This dissertation examines the subtleties inherent to socioeconomic models of health. The relative contributions of unobserved heterogeneity and state dependence in explaining the dynamics of health in the U.S. are controlled for as the average partial effects of socioeconomic status are estimated. In addition to these investigations, I use anchoring vignette data from the Health and Retirement Study to investigate health perceptions in the U.S. and how they affect the comparability of self assessments of health across the population.

The results imply that just over 30% of the unexplained variation in self-assessed health is attributable to unobserved heterogeneity in the form of a time invariant individual effect. Although attrition bias is detected in the panel, estimates obtained using an inverse probability weighted specification reveal that attrition has virtually no effect on the estimated relationships between health and socioeconomic status.

Regarding the vignette analyses, many of the characteristics that affect health status are also found to affect perceptions of health. For example, I find that respondents in lower income groups tend to be more tolerant of mental health problems than the rest of the population and that recipients of federal disability insurance may be more likely to rate a given health problem as being disabling. Upon identifying which personal characteristics influence reporting behavior, I estimate models of self-reported health that enhance comparability of survey responses across the population. I find that failing to account for reporting heterogeneity ultimately leads to a distortion of the estimated relationships between health and socioeconomic status.