Measuring the distribution of self-reported health

Many questions in health policy require an understanding of the distribution of health across a given population and how it changes as a result of policy interventions. Since objective data on individual health status are often unavailable or incomplete, especially for populations with very low mortality, increasing use has been made of Self-Reported Health Status (SRHS) data, which record people's own perceptions of their health status. The qualitative or categorical nature of SRHS data prevents the straightforward use of traditional tools of distributional analysis such as the Lorenz curve and the mean. This dissertation explores a number of alternatives for measuring the distribution of self-reported health.;Chapter II examines a prominent method introduced in Wagstaff and van Doorslaer (1994) that employs numerically scaled SRHS data to measure socioeconomic inequalities in health. The chapter demonstrates several potential limitations of this method, especially when applied in an international context.;Chapter III presents a new methodology for evaluating qualitative distributions that generates results that do not depend on the numeric scaling assigned to the categories. A partial inequality ordering is defined that indicates when a distribution is more "spread out" than another; a second partial ordering (first order stochastic dominance) is used to indicate when the overall health level rises. It has been argued that reasonable scalings of SRHS data are concave, emphasizing distances between lesser health categories. Chapter IV introduces partial distributional orderings (using second order dominance) that apply to all reasonable scalings, of the data, so that results do not depend on which concave scale is used.;The methods introduced in chapters III and IV are illustrated using SRHS data from the 1990--1994 National Health Interview Survey State Data Files. Results indicate that southern states generally have less overall health, higher health poverty rates, and a more unequal distribution of health than states in other regions.;The dissertation concludes in chapter V with a brief empirical analysis that identifies key determinants of interstate differences in health and demonstrates how these results can be used to evaluate the mechanism for distributing federal aid to states via the Medicaid program.