Content area
Abstract
It is frequently mentioned in the retirement literature that survey respondents may misreport their health status at retirement because retiring due to poor health is more socially acceptable than other reasons. This dissertation investigates health status measurement in retirement studies. The quality of survey measures of health status is studied with a focus on whether individuals misrepresent health status to conceal actual retirement motivations. What difference the choice of a health measure makes in retirement models is investigated. Finally, how to take health status into account when a data set contains poor or nonexistent health information is examined.
Misreporting health status to conceal actual retirement motivations is studied by looking at age distributions of self-reported measures of health status in the Longitudinal Retirement History Study. Because retirements are most prevalent at ages 62 and 65, it seems likely that the sample would reflect unusually high levels of poor health at these ages if people are falsely claiming to be in poor health at retirement. The age distributions, however, do not provide any support for this hypothesis.
Retirement behavior for 58 to 70 year old white males who are not self-employed or farming is studied in a discrete-time hazard framework. The hazard model estimates of a strong health effect are not sensitive to the choice of the health measure. Nonhealth coefficients are not sensitive to either the choice or the existence of a health measure. An instrumental variables technique involving a mixing distribution that generates consistent estimates in maximum likelihood estimations is developed and implemented. This model does not generate reasonable health effect estimates.
Athough there are problems with retirement survey measures of health status, these measures are useful. The strong influence of poor health on retirement behavior found elsewhere is supported and instrumentation of health status does not work well. Therefore, ignoring some health measures available in retirement surveys results in a loss of valuable information.
