There are a variety of studies that show that absolute income is positively correlated with individual well-being, but find at the same time that average income of the reference group (comparison income) affects individual well-being most often negatively (Clark et al., 2008). Although the results allover the literature are quite consistent, there is a large variety how the reference group is defined. For example, some authors assume that people compare themselves with people living in the same area (Luttmer, 2005; Graham & Felton, 2006) or with people inside the same age range (McBride, 2001). Others define the reference group more precisely and assume that people compare themselves with people of same age, same education and same area of living (Ferrer-i-Carbonell, 2005). However, to the best of my knowledge, there is no systematic empirical research on the impact of different reference group specifications on life satisfaction in happiness regressions. Therefore, I investigate in this master thesis to what extent different reference group specifications alter the statistical impact of comparison income on happiness regarding sign, magnitude and statistical significance. The results show that the specification of the reference group matters, since some specifications produce significant and others produce insignificant coefficients. However, the results also show that the sub-sample treated has a considerable impact on sign and statistical significance of the reference groups defined.

There are a variety of studies that show that absolute income is positively correlated with individual well-being, but find at the same time that average income of the reference group (comparison income) affects individual well-being most often negatively (Clark et al., 2008). Although the results allover the literature are quite consistent, there is a large variety how the reference group is defined. For example, some authors assume that people compare themselves with people living in the same area (Luttmer, 2005; Graham & Felton, 2006) or with people inside the same age range (McBride, 2001). Others define the reference group more precisely and assume that people compare themselves with people of same age, same education and same area of living (Ferrer-i-Carbonell, 2005). However, to the best of my knowledge, there is no systematic empirical research on the impact of different reference group specifications on life satisfaction in happiness regressions. Therefore, I investigate in this master thesis to what extent different reference group specifications alter the statistical impact of comparison income on happiness regarding sign, magnitude and statistical significance. The results show that the specification of the reference group matters, since some specifications produce significant and others produce insignificant coefficients. However, the results also show that the sub-sample treated has a considerable impact on sign and statistical significance of the reference groups defined.:Chapter Page Outline II List of Figures and Tables III Abbreviations III 1 Introduction 1 2 Measurement of Happiness in Economics 2 3 The Economics of Happiness 6 3.1 The Relation between Absolute Income and Happiness 6 3.2 Unemployment, Inflation and Inequality 12 4 The Role of Relative Income 15 4.1 Empirical Evidence 16 4.1.1 Empirical Evidence for the \''Social Comparison Effect\'' 16 4.1.2 Empirical Evidence for the \''Tunnel Effect\'' 21 4.1.3 Derived Empirical Regularities 23 4.2 Theoretical Considerations 24 4.2.1 Modeling of Utility Functions 24 4.2.2 A Contribution in Explaining the Easterlin-Paradox? 26 4.2.3 Concluding Remarks 29 5 Specifications of the Reference Group 30 5.1 The Reference Group as Exogeneous Variable 30 5.2 The Reference Group as Endogeneous Variable 33 6 Different Reference Group Specifications and Life Satisfaction 35 6.1 Data Description and Choice of Variables 36 6.2 Methodology 38 6.2.1 Data Preparation 38 6.2.2 Estimation Procedure 39 6.2.3 Further Econometric Issues 44 6.3 Relative Income and Reference Group Specifications 46 6.4 Results 49 6.4.1 Whole Sample 49 6.4.1.1 Looking for the Tunnel Effect: Young and Old respondents 54 6.4.1.2 Looking for the Tunnel Effect: Transition Countries 56 6.4.2 German Sub-Samples 57 6.4.2.1 Whole German Sample 57 6.4.2.2 Looking for the Tunnel Effect: Young and Old respondents 58 6.4.3 Summary of the Empirical Results 60 6.5 Problems and Shortcomings of the Study 61 7 Conclusion 63 Appendix A – List of Variables 65 Appendix B – Correlations 67 References 68 Statement of Authorship 74