A DYNAMIC STOCHASTIC MODEL OF INVESTMENT IN HEALTH

拟议的项目对个人在投资其健康方面的动态行为进行建模，以便量化各种行为对健康结果的影响，从而评估替代公共卫生政策对个人健康的影响。健康被认为既是一种消耗性商品，也是一种人力资本，可以因个人的行为而产生或退化。在该模型中，个人对医疗保险、吸烟、锻炼、饮酒和医疗保健用途做出选择。这项研究将使用健康和退休研究(HRS)的数据来解释收入、人口统计、医疗保险和个人行为差异对健康结果的影响。相反，将计算健康结果、医疗保险、收入和人口结构对个人行为影响的估计。特别是，将获得性别、种族、收入、教育、年龄和健康状况对在医疗保险、吸烟、锻炼、饮酒和医疗保健使用方面做出选择的概率的重要性的估计。估计方法将需要解决离散随机动态规划问题。利用该问题解的最优决策规则，在对模型中的状态变量和随机不可观测变量进行分布假设后，构造一个非线性似然函数。因此，模型的结构参数将通过非线性似然函数的全信息最大化来获得。对(个人)健康投资的动态随机模型的结构性估计将使各种反事实的公共卫生政策的比较成为可能，包括涉及医疗保险覆盖范围变化和医疗费用补贴的政策。具体地说，该模型的结构估计将用于生成(假设的)人群健康结果概况，这些结果将导致(假设的)医疗保险覆盖范围、医疗保健成本和关于吸烟、锻炼和饮酒的行为选择的变化。

The proposed project models the dynamic behavior of an individual in investing in his/her health in order to quantify the effects of the various behaviors on health outcomes and to thereby evaluate the impacts of alternative public health policies on individual health. Health is considered to be both a consumable commodity as well as human capital that can be produced or degraded by an individual's actions. In the model, individuals make choices about health insurance, smoking, exercise, alcohol use and medical care use. The research will use data from the Health and Retirement Study (HRS) to explain the effects of differences in income, demographics, health insurance, and individual behaviors on health outcomes. Conversely, estimates of the effects of health outcomes, health insurance, income and demographics on the individual behaviors will be computed. In particular the estimates of the importance of gender, race, income, education, age and health status on the probabilities of making choices with regard to health insurance, smoking, exercise, alcohol use and medical care use will be obtained. The estimation methodology will require solving a discrete stochastic dynamic programming problem. Using the optimal decision rules from the solution to this problem a non-linear likelihood function will be constructed after making distributional assumptions about the state variables and the random unobservables in the model. The structural parameters of the model will consequently be obtained by the full information maximization of the non-linear likelihood function. The structural estimation of a dynamic stochastic model of (individual) investment in health will make feasible the comparison of various counterfactual public health policies, including those that involve changes in health insurance coverage and subsidies for medical care costs. Specifically, the structural estimates of the model will be used to generate (hypothetical) population health outcome profiles that will result as a consequence of (hypothetical) changes in health insurance coverage, medical care costs and the behavioral choices about smoking, exercise and alcohol use.