A Stochastic Frontier Model for Discrete Ordinal Outcomes: A Health Production

Xueyan Zhao,William E. Griffiths

2010-01-01

SSRN Electronic Journal

Abstract:Since its introduction in Aigner, et al (1977) and Meeusen (1977), the stochastic frontier model has been widely used in the analysis of productivity and firm efficiency, leading to the development of an enormous literature on its theory and application. Included in the literature are recent applications to the analysis of health care production and hospital efficiency. Another area of interest for health economists is the production of individuals’ health at the household level. While much of the literature on health production has focused on health production at a macro level using country data, there has been many empirical studies examined health production at an individual level using survey data. A commonly used measure for an individual’s health status is the self-assessed overall health grade obtained from survey questions with ordered multiple discrete choices. Although the self-assessed grade is subject to measurement errors as all other self-reported variables, empirical evidence suggests that it is a reliable measure of overall health status. Econometric models specifically designed for ordered discrete dependent variables, such as the ordered probit model, are more suitable for analysing health production functions of this type. In this paper, we extend the traditional stochastic frontier model for continuous dependent variables to accommodate output measured as a discrete ordinal outcome variable. Conditional on the inefficiency error, the assumptions of the ordinal ordered probit model are adopted for the logarithm of output when it is observed only as an ordered categorical variable. Bayesian estimation utilizing a Gibbs sampler with data augmentation is applied to a convenient re-parameterisation of the model. The model and estimation techniques are applied to a health production function using panel data from an Australian longitudinal survey. Demographic and socioeconomic characteristics are specified as inputs to health production, whereas production efficiency is made dependent on lifestyle factors, such as exercise level, alcohol consumption, smoking habits, and social networks. Posterior summary statistics are obtained for selected health status probabilities and efficiencies, and for the marginal effects of individual characteristics and lifestyle factors on health status probabilities and efficiencies.