or less than 20 minutes of vigorous activity three times per week, or their equivalent) and consumption of fruits and vegetables (in kilograms per capita per year) in 2008. These two values were obtained from the World Health Organization Global Infobase32 and from the GMID, respectively.28 Finally, as previous studies have revealed that obesity and the availability of cheap, energy-dense food tend to be higher in societies with greater economic inequality,33,34 we adjusted for the Gini index, a measure of inequality in household disposable income. Data on the Gini index were taken from the Standardized World Income Inequality Database.35,36

Our analyses also include three potential mediators of the association between fast food and BMI: consumption of animal fats (in kcal per capita per day); total caloric intake (in kcal per capita per day); and soft drink consumption (in litres per capita per year). The first two values were obtained from the Statistics Division of the Food and Agriculture Organization;37 the last one came from the GMID.28

Statistical analyses

To study the association between fast food consumption and BMI we used longitudinal panel analyses, which allow the dynamics of change over time to be explored.38 Our regression models included corrections for fixed aspects of initial country conditions and other characteristics that could influence the level of fast food consumption – and hence average BMI – in a given country.39,40 By assessing within-country annual variations in fast food and obesity over time and adjusting for fixed, country-level characteristics, these conservative models effectively address the problem of confounding of study results. Robust standard errors –– clustered by region to adjust for the non-independence of time series data – were calculated in all models.8 Regressions were analysed using Stata version 12.0 (StataCorp. LP, College Station, United States of America).

We formulated the following fixed effects models: (1) (2)where i is the country, t is the year, β1 is the regression coefficient for per capita fast food transactions, β2 is the regression coefficient for GDP, υi is an error term denoting country-specific heterogeneity, εit indicates an identically distributed random error term or measurement error and α is a constant.

(1).
(2).

Results

Fast food consumption and BMI

As shown in Table 1 (available at: http://www.who.int/bulletin/volumes/92/2/13-120287), between 1999 and 2008, the average number of annual fast food transactions per capita increased from 26.61 to 32.76. During the same period, age-standardized mean BMI increased from 25.8 to 26.4 kg/m2. There was a strong and positive association between fast food consumption and age-standardized mean BMI (unadjusted r = 0.658; P < 0.001). When considering changes between 1999 and 2008 (Fig. 1), the average annual number of fast food transactions per capita was positively associated with age-standardized mean BMI (unadjusted r = 0.503; P < 0.01). The highest increases in the average number of annual fast food transactions per capita were observed in Canada (16.6), Australia (14.7), Ireland (12.3) and New Zealand (10.1), while the lowest increases occurred in Italy (1.5), Greece (1.9), the Netherlands (1.8) and Belgium (2.1).

Fig. 1. Change in age-standardized mean body mass index (BMI) as a function of change in average annual fast food transactions per capitaa in 25 high-income countries of the Organisation for Economic Co-operation and Development, 1999–2008
Fig. 1. Change in age-standardized mean body mass index (BMI) as a function of change in average annual fast food transactions per capita<sup>a</sup> in 25 high-income countries of the Organisation for Economic Co-operation and Development, 1999–2008
United Kingdom, United Kingdom of Great Britain and Northern Ireland; United States, United States of America.
a Meals and refreshments sold annually per capita in local and transnational fast food outlets, including chain restaurants, independent eateries and convenience stores.
Note: The figure illustrates the positive correlation between changes in age-standardized mean BMI and changes in the number of annual fast food transactions per capita between 1999­ and 2008.
Sources: Age-standardized mean BMI: Global Burden of Metabolic Risk Factors of Chronic Diseases Collaborating Group;1 average annual fast food transactions per capita: Euromonitor’s Passport Global Market Information Database.28

Table 2 presents the results of multivariate panel analyses in which age-standardized mean BMI was the dependent variable. Fast food consumption was positively and significantly associated with BMI (unadjusted β: 0.0657; 95% confidence interval, CI: 0.0433–0.0881). After correcting for income, urbanization, population size, openness to trade and FDI, the estimated relationship weakened but remained strongly significant (β: 0.0329; 95% CI: 0.0136–0.0522), so that each 1-unit increase in the average number of annual fast food transactions per capita was associated with an increase of 0.0329 kg/m2 in age-standardized BMI.

Robustness checks

Before analysing the influence of market deregulation and the possible mediators between fast food consumption and BMI, we performed a series of robustness checks. When we excluded Anglo-Saxon economies from the model while controlling for the same confounders, we found no significant differences in the magnitude of the association between fast food consumption and BMI (P > 0.05 when testing effect heterogeneity). Similar results were found when we included Asian countries in the models. We then used first-difference methods to estimate the same basic model developed in Table 2, results confirmed the robustness of the fixed effects estimates (β: 0.0148; 95% CI: 0.0017–0.0279). We also disaggregated the analysis by sex and found no significant differences between males (β: 0.0294; 95% CI: 0.0077–0.0512) and females (β: 0.0360; 95% CI: 0.0183–0.0537) in the size of the estimated association (P > 0.05 when testing for effect heterogeneity). Similar results were obtained when we used per capita transactions only at chain food service outlets as an alternative measure of fast food consumption (β: 0.0271; 95% CI: 0.0114–0.0427). After the inclusion of three additional covariates – insufficient physical activity, motor vehicle use per 1000 people and fruit and vegetable consumption – the association between fast food and BMI remained statistically significant (β: 0.0140; 95% CI: 0.0058–0.0222). Finally, when we included the Gini index of within country income inequality in the model, the association between fast food consumption and BMI remained strongly significant (β: 0.0293; 95% CI: 0.0130–0.0456).

Soft drinks, animal fats and total calories

Table 3 shows the results of a series of separate regression models using mediators known to be associated with both fast food consumption and BMI. If the association between fast food consumption and BMI is mediated by soft drinks, animal fats and total calories, as we hypothesized, holding these mediators constant should attenuate the observed relationship. Only soft drink consumption, however, appeared to be a plausible partial mediator, by slightly reducing the effect size of the association between fast food consumption and BMI, after correcting for covariates (β: 0.0302; 95% CI: 0.0101–0.0504). Neither the intake of animal fats nor total caloric intake changed the effect size of the observed relationship substantially.

Market deregulation, fast food consumption and BMI

In spite of the robustness checks, our results could have been driven by third factors affecting both fast food consumption and BMI, such as changes in the macroeconomic environment. Although fixed effects models can cancel out the possible confounding effect of initial, time-invariant, country-specific characteristics, they do not correct for time-varying confounders. To address this problem, we employed two-stage least squares regression models using economic freedom as an instrumental variable. These models allowed us not only to put to further testing the robustness of the fixed-effects estimates in Table 2, but also to investigate the role of market deregulation as a determinant of BMI through fast food consumption. Instrumental variables are believed to simulate a natural experiment, and act as a randomization device in dealing with unobserved covariates that, in our case, may be correlated with both fast food consumption and BMI.41 Valid instruments have at least two major properties. First, they affect the exposure variable we want to test, in this case fast food consumption. Second, they must have no direct effect on the outcome measure, in our case BMI.41 Table 4 presents estimates of fixed-effects regression models investigating the associations between the IEF (market deregulation) and fast food consumption and BMI. After adjustment for fast food consumption, the association between the IEF and BMI weakened to non-significance (P > 0.05), qualifying the IEF as a valid instrument.

Table 5 shows the first-stage and two-stage least square regression models for the effect of fast food consumption on BMI, with the IEF used as an instrument, after adjustment for other covariates. The first-stage regression confirmed that market deregulation is a strong predictor of higher fast food consumption (β: 0.2714; 95% CI: 0.1644–0.3785), after correction for confounders. Each 1-unit increase in the IEF was associated with an increase of 0.2714 in the average number of per capita annual transactions at fast food outlets. The second-stage regression indicated that, when the IEF was used as an instrumental variable for fast food consumption and after correction for confounders, each 1-unit increase in fast food consumption was associated with an increase of 0.0232 kg/m2 in BMI (95% CI: 0.0011–0.0452).

Discussion

Our study shows that fast food consumption is independently and positively associated with mean BMI in high-income countries. While the consumption of soft drinks explains a small proportion of the variation in the association between fast food consumption and BMI, the intake of animal fats and total caloric intake do not seem to be significant mediators of the association. This is puzzling. The fat and calories in fast food meals are usually blamed for the unhealthful effect of fast food.42 Although we cannot exclude the possibility of measurement errors, factors other than calories and fat content may explain why fast food makes people fat. Researchers need to investigate, for example, the metabolic effects of long-term exposure to fast foods produced from the meat of animals fed on corn, kept in confinement and exposed to excessive fertilization.43 Researchers should also examine the health effects of a poor diet, which can lead not only to obesity but also to the development of noncommunicable diseases. More research is also needed to study the effects of the degree of processing of food items and not just their nutrient and caloric content.44

In line with previous research,17 our study shows that countries adopting what are considered market-liberal policies experience faster increases in both fast food consumption and mean BMI. These results are in accord with previous research showing that more stringent trade restrictions – including better protection of agricultural producers45 – the frequency of price controls26 and stricter government regulations46 are negatively correlated with obesity. The mechanisms explaining the influence of economic freedom on fast food and obesity have not been sufficiently studied. One possibility is that indiscriminate market deregulation favours global food chains at the expense of smaller farmers and local food systems.47 In effect, additional analyses (available from the corresponding author upon request) showed that, while per capita transactions at chain food service outlets were positively and significantly correlated with mean BMI, this was not the case for per capita transactions at independent food service outlets.

Our results must be interpreted with caution. First, the IEF reflects perceptual biases because it disproportionately relies on the perspective of investors and the business community.48 Moreover, it does not necessarily reflect the extent to which market deregulation is applied to the agricultural sector. Our data show, however, that the most “market-friendly countries, including Australia, Canada, New Zealand and the United States have less restrictive agricultural regulations and provide substantially lower farm subsidies than European countries such as France, Italy and Greece.45 Another limitation has to do with the dependent variable, age-standardized mean BMI, which is based on estimates from a Bayesian hierarchical model involving a complex dependence structure for which we could not adjust.1 In spite of this, the correlation between the BMI measure used in this study and obesity prevalence as obtained from the Global Health Observatory database was very strong. (r = 0.953; P < 0.001) (Appendix D, available at: http://goo.gl/ElLR0z) Although mean BMI may be a biased measure of overweight and obesity, especially because the prevalences of underweight and malnutrition can influence its interpretation, such bias is more likely to affect BMI estimates for low- and middle-income countries. Moreover, a continuous variable like BMI is a more practical indicator than a categorical variable such as obesity because its associations with most health outcomes are continuous, rather than characterized by a specific threshold. An additional limitation relates to the ecological and observational nature of the data. Although confounding can never be completely ruled out, our findings remained robust following numerous estimation methods and statistical checks. Finally, although the magnitude of the association between fast food consumption and BMI weakened substantially under instrumental variable specification, it remained statistically significant.

Conclusion

Our study provides novel findings on the association between fast food consumption and mean population BMI and on the influence of market deregulation as a contributor to higher fast food consumption and BMI. The study has important implications for policy. In particular, they suggest that government regulations hindering the spread of fast food consumption might help to mitigate the obesity epidemic. Indeed, although all countries included in our sample have experienced increases in fast food consumption and mean BMI over the period studied (1999–2008), nations that have adopted more stringent market regulations have experienced slower increases in both. More research is needed to confirm whether deregulation is a significant contributor to body weight and to determine what types of government interventions could mitigate the obesity epidemic and curb the spread of transnational fast food companies.


Acknowledgements

The authors thank David Stuckler and Sanjay Basu who provided feedback and suggestions on earlier versions of this manuscript.

Funding:

RDV is supported by a grant from the Economic and Social Research Council (RES-070-27-0034). No funding bodies had any role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Competing interests:

None declared.

References

Share