Robust Integration Modeling of Health, Wealth, and Disability

Rand Corporation, Santa Monica CA

Investigators

Linked publications & trials

Abstract

DESCRIPTION (provided by applicant): Modeling the interaction between health, disability, wealth, family structure, and institutions in one comprehensive framework poses formidable challenges. Existing approaches vary from relatively simple multivariate structures ("reduced form models") that can handle a large number of variables, to highly structured non-linear specifications with considerable detail about the choices faced by individuals and households. Acknowledging that currently no single model can provide a fully satisfactory description of the interrelation between health, disability, wealth, family structure, and institutions, we propose to consider three different models. Each of these models is comprehensive, but has different comparative advantages. First of all, we propose to use RAND's Future Elderly Model as the basis of a comprehensive reduced form simulation model. A second model we consider is a full fledged dynamic programming model that we will calibrate, but not estimate. A third model is in between the reduced form model and the DP model. It simplifies the environment under which individuals and households make decisions, and derives largely analytic solutions for the utility maximization problem faced by individuals and households. Each of the models provides insight in how institutions affect the relation between health, disability, wealth, family structure, and institutions. For each of the models, the measurement of health is a crucial ingredient. The datasets we plan to use (HRS, ELSA, SHARE) contain a rich set of health and functional limitations indicators. We will build a Multiple Indicators Multiple Causes (MIMIC) model of health, which will next be incorporated in the three models mentioned. Finally, although most of the modeling is based on main stream economic theory, we do investigate important deviations from the standard rational model, by exploring the consequences of non-standard assumptions, like hyperbolic discounting. The resulting menu of models allows for an assessment of the robustness of empirical relationships to the choice of specification that one is likely to see in practice. In addition we will formally compare and test the statistical performance of the models, both within and out of sample. Various policy simulations will illustrate the relation between health, disability, wealth, family structure, and how these are impacted by institutions.