In this article we develop a finite mixture negative binomial count mo
del that accommodates unobserved heterogeneity in an intuitive and ana
lytically tractable manner. This model, the standard negative binomial
model, and its hurdle extension are estimated for six measures of med
ical care demand by the elderly using a sample from the 1987 National
Medical Expenditure Survey. The finite mixture model is preferred over
all by statistical model selection criteria. Two points of support ade
quately describe the distribution of the unobserved heterogeneity, sug
gesting two latent populations, the 'healthy' and the 'ill' whose fitt
ed distributions differ substantially from each other. (C) 1997 by Joh
n Wiley & Sons.