Obesity and Happiness Marina-Selini Katsaiti United Arab Emirates University Working Paper 2009-44R October 2009, Revised April 2011 Obesity and Happiness Marina-Selini Katsaiti ∗ April 22, 2011 Abstract This paper provides insight on the relationship between individual obesity and happiness levels. Using the latest available panel data from Germany (GSOEP), UK (BHPS), and Australia (HILDA), we examine whether there is statistical evidence on the impact of overweight on subjective well being. Instrumental variable analysis is utilized under the presence of endogeneity, stemming from several explanatory variables. Results indicate that in all three countries obesity has a negative effect on the subjective well being of individuals. The results also have important implications for the effect of other socio-demographic, economic and individual characteristics on well being. JEL codes: D60, I31 Keywords: Happiness, Obesity, Instrumental Variable Analysis, Subjective Well Being Faculty of Business and Economics, Department of Economics and Finance, United Arab Emirates University, P.O. BOX 17555, Al Ain, UAE; E-mail: selini.katsaiti@uaeu.ac.ae ∗ 1 1 Introduction Happiness is one of life’s fundamental goals. Whether people pursue better jobs or higher income, try to achieve better health or a stable family life, want to win an Olympic medal or the Nobel prize, the motivation behind their effort is normally happiness. People may engage in risky behavior, such as smoking or racing, because they derive temporary satisfaction from this. Similarly, people derive instant pleasure from food consumption. As with numerous habits and consumption patterns, the effect of food consumption is usually immediate gratification, however in the long run, consumption of food in excess of daily calorific needs leads to excessive weight gain, which in turn can lower subjective well-being. Happiness can be defined as the degree to which people positively assess their life situation (Veenhoven (1996)) and depends on a variety of individual and social characteristics. These characteristics differ in how important they are to each individual and are measured by ordinal ranking. Happiness is often defined in terms of living a good life, rather than a simple emotion. Happiness is naturally the subject of psychological and sociological research as well as medicine, and is often associated with good health. Economics research has connected happiness with the concept of utility since the 18th century and the works of Bentham and Jevons. This multidisciplinary research has identified several determinants of happiness. The most important ones include demographics, socioeconomic traits, education, and health related characteristics. Empirical work in economics has shed light on significant determinants of individual well being. Age, gender, income, employment status, marital status and education are among them. Body Mass Index (BMI) has recently been added to 2 the list of factors that can explain life satisfaction levels. BMI can influence happiness through deterioration in health, lower self-esteem, or lower social acceptance. In addition, it may affect self confidence, personal and social relationships, and attitude. Though not perfect, BMI is a well established measure of obesity, employed by the Centers for Disease Control (CDC) and by the World Health Organization (WHO). Individuals with BMI i) between 18-25 are indexed as normal weight, ii) between 25 and 30 are categorized as overweight, and iii) over 30 are classified as obese. Moods often affect consumption patterns and are associated with eating habits and disorders. In addition, it is intuitive that subjective well being itself influences numerous other aspects of life, both in the short and in the long run. Examples of the factors arguably influenced by happiness levels are, among other things, education and income levels, marital status, and employment. Thus, in the empirical estimation several explanatory variables are endogenous and do not obey the standard assumptions, since the causality could be running in both directions. This issue, not adequately addressed in the happiness literature, cannot be neglected as it may affect the robustness of the results, when the estimator is inconsistent. Stutzer (2007) is the only study that addresses this issue, however, acknowledging and treating for reverse causality only between happiness and BMI. The purpose of this study is to examine the impact of BMI on individual wellbeing. It contributes to the literature in the following ways. Firstly, it analyzes the most recently available panel data from Germany, Australia and the UK. In addition, it is the first study to examine the Australian case. Last, it identifies the endogeneity issues arising from dual causality in the model and addresses them 3 appropriately. The paper is structured as follows. Section 2 reviews the relevant litera- ture. Section 3 describes the estimation methodology and the data, and Section 4 presents and examines the empirical results. Section 5 summarizes the primary findings and offers some final remarks. 2 The Literature The medical literature provides diverse conclusions about the relationship between obesity and depression. Roberts, Kaplan, Shema, and Strawbridge (2000) use data from Alameda County, California, to investigate whether the obese are at greater risk for depression. They conclude that, among other groups, the obese, females, and those with two or more chronic health conditions are at higher risk for depression. In addition, they find that, when all individuals with depressive symptoms in the previous year are excluded, there is greater relative risk for future depression for the obese than for the non-obese. This result holds in specifications that control for a number of variables affecting the risk of depression. Based on their results and on the results of other studies, they conclude “that the obese may be at increased risk for depression.” Reed (1985) uses data from the First National Health and Nutrition Examination Survey (NHANES I) and identifies young, more educated, obese females as a subgroup of worse mental health condition. Several studies find strong evidence of the relationship between overweight/obese individuals and depression in females (Noppa and Hällström (1981), Palinkas, Wingard, and Barrett-Connor (1996), Reed (1985)). Larsson, Karlsson, and Sullivan (2002) analyze the effect of 4 overweight and obese on health-related quality-of-life (HRQL) in Sweden. Using data from a cross-sectional survey on 5633 men and women aged 14-64, their regression analysis finds the following: overweight and obesity for young men and women(16-34 years) leads to poor physical health, but not mental health. For middle-aged (35-64 years) individuals, obese men and women report health impairments, however only women report mental health problems. The same result for females is supported by a study of adolescents aged 11 to 21 years. Needham and Crosnoe (2005) find evidence that relative weight is associated with depressive symptoms for girls but not for boys. Greeno, Jackson, Williams, and Fortmann (1998) also confirm that females with lack of perceived eating control and higher BMI are associated with lower life satisfaction levels. For men only the lack of perceived eating control explains lower happiness levels. Stutzer (2007) investigates i) the probability of being obese given certain socioeconomic and demographic characteristics, ii) the effect of obesity on happiness taking into account self-reported self-control levels. His intuition stands on the hypothesis that only individuals who feel unable to control their food consumption should have lower happiness levels due to obesity. Using Swiss data, he finds that lower self-control is associated with lower happiness levels given the presence of obesity. Stutzer (2007) checks for reverse causality. He finds no evidence that eating due to stress leads to lower happiness levels of obese individuals with limited self control. A similar study by Oswald and Powdthavee (2007) examines cross sectional data from the UK and Germany, using regression analysis to identify the relationship between BMI and self-reported life satisfaction. For the British data they also 5 explore the impact of BMI on psychological distress and on self-reported “perception of own weight”. Under all univariate and multivariate specifications in both datasets, BMI has a negative and significant effect on subjective well-being. Moreover, for the British regressions they find that BMI increases psychological distress and is positively associated with perception of own weight. Employment status, age, education, income, marital status, and disability status stand out as significant determinants of individual happiness under most specifications. However, Oswald and Powdthavee (2007) do not correct for endogeneity. 3 Empirical Estimation 3.1 Data The data for Germany come from the German Socio-Economic Panel (GSOEP), a representative longitudinal study of individuals and households. The aim of the GSOEP survey is to collect data on living conditions, together with demographic, economic, sociological, political, and other individual and household characteristics. The data contains information about German citizens, foreigners, and immigrants to Germany. Weight and height data, are available only for the years 2002, 2004, 2006, and 2008. Most other variables included in our specifications are available for all years, with no breaks. For UK, the data come from the British Household Panel Survey (BHPS). This survey includes households from England, Scotland, Wales and Northern Ireland. It surveys approximately 22, 000 individuals yearly, and provides information on demographics, economic situation, household characteristics, and individual 6 health. The main information of interest here, the weight and height data, are available for 2005 and 2007. Once again most other variable information included in the analysis is available for all years. For Australia the data source is the Household, Income and Labour Dynamics in Australia (HILDA) Survey. BMI information is available for years 2006, 2007 and 2009. HILDA provides similar or equivalent information with that of BHPS and GSOEP. In the Australian data the financial information variables are only available for one year, and this fact makes it not possible to contain this information in the panel regressions. Descriptive statistics on German, British and Australian data are presented in Tables 1, 3, and 5 respectively. Correlation matrices for the variables of interest are shown in Tables 2, 4, and 6 respectively. Besides Body Mass Index (BMI) the following variables are included in the multivariate specifications: age, gender, years of education, income, employment status ∗ , marital status, number of children, disability, and household size. When data is available, some additional variables are also included in the analysis: political party membership, house ownership, saving habits, whether one has a second job, smoking habits, labor union membership, religion, region and nationality. BMI is used to control for individual obesity level. Happiness is measured using the self-reported life-satisfaction index. Here we have to acknowledge that individual happiness and self-reported life satisfaction may not be perfect substitutes and in fact, as the literature has concluded, the two are distinct. However, due to the fact that life satisfaction levels are reported, and there is no clear existing ∗ For BHPS employment status contains information on whether individuals are employed or unemployed. For GSOEP and HILDA data, the information is on whether individuals are employed or not employed. This requires attention in the interpretation of the results, as for Germany and Australia the results do not refer to the impact of unemployment on life satisfaction 7 alternative variable that could be used as a proxy for happiness, we feel confident that for the purposes of the present study the use of life satisfaction measure can offer a good approximation of individual happiness and well being levels. In the German and the Australian data happiness indicators are measured using an eleven point index from 0 “completely dissatisfied” to 10 “completely satisfied”. The question is: “How satisfied are you with your life, all things considered?”. For British data, the satisfaction index is measured on a 0 to 7 scale. Subjective survey data, like that used in the present study, could be prone to several systematic or non-systematic biases (Kahneman, Diener, and Schwarz (1999)). However as Frey and Stutzer (2005) reports, “the relevance of reporting errors depends on the intended usage of the data”. Thus, when the purpose is not to measure or to compare levels in an absolute sense, the bias does not seem to be relevant. So, for the purpose of identifying parameters that influence happiness, these measures are valid. 3.2 Methodology Although availability of data is often not an issue, existing studies do not exploit all the available information, neglecting the strength of panel analysis. The present study, utilizes panel methodology in order to exhaust the possible sources of information and enhance the explanatory power of the model. Differences across individuals are expected to have some influence on the dependent variable, and thus a Random Effects (RE) model is used. RE here allow to control for time invariant variables, i.e. gender, disability status etc. In order to test whether our model of choice, that is RE versus Fixed Effects (FE), is the appropriate one, we 8 run a Hausman Test. The results indicate that RE should be used. The choice of explanatory variables used in the regression analysis follows on i) our intuition regarding the possible determinants of individual happiness given natural limitations in the data, and ii) the literature on this topic (Oswald and Powdthavee (2007), Cornlisse-Vermatt, Antonides, Ophem, and den Brink (2006)). Surprisingly, existing literature, with the exception of Stutzer (2007), examining the relationship between happiness and obesity does not address the issue of endogeneity that could be resulting from reverse causality running from dependent and independent variables. Endogeneity could stem from multiple sources here since happiness influences and is being influenced by a series of factors. In addition to obesity, several other factors included in happiness regressions, i.e. employment status, marital status, income, and which arguably have an impact on individual well-being, are at the same time influenced by it. As a consequence, dual causality might run in these types of specifications for more than one variable, in fact it could run for most regressors which are not exogenous by nature (such as age and gender). In the presence of endogeneity we build the following model: yit = αi + Xit′ β + uit (1) Here X is an n×K matrix of control variables (Cornlisse-Vermatt et al. (2006), Frey and Stutzer (2000), Blanchflower (2008)), some of which are endogenous and thus EX ′ u 6= 0. Given the panel structure of our dataset, for all potentially endogenous variables, excluding BMI, we instrument using their first lag. The availability of data 9 for all years, i.e. income, employment status, marital status, etc., makes the use of lags as instrumental variables the best option. Following the existing theory (Cameron and Trivedi (2009)) lags of endogenous variables can offer consistent estimators of the coefficients of interest when they serve as excluded instruments and are by nature exogenous. For BMI the instrument of choice is individual height. BMI is correlated with the instrument by definition since height is used in the construction of BMI. Hence the first IV assumption, cov(Z, y2 ) 6= 0, where Z is the IV and y2 is BMI, holds. The second critical assumption is that EZ ′ u = 0. In order to provide necessary and appropriate justification that the instrument of choice serves the second assumption too, we test whether it is uncorrelated with the error term u in the main equation. The correlation results show that the second critical assumption for consistent IV, that is EZ ′ u = 0, holds. Recent literature analyzing the relationship between height and happiness cannot be neglected at this point. The findings of Deaton and Arora (2009) reveal a positive relationship between height and happiness levels. However, after controlling for income this relationship is not statistically different from zero. The effect captured in this case is the one of height on wages, and thus indirectly on happiness, and not a direct effect of height on happiness. For this reason we argue that, both intuitively and statistically, height is exogenous to happiness and thus can be used to instrument for BMI in the main equation. 10 4 Results 4.1 Results for Germany All results for Germany are shown in Table 7. Below, we analyze only the instrumental variable results (IV REG1 through IV REG7). OLS results are presented in columns 1 and 2. The regression results for Germany point to a clear negative and statistically significant relationship between obesity and happiness. Under all specifications, IV REG1 through IV REG6 in Table 7 the coefficient on BMI is in the range (−0.0729, −0.0797), significant at the 1% level. Given the size of the coefficient and its robustness in the multiple specifications used, one can conclude that higher levels of BMI are associated with lower levels of self reported life-satisfaction. Regarding the rest of the explanatory variables, females report to be less “satisfied” with life compared to men. All results on gender are statistically significant at the 5% or the 1% levels. As expected, disability reduces life-satisfaction by approximately 0.5 units under all specifications. This result is statistically significant at the 1% level. Income is associated with higher levels of happiness. The coefficient on income suggests that individuals with 20% higher income report life-satisfaction levels one unit higher than others, ceteris paribus. Educational attainment is positively associated with individual well-being. The sign of the coefficient is consistently positive, significant at the 1% level across specifications. The size varies in the range between 0.0238 and 0.0472. With respect to marital status, only being single or being divorced appears to be statistically different from 0. In agreement with intuition, as well as past research, the number of children 11 increases life-satisfaction. One additional children in the family appears to be associated approximately with a 0.09 unit increase self reported happiness levels. These results are significant at the 1% level across all specifications. Individuals who live in “crowded” homes, seem to suffer a loss in their well being, equivalent to almost 0.1 of a unit, for every additional person added to the household. Agents who report to be members of a political party, report self satisfaction levels approximately 0.3 units higher than those who report the opposite. 4.2 Results for the UK Regressions output for Britain are shown in Table 8. Columns IV REG1 through IV REG7 present the instrument variable regression results. The coefficient on BMI has the expected sign. However, except for IV REG1, the results across specifications are not statistically significant. Life satisfaction is decreasing with age, at an increasing rate. The results on age are all significant. Disability status appears to decrease life satisfaction by 1 whole unit, under all specifications, a result that is significant at less than 1% level. Surprisingly, the coefficient on income is negative. However, across specifications IV REG3 through IV REG6 this result is not statistically significant. The results regarding the relationship between education and happiness are not robust. For the most part, they are not statistically significant, leading to no-single firm conclusion. Being separated, widowed or divorced are all received as negative shocks to individual life satisfaction. Similar to the German results, being divorced or separated appears to have the most severe negative impact on personal well being, among different marital statuses. All marital status results are statistically 12 significant. Smoking appears to negatively affect well being. In particular, under specifications IV REG5, IV REG6 and IV REG7, the coefficient ranges between −0.22 and −0.28, significant at the 1% level. Regarding financial information, our findings indicate that people who save and people who own their own home are happier, ceteris paribus. 4.3 Results for Australia The Australian regression results are presented in Table 9. Here, like in the German data, the regression analysis reveals a negative and highly statistically significant relationship between BMI and self reported life-satisfaction. In particular, in the multivariate instrumental variable specifications IV REG4 through IV REG7 the coefficient on BMI is approximatly −0.04 at the 1% level of significance. Older age decreases well being, at an increasing rate. Disability is found to lower life satisfaction slightly more than half a unit, on the 0-10 scale. These findings are highly statistically significant. As expected, income is associated with higher levels of individual happiness. Individuals with 10% higher income are expected to report higher levels of life satisfaction of approximately 0.3 units. Educational attainment has a negative and statistically significant coefficient. Individuals with more years of education are expected to report a 0.04 lower happiness levels for every extra year of education they have acquired. With respect to marital status, individuals who are single, separated, widowed or divorced are expected to report lower self satisfaction levels than married ones, ceteris paribus. Once again, the most severe effect appears to come from being separated, where the coefficient is close to −0.85 under all specifications. The number of children in the Australian 13 regressions does not exhibit statistically significant results, unlike the German and the British regressions. On the contrary, the size of the household is negatively related with individual self reported life satisfaction levels. In particular, an one member difference in the size of a household is expected to result in a 0.07 difference in the individual happiness levels. The results are significant at the 1% level. 5 Conclusions This study investigates the impact of obesity on individual happiness using panel analysis for Germany, the United Kingdom, and Australia. The contribution to the literature is three fold: first, to our knowledge, this is the first study to explore the panel dimension of the existing data in the investigation of the addressed research question. Secondly, this is the first study that examines the Australian data to identify the possible relationship between obesity and happiness. In addition, this study addresses the potential endogeneity problems that arise from most variables included in the specifications used, as a result of reverse causality. These endogeneity issues are tackled using the panel elements of the data which offer the necessary exogenous instruments. Last, but not least, we identify other significant determinants of life-satisfaction and discuss them. For Germany and Australia, BMI has a negative and statistically significant relationship with self reported life-satisfaction levels. For Britain, although the coefficient on BMI is negative in all specifications, the results are not statistically different from 0. The findings across specifications for all three countries point to some common conclusions. First, disability severely impacts individual 14 happiness more than any other individual characteristic. Secondly, being separated or divorced (compared to being married) reduces well being at a statistically significant level. Other results indicate that for Germany and Australia income is positively associated with happiness, as expected. For education, the results are mixed: for Australia the relationship is negative and significant, whereas for Germany the opposite holds. For Britain, house ownership and saving habits appear to be beneficial for individual happiness whereas smoking impairs well being. Household size, measured as number of people living in a household, decreases life satisfaction at a statistically significant level, both in Germany and Britain. In Australia, females appear to be happier whereas in Germany they are found to be less happy compared to males. 15 References Blanchflower, D. G. (2008). International Evidence on Well-being. Iza discussion papers 3354, Institute for the Study of Labor (IZA). Cameron, C., & Trivedi, P. (2009). Microeconometrics using Stata (1st. edition)., p. 692. Stata Press. Cornlisse-Vermatt, J. R., Antonides, G., Ophem, J. V., & den Brink, H. M. V. (2006). Body Mass Index, Perceived Health, and Happiness: their determinants and structural relationships. Social Indicators Research, 79, 143–158. Deaton, A., & Arora, R. (2009). Life at to top: the benefits of height. Nber working papers 15090, National Bureau of Economic Research, Inc. available at http://ideas.repec.org/p/nbr/nberwo/15090.html. Frey, B., & Stutzer, A. (2000). Maximising Happiness?. German Economic Review, 1 (2), 145–167. Frey, B., & Stutzer, A. (2005). Happiness Research: State and Prospects. Review of Social Economy, 62 (2), 207–228. Greeno, C., Jackson, C., Williams, E., & Fortmann, S. (1998). The Effect of Perceived Control over Eating on the Life Satisfaction of Women and Men: Results from a Community Sample. International Journal of Eating Disorders, 24 (4), 415–419. Kahneman, D., Diener, E., & Schwarz, N. (1999). Well-being: The Foundations of Hedonic Psychology, pp. 61–84. Russel Sage Foundation: New York. 16 Larsson, U., Karlsson, J., & Sullivan, M. (2002). Impact of overweight and obesity on health-related quality of life - a Swedish population study. International Journal obesity, 26, 417–424. Needham, B., & Crosnoe, R. (2005). Overweight Status and Depressive Symptoms During Adolescence. Journal of Adolescent Health, 36 (1), 48–55. Noppa, H., & Hällström, T. (1981). Weight gain in adulthood in relation to socioeconomic factors, mental illness, and personality traits: a prospective study of middle-aged women. Journal of Psychosomatic Research, 25, 83–89. Oswald, A., & Powdthavee, N. (2007). Obesity, Unhappiness, and the Challenge of Affluence: Theory and Evidence. Economic Journal, 117, F441–F459. Palinkas, L., Wingard, D., & Barrett-Connor, E. (1996). Depressive symptoms in overweight and obese older adults: a test of the ”jolly fat” hypothesis. Journal of Psychosomatic Research, 40, 56–60. Reed, D. (1985). The relationship between obesity and psychological general wellbeing in United States women. Journal of Psychosomatic Research (Abstract), 46, 3791. Roberts, R., Kaplan, G., Shema, S., & Strawbridge, W. (2000). Are the Obese at Greater Risk of Depression?. American Journal of Epidemiology, 152 (2), 163–170. Stutzer, A. (2007). Limited Self-Control, Obesity and the Loss of Happiness. Iza discussion papers 2925, Institute for the Study of Labor (IZA). 17 Veenhoven, R. (1996). Happy Life-Expectancy. Social Indicators Research, 39, 1–58. 18 Table 1: Descriptive Statistics for German Data Variable German Data - GSOEP - Years: 2002, 2004, 2006, 2008 Mean Std. Dev. Min Max Observations Age overall between within 46.053 18.242 18.923 2.016 15 15 41.553 100 99 50.553 N = 180714 n = 33272 T-bar = 5.43141 Household size overall between within 3.023 1.370 1.306 0.448 1 1 -6.23 13 13 9.82 N = 215766 n = 39311 T-bar = 5.48869 No children overall between within 0.820 1.088 1.042 0.357 0 0 -4.805 9 7.875 4.820 N = 215766 n = 39311 T-bar = 5.48869 Education overall between within 12.068 2.676 2.654 0.331 7 7 6.924766 18 18 17.21048 N = 157209 n = 28833 T-bar = 5.4524 Life Satisfaction overall between within 6.964 1.785 1.491 1.101 0 0 -1.61 10 10 14.11 N = 165630 n = 30615 T-bar = 5.41009 Height overall between within 1.713 0.093 0.092 0.016 0.82 1.31 0.96 2.1 2.09 2.11 N = 83227 n = 28545 T-bar = 2.91564 Weight overall between within 75.433 15.514 15.039 3.962 32 35 7.43 230 200 156.18 N = 82681 n = 28452 T-bar = 2.90598 BMI overall between within 25.619 4.589 4.338 1.556 11.63 12.86 -20.66 197.23 73.46 152.64 N = 82644 n = 28449 T-bar = 2.90499 ln Income overall between within 10.384 0.641 0.591 0.304 0 0 0.80 15.62 13.83 15.82 N = 215763 n = 39310 T-bar = 5.48876 Female overall between within 0.510 0.500 0.500 0.000 0 0 0.510 1 1 0.510 N = 215766 n = 39311 T-bar = 5.48869 Widowed overall between within 0.050 0.219 0.209 0.063 0 0 -0.825 1 1 0.925 N = 215766 n = 39311 T-bar = 5.48869 Divorced overall between within 0.054 0.226 0.206 0.087 0 0 -0.821 1 1 0.929 N = 215766 n = 39311 T-bar = 5.48869 Separated overall between within 0.013 0.115 0.088 0.080 0 0 -0.862 1 1 0.888 N = 215766 n = 39311 T-bar = 5.48869 Unemployed overall between within 0.367 0.482 0.426 0.247 0 0 -0.508 1 1 1.242 N = 215766 n = 39311 T-bar = 5.48869 Disabled overall between within 0.088 0.283 0.249 0.120 0 0 -0.787 1 1 0.963 N = 208742 n = 39030 T-bar = 5.34824 Political party member overall between within 0.448 0.497 0.413 0.297 0 0 -0.427 1 1 1.323 N = 166048 n = 30638 T-bar = 5.41967 Has a second job overall between within 0.027 0.162 0.118 0.120 0 0.000 -0.848 1 1 0.902 N = 166048 n = 30638 T-bar = 5.41967 German overall between within 0.926 0.261 0.263 0.045 0 0.000 0.0513 1 1 1.8013 N = 166048 n = 30638 T-bar = 5.41967 19 Table 2: Correlation Matrix for GSOEP variables Age Age Household size No children Education Life Satisfaction Height Weight BMI ln Income Female Widowed Divorced Separated Not Employed Disabled Political party member Has a second job German BMI ln Income Female Widowed Divorced Separated Not Employed Disabled Political party member Has a second job German No children Education 1 0.7934* 0.0089* 0.0670* 0.0991* -0.0102* -0.0748* 0.4198* -0.0387* -0.2648* -0.1780* -0.0850* -0.1934* -0.1875* -0.0756* 0.0150* -0.1210* 1 0.0355* 0.0460* 0.0675* -0.0170* -0.0655* 0.1791* -0.0055* -0.1574* -0.0936* -0.0322* -0.2561* -0.1849* -0.0641* 0.0112* -0.1009* 1 0.1370* 0.1900* -0.0005 -0.1230* 0.3403* -0.0805* -0.1325* -0.0164* 0.0119* -0.2486* -0.1025* 0.1989* 0.0455* 0.1703* BMI ln Income Female 1 -0.0784* -0.1447* 0.0656* 0.002 -0.0031 0.0332* 0.1182* 0.0287* -0.0144* -0.0209* 1 -0.0605* -0.2051* -0.1438* -0.0732* -0.2330* -0.1104* 0.1386* 0.0288* 0.0495* 1 0.1264* 0.0372* 0.0045* 0.0907* -0.0277* -0.0789* -0.0061* 0.0076* Political party member Has second job 1 0.0177* 0.1299* 1 0.0048 1 -0.4380* -0.3878* -0.0892* -0.0631* -0.2298* 0.1055* 0.2647* -0.1515* 0.0281* 0.3583* 0.0672* 0.0084* 0.2638* 0.2815* 0.1741* -0.0567* 0.0698* Disabled Disabled Political party member Has a second job German 1 0.0432* -0.0269* 0.0234* Household Size *significant at 5% 20 Height Weight 1 0.0665* -0.0348* -0.0835* 0.2179* -0.0005 -0.0471* -0.0812* -0.0544* -0.0665* -0.1618* 0.1021* 0.0145* 0.0272* 1 0.5266* -0.0181* 0.1650* -0.6696* -0.1921* -0.0260* 0.0007 -0.2079* -0.0564* 0.0663* 0.0300* 0.0814* 1 0.8298* 0.0253* -0.4832* -0.0493* -0.0108* -0.0015 -0.0893* 0.0708* 0.0634* 0.0037 0.0284* Widowed Divorced Separated Not Employed 1 -0.0551* -0.0268* 0.2285* 0.1092* 0.0228* -0.0280* 0.0438* 1 -0.0279* -0.0284* 0.0371* -0.0188* 0.0144* 0.0176* 1 -0.0259* 0.0093* -0.0038 0.0118* -0.0039 1 0.2117* -0.0082* -0.0799* -0.0077* German 1 Life Satisfaction Table 3: Descriptive Statistics for British Data Variable British Data - BHPS - Years: 2005 and 2007 Mean Std. Dev. Min Max Observations Age overall between within 45.958 18.649 19.267 1.073 15 15 41.46 99 99 50.46 N = 63036 n = 18961 T-bar = 3.32451 Household size overall between within 2.870 1.405 1.382 0.402 1 1 -3.880 14 13.5 8.870 N = 63038 n = 18961 T-bar = 3.32461 No children overall between within 0.499 0.914 0.884 0.182 0 0 -1.834 7 7 3.166 N = 46800 n = 17675 T-bar = 2.64781 Education overall between within 11.329 5.052 4.999 0.568 2 2 2.329 20 20 20.329 N = 28575 n = 15968 T-bar = 1.78952 Height overall between within 1.646 0.112 0.103 0.046 0.55 0.85 1.05 2.275 2.125 2.25 N = 28522 n = 16088 T-bar = 1.77287 Weight overall between within 76.051 15.802 15.969 2.576 12.7 12.7 37.05 184.15 184.15 115.05 N = 23499 n = 14768 T-bar = 1.59121 BMI overall between within 27.925 5.913 5.526 2.316 5.161 6.040 -74.710 227.769 125.133 130.560 N = 23249 n = 14652 T-bar = 1.58675 Life Satisfaction overall between within 5.228 1.280 1.104 0.709 1 1 0.727852 7 7 9.727852 N = 58402 n = 18066 T-bar = 3.2327 Female overall between within 0.535 0.499 0.495 0.066 0 0 -0.215 1 1 1.035 N = 63038 n = 18961 T-bar = 3.32461 Widowed overall between within 0.076 0.265 0.254 0.061 0 0 -0.674 1 1 0.826 N = 63038 n = 18961 T-bar = 3.32461 Divorced overall between within 0.080 0.271 0.254 0.083 0 0 -0.670 1 1 0.830 N = 63038 n = 18961 T-bar = 3.32461 Separated overall between within 0.021 0.142 0.121 0.076 0 0 -0.729 1 1 0.771 N = 63038 n = 18961 T-bar = 3.32461 Unemployed overall between within 0.032 0.176 0.150 0.120 0 0 -0.718 1 1 0.782 N = 63038 n = 18961 T-bar = 3.32461 Disabled overall between within 0.078 0.269 0.217 0.169 0 0 -0.672 1 1 0.828 N = 63038 n = 18961 T-bar = 3.32461 ln Income overall between within 8.883 2.031 2.070 1.009 0 0 1.212 13.99 12.23 15.728 N = 59036 n = 17902 T-bar = 3.29773 Has a second job overall between within 0.058 0.234 0.188 0.147 0 0 -0.692 1 1 0.808 N = 63038 n = 18961 T-bar = 3.32461 Political party member overall between within 0.264 0.441 0.315 0.309 0 0 -0.486 1 1 1.014 N = 63038 n = 18961 T-bar = 3.32461 Smoker overall between within 0.240 0.427 0.405 0.148 0 0 -0.510 1 1 0.990 N = 63038 n = 18961 T-bar = 3.32461 Labor Union member overall between within 0.151 0.358 0.320 0.142 0 0 -0.599 1 1 0.901 N = 63038 n = 18961 T-bar = 3.32461 Saves overall between within 0.392 0.488 0.390 0.304 0 0 -0.358 1 1 1.142 N = 63038 n = 18961 T-bar = 3.32461 House Owner overall between within 0.743 0.437 0.425 0.157 0 0 -0.007 1 1 1.493 N = 63038 n = 18961 T-bar = 3.32461 21 Table 4: Correlation Matrix for BHPS variables Age Age Household size No children Education Height Weight BMI Life Satisfaction Female Widowed Divorced Separated Unemployed Disabled ln Income Has a second job Political party member Smoker Labor Union Member Saves House Owner 1 -0.4448 -0.248 0.3231* -0.1569 0.0183* 0.1256* 0.0709* 0.0233* 0.4268* 0.0678* -0.0125* -0.1169* 0.1771* 0.1984* -0.0976* 0.2346* -0.1386* -0.0919* -0.0187 0.0912* Life Satisfaction Female Widowed Divorced Separated Unemployed Disabled ln Income Has a second job Political party member Smoker Labor Union Member Saves House Owner Life Satisfaction 1 -0.0093 0.0166* -0.0928 -0.0649 -0.0841 -0.1683* -0.0124 0.0031 0.0267* -0.1337* 0.0094* 0.0945* 0.1262* ln Income ln Income Has a second job Political party member Smoker Labor Union Member Saves House Owner 1 -0.0424 0.0764* -0.0032 0.2040* 0.1505* 0.0952* Household size No children Education Height Weight BMI 1 0.5744* -0.0854* 0.0528* 0.0129* -0.0265* -0.0278* -0.0176* -0.2946* -0.1277* -0.0453* 0.0454* -0.1089* -0.1433* 0.0561* -0.1067* 0.0362* 0.0354* -0.0451* 0.0685* 1 -0.1227* 0.014* 0.0327* 0.0134* -0.0447* 0.0317* -0.1449* -0.0178* 0.0397* 0.0018 -0.0605* 0.1413* 0.0105* -0.0833* 0.0523* 0.0746* -0.0438* -0.0017 1 -0.1488* -0.0479* 0.0554* -0.0086 0.0556* 0.2199* 0.0167* 0.007 0.0452* 0.1368* -0.1935* -0.1025* 0.0618* 0.1264* -0.2149* -0.1716* -0.1913* 1 0.4023* -0.2978* 0.0108 -0.5426* -0.1382* -0.0419* -0.0159* 0.0195* -0.0253* 0.0614* 0.0381* -0.0107 0.0087 0.0163* 0.0215* 0.0507* 1 0.7281* -0.0394* -0.4119* -0.0998* 0.0014 0.0004 -0.0057 0.0464* 0.1622* 0.0101 0.0396* -0.0706* 0.0453* 0.0038 0.0353* 1 -0.0446* -0.0701* -0.0086 0.0307* 0.0082 -0.0167* 0.0675* 0.1161* -0.0207* 0.0456* -0.0731* 0.0274* -0.0117 -0.0043 Female Widowed Divorced Separated Unemployed Disabled 1 0.1236* 0.0563* 0.022* -0.0443* 0.0072 -0.1207* 0.0011 -0.0259* -0.0039 0.0224* 0.002 -0.0308* 1 -0.0842* -0.0416* -0.0449* 0.1001* 0.0325* -0.0551* 0.0843* -0.0524* -0.0964* -0.0240* -0.0389* 1 -0.0428* 0.0115* 0.0615* 0.0686* 0.0074 -0.0090* 0.0929* 0.0225* -0.0233* -0.0895* 1 0.0236* 0.0094* 0.0412* -0.0056 -0.0180* 0.0666* 0.0113* -0.0254* -0.0641* 1 -0.0295* -0.1133* -0.0066 -0.0407* 0.1158* -0.0738* -0.0984* -0.1232* 1 0.0004 -0.0599* 0.0048 0.0613* -0.0993* -0.0706* -0.1269* Has second Job Political party member Saves House Owner 1 -0.0135* -0.0031 0.0339* 0.0405* 0.0391* 1 -0.0529* 0.0152* 0.0402* 0.0788* 1 0.1670* 1 *significant at 5% 22 Smoker 1 -0.0361* -0.1051* -0.1835* Labor Union member 1 0.1622* 0.1378* Table 5: Descriptive Statistics for Australian Data Variable Australian Data - HILDA - Years: 2006, 2007 and 2009 Mean Std. Dev. Min Max Observations Age overall between within 43.85421 18.59439 19.05329 1.320974 15 15 38.35421 93 93 49.35421 N = 67729 n = 17315 T-bar = 3.91158 Household size overall between within 3.207649 1.462235 1.363177 0.657998 1 1 -4.99235 14 13.5 10.20765 N = 82649 n = 21265 T-bar = 3.88662 No children overall between within 1.241713 1.394873 1.23369 0.747201 0 0 -5.95829 14 9.6 12.44171 N = 81787 n = 20736 T-bar = 3.9442 Education overall between within 11.58874 2.391224 2.023765 1.260333 0 0 3.088743 18.5 18.5 19.58874 N = 63909 n = 16348 T-bar = 3.90929 Height overall between within 1.704411 0.104764 0.10352 0.022699 0.82 1.27 1.204411 2.29 2.29 2.204411 N = 28695 n = 13759 T-bar = 2.08554 Weight overall between within 76.87511 17.98581 17.51675 4.691137 28 28 7.541781 260 236.6667 160.5418 N = 32870 n = 14216 T-bar = 2.31218 BMI overall between within 26.32283 5.573583 5.379719 1.74933 12.12121 13.06122 -38.4166 163.5931 98.85366 91.06226 N = 28248 n = 13650 T-bar = 2.06945 Income overall between within 65529.32 51454.74 42482.27 30003.61 1 1 -354404 611361 562353 474042.1 N = 81902 n = 21209 T-bar = 3.86166 ln Income overall between within 10.30597 2.517529 1.48586 2.129551 0 0 0.327107 13.32344 13.23989 16.8477 N = 81902 n = 21209 T-bar = 3.86166 Female overall between within 0.514283 0.499799 0.499899 0 0 0 0.514283 1 1 0.514283 N = 86816 n = 20710 T-bar = 4.19198 Married overall between within 0.456497 0.498107 0.463687 0.174039 0 0 -0.3435 1 1 1.256497 N = 82649 n = 21265 T-bar = 3.88662 Single overall between within 0.191339 0.393358 0.362321 0.161919 0 0 -0.60866 1 1 0.991339 N = 82649 n = 21265 T-bar = 3.88662 Widowed overall between within 0.041936 0.200445 0.186951 0.055324 0 0 -0.75806 1 1 0.841936 N = 82649 n = 21265 T-bar = 3.88662 Divorced overall between within 0.052064 0.222157 0.195576 0.091528 0 0 -0.74794 1 1 0.852064 N = 82649 n = 21265 T-bar = 3.88662 Separated overall between within 0.022335 0.147773 0.122422 0.083549 0 0 -0.77766 1 1 0.822335 N = 82649 n = 21265 T-bar = 3.88662 Not employed overall between within 0.469455 0.499069 0.427514 0.283706 0 0 -0.33054 1 1 1.269455 N = 82649 n = 21265 T-bar = 3.88662 Disabled overall between within 0.152379 0.35939 0.293779 0.204189 0 0 -0.64762 1 1 0.952379 N = 82649 n = 21265 T-bar = 3.88662 23 Table 6: Correlation Matrix for Australian variables Age Age Household size No children Education Life Satisfaction Height Weight BMI ln Income Female Single Widowed Divorced Separated Not Employed Disabled Household Size No children Education Life Satisfaction Height Weight 1 -0.3244* -0.1221* -0.1421* 0.0623* -0.1473* 0.0698* 0.1675* -0.2022* 0.0284* -0.5084* 0.3776* 0.1353* 0.0419* 0.3009* 0.3188* 1 0.7610* 0.0273* 0.0237* 0.0442* 0.0039 -0.0216* 0.1604* -0.0212* -0.0787* -0.2125* -0.1736* -0.0967* 0.1431* -0.1988* 1 -0.0693* -0.0104* 0.0103 0.0380* 0.0279* -0.0986* 0.0100* -0.1751* -0.0878* -0.0575* -0.0164* 0.1070* -0.1197* 1 -0.0453* 0.0827* -0.0063 -0.0483* 0.2714* -0.0281* -0.0429* -0.1555* -0.0258* -0.0048 -0.1721* -0.1732* 1 -0.01 -0.0350* -0.0335* 0.0498* 0.0250* -0.0353* 0.0405* -0.0804* -0.0935* 0.0407* -0.1582* 1 0.4728* -0.0731* 0.0643* -0.6629* 0.0890* -0.1449* -0.0558* -0.0093 -0.1327* -0.0751* 1 0.8314* 0.0230* -0.3986* -0.1085* -0.0676* 0.003 0.0126* -0.0640* 0.0586* BMI ln Income Female Single Widowed Divorced Separated BMI ln Income Female Single Widowed Divorced Separated Not Employed Disabled 1 0.0012 -0.0570* -0.1721* 0.0132* 0.0352* 0.0160* 0.0157* 0.1097* 1 -0.0489* -0.0488* -0.1686* -0.0761* -0.0333* 0.1530* -0.2413* 1 -0.0398* 0.1284* 0.0507* 0.0126* 0.0305* 0.0304* 1 -0.1018* -0.1140* -0.0735* -0.1963* -0.0346* 1 -0.0490* -0.0316* 0.1011* 0.1901* 1 -0.0354* -0.0672* 0.1083* 1 -0.0464* 0.0393* Not Employed Disabled Not Employed 1 0.0719* Disabled 1 *significant at 5% 24 Table 7: Regression Results for Germany Independent OLS1 Variable Age Age 2 Female OLS2 IVREG1 IVREG2 IVREG3 IVREG4 IVREG5 IVREG6 IVREG7 -0.0164 ( -7.11) 0.000 (2.03) 0.0077 (0.46) -0.0149 (-5.62) 0.0001 (2.85) -0.0335 (-1.83) -0.0141 (-6.58) 0.0068 (1.22) -0.0001 (-2.46) -0.1132 (-3.39) -0.0776 (-4.44) 0.0071 (1.41) -0.0001 (-2.01) -0.1309 (-5.21) -0.0797 (-5.29) -0.5352 (-13.2) -0.0034 (-0.45) 0.00 (0.59) -0.0874 (-3.16) -0.0739 (-4.98) -0.5667 (-13.32) 0.3232 (7.69) 0.0472 (9.58) -0.1277 (-1.60) 0.0965 (1.74) -0.1182 (-3.26) -0.2438 (-5.16) 0.0199 (0.57) -0.0086 (-1.15) 0.0001 (1.22) -0.0915 (-2.93) -0.0733 (-4.07) -0.5889 (-15.9) 0.4557 (8.29) 0.0353 (5.41) -0.1886 (-1.94) 0.02 (0.36) -0.1423 (-3.94) -0.294 (-5.99) 0.0534 (1.66) 0.0935 (5.30) -0.1188 (-4.91) -0.0111 (-1.20) 0.0001 (1.07) -0.0678 (-1.95) -0.0728 (-3.81) -0.6031 (-14.31) 0.4166 (7.13) 0.0238 (4.28) -0.2155 (-1.85) 0.017 (0.27) -0.1543 (-3.94) -0.2993 (-6.04) 0.0435 (1.05) 0.0852 (5.07) -0.1052 (-4.60) 0.2368 (1.51) 0.3678 (8.75) 7.63 (151.65) No 165630 7.83 (112.36) No 82466 8.96 (27.43) No 82466 8.99 (32.58) No 82423 5.06 (9.06) No 70287 4.25 (6.99) No 70287 4.66 (6.22) No 70247 -0.0102 (-1.30) 0.0001 (1.11) -0.0688 ( -2.35) -0.074 (-4.38) -0.599 (-15.39) 0.4127 (7.63) 0.0249 (4.94) -0.2127 (-1.66) 0.0227 (0.36) -0.1503 (-4.07) -0.2954 (-5.09) 0.0441 (1.19) 0.084 (4.33) -0.1041 (-4.69) 0.2379 (1.79) 0.3707 (8.89) -0.0641 (-1.43) 4.76 (6.93) No 70247 -0.0114 (-1.30) 0.0001 (1.09) -0.0726 (-2.75) -0.0619 (-4.12) -0.5985 (-15.55) 0.4208 (7.89) 0.0316 (6.36) -0.2263 (-1.66) 0.0582 (0.90) -0.1443 (-4.04) -0.2603 (-4.57) 0.0418 (0.97) 0.0873 (4.25) -0.1067 (-4.79) 0.226 (1.52) 0.2973 (5.20) -0.1006 (-1.97) 4.49 (7.43) Yes 65338 0.0103 0.0089 0.0044 692.63 0.000 0.0007 0.121 0.0075 377.55 0.000 0.000 0.0121 0.0078 349.34 0.000 0.0001 0.0285 0.0194 614.08 0.000 0.0005 0.0778 0.0555 1339.96 0.000 0.0008 0.0868 0.0624 1549.52 0.000 0.0014 0.094 0.0668 1866.61 0.000 0.0013 0.0929 0.066 2366.19 0.000 0.0017 0.121 0.081 2455.42 0.000 BMI Disabled ln Income Education Separated Widowed Single 25 Divorced Not employed No children Household size Second job Politics German Constant Religion Dummies N R-squared Within Between Overall Wald χ2 P-value Robust t-stats in parentheses Table 8: Regression Results for Britain Independent Variable Age Age2 Female OLS1 OLS2 IVREG1 IVREG2 IVREG3 IVREG4 IVREG5 IVREG6 IVREG7 -0.0225 (-10.11) 0.0002 (11.97) -0.032 (-1.98) -0.0174 (6.32) 0.0002 (8.45) -0.041 (-2.09) -0.007 (-5.30) -0.0182 (-6.64) 0.0002 (8.20) -0.0391 (-1.65) -0.0055 (-1.96) -0.0107 (-4.00) 0.0002 (5.85) -0.0384 (-1.75) -0.004 (-1.36) -1.055 (-18.09) -0.0078 (-1.88) 0.0001 (3.53) -0.0202 (-1.01) -0.0038 (-1.29) -1.0458 (-16.26) -0.0116 (-1.04) -0.0024 (-1.00) -0.3256 (-2.94) -0.2042 (-4.60) -0.37 (-7.75) -0.833 (-3.76) -0.0067 (-1.54) 0.0001 (2.84) -0.0162 (-0.76) -0.0036 (-1.41) -1.0533 (-18.46) -0.0083 (-0.72) -0.002 (-0.90) -0.3261 (-2.82) -0.2025 (-3.20) -0.377 (-8.21) -0.842 (-3.44) -0.032 (-2.29) -0.003 (-0.31) -0.0063 (-1.74) 0.0001 (2.83) -0.0266 (-1.49) -0.0042 (-1.75) -1.0352 (-16.84) -0.0105 (-0.87) 0.0032 (1.16) -0.2626 (-1.95) -0.1838 (-3.32) -0.33 (-6.48) -1.109 (-4.28) -0.0348 (-2.65) -0.0111 (-2.92) 0.0002 (4.08) -0.0201 (-0.78) -0.0026 (-0.82) -0.9978 (-16.32) -0.0179 (-1.43) 0.0078 (2.47) -0.2998 (-2.47) -0.1944 (-3.89) -0.282 (-6.43) -0.469 (-2.20) -0.0171 (-1.56) -0.0111 (-3.06) 0.0002 (4.10) -0.0262 (-1.06) -0.0028 (-0.82) -1.0018 (-12.85) -0.0194 (-1.81) 0.0072 (2.97) -0.3094 (-2.79) -0.1971 (-3.93) -0.264 (-5.92) -0.459 (-1.96) -0.0203 (-1.85) -0.28 (-9.03) -0.218 (-6.55) -0.053 (-1.42) -0.018 (-0.22) 0.001 (0.01) 0.315 (7.16) 0.191 (6.15) 5.49 (44.87) Yes BMI Disabled ln Income Education Separated Widowed Divorced Unemployed 26 No children Household size Smoker 5.59 (110.90) No 5.69 (83.28) No 5.64 (69.46) No 5.48 (60.14) No 5.59 (43.94) No 5.57 (43.54) No 5.65 (45.04) No -0.232 (-7.73) -0.049 (-1.59) -0.002 (-0.04) 0.004 (0.05) 0.322 (7.87) 0.181 (6.13) 5.40 (33.79) No 43504 22390 22390 22390 20550 20550 20550 20550 20154 0.000 0.0119 0.0105 209.80 0.000 0.0001 0.0134 0.0141 205.07 0.000 0.0001 0.0131 0.0139 126.39 0.000 0.0032 0.0586 0.0532 491.24 0.000 0.0047 0.0785 0.0703 470.69 0.000 0.0046 0.0792 0.0708 616.49 0.000 0.0042 0.0843 0.0741 1131.16 0.000 0.0041 0.0958 0.0842 1320.17 0.000 0.0041 0.098 0.0863 1420.62 0.000 Union Second job Politics Saves Home Owner Constant Religion Dummies N R-squared Within Between Overall Wald χ2 P-value Robust t-stats in parentheses Table 9: Regression Results for Australia Independent Variable Age Age2 Female OLS1 OLS2 IVREG1 IVREG2 IVREG3 IVREG4 IVREG5 IVREG6 IVREG7 -0.042 (-13.26) 0.0005 (15.54) 0.044 (1.87) -0.044 (-17.78) 0.0005 (19.5) 0.058 (2.91) -0.005 (-2.27) -0.033 (-7.82) 0.0004 (11.01) 0.036 (1.64) -0.03 (-2.74) -0.034 (-8.10) 0.0005 (11.55) 0.036 (1.59) -0.0306 (-2.66) -0.570 (-15.8) -0.034 (-5.85) 0.0005 (8.41) 0.063 (2.49) -0.0439 (-3.04) -0.524 (-12.46) 0.232 (3.52) -0.036 (-6.73) -0.241 (-4.55) -0.826 (-4.99) -0.198 (-2.46) -0.326 (-3.95) 0.132 (1.60) -0.036 (-4.31) 0.005 (6.11) 0.061 (1.89 -0.0412 (-2.53) -0.531 (-11.19) 0.311 (2.76) -0.041 (-5.57) -0.283 (-5.30) -0.853 (-5.47) -0.214 (-2.38) -0.358 (-4.78) 0.221 (2.31) -0.036 (-4.23) 0.0005 (5.99) 0.060 (2.64) -0.0412 (-2.10) -0.531 (-9.65) 0.311 (3.28) -0.040 (-6.33) -0.282 (-5.54) -0.853 (-5.47) -0.214 (-2.57) -0.358 (-5.42) 0.221 (2.34) -0.039 (-5.52) 0.0005 (6.98) 0.069 (2.55) -0.037 (-2.21) -0.534 (-8.59) 0.349 (2.53) -0.034 (-4.03) -0.260 (-5.35) -0.850 (-5.03) -0.189 (-2.14) -0.327 (-4.26) 0.217 (1.69) -0.068 (-2.35) 6.86 (5.46) -0.068 (-2.78) 6.86 (5.67) -0.074 (-2.16) 6.29 (4.02) 8.68 (112.19) 8.63 (161.2) 9.21 (38.84) 9.20 (37.64) 7.47 (8.44) -0.037 (-3.94) 0.0005 (5.37) 0.062 (2.45) -0.0409 (-2.16) -0.533 (-8.41) 0.326 (2.13) -0.041 (-4.76) -0.271 (-5.13) -0.875 (-6.28) -0.241 (-2.56) -0.372 (-5.64) 0.2135 (1.78) 0.053 (0.97) -0.104 (-1.47) 6.77 (4.14) No No No No No No No No Yes 51209 21887 21887 21887 19739 19739 19739 19739 19739 0.000 0.0283 0.0212 0.003 0.0267 0.0253 105.23 352.09 0.0000 0.0013 0.0583 0.0495 0.0013 0.0583 0.0495 29.7873 1053.32 0.0000 0.0045 0.0708 0.0619 0.0047 0.0697 0.0614 0.0048 0.0695 0.0614 0.0048 0.0696 0.0614 0.0047 0.0777 0.0677 882.29 0.0000 839.96 0.0000 1089.32 0.0000 1104.06 0.0000 1574.61 0.0000 BMI Disabled ln Income Education Single Separated 27 Widowed Divorced Not Employed No children Household size Constant Regional Dummies N R-squared Within Between Overall F Wald χ2 P-Value Robust t-stats in parentheses 427.15 0.0000 714.41 0.0000 Disclaimer This paper uses unit record data from the Household, Income and Labour Dynamics in Australia (HILDA) Survey. The HILDA Project was initiated and is funded by the Australian Government Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) and is managed by the Melbourne Institute of Applied Economic and Social Research (Melbourne Institute). The findings and views reported in this paper, however, are those of the author and should not be attributed to either FaHCSIA or the Melbourne Institute. The data used in this publication was made available to us by the German Socio-Economic Panel Study (SOEP) at the German Institute for Economic Research (DIW Berlin), Berlin. 28