| The first chapter provides solutions in a Bayesian framework for the problems that arise in estimating health care utilization. In particular, I am interested in determining how consumption of medical care especially long term care changes with changes in the informal care they receive from children. A two-part model for count data is developed with informal care explicitly modeled as an endogenous regressor. The model is fitted using Bayesian Markov Chain Monte Carlo methods. The results indicate that informal care substitutes for nursing homes and hospital inpatient stay, but does not affect home health care significantly. Subsidies that can induce part-time caregivers to become full-time for the existing care receivers are the most effective way to reduce government spending.;The third chapter focuses on estimation methods in a nonlinear model setting with endogenous regressors. In particular, a model with categorical endogenous regressors and binary outcome is examined in a Monte Carlo simulation exercise using two-stage residual substitution (2SRI), linear IV method and probit. In this model setting, 2SRI is shown to yield consistent results, but not those of the linear IV and probit. 2SRI method is further applied to an example to estimate the causal effect of choices of health insurance plans on utilization of ambulatory and hospital inpatient services. We found a larger role of PPOs in increasing the probability of using ambulatory services and reducing the probability of using inpatient services.;The second chapter explores and compares different estimators for a panel dataset. During the 1990s, Chinese state-owned enterprises (SOEs) and collective-enterprises consistently decreased coverage of pubic health insurance to their employees. This chapter investigates the changing pattern of health insurance coverage in China using the China Nutrition and Health Survey. The fixed effects linear model clustering at the province level is used, and results are compared to alternative models: pooled OLS, random effects GLS and fixed effects logit model. Empirical evidence is found that unemployment as a side effect of the Open Door Policy, and the deregulation of SOE and collective-enterprises were the main causes for the decreasing trend. |
