Three essays on count data estimation methods with applications to health economics

The purpose of this dissertation is to propose and estimate count data microeconometric models that take into account special features of health care data. Three problems in health economics are investigated. The first essay analyzes the effect of managed care insurance plans on the demand for outpatient physician visits. It provides a Bayesian method to empirically separate selection effects due to individual choice of coverage and incentive effects known as moral hazard. The proposed Endogenous Hurdle Poisson log-Normal model addresses two important econometric issues: the large proportion of zero outpatient visits and the endogeneity of managed care insurance status to utilization of outpatient services. The data are obtained from the Medical Expenditure Panel Survey, (MEPS), and the sample consists of privately insured individuals, aged 21-64, all of whom are employed but not self-employed.The analysis indicates both favorable selection bias and important moral hazard effects in the decision whether to utilize an outpatient visit or not. The second essay investigates the impact of health insurance coverage on utilization of doctor visits. A Bayesian panel count data model with correlated random effects and endogenous treatment of the insurance variable is applied to a subset of the Health and Retirement Study (HRS), a national longitudinal survey of individuals over 50 years old and their spouses. The age limit for Medicare eligibility at age 65 serves as an exclusion restriction. After controlling for selection effects insured individuals present higher utilization of doctor visits. The third essay analyzes the demand for cigarettes using data from the 1994-1996 Continuing Survey of Food Intakes by Individuals (CSFII).Since individuals tend to round their consumption of cigarettes in packs smoked per day the dependent variable - number of cigarettes - exhibits pile-ups of counts. The mixed binary-ordered probit approach accommodates this feature of the data and models the starting smoking, the quitting smoking and the how much to smoke decisions. The analysis is performed separately for men and women and provides strong evidence of gender differences in cigarette consumption.