Does bodyweight have an influence on Labour
Income?
A special case for European citizens
Bachelor Thesis Economics and Business Economics Student: Renzo Koolhaas Student number: 10665781 Year 2016-2017
Statement of Originality
This statement is written by student Renzo Koolhaas who declares to take full
responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The faculty of Economics and Business of the University of Amsterdam is responsibly solely for the supervision of completion of the work, not for the contents.
Excessive bodyweight:
an handicap to social
advancement
Brunello and D’Hombres, 2007, p.2
Abstract
In this research, the relation between bodyweight and labour income is being studied. Using the European Health Interview Survey by Eurostat, which contains descriptive data of percentages of the population of most European countries belonging to certain age- and genderclasses, an answer to the question ‘’Does bodyweight have an influence on labour income?’’ will be elaborated on. By looking at those 28 different countries in Europe, I have found that, in general, there exists a correlation between bodyweight and labour income. Most correlations have been found for females aged 25 years and older. With the data used, only a correlation has been proved; further research should make sure whether bodyweight is of influence on labour income or the other way around.
List of Contents
Part I – Introduction and Background……… 5 Introduction………...…… 5 Background of the study ………...…... 5 This Research……….6 Part II – Literature Review………..……... 8 Part III – Emprical Framework………... 11 Hypothesis……….. 11 Methodology……….. 12 Part IV – Data……….. 14 Dataset………... 14 Summary Statisticis………... 15 Part V – Results……….. 20 General Results for Europe………...20 Results for countries seperately……….... 21 Part VI – Discussion……… 25 Discussion of the Results………... 25 Discussion of this Research………... 26 Part VII – Conclusion………..28 Reference List………30 Appendix……….. 31Part I – Introduction and Background
Introduction
‘’Little and fat people are more likely to be poor’’; an expression of Margeet Vermeulen in the Volkskrant (2016). According to an investigation in the United Kingdom, this means that, especially for women, labour income can be lower for people who are overweight than people with a normal bodyweight, holding other factors constant.
By first time of hearing, such a personal characteristic, a priori not saying anything about intelligence, schooling and workexperience having an influence on labour income, seems to be discrimination. Why would someone with more bodyfat have a higher chance to be poor?
A lot of research on the relation between bodyweight and labour income has already be done. For example, Cawley (2004, pp. 451-474) agrees to see a relation between labour income and bodyweight. He describes three explanations why a correlation between bodyweight and labor market outcomes exists. First, low wages or employment rates could cause obesity, for instance as a result of poorer people consuming cheaper, more fattening foods. Second, lack of self-confidence could cause both obesity and low wages or employment rates. Third, obesity might lower wages or employment rates, for instance by lowering productivity or because of employer discrimination.
Considering those three explanations, actually the relationship, if any, seems to be twofolded. Obesity seems to be a cause of lower labour income, however, a lower labour income seems to have obesity as a consequence as well. So, despite this research will be mostly about the direction of bodyweight to labour income, for understanding the complete relationship between bodyweight and labour income, it seems both directions should be kept in mind. But why, at all, is it interesting to do a research about this relationship?
Background of the study
Since obesity has become an issue of overwhelming concern around the world due to its unfavourable economic and social consequences (Majumder, 2013, p. 200), an answer tothe question whether bodyweight is of influence on labour income seems to be important for many people related to the labour market.
To start with employees themselves, it might be interesting to know if a personal characteristic like being overweight has an influence on wages. Either if the relationship goes from bodyweight to labour income or from labour income to bodyweight, an employee who is overweight would like to be aware of it. He or she wants to know if there is something going on which has to due with self-confidence, the consumption of cheap food or, in the worst case, something with discrimination. In every case, awareness may improve the situation of the employee by taking action by the employee itself.
Also, for employers, awereness about the relationship is important. For them, it is good to know if they indeed discriminate on people only basing on their bodyweight. If so, they can find out for themselves if they are doing this for reasonable and right facts, or because of personal preferences. In the last case, they need to rethink about their decisions.
Moreover, in some cases, the government will be interested too. For example, the government may feel the obligation to intervene when discrimination by employers in selecting, promoting and decisions of payment of their employees on the basis of bodyweight, is proven. But, not only in this extreme case, also other knowledge about the relationship might be useful to know about for the government. If labour income causes obesity, and thereby poorer people may buy less healthy food because less healthy food is cheaper than healthy food, the government may intervene to improve this situation.
Concluding, this research is important for all people who are related to the labour market. The relationship between bodyweight and labour income is interesting for employees as well as employers and the government.
This Research
This thesis will discuss the influence of bodyweight on labour income. In a general way, by doing empirical research, there will be tried to give an answer to the question: ‘’Does bodyweight have an influence on labour income?’’After finding an answer to this question, it will be discussed whether changing the variables of age, gender and country of living may change the results. Also, a
comparison between this research and other research will be made to find important similarities and differences.
So, first, in Part II other research in this field will be described. Then, the empirical framework will be constructed in Part III. Within this part, the hypothesis and methodology of this empirical research will be explained. Afterwards, in Part IV, the data being used will be described and the first summary statisticals will be given. In Part V, the results themselves will be described. Then, in Part VI, a discussion about the results and about the research will be started. This will end in the conclusion, where I will show the main findings and do a suggestion for further research.
Part II – Literature Review
This part describes the existing literature and research done in this field. In the last decades, a lot of researchers have questioned whether there might be a relation between bodyweight and labour income. The work of those researchers will be looked at, whereby similarities as well as the differences between their findings will be described. In general, most researchers do agree with each other to find a negative relation between bodyweight and labour income. However, as the quote of Cawley (2004, pp. 451-474) in the introduction makes clear, the causal direction between bodyweight and labour income is doubtful. This means that there are two possible causal directions: when bodyweight goes up, labour income goes down, or, when labour income goes up, bodyweight goes down. Following in the next parts, most research is about the first relationship, including this study itself.
According to Baum and Ford (2004, p. 897), obese workers suffer from a wage penalty in the range of 0.7-6.3%, whereby obese females suffer more from a wage penalty than obese males. Han, Norton and Powell (2011, p. 381) give a more concrete percentage; they find that a one-unit increase in BMI is directly associated with 1.83% lower hourly wages. Moreover, Katsaiti and Shamsuddin (2016, p. 4167) who base their research about this relation in Germany on a report of Gallup (2011), indicate that Germans who are most likely to be obese, are having a low income: ‘’more than 1 in 5 Germans with a monthly income of less than 1400 euros are obese compared to 1 in 10 Germans with a monthly income of 5451 euros or more’’. Lastly, Brunello and D’Hombres (2007, p. 1) find that a 10% increase in the average body mass index reduces the real earnings of males and females by 3.27% and 1.86% respectively.
So, generally, there can be fixed that there exists a negative relationship between bodyweight and labour income. However, as mentioned, the strenght of the relationship is somewhat different between the research done. This is mostly due to a different way of measuring. Also, some researchers find a strong relationship for all kind of people; other researchers find differences between ages, gender, occupations and cultures. These differences need to be made clear.
Starting by age, young people are physically less likely to be obese (Baum and Ford, 2004, p. 887). But still, the wage effect of obesity among young workers may be eminent. For example, Register and Williams (1990, p. 130), who base their research on
the NLS Young Cohort, find no significant effect for males in the ageclass of 18-25 years old, but for females in the same age category, they find the significant effect of 12% wage difference between obese and nonobese.
These differences between males and females do not only hold for this particular ageclass. Register and Williams (1990, p. 130) state that obese females in general earned an average of 59 cents less per hour than nonobese females, while obese males in general earned even about 30 cents more than nonobese males. Baum and Ford (2004, p. 885) also find this obesity penalty for females, but, according to them, this obesity penalty also exists for males; the effect is just smaller. An other interesting fact is that there is a difference in callback rates for obese men and obese women, regarding the procedure of getting employed (Rooth, 2008, p. 718). Moreover, for females, it becomes less probable to be in an occupation requiring social interactions to the extent that a one-unit increase of late-teen BMI reduces the probability of being in such a occupation by 0.12 percentage points (Han, Norton and Powell, 2011, p. 390). While these researchers find more effect for females, some researchers find a contrary result. According to Brunello and D’Hombres (2007, p. 14), the effect of increasing bodyweight on the reduction of wages is of more strenght for males than for females. García Villar and Quintana-Domeque (2009, p. 77) also find more influence for men; they state that for high-income families it is more likely for men to be overweight, while the reverse is true for females. The different results regarding the different effects of gender, may be due to the data and methodology used by the different researchers. For example, Brunello and D’Hombres (2007, p. 4) base their research on a few European countries, while Baum and Ford (2004, p. 885) base their research on the United States. In methodology there is also a difference: whereas Brunello and D’Hombres (2007, p. 6) only use a model using instrumental variables, Baum and Ford (2004, pp. 891-897) use several ways to prove their statement. Therefore, their results may be more convincing; it is more convincing the relation is stronger for males.
As mentioned, there may also be differences between occupations; especially for occupations requiring social interactions a negative relation between bodyweight and labour income may exist. For example for sales occupations, which require customer contact, a one standard deviation increase in the unattractiveness rating (of which bodyweight is part of) lowers the probability of a callback for interview by 6-8% (Rooth, 2008, p. 725). Baum and Ford, who also indicate that obese workers in occupations
involving direct public contact may experience a higher wage penalty than in other occupations, add here this might be caused by customers being averse to interacting with those who are obese (Baum and Ford, 2004, p. 886).
Despite differences in ages, gender and occupations, researchers also make distinctions between cultures and countries. According to Baum and Ford (2004, p. 891) obese workers are more likely to be black or Hispanic and to live in the South of the United States of America. That there may be differences like those, is something which is also approved by Brunello and D’Hombres (2007, p. 2). They state that living in a region with higher than national average BMI has a moderate but statistically effect on the relationship between BMI and pay. When we zoom in into Europe, the impact of obesity on wages varies across the countries (Brunello and d’Hombres, 2005, p. 2). According to them, obesity affects wages negatively in countries with lower GDP per capita and positively in countries with higher GDP. They give as an example that being overweight is an asset in Dublin, while in Madrid it carriers a penalty (2005, p. 14).
Having looked at all different research done within this field, actually there is one remaining question: why would someone with more bodyweight get less labour income? According to Rooth (2008, p. 711), this could be because of employer discrimination. Obese people are expected to be less productive; obesity is correlated with bad health and with higher absenteeism. Therefore, obese people may be less desirable as a candidate. This could also be because of the feeling that employers have to compensate for higher expected medical costs (Katsaiti and Shamsuddin, 2016, p. 4167). This is approved by Baum and Ford (2004, p. 886). Also, as mentioned, preference-based customer discrimination could be a cause. Moreover, according to Register and Williams (1990, p. 131), who base their research on a number of other research, unattractive individuals have been considered to be less intelligent, less succesfull in social relationships, more stupid, lazier and less mentally healthy than their attractive counterparts. According to Baum and Ford, obesity may limit the kind of work they can perform (2004, p. 886).
In this study, other data will be used to examine the relationship between bodyweight and labour income. I will focus on 28 countries in Europe, and look whether the results of the exisiting literature will differ from this study by changing variables like age, gender and country.
Part III – Empirical Framework
In this part, the empirics used in this research will be clarified. First, the hypothesis will be given. Based on the existing literature, a lot of questions about the relationship between bodyweight and labour income come up. Therefore, it has to become concrete what this study is exactly about. Second, the steps being followed through the research, will be given, in the form of a methodology.
Hypothesis
First, I will examine wheter there, indeed, holds a relationship between bodyweight and labour income. Therefore, the following relationship has to be true: I = α + βO + ζ Equation 1: The relationship between Income and Obese Here, I stands for income, O for obese and ζ is an error term. This relationship will be studied for Europe as a whole and for the 28 countries seperately. The countries of Europe will be examined seperately too, because of the literature concluding there may be differences between the countries as well.Moreover, age and gender will be used as sample restrictions. This will be done, because the existing literature tells us that there will be important differences by changing those variables. Interesting here is that the results of the other researchers vary a bit.
In the regressions for Europe as a whole, pre-obese and obese are added as dummy variables. This should possibly make the estimates more precise. A more extensive explanation of every variable, will be given in Part IV.
Based on the existing literature and the research already done, the following hypothesis is being made:
H0: β=0
vs. H1: β<0
So, besides having the hypothesis of bodyweight having a negative influence on labour income there will also be done some subsamples to look whether age and gender are of different influence.
Methodology
Knowing the hypothesis, it is necessary to clarify how the empirical research is constructed.
Based on equation 1, it is possible to run OLS regressions. To give a answer to the questions put in the hypothesis, it is necessary to do more than just one OLS regression. First of all, I will look at the most general situation, whereby most European countries are included. In this situation, six regressions will be made in three different ways, which sums up to a total of 18 regressions. These different ways are being done to get more certainty about results which will be found. The sets of six regressions come about because of different sample restrictions by gender and agegroup. There will be six, because of combining three gendergroups, which are male, female and total group, with the agegroup of 15-75 years old and 25-34 years old.
The first six of those regressions will look at the correlation between quintile and obese only, without using country in the regession. The second six of them add the variable country as a dummby variable. The last six regressions add the variable country as a dummy variable as well, but are also controlling for underweight and pre-obese (in this specification normal weight will be the base category, because it is the most neutral).
By doing those regressions, it becomes clear whether bodyweight is of influence on income for most European countries together. Also, there will be found out whether there will be different results looking at different variables of gender and age.
Moreover, extra regressions will be made for all countries separately. This will be done, because, regarding the literature review, the hypothesis is that there will be differences between the European countries as well.
So, second, there will be done 168 regressions. For every of the 28 countries, six regressions are done. These six regressions will look at the correlation between quintile and obese only. In contrast to the regressions for Europe as a whole, I will not do six extra regressions controlling for underweight and pre-obese per country. This has to do
with the econometrical problem of overfitting; estimating 4 parameters from 5 observations may incorrectly affect the p-values and estimates. So, there will be continued with only six regressions per country, because of combining three gendergroups, which are male, female and total group, with the agegroup of 15-75 years old and 25-34 years old. After doing all these regressions, I will look at the p-values which are significant. By making a distinction between significant and insignificant values, and comparing the results between all regressions and looking at the size of the estimates, an answer to the research questions can and will be given.
Part IV – Data
This part describes the data used in this research. First, the data itself will be described. Second, I will give some summary statistics to get to know the data being used.The dataset
The dataset used in this research is uploaded from the website of Eurostat, a statistical office in the European Union, on December 11, 2017. The title of the dataset is ‘’Body Mass Index (BMI) by sex, age and income quintile’’. The data is quite recent; all values are from the year 2014. The data is based on the European Health Interview Survey (EHIS), which aims at collecting data about health status, health determinants, use and limitations in access to health care services in Europe, and backround variables like labor force and income.The data found in this dataset, is descriptive data. It are all percentages of a (part of) the population. So, the real data per individual are not given; however, these real data per individual are already converted to percentages of the population belonging to certain classes of the population. So, intrinsically, the values used in this thesis are based on individual values. In total, the datatset contains 840 obeservations, of which 793 remain because of missing values. Because of every observation having four percentages (one for underweight, one for normal weight, one for pre-obese and one for obese), in the end the research is based on 3172 percentages.
The dataset contains five variables: country, quintile, gender, age and bodyweight.
As mentioned, 28 countries are included in the dataset: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Luxembourg, Malta, the Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Slovakia, Spain, Sweden and the United Kingdom. Every country gets a number in the dataset, ranging from 1 for the Netherlands to 28 for the United Kingdom.
Quintile is the income quintile group on the base of the total equivalised disposable income attributed to each member of the household. These persons are ordered according to the value of the total equivalised disposable income. Four cut-point values of income have been found, dividing the survey population into five groups
equally represented by 20% of individuals each: First quintile, Second quintile, Third quintile, Fourth quintile and Fifth quintile.
Gender has three classes: male, female and a total class of males and females. In the dataset, the total class gets number 1, females number 2 and males number 3.
Age has two classes: young people (25-34 years old) and a total class of ages (15-75years old), whereby, in the dataset, the total class gets number 1 and the class of young people gets number 2. Body Mass Index (BMI) measures bodyweight: the weight in kilos divided by the square of the height in meters. There are four classes: Underweight (BMI less than 18.5), Normal weight (BMI between 18.5 and less than 25), Pre-obese (BMI between 25 and less than 30) and Obese (BMI equal or greater than 30). Most of the regressions only use obese; some of the regressions, for example within the case of Europe as a whole, also control for the other classes (leaving out the normal class, because of the dummy variable trap).
Summary statistics
Now, to get to know the data a bit better, in this part a few tables and graphs with general reults will be given.
First of all, it is interesting to know the spread of bodyweight in the European countries. Therefore, the following figures are given:
Country Underweight Normal Pre-obese Obese
Austria 2.7 50.4 32.6 14.3 Belgium 3.3 48.8 34.2 13.7 Bulgaria 2.6 44.6 38.4 14.4 Croatia 2.3 41.9 37.8 18.0 Cyprus 4.3 49.2 32.6 13.9 Czech Republic 1.3 43.3 36.7 18.7 Denmark 2.9 51.1 31.6 14.4 Estonia 2.7 45.0 32.6 19.7 Finland 1.8 44.7 35.7 17.8 France 4.2 50.3 30.8 14.7 Germany 2.4 47.0 34.3 16.4 Greece 2.3 42.2 38.6 16.9 Hungary 3.4 42.6 33.3 20.6 Italy 3.7 52.5 33.3 10.5 Latvia 1.9 42.9 34.4 20.8 Lithuania 2.4 44.3 36.7 16.6
Luxembourg 3.2 50.4 31.3 15.1 Malta 2.4 38.0 34.4 25.2 Netherlands 2.3 50.0 34.8 12.9 Norway 2.1 50.0 35.2 12.6 Poland 2.9 43.8 36.6 16.7 Portugal 2.5 45.3 36.1 16.1 Romania 1.6 44.5 44.8 9.1 Slovenia 1.9 43.0 36.5 18.6 Slovakia 2.7 44.3 37.1 15.9 Spain 2.6 46.4 34.8 16.2 Sweden 2.4 49.8 344 13.4 United Kingdom 2.3 42.7 35.2 19.8 Figure 1: Spread of bodyweight in percentages of the total population of the countries separately in Europe Figure 2: Spread of bodyweight in percentages of the total population of Europe as a whole As can be seen in figure 2, about 35% of the total population of Europe is pre-obese and about 15% is obese. As a consequence, only almost half of the people in Europe are normal weight or underweight.
Moreover, looking at figure 1, in Malta relatively most people are obese. In Romania, least people are obese. Italy is the country with most people normal weight, while again; Malta is the country with least people normal weight. Comparing Northern and Southern countries, there are no big different patterns seen. However, comparing Western and Eastern countries, it can be seen most of the Western countries have a percentage above 47% of normal weight people, while most of the Eastern countries
0,0 5,0 10,0 15,0 20,0 25,0 30,0 35,0 40,0 45,0 50,0 Percentage of the total population in Europe
have a percentage below 47% of normal weight people in Europe.
Because this research will also make a distinction between the ageclass of 25-34 years old and the total class of 15-75 years old, it is interesting to know the spread of bodyweight of the population concerning this characteristic. Therefore, the following figure is given:
Figure 3: Spread of bodyweight in Europe in the ageclass of 25-34 years old
Comparing figure 2 and figure 3, it can be seen that, overall and relatively, younger people are more often normal weight than the population in total. This result agrees with Baum and Ford (2004, p. 887). According to them, young people are physically less likely to be obese.
Last, a comparison between males and females has to be made. Therefore, the following figures are given: 0,0 10,0 20,0 30,0 40,0 50,0 60,0 70,0 80,0 Percentage of the population in Europe in the ageclass of 25-34 years old
18 Figure 4: Spread of bodyweight in Europe of males only
Figure 5: Spread of bodyweight in Europe of females only 0,0 10,0 20,0 30,0 40,0 50,0 60,0 Percentage of the male population in Europe 0,0 5,0 10,0 15,0 20,0 25,0 30,0 35,0 40,0 45,0 Percentage of the female population in Europe
Underweight Normal Pre-obese Obese
As can be seen by comparing figures 4 and 5, it seems that, overall and relatively, males are more often normal weight than females. This is mainy due to the category of pre-obese; the difference is almost 15%-points. Having deepened our knowledge into the variables in this part, we can start the following part: the results.
Part V – Results
In this part, the results of this research will be shown. First, there will be looked at the general results for Europe as a whole and there will be tried to find an answer to the question: ‘’Does bodyweight have a influence on labour income?’’. This will be done by looking at the p-values of the regressions done and by looking at the size of the estimates, as decribed in Part III. A significance level of 10% will be used. Second, there will be looked at the different countries seperately. The same way of getting results will be followed. Third, it will be discussed whether changing the variables of age and gender may influence the results.
General Results for Europe
For Europe as a whole, we have to look at 18 regressions. Summing op those regressions, we get the following results:
15-75 years old 25-34 years old
Total Female Male Total Male Female
Obese -0.143* -0.177* -0.046 -0.101* -0.121* -0.024* (0.028) (0.022) (0.029) (0.026) (0.022) (0.027) R2 0.1611 0.3235 0.018 0.096 0.1920 0.0063 n 139 139 138 139 124 114 Figure 6: Results for Europe regressing Obese on Quintile 15-75 years old 25-34 years old
Total Female Male Total Male Female
Obese -0.393* -0.325* -0.198* -0.182* -0.209* -0.081* (0.040) (0.025) (0.059) (0.036) (0.031) (0.046) R2 0.4651 0.6071 0.0961 0.188 0.3422 0.0948 n 139 139 138 139 124 114 Figure 7: Results for Europe regressing Obese on Quintile controlling for Country
15-75 years old 25-34 years old
Total Female Male Total Male Female
Obese -0.379* -0.328* -0.183* -0.200* -0.176* -0.024* (0.032) (0.029) (0.042) (0.035) (0.029) (0.048) R2 0.6766 0.6722 0.5691 0.2786 0.4779 0.2148 n 139 139 138 139 124 114 Figure 8: Results for Europe regressing Obese on Quintile controlling for Country, Pre-obese and Underweight As can be seen, in 17 out of 18 regressions, which is 94%, there has been found a significantly relation between bodyweight and labour income. The size of the estimates of the variable obese do not differ that much between the the three different ways of regressing. This gives a good reason to believe there exists indeed a relationship between bodyweight and labour income, at least for Europe as a whole.
Results for countries seperately
Having the results for Europe, it is already possible to give a answer to the main questions. However, it is interesting to know whether there may be similarities and differences between countries as well. So, when enlarging the research by doing the regressions for every country seperately, as described in part III, it has been found that for many of the 28 countries in Europe a relation between bodyweight and labour income exists. After doing all 168 regressions, the following relations have been found:
15-75 years old 25-34 years old
Country Variable Total Female Male Total Female Male
Austria Obese -0.505 -0.406* -0.359 -0.155 -0.263 -0.093 Belgium Obese -0.448* -0.399* -0.491* -0.315 -0.139 -0.447* Bulgaria Obese -0.262 -0.498* 0.121 -0.376 -0.408* -0.053 Croatia Obese -0.694 -0.303 -0.516 0.084 . . Cyprus Obese -0.761* -0.542* -0.343 -0.401 -0.368 -0.192 Czech Republic Obese -0.179 -0.268 0.217 -0.241 -0.227 . Denmark Obese -0.301 -0.241 -0.355 -0.111 0.761 . Estonia Obese -0.505 -0.332* 0.317 -0.167 -0.156 -0.045 Finland Obese -0.870 -0.450 -1.677 -0.123 -0.028 0.223 France Obese -0.472* -0.348* -0.555* -0.424 -0.316* -0.556* Germany Obese -0.655* -0.543* 0.578 -2.033* -0.586 -0.248 Hungary Obese -0.537* -0.417* -0.269 -0.244 -0.190 -0.285 Italy Obese -1.097* -0.806* -1.474* -0.851 -0.707 -0.881 Latvia Obese -0.361* -0.236* 0.189 0.637* -0.087 0.373* Lithuania Obese -0.389* -0.294* -0.229 -0.102 -1.556 . Luxembourg Obese -0.555 -0.357 -1.540* -0.084 -0.144 . Malta Obese -0.384* -0.330* -0.349 0.115 0.219* . Netherlands Obese -0.357 -0.319 -0.336 -0.484* -0.412 -0.457* Norway Obese -0.949 -0.567* 0.908* -0.227 -0.507 -0.118 Poland Obese -0.120 -0.441 0.454 -0.445 -0.451* 0.026 Portugal Obese -0.526* -0.352* -1.128* -0.262 -0.185 -0.284 Romania Obese -0.245 -0.567 0.396 -0.179 -0.634 0.115 Slovenia Obese -0.353* -0.289* -0.401* -0.450 -0.385 0.454 Slovakia Obese -0.229 -0.219 -0.039 -0.576 -0.465* -0.230 Spain Obese -0.345* -0.249* -0.377 -0.270* -0.181* -0.313 Sweden Obese 0.003 -0.120 0.297 -0.499* -0.204* 0.347 United Kingdom Obese -0.488* -0.528* -0.400* -0.412* -0.515* -0.167 Figure 9: Results of the coefficents for most European countries seperately regressing Obese on Quintile
The red values are significantly negative relations between bodyweight and labour income. A dot means there were too less data to do the regression required. A few relations are significantly positive, highlighted by a green color. The whole outputs, including the standard errors, the R2s and the number of obeservations, can be found in
Appendix 1.
Summing up the significantly negative relations found, we get the following number of countries with significantly negative relations divided by the certain classes:
Age Gender Number of Countries 15-75 years old Total 13 Female 17 Male 7 25-34 years old Total 5 Female 7 Male 3 Figure 10: Number of countries having a significantly negative relation within certain classes Thus, looking at all genders and all ages, for 13 out of 28 countries, a significantly negative relation between bodyweight and labour income can be found. This means that for 46% of the countries in Europe, a signifantly negative correlation exists.
For females in the ageclass of 15-75 years old, most significant relations have been found. In 17 out of 28 countries, females experience a negative relation between bodyweight and labour income, which counts for 61% of all countries. Compared to only 7 countries for males in the same ageclass, which counts for 25%, we can conclude that the relation for females is much more common than for males.
Moreover, in only 3 out of 28 countries, which is 11%, a significantly negative relation has been found for males in the ageclass of 25-34 years old. For females in the same ageclass, it has been found for 7 out of 28 countries, which is 25%. Compared to the total ageclasses of 15-75 years old, these percentages are low. As mentioned, for males in ageclass of 15-75 years old, a percentage of 25% of all countries has been found and for females a percentage of 61%. A difference for males of 14%-points and for females of 36%-points, seems to conclude that the correlation for the ageclass of 25-34 years is less common than for the total ageclass of 15-75 years old. This is partly due to the fact that some data was missing to regress obese on quintile controlling for male and ageclass 25-34 years old. However, still, it may be a sign of a lesser probability of a relation between bodyweight and labour income for that ageclass.
When zooming in on the countries themselves, for a few countries there have been found no significantly negative relations at all. These countries are Denmark, Croatia, Czech Republic, Finland, Greece and Romania. Also, there are some countries for which there has been found only one or two out of six significant relations: Austria (1), Estonia (1), Luxembourg (1), Norway (1), Poland (1), Slovakia (1), Bulgaria (2), Cyprus (2), Hungary (2), Latvia (2), Lithuania (2), Malta (2), the Netherlands (2), and
Sweden (2). For all other countries, more than three out of six significantly negative relations have been found: Germany (3), Italy (3), Portugal (3), Slovenia (3), Belgium (4), Spain (4), France (5) and United Kingdom (5).
Interesting to see is that for the ageclass 25-34 years old, the significantly negative correlations have been found mostly for Northern European countries. Looking at males, all three countries for which negative correlations, are Northern countries. For females, only two out of seven countries for which negative correlations have been found, are Southern countries.
However, no other big different patterns have been found between Northern and Southern countries. This also holds for differences between Eastern and Western countries. The only difference found is for males in both ageclasses. Only one Eastern country of the respectively seven and three countries, which is Slovenia, has significantly negative correlations. For Latvia, an Eastern country, there have be found even significantly positive relations between bodyweight and labour income for males in both ageclasses. Therefore, it seems that for Europe in total, the significantly negative correlations found for males are mostly driven by Western countries. Comparing this to the summary statistics given in Part IV, it is quite surprising that, on average, relatively more people in Western countries are normalweight compared to Eastern countries, while the relation between bodyweight and labour income is stronger for Western countries.
A last note on the results is given by looking at the size of the significant estimates. Those estimates, the (conditional) correlations, are mostly varying negatively between -0.200 and -0.600. This means that the correlations found are quite strong: every increase of one unit of bodyweight (labour income), results in a decrease of 0.2-0.6 unit of labour income (bodyweight). Stronger correlations, probably outliers, have been found in Western Europe. In Germany, Italy, Luxembourg and Portugal there have been correlations stronger than 1.
Part VI – Discussion
In this part, there will be continued what has been started in the last section – elaborating on the results. First, the results will be compared to some of the results found by the researchers described in Part II.
Second, the strength and weakness of this study will be discussed. Then, bearing in mind this strength and weakness, I will do a suggestion for further research.
Discussion of the Results
In Part II, a lot of research has been discussed. Generally, it was concluded there is a negative relationship between bodyweight and labour income. According to the results in this study, it can be seen that for 13 out of 28 countries in Europe there has been found a significantly negative correlation between bodyweight and labour income. Looking at the regressions of Europe as a whole, there has been found only significantly negative relations. So, taking into account the results of this study, there can be concluded there exists a correlation between bodyweight and labour income for at least almost half of the countries of Europe.
As mentioned in Part II, researchers were not in consensus about the different effect by different genders. The result of this study shows that for females most significantly negative relations have been found. This result is in line with Baum and Ford (2004, p. 885), who found out that the relation between bodyweight and labour income is much stronger for females, and with Register and Williams (1990, p. 130), who even found out that there doesn’t exist a significant relation for males.
Another result which is in line with those researchers, is that for younger people, there will be found less significantly negative correlations (Register and Williams, 1990, p. 130). According to them, there are some negative correlations; but not all of them are significant. The results of this study show exactly the same: there have been found less significant correlations for the younger ageclasses. Looking at the ageclass of 15-75 years old, the difference is 8 out of 28 countries of significantly negative correlations. This is a difference of about 29%-points.
So, concluding, the results of this study mostly agree to found results in older research. However, it also adds some new views into this field. First, there have been found no big differences between Northern and Southern countries, except for the
ageclass 25-34 years old. Here, there have been found more significantly negative correlations for Northern European countries than for Southern European countries. This is an interesting result; Brunello and d’Hombres (2005, p. 17) found a contrary result. According to them, for Northern countries, there exists a significantly positive relation between bodyweight and labour income, while for Southern countries, there exists a significantly negative one. I did find some positive coefficients, but these (few) coeffcients were found in Northern as well as in Southern countries.
Also, between Eastern and Western countries there are barely differences. The only different result has been found for males. Therefore, it seems that for Europe as a whole, the significantly negative correlations found for males are mostly driven by Western countries. For further research, there had to be found out why these (few) differences between Northern and Southern and Eastern and Western countries hold.
Discussion of this Research
First of all, there will be looked at the strength of this research. After that, the critics will be discussed and eventually there will be elaborated on some questions for further research.
To start with something good, is that there has been found data for almost every country in Europe. Despite some missing values, general results and general differences between gender and ages can be looked at. This will help to get a broad sense of the characteristics of the relation between bodyweight and labour income.
However, some critics must be given too. First, as mentioned, with the data used, the direction of the relation will not be clear. Bodyweight can have an influence on labour income, but, the other way around, can hold too: labour income having an effect on bodyweight. Therefore, it gets really hard to answer the question by which of the two direction holds: it can be one of them, but also both. This simultaneous causality is really hard to deal with. Another critic which must be given is that there might be endogeneity: BMI might be correlated with the error term. This may cause biased results. The reason why this could be the case, is because some important variables are not controlled for in the dataset. For example, schooling, occupation and place of living (nearby a supermarket or fastfoodrestaurant) might be correlated with BMI and labour income. This may make comparing the size of the correlations found in this study and the other studies really
difficult. The correlations seem to be much stronger, but may be biased by variables for which there was not controlled for.
So, concluding, this research is good for withdrawing some general results and correlations in a broad sense, but, because of simultaneous causality and endogeneity, the results may be biased. Therefore, it is not possible to know into what direction the relation holds and it is difficult to know the size of the estimates perfectly.
For further research, it is therefore necessary to find a way to deal with the problems of simultaneuous causality and with endogeneity. More data should be collected to deal with endogeneity, so adding variables like schooling. Afterwards, to deal with simulatenous causality, using an instrumental variable might be the solution. An example of this would be instrumenting BMI with the BMI of a biological family member, as is done by Brunello and D’Hombres (2007, p. 6).
Having fulfilled those conditions, it may be possible to investigate which direction of the relation holds. If it is truly proved bodyweight has a negative effect on labour income, it should be investigated wheter employer discrimination is of influence. If so, there should be done research whether government intervention might solve this problem. If it is proved that labour income is of negative effect on bodyweight, there may also be need for government intervention. For example foodprices should then be reconsidered.
Part VII – Conclusion
In this research, I have tried to give an answer to the question: ‘’Does bodyweight have an influence on labour income? By doing empirical research and using descriptive data of Eurostat based on the European Health Interview Survey, a total of 186 regressions has been done. Looking at the results of those 186 regressions, there can be concluded that, in general, there has been found at least a correlation between bodyweight and labour income. For 13 out of 28 countries, a significantly negative correlation between bodyweight and labour income holds. Because of the problems concercing simultaneous causality, however, it is not possible to conclude wheter bodyweight is of negative influence on labour income or the other way around.Specifying it towards several countries and classes, there can be concluded that most significantly negative correlations have been found for older females. For 17 out of 28 countries in Europe, such a significantly negative relation has been found. Least significant relations have been found for males aged 25-34 years old: for only 3 out of 28 countries, which is 11%, such a significantly negative relation holds. For females in the same ageclass, it has been found for 7 out of 28 countries, which is 25%. Compared to the total ageclasses of 15-75 years old, these percentages are low. A difference for males of 14%-points and for females of 36%-points, concludes that the correlation for the ageclass of 25-34 years is less common than for the total ageclass of 15-75 years old. This is partly due to the fact that some data was missing to regress obese on quintile controlling for male and ageclass 25-34 years old. However, still, it may be a sign of a lesser probability of a relation between bodyweight and labour income for that ageclass.
Comparing these results with results of other research within this field, no results within this study differ that much from other studies. The fact that the relation between bodyweight and labour income is mostly found for females, and mostly not found for younger ages, are both results other researchers have found as well. However, as mentioned, it adds some new views. These new views come about because of the broad dataset used in this study. First, by looking at different regions in Europe, no big different patterns have been found, except for the ageclass 25-34 years old. Here, the significantly negative correlations are more found for Northern European countries than Southern European countries. Between Eastern and Western countries, there only has
been found a difference for males; more significantly negative relations have been found for males in Western countries than in Eastern countries.
For further research, there has been recommended to find out why these (few) differences between Northern and Southern and Eastern and Western countries hold. Second, there has to be found a way to deal with the problems of simultaneuous causality and with endogeneity. Having results found which can deal with these problems, it may be possible to investigate which direction of the relation holds. By knowing which direction holds, it becomes possible for the government to intervene the way it should.
Reference List
Baum II, C.L. and Ford, W.F. The wage effects of obesity: a longitudinal sturdy. Health Economics, 2004, 13, 885-899. Brunello, G. and D’Hombres, B. Does body weight affect wages? Evidence from Europe. Economics and Human Biology, 2007, 5, 1-19. Cawley, J. The impact of obesity on wages. Journal of Human Resources, 2004, 39, 451- 474. Brunello, G. and D’Hombres, B. Does obesity hurt your wages more in Dublin than in Madrid? Evidence from EHCP. IZA Discussion Paper, 2005, 4, 1-26. Gallup. Gallup-Healthways Well-Being Index. Technical Report. Franklin, TN: Healthways, 2011. García Villar, J., Quintana-Domeque, C. Income and body mass index in Europe. Economics and Human Biology, 2009, 7, 73-83. Han, E., Norton, E.C. and Powell, L.M. Direct and indirect effects of body weight on adult wages. Economics and Human Biology, 2011, 9, 381-392 Katsaiti, M. and Shamsuddin, M. Weight discrimination in the German labour market. Applied Economics, 2016, 48, 4167-4182 Majumder, A. Does Obesity Matter for Wages? Evidence from the United States. Economic Society of Australia, 2013, 2013, 32, 200-217 Register, A.C. and Williams, D.R. Wage effects of Obesity among Young Workers. Social Science Quarterly, 1990, 71, 130-142.Rooth, D. Obesity. Attractiviness, and Differential Treatment in Hiring. The Journal of Human Resources, 2009, 44, 3, 712-735 Vermeulen, M. Kleine en dikke mensen hebben grotere kans op armoede. De Volkskrant, 2016.
Appendix
Appendix 1
In this appendix, all the outputs of the regressions done for the seprate countries in Europe are given. While in the text itself only the coefficients of the variable obese are given, here the standard erros, R2s and number of observations are shown as well. The
meaning of the colors and dots hold.
15-75 years old 25-34 years old
Country Variable Total Female Male Total Female Male
Austria Obese -0.505 -0.406* -0.359 -0.155 -0.263 -0.093 0.292 0.158 0.428 0.287 0.299 0.260 R2 0.500 0.687 0.190 0.088 0.633 0.041 n 5 5 5 5 5 5 Belgium Obese -0.448* -0.399* -0.491* -0.315 -0.139 -0.447* 0.050 0.020 0.110 0.183 0.170 0.145 R2 0.964 0.993 0.870 0.498 0.183 0.760 n 5 5 5 5 5 5 Bulgaria Obese -0.262 -0.498* 0.121 -0.376 -0.408* -0.053 0.381 0.178 0.298 0.261 0.122 0.312 R2 0.136 0.722 0.052 0.410 0.850 0.010 n 5 5 5 5 4 5 Croatia Obese -0.694 -0.303 -0.516 0.084 . . 0.585 0.5341 0.401 0.207 . . R2 0.319 0.097 0.356 0.052 . . n 5 5 5 5 . . Cyprus Obese -0.761* -0.542* -0.343 -0.401 -0.368 -0.192 0.160 0.095 0.393 0.271 0.283 0.261 R2 0.883 0.916 0.202 0.421 0.361 0.213 n 5 5 5 5 5 4 Czech Republic Obese -0.179 -0.268 0.217 -0.241 -0.227 . 0.345 0.201 0.368 0.177 0.145 R2 0.082 0.3723 0.104 0.381 0.552 . n 5 5 5 5 4 . Denmark Obese -0.301 -0.241 -0.355 -0.111 0.761 . 0.247 0.224 0.283 0.106 0.721 . R2 0.331 0.277 0.345 0.267 0.53 . n 5 5 5 5 3 . Estonia Obese -0.505 -0.332* 0.317 -0.167 -0.156 -0.045 0.373 0.084 0.254 0.494 0.148 0.157
R2 0.379 0.839 0.342 0.037 0.270 0.077 n 5 5 5 5 5 3 Finland Obese -0.870 -0.450 -1.677 -0.123 -0.028 0.223 0.388 0.235 1.266 0.174 0.155 0.254 R2 0.627 0.550 0.370 0.142 0.011 0.435 n 5 5 5 5 5 3 France Obese -0.472* -0.348* -0.555* -0.424 -0.316* -0.556* 0.057 0.076 0.117 0.125 0.104 0.210 R2 0.958 0.874 0.882 0.792 0.755 0.700 n 5 5 5 5 5 5 Germany Obese -0.655* -0.543* 0.578 -2.033* -0.586 -0.248 0.118 0.024 0.282 0.784 0.621 0.496 R2 0.911 0.994 0.584 0.691 0.228 0.077 n 5 5 5 5 5 5 Greece Obese -0.443 -0.297 0.053 0.165 -0.475 0.162 0.618 0.312 0.540 0.295 0.340 0.149 R2 0.146 0.232 0.003 0.094 0.394 0.284 n 5 5 5 5 5 5 Hungary Obese -0.537* -0.417* -0.269 -0.244 -0.190 -0.285 0.136 0.066 0.314 0.136 0.104 0.187 R2 0.837 0.929 0.196 0.517 0.527 0.435 n 5 5 5 5 5 5 Italy Obese -1.097* -0.806* -1.474* -0.851 -0.707 -0.881 0.170 0.086 0.567 0.449 0.469 0.416 R2 0.933 0.967 0.693 0.545 0.431 0.599 n 5 5 5 5 5 5 Latvia Obese -0.361* -0.236* 0.189 0.637* -0.087 0.373* 0.122 0.032 0.291 0.237 0.264 0.018 R2 0.743 0.947 0.123 0.701 0.974 0.995 n 5 5 5 5 5 4 Lithuania Obese -0.389* -0.294* -0.229 -0.102 -1.556 . 0.161 0.092 0.405 0.161 0.669 . R2 0.661 0.771 0.096 0.118 0.844 . n 5 5 5 5 3 . Luxembourg Obese -0.555 -0.357 -1.540* -0.084 -0.144 . 0.317 0.189 0.190 0.120 0.071 . R2 0.505 0.542 0.970 0.140 0.673 . n 5 5 4 5 4 . Malta Obese -0.384* -0.330* -0.349 0.115 0.219* . 0.072 0.047 0.132 0.233 0.028 . R2 0.934 0.961 0.779 0.108 0.984 . n 4 4 4 4 3 .
Netherlands Obese -0.357 -0.319 -0.336 -0.484* -0.412 -0.457* 0.200 0.142 0.328 0.123 0.220 0.044 R2 0.514 0.628 0.259 0.838 0.540 0.973 n 5 5 5 5 5 5 Norway Obese -0.949 -0.567* 0.908* -0.227 -0.507 -0.118 1.098 0.161 0.176 0.190 0.250 0.139 R2 0.199 0.806 0.899 0.322 0.578 0.192 n 5 5 5 5 5 5 Poland Obese -0.120 -0.441 0.454 -0.445 -0.451* 0.026 0.550 0.302 0.357 0.352 0.156 0.533 R2 0.016 0.415 0.350 0.347 0.736 0.001 n 5 5 5 5 5 5 Portugal Obese -0.526* -0.352* -1.128* -0.262 -0.185 -0.284 0.060 0.041 0.359 0.214 0.135 0.537 R2 0.962 0.961 0.689 0.333 0.387 0.085 n 5 5 5 5 5 5 Romania Obese -0.245 -0.567 0.396 -0.179 -0.634 0.115 0.780 0.506 0.623 0.772 0.805 0.371 R2 0.032 0.295 0.119 0.000 0.171 0.031 n 5 5 5 5 5 5 Slovenia Obese -0.353* -0.289* -0.401* -0.450 -0.385 0.454 0.066 0.074 0.052 0.482 0.296 0.190 R2 0.906 0.835 0.951 0.225 0.458 0.739 n 5 5 5 5 4 4 Slovakia Obese -0.229 -0.219 -0.039 -0.576 -0.465* -0.230 0.211 0.122 0.327 0.272 0.189 0.245 R2 0.282 0.520 0.005 0.600 0.669 0.228 n 5 5 5 5 5 5 Spain Obese -0.345* -0.249* -0.377 -0.270* -0.181* -0.313 0.067 0.039 0.177 0.073 0.050 0.171 R2 0.898 0.932 0.604 0.821 0.814 0.983 n 5 5 5 5 5 5 Sweden Obese 0.003 -0.120 0.297 -0.499* -0.204* 0.347 0.291 0.217 0.323 0.127 0.011 0.427 R2 0.000 0.092 0.220 0.838 0.995 0.180 n 5 5 5 5 4 5 United Kingdom Obese -0.488* -0.528* -0.400* -0.412* -0.515* -0.167 0.056 0.074 0.093 0.127 0.063 0.137 R2 0.962 0.944 0.860 0.778 0.957 0.332 n 5 5 5 5 5 5
