This thesis examines the interrelationships between health, income, and employment using empirical methods that integrate economic, behavioural, and genetic perspectives. It consists of three chapters, each addressing a key question at the intersection of health and labour market outcomes, with a strong focus on causal inference.
The first chapter investigates the impact of mental health on early retirement among individuals aged 50–64 in Australia, using data from the Household, Income and Labour Dynamics in Australia (HILDA) survey. Linear models, survival analysis, and instrumental variable methods are applied to address endogeneity, with mental health instrumented using the death of a family member in the previous 12 months. The results show a positive and statistically significant association between poor mental health and the probability of early retirement, highlighting the importance of mental health as a determinant of labour market exit.
The second chapter examines the relationship between income and health in Israel, distinguishing between absolute and relative income effects. Relative income is measured both at the population level and within major religious groups. The analysis applies OLS, Probit models with fixed and random effects, Hausman–Taylor estimation, and semiparametric techniques. The findings indicate a significant positive relationship between absolute income and health, with stronger marginal effects observed for relative income. The results also reveal heterogeneity across religious groups, suggesting that social and group dynamics shape the income–health relationship.
The third chapter explores the causal effect of obesity on employment in the United Kingdom using UK Biobank data. Mendelian Randomisation is applied, using a genetic risk score for body mass index as an instrumental variable for obesity. To account for unobserved heterogeneity, marginal treatment effect and person-centered treatment effect methods are applied. The findings show a significant negative effect of obesity on employment probability, with PeT estimates indicating stronger effects than conventional OLS and broadly consistent with MR results.
