Journal of Public Health

, Volume 16, Issue 6, pp 403–411 | Cite as

Income-related inequality in the distribution of obesity among Europeans

Original Article

Abstract

Aim

The current study concentrates on the issue of income related inequality in obesity for the case of European Union, an association, which has not been thoroughly examined in the literature.

Subjects and methods

Ten European countries for a period of 4 consecutive years (1998–2001) are under consideration, with the information deriving from the “European Community Household Panel” (ECHP) dataset. In order to elaborate on the above association, the concentration index was selected as a means for measuring quantitatively the degree of inequality. Furthermore, an alternative method was introduced, known as the “indirect standardization method,” so as to examine if the observed level of inequality was over-reported.

Results

Treating the European Union as a whole, income inequality in obesity appears to be a burden for the less affluent. Investigation of each country separately reveals that inequality is of most importance for the female population, and especially for the middle-aged one, while no clear association was found for the males. Furthermore, negligence to adjust the models for the education level and the employment status could lead to an over-estimation of the inequality in obesity.

Conclusion

Our primary results attest to the existing literature, showing that a BMI with a value greater than 30 is most likely to be an encumbrance for those of low socioeconomic profiles. However, the extent of inequality in the European Union is found to be low. Effective preventive policies should address the low socioeconomic status female population in Europe, and special attention should be given to the middle-aged.

Keywords

Income inequality Obesity Socioeconomic status Concentration index Indirect estimation method 

Introduction

Various studies have established that the health status of an individual is closely related to the specific socioeconomic conditions that he or she comes across during their lifetime. To that extent, the most widely used indicators, as reviewed by Sobal and Stunkard (1989) and McLaren (2007), are income, educational attainment and employment status. However, despite the fact that these individual characteristics can inform the researchers about their impact on health status, a new tendency has appeared in the literature. The innovative theory is called the “income inequality hypothesis,” which replaces the three above-mentioned conditions as indicators of health with the way the income is distributed in a society. The notion that the health status of each individual is affected by the income inequality-and not the income by itself-was supported by Wilkinson (1997). The central thesis of that hypothesis has to do with the distribution of wealth among citizens within a specific place or country, and among countries at an international level. The more the economic differences among social parties, the worse the health conditions of the society will be.

The harbinger of the income inequality hypothesis comprised studies that concentrated on the mortality and revealed a positive association between income inequality and mortality at the aggregate level of an economy (Kaplan et al. 1996; Kennedy et al. 1998; Kawachi and Kennedy 1997), a negligible relation between the two variables at the individual level (Daly et al. 1998; Soobader and LeClere 1999) or a positive and statistically significant connection of the two at either level (Kaplan et al. 1996; Kennedy et al. 1998). However, the vast majority of the literature has studied the association of inequality and bad self-assessed health condition, either ascertaining the negative effect of poor health on the income inequality and vice versa (van Doorslaer et al. 1997; Kennedy et al. 1998; Soobader and LeClere 1999; Subramanian and Kawachi 2004) or claiming that such a connection is imaginary (Daly et al. 1998).

At the same time, during the last few decades a new health condition has made its appearance throughout the world: the obesity epidemic. Obesity is a leading cause of preventable disease and death next to smoking for many countries. The prevalence of obesity is vividly presented by the estimations of the OECD Health Data (2005), which report a tripling in the numbers of those affected in Europe, as compared to the 1980s. The situation is even worse in the USA, which is ranked in the first place as far as the accumulation of obese people is concerned, followed by Mexico and the UK. This unflattering premiership of the US accounts for the fact that the vast majority of the published research on obesity comes from this country.

Lissau-Lund-Sorensen and Sorensen (1992) suggested that it is possible for the socioeconomic inequality in obesity to stem from the childhood, corroborating the previous work of Power and Moynihan (1988). Others have gone a step further to examine whether time alters the observed socioeconomic inequality or not (Miech et al. 2006; Zhang and Wang 2007). Nevertheless, the results did not reveal an outright pattern, as Sobal and Stunkard (1989) and McLaren (2007) have noted. Some works claim that the evidence does not support the association between socioeconomic inequality and obesity (Gordon-Larsen et al. 2003), while some allege that the effect decreases as time goes by (Zhang and Wang 2007).

That disparity in the findings can be justified if gender differences are taken into account in the analysis (Sobal and Stunkard 1989). It is widely believed that a lower socioeconomic condition is associated with a greater body mass index primarily for women, whereas the findings can be ambiguous for men and children. For example, Kinra et al. (2000) confirm that the income inequality in obesity is greater for the female population as compared to the male one, with the former referring to both adolescents and adults, while the latter includes only the adults. However, the results are bound to differ according to the country of reference. As Wang (2001) suggested, for the low-income countries, a positive association between income and obesity appears, which flattens out for the middle-income countries, until the high-income countries are met, where a negative association prevails.

Despite the fact that obesity has been on the agenda of health economists in the European continent, the existence of inequalities in the obese population has not been widely examined in the literature. To our knowledge, the existent studies either concentrate only on one country (Costa-Font 2005) or resort to different econometric methods for the measurement of inequality in obesity than the one used in this study (Lorant and Tonglet 2000; Pickett et al. 2005). Given the lack of adequate literature for the European Union, the main goal of the current paper is to quantify the degree of income inequality in the distribution of obesity, using the European Community Household Panel and applying the most appropriate index of inequality-as discussed in the following section.

Methods

Several indices have been developed by health economists in order to quantify the degree of income inequality in health and to avoid certain limitations of the logistic and linear regressions (Zhang and Wang 2007), which are usually used in similar studies. The most characteristic of them are the range, the index of dissimilarity, the slope index of inequality and the relative index of inequality (Kakwani et al. 1997), the Gini coefficient (Kennedy et al. 1998; Soobader and LeClere 1999; Pickett et al. 2005) and the concentration index (Kakwani et al. 1997; van Doorslaer et al. 1997; Zhang and Wang 2004; Costa-Font 2005; Zhang and Wang 2007). At the same time, other measures have appeared, such as the poverty income ratio (Kaplan et al, 1996; Miech et al. 2006; Zhang and Wang 2007), the ratio of the most affluent 20% of the population to the corresponding poorest part of the population (Pickett et al. 2005) or the Robin Hood Index (Lorant and Tonglet 2000).

To this end, the work of Wagstaff et al. (1991) that discusses and compares those measures is invaluable. According to their findings, the slope, the relative index of inequality and the concentration index (CI) seem to be the most complete, as they satisfy the three prerequisites the authors pose: first, they can interpret the health inequalities that arise from the socioeconomic characteristics; second, they are representative of the whole population; finally, they are “sensitive to the distribution of the population across socioeconomic groups.” In the literature, the two most commonly used measures are the Gini coefficient and the CI. Nevertheless, the Gini coefficient has a deficiency compared to the CI, which is used in the current paper, because it does not satisfy the first requirement.

The CI is the quantitative expression of the concentration curve (Fig. 1), which is a graphical presentation of the level of inequality (Wagstaff et al. 1991; Zhang and Wang 2004). The cumulative percentage of the income of the population is placed on the horizontal x-axis in increasing order, while on the vertical y-axis the cumulative percentage of the variable in question is presented, which in this case is obesity. Three possible places for the L (p) curve can appear; if it coincides with the line of 45ο degrees, then the distribution of obesity is equal among different income levels (CI=0). In other words, we refer to an egalitarian society where no inequalities due to obesity exist. In the case where the concentration curve is above the line of equality ([L1(p)], that would mean a disproportionate concentration of obese people in the low income levels (CI<0). Adversely, had the curve been placed below the diagonal ([L2(p)], obese individuals would have been concentrated on the high-income levels (CI>0).
Fig. 1

Concentration curve for obesity

Theoretically, the CI can take values from –1 to 1 and can be computed by the “convenient covariance” method, which was proposed by Jenkins (1988) and Lerman and Yitzaki (1989):
$$C.I. = \frac{{2\operatorname{cov} \left( {y_i ,R_i } \right)}}{\mu }$$
where the index i refers to each individual in the population, y stands for the measurement of the health condition, and thus obesity, R, is the place of each person in the distribution of the income, cov(yi,Ri) expresses the covariance between the two variables, and μ represents the mean value of the health variable in the population. However, this method does not provide any information about the statistical significance of the computed indices. Kakwani et al. (1997) alleged that the previous covariance could be used to estimate a “convenient regression” of the form:
$$2\sigma _R^2 \left( {\frac{{y_i }}{\mu }} \right) = \alpha + \beta R_i + e_i $$
where the interpretation of the variables is the same, and \(\sigma _R^2 \) corresponds to the variance of the rank variable. Since this equation is estimated, the coefficient β will give the value of the CI and its standard error, allowing for statistical inferences. Another alternative is to use the method suggested by Newey and West (1994), whose estimator takes into account the autocorrelation as well as the heteroscedasticity problems. In this study the CI was calculated using both methods. However, the results where almost equal, and thus, the presented findings come from the Newey-West method, due to the corrected standard errors.

At the same time, it is quite possible that obesity is being influenced by other demographic and socioeconomic factors. In order to adjust for these effects, the analysis is separated into two parts. The first method uses the CI as computed by the Newey-West estimation. The findings for the whole population are reported, which do not account for other factors. Thereafter, the population is stratified by gender, and each gender is separated according to its age group.

The second method is the indirect standardization (Kakwani et al.1997). In summary, it describes the distribution that it would have aroused had each individual had his or her own age, but all of them had a common gender-age effect as the whole population. The proposed formula takes the form:
$$y_i = a + \sum\limits_j {\beta _j x_{ji} + \sum\limits_k {\gamma _k w_{ki} + } e_i } $$
where yi is the variable of obesity, xji stands for the j variables for which we want to standardize (such as the age and the gender), and wki reflects the k variables for which we do not wish to standardize, but want to examine their impact (education, marital status and employment). The question is how the obesity would be distributed among the income clusters regardless of the distribution of the xji variables. Thus, the expression:
$$\hat y_i = y_i + \bar y - \hat y_i^\prime $$
(1)
has to be estimated. The only unknown term is the last one describing the function: \(\hat y_i^\prime = \hat a + \sum\limits_j {\hat \beta _j x_{ji} + \sum\limits_k {\hat y_k \bar w_k } } \). Only after the observed values of the variables for which we want to standardize (xji) and the mean values of the variables that we only want to examine \(\left( {\bar w_k } \right)\) are taken into account can the predicted values of obesity be obtained, which, in turn, will be substituted in equation (1).

Data

Our data originate from the European Community Household Panel (ECHP), UDB-version December 2003, which constitutes a unique dataset generated by the European Union Statistical Office (Eurostat). It covers a wide range of information for individuals older than 16 years old, such as demographic characteristics, income, employment, education and training, housing conditions, health, immigration and financial situation. The data were collected from the 15 members of the European Union-Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, The Netherlands, Portugal, Spain, Sweden and the UK, covering the period 1994–2001. However, our sample is limited to ten countries and to the last 4 years (1998–2001), since for the cases of France, Germany, Luxembourg, The Netherlands and the UK, no information was available about the body mass index and for the first 4 years the questions concerning the weight, height and BMI were deficient.

Our sample consists of a total of 336,386 persons of whom 30,849 or 9.2% are characterized as obese. The number of the respondents is almost evenly distributed between the two genders, with 48.35% (162.647) being men and the remaining 51.65% women (173.739).

Two basic variables are used in this study: obesity and income as a proxy for the socioeconomic status. According to the World Health Organization, obesity is just one of the categories of body mass index (BMI). Division of a person’s weight, expressed in kilograms, by height in a quadratic form gives the formula of the BMI, or mathematically as [kg]/height2[m2]. Different values translate into different health states, which vary from underweight and normal, to overweight and obesity. Since the last cluster is the one of interest, obesity, it contains individuals with a BMI greater than 30.

In order to calculate the CI, a measure is needed that will help to rank the population. The most commonly used measures for the socioeconomic status (SES) are income, educational attainment and occupational status. The first variable prevails over the other two, mainly because income is a continuous variable allowing the ranking of the sample from the least wealthy persons to the wealthiest ones. Conversely, education and occupational status are categorical variables as well as they tend to remain stable (Zhang and Wang 2004). The ECHP contains information on the total net household income, which is expressed in the national currency of each country and flawed by the level of yearly inflation. After we correct for the above two problems, we take the ratio of the customized income to the equivalized number of members in each household. Furthermore, for the first half of the analysis two demographic variables, age and gender, are used to stratify the sample. Five age groups were created: 17–29, 30–44, 45–54, 55–64 and over 65 years old.

The second half of this study examines the impact of education, marital and employment status on the income inequality in obesity. Education level of the participants was calculated using the highest level of the degree they have obtained. The dummy primary education refers to people who have completed at most the first stage of secondary education. As a median level of education, we consider the second level of the secondary education expressed from middle education, whereas high education contains the respondents who have graduated from a third-level institution. Also, three marital status categories were created, and the participants were split into married, single and those having been sometime in their lifetime married (divorced, separated or widowed). Finally, employment status led to four distinct categories: employed, self-employed, unemployed and inactive. In the last group we include those who are unemployed, but are not willing to find a job, students, house workers, pensioners and the army employees. The distribution of the obese population in the above-mentioned socioeconomic characteristics, as a percentage of the whole population, is presented in Table 1.
Table 1

Characteristics for the obese population as percentage of the sample (per gender)

Variables/countries

Austria

Belgium

Denmark

Finland

Greece

Ireland

Italy

Portugal

Spain

Sweden

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Male

Female

Age 17–29

3.08

1.87

3.23

4.99

5.66

6.03

4.19

4.21

2.82

2.08

2.58

4.19

2.20

1.05

3.48

2.81

4.71

2.87

4.09

2.84

Age 30–44

9.16

6.79

10.33

6.14

7.93

6.89

10.75

9.16

8.95

7.51

10.14

8.39

6.36

3.77

8.95

9.29

13.32

7.19

9.10

7.83

Age 45–54

17.35

13.91

14.69

13.39

10.94

11.53

14.15

16.06

13.81

12.25

13.67

11.50

12.45

9.75

13.22

16.23

16.54

14.70

10.39

9.92

Age 55–64

17.41

16.85

14.75

15.83

15.39

11.93

18.81

23.81

15.86

14.66

10.08

12.20

14.93

13.91

13.12

16.00

19.79

22.99

12.74

12.24

Age over 65

14.59

15.57

12.88

16.34

11.25

10.61

12.74

17.74

10.89

14.40

7.55

9.64

12.09

13.49

11.87

12.57

16.59

22.08

10.05

12.88

Primary educ

11.17

13.20

13.21

15.81

13.31

12.02

13.01

17.09

10.60

13.15

9.31

10.85

10.54

10.38

9.93

12.29

15.44

17.32

10.35

11.35

Middle educ

11.77

8.56

11.41

9.26

9.58

8.23

10.18

12.43

8.80

4.79

6.99

7.23

5.74

2.98

5.14

2.87

8.77

4.20

9.87

9.01

eigh Educ

3.94

2.76

7.24

5.73

6.02

6.29

10.95

8.40

9.40

4.53

7.07

4.64

5.68

2.69

6.78

4.24

9.27

3.13

5.48

5.59

Married

13.77

11.39

12.91

10.94

11.08

9.99

13.48

14.27

12.38

10.85

10.06

10.30

10.95

8.02

11.36

11.93

16.77

14.07

9.99

9.53

Singles

5.40

3.46

5.32

5.45

5.90

5.82

7.12

7.00

3.54

2.73

6.00

4.69

3.33

2.31

3.85

3.65

6.03

4.36

7.07

6.03

Div/sep/wid

14.17

15.28

10.63

13.81

12.60

10.13

12.77

16.75

8.33

13.80

12.36

9.76

10.11

13.27

10.43

13.97

13.67

21.00

11.39

11.60

Employed

8.80

5.23

10.49

6.36

9.24

8.12

10.52

9.96

8.16

5.76

7.58

7.03

7.03

3.66

8.01

6.97

11.78

5.29

8.56

7.22

Self-employed

13.53

12.39

9.06

6.45

7.51

11.93

13.71

13.97

13.34

10.74

10.88

11.16

10.79

5.11

11.75

11.14

16.97

8.88

9.32

6.50

Unemployed

12.85

8.39

21.11

13.36

13.49

9.34

8.98

16.69

5.96

5.07

7.75

7.76

4.29

3.07

7.88

9.10

11.46

8.10

12.26

10.87

Inactive

13.81

13.42

10.22

13.62

10.20

9.74

11.76

15.10

9.57

11.84

7.44

9.41

10.04

9.57

9.81

13.08

13.38

16.16

8.65

10.69

Total *

11.06

10.16

10.75

10.36

9.52

8.85

11.27

12.60

9.89

9.94

8.14

8.52

8.45

7.29

9.21

10.58

12.98

12.53

8.84

8.70

Mean income**

12.78

12.59

13.76

13.51

10.03

9.96

11.78

11.66

15.32

15.15

9.76

9.62

10.39

10.26

14.79

14.57

14.83

14.70

9.98

9.95

*Percentage of obese persons per gender to the total number of individuals per gender. **Mean of logarithmic equivalized household income for the obese

Results using the concentration index

The CIs for each of the ten examined countries are presented in Table 2. Starting with the population as a whole, the CIs for Austria, Greece and Sweden do not appear to be different from the zero point, stating the nonexistence of income inequality due to obesity, while the results for Denmark and Finland indicate a positive association between obesity and SES. The remaining five countries exhibit the expected negative sign, where the greatest level of income inequality among the obese appears in Belgium (-0.084), while the lowest in Portugal (-0.032). For Spain, the results are in accordance with those in the survey of Costa-Font (2005), although the level of inequality is greater in the latter study (-0.063 versus -0.107).
Table 2

Concentration indices for the obese population by gender and age groups for Austria, Belgium, Denmark, Finland, Greece, Ireland, Italy, Portugal, Spain and Sweden

 

Austria

Belgium

Denmark

Finland

Greece

Ireland

Italy

Portugal

Spain

Sweden

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

N

CI

CI+

Total

23929

-0.014

-0.040, 0.011

18838

-0.084 ***

-0.112, -0.056

15452

0.036 **

0.005, 0.068

24760

0.043 ***

0.019, 0.066

37328

-0.003

-0.024, 0.017

19935

-0.049 ***

-0.080, -0.018

58002

-0.037 ***

-0.055, -0.018

44008

-0.032 ***

-0.052, -0.013

47727

-0.063 ***

-0.078, -0.048

14934

-0.005

-0.035, 0.024

Men

11573

0.010

-0.027, 0.047

8801

-0.054 **

-0.097, -0.013

7610

0.014

-0.030, 0.059

12273

0.084 ***

0.048, 0.112

17657

0.016

-0.013, 0.047

9791

-0.011

-0.056, 0.034

28330

-0.033 **

-0.059, -0.006

20813

0.027 *

-0.002, 0.057

23128

-0.039 ***

-0.061, -0.018

7261

0.006

-0.036, 0.048

17–29

2676

0.127 *

-0.013, 0.268

1733

-0.016

-0.197, 0.165

1605

0.011

-0.123, 0.145

3000

0.101 *

-0.015, 0.217

3697

0.147 ***

0.037, 0.258

2748

0.321 ***

0.152, 0.491

6932

0.016

-0.083, 0.117

5795

0.066

-0.026, 0.159

6062

-0.006

-0.078, 0.066

1093

0.086

-0.067, 0.241

30–44

3175

-0.029

-0.107, 0.049

2872

-0.142 ***

-0.220, -0.064

2408

-0.010

-0.106, 0.084

3356

-0.045

-0.114, 0.024

4500

-0.062 *

-0.125, 0.000

2426

-0.004

-0.079, 0.070

8171

-0.142 ***

-0.197, -0.085

4925

-0.072 **

-0.134, -0.009

6294

-0.078 ***

-0.118, -0.038

2132

-0.042

-0.119, 0.033

45–54

1768

0.004

-0.068, 0.076

1631

-0.064

-0.142, 0.014

1417

-0.002

-0.095, 0.092

2562

0.052

-0.017, 0.123

2885

0.042

-0.019, 0.104

1592

-0.071

-0.158, 0.015

4588

-0.023

-0.075, 0.028

2987

0.059 *

-0.003, 0.122

3215

-0.035

-0.085, 0.014

1470

0.031

-0.055, 0.116

55–64

1747

-0.121 ***

-0.193, -0.048

987

-0.029

-0.123, 0.064

1018

-0.180 ***

-0.271, -0.090

1770

0.016

-0.047, 0.080

2532

0.052 *

-0.008, 0.112

1330

-0.069

-0.168, 0.029

4023

-0.082 ***

-0.133, -0.030

2770

-0.015

-0.077, 0.047

2750

-0.068 ***

-0.114, -0.022

1158

-0.061

-0.146, 0.023

65+

1973

0.029

-0.043, 0.102

1577

-0.036

-0.125, 0.052

1160

0.029

-0.066, 0.126

1585

0.024

-0.065, 0.115

4043

-0.019

-0.076, 0.036

1695

-0.028

-0.134, 0.078

4616

-0.037

-0.091, 0.016

4336

0.090 ***

0.034, 0.147

4489

-0.037 *

-0.081, 0.006

1405

-0.117 ***

-0.205, -0.029

Women

12356

-0.039 **

-0.074, -0.004

10037

-0.109 ***

-0.146, -0.072

7842

0.058 **

0.013, 0.103

12487

0.009

-0.022, 0.041

19671

-0.022

-0.050, 0.006

10144

-0.081 ***

-0.123, -0.039

29672

-0.042 ***

-0.069, -0.015

23195

-0.077 ***

-0.103, -0.052

24599

-0.086 ***

-0.106, -0.066

7673

-0.017

-0.058, 0.023

17–29

2543

-0.213 **

-0.381, -0.045

1897

-0.123 *

-0.249, 0.003

1684

0.120 **

0.009, 0.232

3038

-0.003

-0.113, 0.107

4053

-0.031

-0.169, 0.107

2717

0.028

-0.090, 0.147

7036

-0.132 *

-0.283, 0.018

5508

-0.282 ***

-0.393, -0.170

5951

-0.203 ***

-0.296, -0.111

1186

0.091

-0.099, 0.282

30–44

3241

-0.091 *

-0.183, 0.000

3249

-0.209 ***

-0.296, -0.123

2364

-0.015

-0.109, 0.078

3406

-0.116 ***

-0.187, -0.046

4658

-0.064 *

-0.137, 0.007

2538

-0.182 ***

-0.266, -0.097

8198

-0.103 ***

-0.181, -0.025

5059

-0.161 ***

-0.217, -0.105

6181

-0.209 ***

-0.267, -0.152

2208

-0.084 **

-0.164, -0.004

45–54

1851

-0.232 ***

-0.307, -0.157

1683

-0.192 ***

-0.272, -0.113

1396

0.014

-0.084, 0.112

2616

-0.115 ***

-0.176, -0.055

2975

-0.139 ***

-0.202, -0.077

1641

-0.141 ***

-0.227, -0.055

4732

-0.154 ***

-0.212, -0.096

3447

-0.106 ***

-0.158, -0.056

3434

-0.122 ***

-0.172, -0.073

1524

-0.088 **

-0.175, -0.002

55–64

1896

-0.159 ***

-0.228, -0.090

1107

-0.234 ***

-0.321, -0.148

1078

-0.053

-0.151, 0.044

1747

-0.092 ***

-0.150, -0.034

2751

-0.036

-0.093, 0.020

1372

-0.082 *

-.0176, 0.010

4042

-0.121 ***

-0.174, -0.068

3314

-0.036

-0.089, 0.016

2960

-0.094 ***

-0.135, -0.054

1150

-0.171 ***

-0.257, -0.086

65+

2600

0.042

-0.016, 0.101

2100

-0.033

-0.097, 0.030

1319

0.027

-0.072, 0.125

1680

-0.087 **

-0.156, -0.019

5234

0.058 ***

0.015, 0.101

1876

-0.015

-0.102, 0.071

5664

-0.098 ***

-0.142, -0.054

5867

-0.006

-0.052, 0.040

5775

-0.079 ***

-0.109, -0.049

1601

-0.101 ***

-0.176, -0.025

Source: Calculations of the authors based on the ECHP. The signs *, ** and *** refer to statistical significance at the levels of 1%, 5% and 10%, respectively

+Confidence Interval

For cross-country comparisons, the t-statistics were also calculated. The results, which are not presented here, were statistically significant for the majority of the countries, even though low. As a matter of fact, income inequality did not seem to be a concern for the obese Scandinavians. Specifically for Sweden, no burden appeared for the obese population due to their SES status, while the positive CIs documented for the cases of Denmark and Finland mean that inequality does not have a negative effect on those who are classified as obese. A plausible explanation for this pattern may be the well-organized welfare state of the Scandinavian countries, as compared to the other EU countries (Kautto et al. 2001).

Since none of the indices surpass the level of -0.1 or 0.1, it is an indication that the inequality in obesity for the European countries is quite negligible, something that is quite plausible from Table 1 as well. Nevertheless, treating the European countries as a whole would have led to a negative CI.

Will the findings continue to be the same if certain subgroups of the population are examined? In order to answer this question, the sample is firstly stratified by sex. For the male group, seven countries follow the same pattern as the entire population, two countries (Denmark, Ireland) report a no longer statistically significant CI, and one country (Portugal) shows a conversion on its sign. The results indicate that obese men in the EU are unlikely to come up with an income inequality as for five countries the CIs are not statistically significant. Had they been citizens in Finland or Portugal, males of high level of earnings would be more probable to be obese, whereas low-income obese males would be more possible to face socioeconomic inequality had they lived in Belgium, Italy or Spain. On the other hand, the female subgroup did not report any significant difference as compared to the whole population. Only for the case of Finland an alteration is marked; for women, inequality does not represent a problem since the CI is no longer significantly different from zero. Six of the examined countries confirmed that there is a lesser likelihood for the females of higher social status to be obese in comparison with their fellow females of lower social status.

Hence, for the majority of the EU countries, the socioeconomic inequality in obesity is more prevalent for the female group and not for the male one. This tendency is a characteristic of the developed countries (Kinra et al. 2000; Wardle and Griffith 2001) owing to the greater emphasis women place on their physical appearance. At the same time, Zhang and Wang (2004), who applied the same method of analysis for the USA, affirm that result, as obesity affects both men (-0.015) and women (-0.082) of lower socioeconomic status.

Turning now to the age of the respondents, no clear association appears among age groups and inequality for men. The value of the CI surpassed zero for the group of 17–29, and it was below that threshold mostly for the age from 30 to 44 as well as from 55 to 64. However, the pattern is more straightforward for the case of women. Where the CIs are statistically different from zero, the reverse relationship between obesity and socioeconomic status is established; more income diminishes the probability of exhibiting a BMI of more than 30 points. In fact, the degree of inequality for the obese women was at its summit for the middle-aged groups covering a period of 2 1/2 decades (30–54). A more comprehensive illustration of this concept appears in Fig. 2, where the CIs for the middle aged women are presented. All the bars are below the x-axis, reflecting that inequality concentrates on the low socioeconomic categories. At the same time, only for Finland and Sweden no great difference is documented for these two age groups. Among other factors, this trend may stem from unhealthy dietary habits and lack of physical exercise (Sarlio- Lähteenkorva et al. 2005) or from psychological conditions such as stress and employment strain, affecting mainly the middle-aged women (Wamala et al. 1997). Once again, the findings can be juxtaposed with those of Zhang and Wang (2004).
Fig. 2

Concentration indices for the middle–aged women (30–44 and 45–54 years old). Comment: Created by the authors, using the ECHP dataset. Denmark is not included as the concentration index was not statistically significant for these age groups

Results using the indirect estimation method

In Table 3, the first column reports the results from the unstandardized CIs whose patterns are in accordance to the ones observed in Table 2. In order to adjust the findings for different characteristics, the age and gender are added (second column), then the educational level (third column), the marital status (fourth column) and finally the employment status of each respondent (last column). This procedure constitutes the standardized method for the calculation of the CI. In other words, the results in the second column represent the CI if each person in the sample had the same age and was of the same sex as the total population.
Table 3

Estimates for the concentration index using the unstandardized and the indirect standardized method

 

Standardized

 

Unstandardized

Age and eex

Age, sex, education

Age, sex, education, marital

Age, sex, education, marital, employment status

Austria

-0.01142

-0.03629***

-0.02411**

-0.02421**

-0.02098**

Belgium

-0.08249***

-0.10182***

-0.07487***

-0.07236***

-0.06318***

Denmark

0.02375*

-0.00769

-0.00163

-0.00068

0.00088

Finland

0.01352*

-0.04241***

-0.03420***

-0.03468***

-0.03276***

Greece

-0.00097

-0.00361

0.00480

0.00613

0.00958

Ireland

-0.05842***

-0.07219***

-0.05023***

-0.04905***

-0.04799***

Italy

-0.03079***

-0.06574***

-0.04949***

-0.04943***

-0.04483***

Portugal

-0.02552***

-0.02544***

-0.01423*

-0.01464*

-0.01492*

Spain

-0.03776***

-0.04883***

-0.03258***

-0.03208***

-0.03000***

Sweden

0.01525

-0.01087

-0.00627

-0.00991

-0.00092

Source: Calculations of the authors based on the ECHP. The signs *, ** and *** refer to statistical significance at the levels of 1%, 5% and 10%, respectively

The juxtaposition of the first two columns allows the classification of countries to three chief clusters. Greece and Sweden comprise one of these groups. Even after the adjustment for age and sex, the CI appears to lie on the egalitarian line, meaning that the probability of being obese is the same either if someone is of higher socioeconomic status or is of the lowest one. The second group contains Finland whose value of the CI drops below zero once the standardized method is pursued. The translation for the reported change is that when the individuals have the mean age-sex of the whole population, the distribution of obesity is more concentrated in the poorest part of the population than in the wealthiest one-as currently observed. Austria is also added in the second group, as the CI becomes statistically significant with a negative sign. Finally, Belgium, Ireland, Italy, Portugal and Spain are classified into the third group as the unstandardized indices fall short of the standardized ones. Had the respondents been of the mean population age and sex, the distribution of obesity would be even more unequal from the one observed. Denmark is a special case floating between the second group, as the CI for the standardized method is positively associated with the inequality, and the first group due to the non-significance of the index once the standardized method is used.

What is worth mentioning is that the ten countries after using the indirect standardized method are classified in two groups; for seven countries inequality has an unfavorable effect on the lowest socioeconomic part of their population, while for Denmark, Greece and Sweden, no indication for such inequality is found. Thus, the discussion that follows will nest to the former group of the seven countries. Gradual addition of the education, marital and employment status variables can reveal whether there is an over-report or an under-report of the level of inequality for the obese part of the population.

Educational attainment is considered to be an indicator that explains the inequality in obesity as it can capture the effects of the conditions of an individual while in childhood and adolescence (Kinra et al. 2000). Once the education level is considered (Table 3, third column) the association between income inequality and obesity is attenuated (Subramanian and Kawachi 2004). The same does not seem to apply for the case of the marital status (column four). Since the values of the CIs remain almost the same, elimination of the co-habilitation status is not expected to create a major bias in the results.

However, the addition of the employment status (Table 3, last column) attenuates further the values of the CI, leading to a more egalitarian society for each one of the seven European countries. Thus, the core conclusion from the indirect standardization method is that certain demographic and socioeconomic characteristics can account for the reported level of inequality in the obese population. The current findings indicate that not paying attention to education and employment status leads to over-reported results, whereas marital status exerts only a minimal impact on inequality. As Wilkinson (1997) had mentioned, controlling for some socioeconomic conditions of an individual is like absorbing the effect of the income inequality.

Caveats

Before summarizing the central findings of the article, it is essential to make a reference to some limitations facing this work. The tendency of the respondents to either over report their height or underreport their weight can lead to biased results for the variable of obesity. Nevertheless, that form of bias is not expected to be significant as the dataset was constructed through personal interviews, diminishing the probability that the respondents will proclaim a measure that diverts from what the interviewer sees. On the other hand, there are no significant discrepancies between a model that corrects for biases resulting from reporting errors and a model that ignores such biases. Moreover, the ECHP dataset does not provide details about the race of each respondent; thus, it was not possible to examine whether ethnic origin is a determining factor for the relation of obesity and income inequality in EU countries or not. With such information, more robust conclusions could have been drawn, despite the fact that Zhang and Wang (2004) do not find that minority groups are bound to show a higher level of income inequality. Finally, Costa-Font (2005) adverts to the endogeneity bias limitation, which is almost certain to hold for every survey that focuses on the inequalities in obesity.

Conclusions

Almost every study engaged with the association between obesity and socioeconomic conditions reports a negative effect for those categorized in the lowest socioeconomic status. In the current paper a different aspect of the matter is examined; does income inequality exist, and if the answer is positive, is the phenomenon of obesity concentrated in the lower income levels? Despite some differences, our main findings are in line with the existing literature, whatever statistical method is used. If the EU is treated as a whole, the income inequality in obesity is a burden for the lowest socioeconomic classes. Investigation of each country separately presented statistically significant discrepancies of inequality in obesity-even though low-among the EU countries, without, though, deviating from the general pattern. The main results can be summarized in the following points: first, the extent of inequality in the EU is found to be low. Second, the inequality is of most importance for the female subpopulation compared to the male one. Third, even if the sample is stratified according to the age of each individual, the inequality is a burden for the women, and especially for the middle aged, compared to the men, as for the latter no clear association was found. Fourth, other individual characteristics such as education, marital and employment status can provide an insight in the understanding of the reported inequality. Among those traits, educational attainment and occupational status seem to have the most important effects on inequality, whereas the marital status has just a minimal impact on it. Therefore, failing to take into account the previous three variables will lead to an over-estimation of the observed inequality in obesity. Finally, preventive policies against obesity, in order to be effective, should target the low SES female population in Europe and special attention should be given to the middle-aged.

Notes

Acknowledgements

None.

Conflict of interest statement

The authors confirm that there are no relevant associations that might pose a conflict of interest.

References

  1. Costa-Font J (2005) Are there socioeconomic inequalities in obesity in Spain? FEDEA 216:1–26Google Scholar
  2. Daly MC, Duncan GJ, Kaplan GA, Lynch JW (1998) Macro to micro links in the relation between income inequality and mortality. Milbank Quart 76:315–339CrossRefGoogle Scholar
  3. Gordon-Larsen P, Adair LS, Popkin BM (2003) The relationship of ethnicity, socioeconomic factors, and overweight in U.S. adolescents. Obes Res 11:121–129Google Scholar
  4. Jenkins S (1988) Calculating income distribution indices from micro data. National Tax J 41:139–142Google Scholar
  5. Kakwani N, Wagstaff A, van Doorslaer E (1997) Socioeconomic inequalities in health: measurement, computation, and statistical inference. J Econometrics 77:87–103CrossRefGoogle Scholar
  6. Kaplan GA, Pamuk ER, Lynch JW, Cohen RD, Balfour JL (1996) Inequality in income and mortality in he United States: analysis of mortality and potential pathways. Brit Med J 312:999–1003PubMedGoogle Scholar
  7. Kautto M, Fritzell J, Hvinden B, Kvist J, Uusitalo H, Nordic Welfare States in the European Context. Routlege, NY; 2001Google Scholar
  8. Kawachi I, Kennedy BP (1997) The relationship of income inequality to mortality: does the choice of indicator matter? Soc Sci Med 45:1121–1127PubMedCrossRefGoogle Scholar
  9. Kennedy BP, Kawachi I, Glass R, Stith DP (1998) Income distribution, socioeconomic status and self-rated health in the United States: multilevel analysis. Brit Med J 317:917–921PubMedGoogle Scholar
  10. Kinra S, Nelder R, Lewendon G (2000) Deprivation and childhood obesity: a cross-sectional study of 20,973 children in Plymouth, United Kingdom. J Epidemiol Commun Health 54:456–460CrossRefGoogle Scholar
  11. Lerman RI, Yitzaki S (1989) Improving the accuracy of estimates of Gini coefficients. J Econometrics 42(1):43–47CrossRefGoogle Scholar
  12. Lissau-Lund-Sorensen I, Sorensen TIA (1992) Prospective study of the influence of social factors in childhood on risk of overweight in young adulthood. Int J Obes Rel Met Disord 16:169–175Google Scholar
  13. Lorant V, Tonglet R (2000) Obesity: trend in inequality. J Epidemiol Commun Health 54:637–638CrossRefGoogle Scholar
  14. McLaren L (2007) Socioeconomic status and obesity. Epidemiol Rev, Department of Community Health Sciences, University of Calgary :1–20Google Scholar
  15. Miech RA, Kumanyida SK, Stettler N, Link BG, Phelan JC, Chang VW (2006) Trends in the association of poverty with overweight among US adolescents, 1971–2004. J Am Med Ass 295:2385–2393CrossRefGoogle Scholar
  16. Newey WK, West KD (1994) Automatic lag selection in covariance matrix estimation. Rev Econ Stud 61(4):631–653CrossRefGoogle Scholar
  17. Pickett KE, Kelly S, Brunner E, Lobstein T, Wilkinson RG (2005) Wider income gaps, wider waistbands? An ecological study of obesity and income inequality. J Epidemiol Commun Health 59:670–674CrossRefGoogle Scholar
  18. Power C, Moynihan C (1988) Social class and changes in weight-for-height between childhood and early adulthood. Int J Obes Rel Met Disord 12:445–453Google Scholar
  19. Sarlio- Lähteenkorva S, Lissau I, Lahelma E (2005) The social patterning of relative body weight and obesity in Denmark and Finland. Eur J Pub Health 16(1):36–40CrossRefGoogle Scholar
  20. Sobal J, Stunkard AJ (1989) Socioeconomic status and obesity: a review of the literature. Psychol Bull 105:260–275PubMedCrossRefGoogle Scholar
  21. Soobader MJ, LeClere FB (1999) Aggregation and the measurement of income inequality: effects on morbidity. Soc Sci Med 48:733–744PubMedCrossRefGoogle Scholar
  22. Subramanian SV, Kawachi I (2004) Income inequality and health: what have we learned so far? Epidemiol Rev 26:78–91PubMedCrossRefGoogle Scholar
  23. van Doorslaer E, Wagstaff A, Bleichrodt H, Calonge S, Gerdtham Ulf-G, Gerfin M, Geurts J, Gross L, Hakkinen U, Leu RE, O’ Donnell O, Propper C, Puffer F, Rodriguez M, Sundberg G, Winkelhake O (1997) Income-related inequalities in health: some international comparisons. J Health Econ 16:93–112PubMedCrossRefGoogle Scholar
  24. Wagstaff A, Paci P, Van Doorslaer E (1991) On the measurement of inequalities in health. Soc Sci Med 33:545–557PubMedCrossRefGoogle Scholar
  25. Wamala SP, Wolk A, Orth-Gomer C (1997) Determinants of obesity in relation to socioeconomic status among middle-aged Swedish women. Prev Med 26:734–744PubMedCrossRefGoogle Scholar
  26. Wang Y (2001) Cross-national comparison of childhood obesity: the epidemic and the relationship between obesity and socioeconomic status. Int J Epidemiol 30:1129–1136PubMedCrossRefGoogle Scholar
  27. Wardle J, Griffith J (2001) Socioeconomic status and weight control practices in British adults. J Epidemiol Commun Health 55:185–190CrossRefGoogle Scholar
  28. Wilkinson R (1997) Health inequalities: relative or absolute material standards. Brit Med J 314:591–595PubMedGoogle Scholar
  29. Zhang Q, Wang Y (2004) Socioeconomic inequality of obesity in the United States: do gender, age, and ethnicity matter? Soc Sci Med 58:1171–1180PubMedCrossRefGoogle Scholar
  30. Zhang Q, Wang Y (2007) Using concentration index to study changes in socioeconomic inequality of overweight among US adolescents between 1971 and 2002. Int J Epidemiol1–10Google Scholar

Copyright information

© Springer-Verlag 2008

Authors and Affiliations

  1. 1.Department of EconomicsUniversity of MacedoniaThessalonikiGreece

Personalised recommendations