OUP user menu

Cross-national comparison of environmental and policy correlates of obesity in Europe

Borsika A. Rabin, Tegan K. Boehmer, Ross C. Brownson
DOI: http://dx.doi.org/10.1093/eurpub/ckl073 53-61 First published online: 3 June 2006


Background: Despite the growing agreement that modern environments fuel increased food consumption and decreased physical activity, few studies have addressed environmental and policy correlates of obesity. This study describes obesity patterns across Europe and identifies macroenvironmental factors associated with obesity prevalence at a national level. Methods: Data on obesity prevalence and indicators of the physical, economic, and policy environment were assembled from international databases for 24 European countries. Coefficient estimates between overall, male, and female obesity prevalence and each independent variable were calculated using linear regression. Results: The obesity prevalence varied widely across countries and between genders with higher values in Central and Eastern European countries and lower values in France, Italy, and some Scandinavian countries. Statistically significant inverse associations were observed between overall and female obesity prevalence and variables from the following domains: economic (real domestic product), food (available fat), urbanization (urban population), transport (passenger cars, price of gasoline, motorways), and policy (governance indicators). There was also a negative association between overall obesity and available fruits/vegetables, and between female obesity and single-member households. Male obesity was inversely associated with available fruits/vegetables and density of motorways. The magnitude of the coefficient estimates suggests stronger associations for female obesity than for male obesity in all cases. Conclusions: This exploratory study suggests a need to conduct additional research examining the role of obesogenic environments in European countries, with a special focus on policy-related variables, and to further study gender-specific differences in obesity and its correlates.

  • cross-national comparison
  • ecological study
  • environment
  • Europe
  • obesity
  • policy

There is a well-established relationship between excess body mass index (BMI) and mortality and morbidity from several chronic diseases, such as cardiovascular diseases, certain cancers, type 2 diabetes, osteoarthritis, work disability, and sleep apnoea.13 Recognizing this relationship, the World Health Organization (WHO) has declared obesity a significant global epidemic, affecting the population of both industrialized and non-industrialized countries.4

Although obesity is less prevalent in most European countries than in the United States, the International Obesity Task Force indicates that the prevalence of obesity has increased during the last decade on the ‘Old Continent’.2 In Europe, the consequences of excess body weight are estimated to account for 7–8% of the overall burden of disease5 and ∼1–5% of total health care expenditures.1

Because of the increasing prevalence and costly consequences, obesity can no longer be considered as a purely medical issue, but rather as a threat to public health, requiring national and global strategies for prevention and management.1,6,7 Unfortunately, existing obesity prevention programmes are low in number and effectiveness.1,8 Several experts claim that the relatively poor success of such interventions is due to their traditional, individual-level, behaviour change approach and a paradigm shift towards an ecological-level understanding of obesity is necessary to reverse the escalating trends.810

An ecological model suggested by Egger and Swinburn defines obesity as the net result of biological, behavioural, and environmental influences, which are mediated by energy intake and expenditure, and moderated by physiological adjustments.9,10 In their model they describe the environment as the ‘public health arm of the obesity problem’ and emphasize the importance of microenvironmental and macroenvironmental attributes in shaping individual behaviour.9 They also categorized the environment according to type, as physical, economic, political, and sociocultural.9,10

Although there is growing agreement among researchers that the modern environment is a major contributor to present obesity trends,1013 there are very few published studies addressing how various environmental factors fuel this epidemic.1418 The existing studies were conducted in the United States and Australia and have mainly focused on the effect of urban form, transportation infrastructure, neighbourhood environment, and lifestyle factors on BMI and obesity at the individual14,1618 and county15 levels. The results of these studies indicate associations between increased likelihood of obesity and negative community perceptions, lack of nearby recreational facilities, absence of sidewalks, spending more time in a car, walking less, not having access to a motor vehicle all the time, being a resident of a less walkable or a more sprawling neighbourhood, and residing in an area with lower levels of land-use mix.1418 However, these studies are limited in scope to the microenvironment of individuals and provide no information on European patterns and correlates.

Despite the widely recognized role of governmental-level factors for many health behaviours,13,1922 the link between policy and obesity trends is not well understood. Policy-level changes can guide individual choices and may be good complementary methods to individual-level interventions, as they can influence the lives of more people, may affect groups who are difficult to reach with traditional approaches, can have longer lasting effects on behaviour change by shaping social norms, and may be cost-effective.10,23 Further research is clearly needed to understand whether policy and governance characteristics influence the obesity epidemic.

The aims of this study are 2-fold: (i) describe geographical patterns of obesity in Europe and (ii) examine correlations between country-specific obesity prevalence rates and macroenvironmental factors including indicators of physical, economic, and policy domains. Given the sparse literature related to the latter aim, this study is considered exploratory.

Materials and methods

Design and data sources

In order to address these research aims, we conducted an ecological, correlational study of European countries. The prevalence of obesity across countries was evaluated in relation to national indicators of the physical, economic, and policy environments. Aggregate country-level data were assembled from several databases (see table 1). European countries were classified as Western European (WE) and Central-Eastern European (CEE) as shown in table 2.

View this table:
Table 1

Description of dependent and independent variables included in the analysis with their sources

Prevalence of obesity
    Total1997–2002% of populationPercentage of population with a body mass index (BMI) ≥ 30 kg/m2NCD InfoBasea, peer-reviewed articles,7,25,26 and personal communication (see text)
    MaleBMI was calculated from self-reported height and weight.
    FemaleData are based on national (both urban and rural) samples except for Russian Federation and Switzerland for whom sub-national, urban samples are included.
Economic variables
    Gross domestic product (GDP)20011000 US$/capitaThe per capita monetary value of all final goods and services produced in a country during a yearHFAb
    Real domestic product (RDP)20011000 US PPP$/capitaThe per capita GDP which is expressed in purchasing power parity (PPP) and adjusted to the relative domestic purchasing power of the national currency as compared with the US dollarHFAb
    Students in tertiary education2000per 1000 populationTertiary education includes post-secondary education leading to: an award not equivalent to a first university degree, a first university degree or equivalent, or a post-graduate university degreeTrends in EU and NAc
    Unemployment rate2000% of total labour forceUnemployed comprises all persons above a specified age who during the reference period were without work, currently available for work, or seeking workHFAb
Food variables
    Available calories2001kcal/person/dayTotal amount of food available for consumption (from food production, imports, exports, and stocks) converted into kilocalories (kcal)HFAb
    Available fat2001% of total energyPercent of total energy available from fatHFAb
    Available fruits/vegetables2001kg/person/yearAverage available fruits and vegetables per person per year in kilogramsHFAb
Urbanization variable
    Urban population2001% of total populationPercentage of population residing in urban areas in each country according to national definitionTrends in EU and NAc
Household variables
    Single-member     households1991–2001% of all householdsHouseholds consisting of one personTrends in EU and NAc
    Large households1991–2001% of all householdsHouseholds consisting of five persons or moreTrends in EU and NAc
Transport variables
    Total passenger cars2001per 1000 populationA motor vehicle, other than a motorcycle, intended for the carriage of passengers and designed to seat no more than nine personsTrends in EU and NAc
    New passenger cars2001per 1000 populationNumber of new passenger cars registered in the given yearTrends in EU and NAc
    Price of unleaded gasoline2000PPP cents/litreThe price per litre of gasoline is expressed in purchasing power parity (PPP) relative to the US dollarTrends in EU and NAc
    Paved roads1999% of total roadsPaved roads are those surfaced with crushed stone and hydrocarbon binder or bituminized agents, or with cobblestones as a percentage of all country's roads measured in lengthWDId
    Density of motorways2000–2001km/100 000 inhabitantsMotorway is defined as a road, specially designed and built for motor traffic, which (i) does not serve properties bordering on it, (ii) has separate carriageways for the two directions of traffic, and (iii) does not cross at level with any road, railway, tramway track, or footpathPanorama of Transporte
Policy variablesIndicators are available for four 2 year long time periods, from 1996 to 2002. An average of the four data points was used in our study. Variables are measured in units ranging from −2.5 to 2.5, with higher values corresponding to better governance outcomes.39Governance Indicatorsf
    Voice and accountability1996–2002Point estimateThe extent to which citizens of a country are able to participate in the selection of governments and the independence of media
    Political stability1996–2002Point estimatePerceptions of the likelihood that the government in power will be destabilized or overthrown
    Government effectiveness1996–2002Point estimateThe quality of ‘inputs’ (e.g. bureaucracy, public service provision, etc.) required for the government to be able to produce and implement good policies and deliver public goods
    Regulatory quality1996–2002Point estimateMeasures of the incidence of market unfriendly policies and perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development
    Rule of law1996–2002Point estimateThe success of a society in developing an environment in which fair and predictable rules form the basis for economic and social interactions, and the extent to which property rights are protected
    Control of corruption1996–2002Point estimatePerceptions of corruption, conventionally defined as the exercise of public power for private gain
    Average governance indicator1996–2002Point estimateCalculated as the arithmetic mean of the six average indicators for each country
  • a: WHO Global NCD InfoBase, World Health Organization, Regional Offices

  • b: European Health for All Database, World Health Organization

  • c: Trends in Europe and North America, United Nations Economic Commission for Europe/Environment and Human Settlements Division

  • d: World Development Indicators, World Bank Institute

  • e: Panorama of transport, statistical overview of transport in the EU, European Commission, Eurostat

  • f: Governance indicators for 1996–2002, World Bank Institute

View this table:
Table 2

Prevalence of obesity for overall, male, and female groups in 24 European countries

Prevalence of obesitya
Western Europe
    The Netherlands8.47.29.5
    United Kingdom1110.912.6
Central–Eastern Europe
    Czech Republic14.21513.5
    Russian Federation15.29.617.5
  • a: Percent of population with a BMI ≥ 30 kg/m2

Obesity prevalence

The dependent variables were self-reported overall and gender-specific prevalence rates of obesity (i.e. percentage of population with BMI ≥ 30 kg/m2) derived from national surveys. Countries were selected based on the availability of obesity prevalence data. The Global Noncommunicable Disease (NCD) InfoBase of the WHO collects country-level data on important non-communicable disease risk factors for all WHO Member States and was used to identify countries for this study.24 A 6-year timeframe was applied (1997–2002) to search the database. We selected the most recent study during this period that was both nationally representative (i.e. both urban and rural samples) and based on self-reported data.

Initially, all European countries fulfilling the above-mentioned criteria and having available prevalence data in the database for at least one of the obesity categories (overall, male, or female) were included in the study. Literature review7,25,26 and personal communication (M.D. Vaz de Almeida and C. Alfonso, October 2004) were used to acquire missing obesity information. Following the completion procedure, countries with missing obesity data in any of the three categories were excluded, leaving a total of 24 countries for analysis.

Independent variables

A search was performed to identify physical, economic, and policy macroenvironmental indicators from databases of international health, economic, and other governmental organizations for the selected countries. The physical environment was further categorized into food, urbanization, household, and transport variables. The available variables of interest are described in table 1, along with their sources.

Statistical method

To describe the geographical patterns of the prevalence of obesity across countries, data were mapped using geographical information system software (ERSI Arc Map 8.2, Redlands CA, 2002). Linear regression models were employed to examine the unadjusted association between each independent variable and obesity prevalence as a continuous dependent variable for overall, male, and female groups separately. Given the exploratory nature of our study, analyses relied on an alpha-level of 0.10 to confer statistical significance. Analyses were run using SPSS for Windows version 12.0.


Obesity patterns

There was a wide variation in the prevalence of obesity across countries and between genders (table 2). Obesity prevalence ranged from 7 to 18% among males, from 6 to 20% among females, and from 6 to 20% for the combined genders. The geographical distribution is displayed in figure 1 using gender-specific quartiles (i.e. the absolute percentage range for each quartile differs for males and females). The highest prevalence rates were reported from CEE countries for both genders with obesity prevalence rates ranging from 9 to 18% with a median value of 15% for males and from 12 to 20% with a median value of 17% for females. Interestingly, relatively high prevalence values for one or both genders were observed in some WE countries like the United Kingdom, Ireland, Denmark, Finland, Switzerland, and Spain. The lowest prevalence rates were found in France, Italy, and some of the Scandinavian countries. In Western Europe, the obesity prevalence for males ranged from 7 to 14% with a median value of 9.5% and for females from 6 to 14% with a median value of 11%.

Figure 1

Geographical distribution of the prevalence of obesity by gender-specific quartiles in 24 European countries

Macroenvironmental correlates

The sample size and distribution of the independent variables of interest are provided in table 3. Overall and gender-specific linear regression coefficients (b) along with the p-values are presented in table 4.

View this table:
Table 3

Summary statistics (median, minimum, and maximum) for independent variables by domain across 24 European countries

IndicatorNumber of countriesUnitMedianMinimumMaximum
Economic variables
    GDP241000 US$/capita22.21.742.0
    RDP241000 PPP$/capita24.25.853.8
    Students in tertiary education24Per 1000 population35.013.052.0
    Unemployment rate24%
Food variables
    Available calories24kcal/person/day3431.02808.53798.9
    Available fat24g/person/day36.824.141.7
    Available fruits and vegetables24kg/person/year188.2131.1417.3
Urbanization variable
    Urban population24% of total population71.255.297.4
Household variables
    Single-member householda22% of total households29.713.840.5
    Large householdb22% of total households7.94.517.8
Transport variables
    Total passenger cars24Per 1000 population415.0139.0635.0
    New passenger cars24Per 1000 population33.07.098.0
    Price of unleaded gasoline19Prices in PPP US cents/litre94.072.0121.0
    Paved roads22% of total roads90.721.0100.0
    Density of motorways20km/100 000 inhabitants12.80.526.2
Policy variables
    Voice and accountability24Point estimate1.3−0.41.6
    Political stability24Point estimate1.0−0.51.6
    Government effectiveness24Point estimate1.6−0.52.2
    Regulatory quality24Point estimate1.2−0.71.7
    Rule of law24Point estimate1.5−0.82.2
    Control of corruption24Point estimate1.4−0.82.4
    Average governance indicator24Point estimate1.3−0.61.9
  • a: Household consisting of one person

  • b: Household consisting of five persons or more

View this table:
Table 4

Relationship between overall, male, and female obesity prevalence and each independent variable across 24 European countries

Economic variables
    Students in tertiary education−0.0110.889−0.0020.975−0.0270.763
    Unemployment rate0.2700.1280.1820.3230.2920.151
Food variables
    Available calories−0.0040.118−0.0020.364−0.0040.132
    Available fat−0.3230.010−0.1410.298−0.3990.004
    Available fruits/vegetables−0.0190.049−0.0220.028−0.0150.199
Urbanization variable
    Urban population−0.0950.080−0.0670.238−0.0990.113
Household variables
    Single-member householda−0.1130.179−0.0320.719−0.1830.049
    Large householdb0.1340.477−0.0410.8350.2740.191
Transport variables
    Total passenger cars−0.0170.000−0.0080.135−0.0200.000
    New passenger cars−0.0810.018−0.0470.200−0.0870.028
    Price of unleaded gasoline−0.0950.042−0.0770.125−0.0960.041
    Paved roads−0.0640.033−0.0380.218−0.0730.032
    Density of motorways−0.2240.022−0.1970.067−0.2270.030
Policy variables
    Voice and accountability−3.3320.013−0.3330.821−4.8820.001
    Political stability−2.9530.023−0.2880.838−4.3650.002
    Government effectiveness−2.1260.006−0.5680.505−2.9850.000
    Regulatory quality−2.4390.0340.0250.984−3.7050.003
    Rule of law−2.1390.004−0.6070.467−2.9610.000
    Control of corruption−1.8310.005−0.5860.417−2.4930.000
    Average governance indicator−2.5280.007−0.5470.598−3.5750.000
  • Note: b: coefficient estimate; p: p-value

  • a: Household consisting of one person

  • b: Household consisting of five persons or more

Statistically significant inverse associations were observed between overall obesity prevalence and gross domestic product (GDP), real domestic product (RDP), available fat, available fruits and vegetables, urban population, and all variables from the transport and policy domains. Associations for female obesity were similar to overall obesity with a few exceptions: coefficients for GDP and available fruits and vegetables were not significant and there was a significant negative coefficient for single-member households. The only statistically significant findings for male obesity were an inverse association with available fruits and vegetables and density of motorways. All food and some of the transport variables were related to obesity in the opposite direction to that expected. The magnitude of the coefficient estimates suggests stronger associations for female obesity than for male obesity in all except one case (available fruits/vegetables). Independent variables not significantly associated with obesity in any of the three groups included students in tertiary education, unemployment rate, available calories, and large household.

Additional analyses of the policy indicators were performed to further explore the large magnitude and robust statistical significance of these coefficients. First, the average governance indicator was transformed into a categorical variable based on quintiles and entered into a linear regression with the overall obesity prevalence as a continuous dependent variable. Second, a logistic regression model was developed including overall obesity prevalence as a dichotomous dependent variable (using the mean value as a cut-off point) and the average governance indicator as a continuous independent predictor. These models provided similar results to those shown in table 4 and are not discussed further.


We examined obesity data for 24 European countries from the period 1997–2002 and found that CEE countries had the highest obesity prevalence rates for both genders. This finding is consistent with previous studies that suggest a Western–Eastern difference in overweight and obesity in Europe.1,27 Several authors explain this difference by a transition in the social, economic, and nutritional environments paralleled by a delayed effect of the Western lifestyle in the countries of the former ‘Eastern block’.5,19,28

Findings regarding the economic characteristics of a country suggest that there is significantly less obesity in countries with higher GDP and RDP. The positive correlation between income and the health status of a country (i.e. lower prevalence of obesity) is well known and explained by the ‘command over many of the goods and services that promote good health, such as better nutrition’.29 Similarly, several studies describe, especially for women in affluent societies, an inverse association between obesity and different proxy indicators of socioeconomic status.19,27,30 Furthermore, our findings are consistent with the idea presented in a study by Philipson and Posner, namely that in technologically advanced countries (as we consider all European countries included in our study), per capita income and weight will have a negative association owing to an increased demand for thinness in more affluent societies.31

Our study found a lower prevalence of obesity in countries with higher per capita availability of fat, and fruits and vegetables, but no association with available calories. These contradictory findings may be due to the limitation of food balance sheets to describe actual consumption.5,32 Data from household surveys or nationally representative food consumption surveys are preferred; however, such data are unavailable for the majority of the countries, especially in CEE.5

There is growing evidence that the land-use characteristics and transportation patterns affect the levels of physical activity and in this way are related to the prevalence of obesity.33,34 Giuliano and Narayan35 claim that most European metropolitan areas have high-density, centralized land-use patterns that promote walking instead of driving a car. In our study, we used the percentage of urban population as a proxy measure of the ‘walkability’ of a country, assuming that European urban areas are more walkable than rural regions. Our results suggest that countries with a higher percentage of urban population have lower obesity prevalence rates.

The concept behind the household variables was to describe the level of residential crowding in each country that could be related to differences in lifestyle habits (e.g. shopping, eating, and household-related physical activity) and also closely linked to income.36 In our study, a greater number of single-member households was negatively associated with obesity among females only.

Our intention with the transport domain was to describe daily transportation patterns and active commuting levels. All variables showed a significant correlation with overall and female obesity in the opposite direction than expected, except for the price of gasoline. Given the important role that economic factors play in chronic disease risk,37 it is important to better understand the role of factors such as price of gasoline in affecting obesity risk.

In our study we included aggregate indicators addressing different aspects of the quality of governance and observed robust significant coefficients for all six indicators, with especially high coefficient estimates and low p-values for women. Our findings suggest lower levels of obesity in countries that can be described among others by more independent media or by higher ‘capacity of the government to effectively formulate and implement sound policies’.38 The complexity of these indicators makes it difficult to interpret our results; however, we hypothesize that better stability and higher effectiveness of a government may provide a better opportunity for policymakers to focus on key public health problems such as obesity.

Laws and regulations were crucial components of public health achievements of the 20th century, however, they are not yet fully recognized or used systematically in the prevention and control of chronic conditions such as obesity.39 Policy-driven structural changes in the environment may shape eating and physical activity habits and consequently affect body weight patterns.21 An economic approach may influence a wide range of intermediate determinants of individual-level choices such as food production and marketing, agriculture, urban design, media, education, and transport by implementing targeted taxes and subsidies.13,21,40 Economic policies that could possibly make active-living choices easier by enhancing public transit, walking, and bicycling include increased tax of gasoline, parking cash-outs, and extra charges for automobiles entering congested city areas.41 For example, a study conducted by Stahl et al.42 in Eastern and Western Germany, and Finland showed a positive correlation between activity-friendly policy orientation, better infrastructure for sport, and physical activity behaviour. To successfully advocate for further policy-level changes, public health professionals need more than anecdotal evidence on the relationship between obesity and governance characteristics and the policy environment, which this study helps to establish.

We found a general trend of higher prevalence rates of obesity and stronger associations between obesity and macroenvironmental indicators among women compared with men. These results support the hypothesis that women may be more susceptible to obesity, especially in former Socialist countries.27,43 In its report on ‘Gender Differences in Susceptibility to Environmental Factors,’ the Institute of Medicine suggests that differences in biological factors such as genetics or hormones may explain why women are more susceptible to certain environmental factors than men and proposes that these environmental factors should be identified and considered in research and health promotion efforts.44 Our study also supports the notion that besides biological differences, women may be more likely to be obese because of their role in the family and household, and, consequently, their easy access to food.27

To our knowledge, no previous study on obesity has included as many countries from both Western, and Central and Eastern Europe or such a wide variety of macroenvironmental indicators, including governmental characteristics. Despite these strengths, the use of an ecological, cross-sectional study design introduces potential biases and cannot establish temporality. Our conclusions are limited to country-level associations, ignoring within-country variations and individual-level associations. National, objectively measured data on obesity in Europe is largely unavailable,43 especially when Central and Eastern European countries are included. We decided to include self-reported obesity data in our study, as available studies with self-reported data tended to be more often nationally representative and more recent than studies with obesity data based on measured height and weight. Although, BMI based on self-reported height and weight is likely to underestimate obesity prevalence and may vary across gender, race, age, and education subgroups,45 it has been found to be sufficiently accurate for comparative and relative measures and should not substantially bias the results of this correlational study.46 However, it is possible that the difference between female and male obesity prevalence can be partially explained by gender-specific variation in the accuracy of self-reported weight and height.45 Furthermore, the quality of data identified from international databases may differ depending upon the accuracy and methodology used by reporting countries. To fully understand the between-country differences in obesity patterns across Europe, an internationally comparable database of environmental and policy variables should be developed, including food production and marketing, media activity, transportation and urban planning laws, and national obesity prevention campaigns. Findings from this study should be used not to draw final conclusions but rather to generate hypotheses for further research utilizing stronger study designs.


This exploratory study seeks to call the attention of experts in the field to the need for cross-nationally comparable obesity data and more sophisticated research in order to understand the role of obesogenic environments, with a special focus on policy-related variables. Furthermore, the higher prevalence rates of obesity and stronger associations between obesity and macroenvironmental indicators among women suggest a need for gender-specific studies.

Key points

  • Our study described obesity patterns across Europe and identified macroenvironmental factors associated with obesity prevalence at a national level.

  • Obesity prevalence varied widely across countries and between genders with higher values in Central and Eastern European countries.

  • Obesity was associated with national indicators from economic, food, urbanization, transport, and policy domains.

  • Associations between obesity and macroenvironmental factors were stronger among females than males in all cases.

  • Further research should examine the role of obesogenic environments, with a special focus on policy-related variables and gender differences.


View Abstract