Entropy-Based Measures of Multidimensional Health Inequality

We propose a new measure of multidimensional health inequality based on a well-known class of entropy-based inequality measures.  We aggregate a set of health attributes into a measure of overall health by minimizing the relative entropic distance between the multivariate distribution of attributes and the distribution of the summary measure. We calculate both general inequality and income-based inequality using generalized entropy, which has many desirable properties in the context of multidimensional inequality measurement.  This methodology offers many advantages in comparison to the Gini and condition index family of inequality measures, including accounting for the multidimensional nature of health, allowing flexibility in the degree of complementarity between health attributes, utilizing the information from the entire distribution of health and income rather than income or health ranks, and making normative assumptions transparent and intuitive.  We demonstrate the contribution of our inequality measure using data from the Health and Retirement Surveys which contains multiple measures of clinical health, self-reported physical and mental health, and measures of income and wealth.