This project will expand a structural model of retirement and saving to incorporate a detailed analysis of the role of health and family in shaping these outcomes. Measures of health status will include the incidence and progression of specific diseases and comorbidities, detailed reports on limitations in ADLs and lADLs, demands of the job, and accommodations at work. The disability insurance application and award process will be modeled. In addition we will extend the model of the family to include the effects of own and spouse health and insurance on joint retirement decisions. At the family level, a first step will be to elaborate on the paths through which the health of one spouse influences the labor market decisions of the other, as the demands on time from one spouse are affected by the health of the other, while the needs for additional income and insurance are affected by the health of one or both spouses, the terms of the health insurance coverage for each spouse, and the course of savings. We also will measure the influence of health insurance offered in different forms to each spouse: only at work, as retiree health insurance, or through other sources of support for purchase of health insurance (including Medicaid). Outcomes in the expanded model are labor force activity and retirement dynamics, including direct and reverse flows among the states of full time work, partial retirement and full retirement, as well as wealth. Expanded state variables will reflect expanded measures of health status, disability status and insurance, and family structure. Avenues through which health affects retirement and wealth include the effect of health on the wage offer, the differential effect of health on the difficulty and unpleasantness of work in different occupations (disutility of work), the complex influence of health on work through the disability program, and the effects of health on wealth and thereby the creation of a wealth effect on retirement. A basic problem is that number of computations is proportional to the state space, and the state space grows exponentially with the number of state variables. Computational feasibility limits our ability to fully estimate at one time a single model that would incorporate the detailed description of each feature noted above. It will be necessary to adopt an incremental strategy for estimation that will be feasible, yet will deal comprehensively with the enriched description of health, disability, family and other factors we will analyze.
