Relationships among socio-demographic characteristics, general assessments of health, and old-age mortality have been well established in developed countries. There is also increasing focus on the connection between early-life experiences and late-life health. This study tests these and other associations using representative survey data from Taiwan. Survey data on the 60 and over population (N=4049) in 1989, 1993, and 1996 are linked to 1989-1999 data from a national death registry. Survival is modeled using Gompertz regression and both fixed and time-varying covariates. Possible effects of Taiwan's Universal Health Insurance Program are tested by coding observations as pre- or post-insurance. Some results replicate findings from developed countries; others are new. Contemporaneous education effects on survival are attenuated after introducing some indicators of health from a previous period. Functional and global assessments of health have stronger associations with survival than reports of other health-related characteristics. Mainlanders have higher survival than others. Increases in survival over the 1990s were concentrated among those with the greatest number and severity of functional limitations. This pattern may be an indication of better access to health care or changes in longer-run, but unmeasured, determinants of old-age mortality.
