Bad health and retirement: A new approach using genetic material

Understanding the role of bad health on retirement decisions presents challenges to researchers due to the inherent endogeneity between health and labor market participation. I exploit the fact that an individual's genotypes are assigned randomly from parent to offspring and are therefore plausibly exogenous to use a genotype as an instrumental variable. I use Apolipoprotein E, a gene that has been associated with variations in cognitive change started at mid-age and has been clinically linked to Alzheimer's disease. Given that genes are randomly assigned and people rarely know their own genotypes, this gene is a valid instrument in that it is correlated with the potentially endogenous variable (cognitive decline) and unrelated to unobserved determinants of the outcome variable (retirement). I consider people's retirement decisions in light of either experiencing cognitive decline or not. The overall equation is: PR(retired)I = α + β₁ HEALTH_i + β₂ WEALTH_i + β₃ Z_i + ε _i With a first-stage equation: HEALTH_i = GenIVi Θ + ν_i I use the Health and Retirement Survey (HRS), which includes key SES and labor market outcomes including wealth, income, and retirement decisions, along with tests of several salient dimensions of cognitive skills. The longitudinal nature of this survey will allow me to track cognitive changes over time. The HRS will soon release genetic data for a subsample of HRS respondents, and I have gained access to this restricted data, which will allow me to jointly consider both phenotypic and genotypic information for a sample of approximately 20,000 respondents.