ESSAYS IN AGING, MARITAL STABILITY, AND MENTAL HEALTH

This dissertation presents three chapters about understudied characteristics of the
older population. As the United States and other developed countries’ populations age,
more dedicated research is needed to understand and implement policies to improve the
welfare of this demographic group. Though there is a vast literature on various life-cycle
outcomes of the elderly, gaps remain. Two such aspects have been examined here: marital
stability and mental health.
Chapter 1 investigates how changes in household wealth affect the likelihood of divorce
among older adults aged 50 and above in the United States. Using panel data from the
Health and Retirement Study (HRS) from 1992 to 2018, I first establish a descriptive
non-linear relationship between wealth, and divorce probabilities. The likelihood is higher
for lower levels of wealth and monotonically declines, but only upto the median - beyond
that it remains fairly flat. Extreme changes in wealth, both positive and negative, are
associated with higher probability of divorce, as compared to moderate changes. To
estimate the causal effects, I use two separate plausibly exogenous shocks to wealth, one
from the stock market and the other from the housing market. A $10,000 predicted
increase in real wealth driven by the S&P 500 contemporaneously raises divorce chances
by 262 percent, marginally significant with a p-value of 0.06. A similar sized lagged
housing market driven wealth increase also boosts divorce likelihood by 300 percent,
significant at 1 percent.
Attention Deficit and Hyperactivity Disorder (ADHD) is a common mental health
condition, usually diagnosed during childhood and often found to be persistent into adulthood as well. Motivated by a lack of reliable, large-scale measure of observable
ADHD status among older adults, in chapter 2 my co-authors and I develop a method
to estimate the likelihood of ADHD among older individuals in the HRS. A small
subsample in the HRS is asked diagnostic questions related to ADHD. We use a series
of machine learning models to predict ADHD status in this subsample as a function of
observables available for tens of thousands more HRS respondents. Our main results
indicate that an 11 percentage point increase in ADHD likelihood (average difference
between those meeting v.s. failing to meet diagnostic criteria) is associated with a 6.7
percentage point lower probability of working for pay, an 11 percent reduction in
earnings conditional on working, an 11 percent reduction in household wealth at
retirement and 0.1 percentage point increase in the risk of first marriage ending in
divorce. Linking the HRS data with the Social Security Administration (SSA) records,
we show substantial differences in the earnings trajectories of low v.s. high ADHD-risk
adults. These differences emerge in the 30s and grow over the life-cycle. The present
discounted value of average earnings differences between these groups over the ages
22-65 is substantial (approximately $180,000).
In chapter 3, I examine whether polygenic risk for ADHD predicts divorce,
leveraging genetic data on approximately 9,600 Americans from the HRS. I track
individuals from ages 20-50 and estimate the association between a polygenic score for
ADHD and marital dissolution risk over this period. A higher genetic risk is associates
with a higher divorce probability. The genetic influence persists adjusting for
demographics and exhibits heterogeneity by birth cohort. Interactions with specific
forms of environment suggest little gene-environment interplay. By linking ADHD
genetic propensity to dissolution, this analysis demonstrates utility of polygenic scores
for elucidating social outcomes related to neuropsychiatric conditions. Incorporating
genetic data on older adults provides unique evidence regarding ADHD’s potential
lifelong impacts on family relationships. Findings emphasize the value of genetic tools
to inform social topics where establishing causality is challenging.