It is standard practice in the literature to assume that an individual’s marginal utility of consumption is independent of health status. If this assumption is not met, it could have important implications for welfare economic analyses. The aim of this paper is to provide new, more comprehensive empirical evidence of state dependence following the approach used by Finkelstein et al. (2013). We use a rich combination of longitudinal survey data and administrative register data. Survey data were obtained from the Survey of Health, Ageing and Retirement in Europe (SHARE) and combined with data from the comprehensive Danish registers, including individual income data and data on health care utilisation based on the universal ICD-10 classification system. Utility is measured in terms of subjective well-being attained from the SHARE survey. To further increase the power of our sample, we used a state-of-the-art prediction algorithm to enrich the register data with information on subjective well-being. With this approach, this paper also makes a general methodological contribution on the use of prediction models for data enrichment and imputation. We define a reduced form equation in which individual subjective well-being is regressed according to income and health with an interaction effect capturing state dependence. Our results show evidence of negative state dependence. From our baseline regression, the estimated magnitude suggests that marginal utility of consumption decreases by 14.6% when an individual becomes sick. The results are robust to different levels of risk aversion and to assumptions regarding the mapping of the latent utility onto observed utility.
|Navn||DaCHE Discussion Papers|