The demand for health care in lower-income nations: Theory, econometrics and empirical evidence from Indonesia.

摘要

This dissertation advances this literature on health care demand in lower income nations on several fronts. The first essay focuses on multinomial provider choice. It begins with a theoretical model where consumers make discrete choices among health care facility alternatives in a dynamic setting. Some intuitive predictions obtain only through numerous parameter restrictions. A flexible semi-nonparametric alternative to the standard logit choice model is introduced to control for any prediction bias emerging from stochastic between indirect utility functions. It involves a error-components structure for the stochastic portion of indirect utility that relaxes the standard independence assumption in a very general way while avoiding many of the pitfalls associated with other models: it does not require further distributional assumptions or unjustifiable exclusion restrictions. Elasticities with respect to key provider characteristics are calculated for alternative samples and specifications. Using insights from recent work, the difference in elasticities for these subgroups is calculated in the technically appropriate but rarely adopted manner.; The second essay compares multinomial models with others that focus on total demand (represented by the number of provider visits) since, in many respects, that is the concept of underlying interest. Analytical expressions for multinomial and total demand elasticities are derived and compared. It is shown that the two demand concepts should yield different elasticity estimates, principally because they weight successive visit count probabilities differently. A dynamic theoretical model of discrete-choice health care demand is developed as a vehicle for considering particular behavioral mechanisms by which the two demand concepts might be expected to yield different elasticity estimates. Innovative empirical models are employed to compare elasticity estimates generated by empirical models of multinomial and total demand for public health clinics. These include a new, extremely flexible estimator for the conditional mean of visit counts, or the expected number of visits. Multinomial models are found to provide reassuringly useful insights into the distribution of total demand for this sample, though the correspondence is less strong for the poorest Indonesians.