We propose to examine the relationship between education and cognitive decline and mortality in old age using longitudinal data from the Health and Retirement Survey (HRS). Although the correlation between education and cognitive decline is well-documented, it is not known if this relationship is causal or due to confounding by individual characteristics such as cognitive ability or family economic status. Furthermore, if we assume this relationship is causal, the key mediators are unknown. We emphasize the importance of innovative analytic approaches in estimating the overall causal effect and in testing specific pathways. Our primary aims are: 1) to derive consistent estimates of the causal effect of education on cognitive change and mortality using instrumental variables analysis; 2) to test whether the effect of education on cognitive outcomes operates through a latency, social trajectory, or cumulative harm model; and 3) to provide unbiased estimates of the importance of three hypothesized pathways linking education and cognitive decline, using new analytic approaches emphasizing the time-varying nature of both mediators and confounders. These hypothesized pathways are: social ties, health behaviors, and cardiovascular health. Rapid changes in state educational policies influenced the educational attainment of birth cohorts in the first half of the 20 th century. Using such natural experiments, instrumental variables analyses can provide estimates of health effects of education, even when important confounders are unmeasured. State policy data on instruments such as compulsory schooling and term length, will be linked to individual level data from the Health and Retirement Survey (HRS). As of the 2002 interview wave, the combined HRS sample included 1 to 5 waves of memory and mental status assessments on over 25,000 participants.We will use innovative methods to test specific pathways while accounting for time-varying covariates. HRS contains detailed information on lifecourse socioeconomic position, including 40 years of Social Security earnings information on over 9,000 original sample members, and physical and behavioral characteristics assessed at up to 6 interview waves. We distinguish between alternative models of how adult socioeconomic trajectories may mediate the effect of education on cognitive decline. Finally, we use gestimation to examine the importance of social ties, health behaviors, and cardiovascular risk in mediating the effect of education on cognitive outcomes.
Showing the most recent 10 out of 15 publications
