This proposal addresses the interplay between health and labor market behavior in the later part of the working life. Although the significance of health for the retirement transition has generally been recognized, existing investigations of this effect have been hampered by the lack of longitudinal data containing adequate information on health status and on the financial constraints individuals face. As a result, important questions remain regarding the magnitude of the effects of health on labor force behavior and regarding the extent to which health status interacts with personal, economic and job characteristics to affect retirement transitions. The proposed research will use the new Health and Retirement Survey (HRS), a national survey of adults aged 50-62 at baseline, to analyze the effect of health on retirement within the context of a well-specified longitudinal economic model. The research will estimate the effects of health and changes in health on labor force exit between Wave 1(1992/93) and Wave 2 (1994) of the HRS. It will pay particular attention to concerns from earlier research regarding the limitations of self-reported health data. Specifically, it will use latent variable techniques to model health status, explicitly addressing issues of endogeneity and measurement error in self- reported data on work disability, general health, functional limitation, morbidity and other health indicators. In addition, it will explore the theoretical and empirical implications of modeling health in different ways. The proposed research will be among the first to fully utilize the rich HRS health measures in longitudinal models of retirement behavior. The HRS was designed with careful attention to the state of the art in measuring health status in self-reported surveys, and includes more detailed health information than has previously been available in labor force surveys. This research is intended to provide additional insight into the quality and possible uses of these data. The findings should improve understanding of labor market behavior, guide future research on health status and retirement behavior, and inform the design of future social science surveys. In addition, the findings may quantify the possible limitations of using datasets with less detailed health measures.
