An econometric analysis of three health economic issues in China
2017-02-14T00:44:56Z (GMT) by
Abstract This thesis investigates three health economic topics relevant to health care policies in China using micro level datasets of nine provinces and a series of econometric models. Studies in this thesis provide comprehensive empirical evidence to quantify the complex relationships among a set of health related variables. We specially study individuals’ health care utilisation, health status, health behaviours, and their determining variables for the Chinese population. A bivariate ordered probit model with sample selection (BVOP), a random effect ordered probit model (RE), and a zero-inflated ordered probit model (ZIOP) are employed to achieve these aims. The thesis consists of three self-contained studies. The focus of the first study in this thesis is on Chinese individuals’ health care utilisation when being ill in both rural and urban regions. Much has been researched on the experience of the medical insurance experiments in China. Fewer studies look at utilisation of medical care services. This work examines jointly the incidence of an individual being sick and the decision of the types of medical care that the individual seeks. We investigate the impacts of socioeconomic, demographic, and life style factors on the probabilities of individuals being sick and seeking various types of medical care, with particular focus on the role of medical insurance status and types of insurance schemes of the individuals. We use individual level data from the China Health and Nutrition Survey (CHNS) that involves randomly selected respondents from nine provinces. We categorise medical services into six levels of cares from seeking no care to high level hospital care in large cities. A bivariate ordered probit model with partial observability is estimated to allow for the correlation of unobservable factors that may affect both the incidence of being sick and the types of care a person seeks. As we only observe the types of medical care for those respondents who have self-reported illness or injury in the past four weeks before the survey, our system equation approach also avoids potential selection bias that may arise from a single equation approach using only the sub-sample of those being sick in the past four weeks. The results from the Full Information Maximum Likelihood (FIML) estimation in the bivariate approach show evidence of selection bias with significant cross equation correlation coefficient, showing clear evidence of non-randomness in the selected sub-sample of those being sick in the past four weeks. The size of potential selection bias is quantified. The results show strong evidence of inequality in being sick and medical care service utilisation associated with socioeconomic factors such as wealth and health insurance types. In the second study, the effects of factors that determine individuals’ self-reported health status and in particular how health inequality is related to income, occupation, and regional differences in China are examined over the period from 1997 to 2006. Self-reported health status is an important indicator of population health status. Inequality of health status which is linked with the growing income gaps is highly concerned by many researchers especially in the context of transition from state-funded universal health care system to market based health care system. This paper examines the factors associated with individuals self-reported health status, using a large scale individuals’ panel dataset from China Health and Nutrition Survey (CHNS), involving nine provinces. We use a random effect ordered probit model to allow for individual heterogeneity. Particular attention is paid to investigating how income, occupation, and geographic difference contribute to health inequality. Factors that contribute to individuals’ self-reported health status and the change of health status are tested. Marginal effects are computed after examining contributory factors to investigate percentages that probabilities for different health status change with a unit change of contributory factors. Individual unobservable heterogeneity is accounted for with an individual random effect and examined significant. Inequalities of self-reported health status are highly concerned in this paper. The results show that income, education level, smoking tobacco, consuming alcohol, and living locations have significant effects on individuals’ self-reported health status, which means that income, education, and geographic inequalities are significantly associated with health status inequalities. The third paper investigates various factors that affect individuals’ choices to smoke and how much to smoke per day in China using micro level data. This paper uses a sample representative of the general population of Chinese adults from nine provinces, with a bigger geographic spread that can give us a better observation on individuals’ choices. Both individuals’ decisions on smoking participation and the level of cigarette consumption are jointly studied in a system framework in this paper. Specially, a zero-inflated ordered probit model (ZIOP) with a probit equation to model smoking participation and an ordered probit equation to study cigarette consumption measured by smoking frequency is employed in this paper. The probit model and the ordered probit model are estimated jointly, with the error terms of the two latent equations being correlated. The unique characteristic of ZIOPC model is that the observed zero consumption is allowed to come from two different sources: non-participation and zero-consumption conditional on participation. The two decisions of participation and level of consumption are driven by different factors, or same factors with opposite effects. This study identifies a number of factors that have a significant effect on individuals’ smoking behaviour. In particular, it identifies several factors, including income, that have opposite effects on the two separate decisions in smoking. Using a less flexible model OP with a single latent equation only provides the combined effects and does not allow identification of two separate effects. Insights into how individuals have potentially made the two decisions differently can help more targeted education programs for controlling cigarette consumption. The compared results show that some factors are very important determinants of participation in smoking and how much to smoke per day, such as income. Since the ZIOPC model takes into account two potential smoking related decisions underlying the observed smoking sample, we have a better insight into how income affect individuals’ decisions in participating in smoking and the amount smoked. The ZIOPC model shows that increasing income may increase the probability of people participating in smoking, but decrease the likelihood of people smoking more once they start to smoke. It is notable that there are difficulties in identifying the true impact of price on cigarette consumption because the large number of tobacco products, which raise difficulties to define a representative price for cigarette. We employ local market price and the prices of several popular cigarette brands in this paper to increase the representativeness of price indices. Results show that prices of branded cigarettes have stronger effects in reducing the level of cigarette consumption. Living location and lifestyle can affect participation in smoking and the amount smoked as well.