The Econometrics of Happiness: A cross-sectional European analysis on the shifting importance of the determinants of happiness

Abstract: This paper tests the hypothesis that determinants affecting an agent's happiness do so differently depending on the happiness level of the agent. What affects a happy person's well-being and to what extent is not necessarily the same as what and how it affects an unhappy person's well-being. The hypothesis is based on the idea that current methodology within the field of happiness economics is flawed. Researchers neglect to test the proportional odds assumption, latent in the ordered logit regression models which are commonly used within the field. This is an implicit assumption of constant betas for the determinants of happiness. We investigate whether the proportional odds assumption holds using a vast, cross-sectional dataset covering large portions of Europe and find that the assumption is indeed violated for several determinants of happiness. We therefore opt to apply a partial proportional odds model, and find statistically significant evidence that the betas of the determinants of happiness vary distinctly across happiness levels for 11 out of 25 variables. Some of the largest effects on an agent's (un)happiness are caused by income, health and unemployment and the effects of all these determinants vary depending on how happy the agent is. We thus find strong indications that the current negligence with regard to the assumptions of the econometric models has large implications for the field of happiness economics.