Divorce is a crucial determinant of health and well-being, with implications for multiple generations –
children of divorced parents are between two and three times as likely to live in poverty compared to those with
married parents. But a long, happy marriage has become a privilege of the highly educated; couples with a four-
year degree are twice as likely to have a marriage that lasts 20+ years compared to those with a high school (or
less) education. Intimate relationships during the transition to adulthood build foundational relation skills that
young people carry forward in their lives. Thus, understanding which relationships endure and dissolve during
young adulthood is crucial for our understanding of adult divorce and the well-being of children.
We propose analyses that are made possible by newly available intensive longitudinal data on intimate
relationships among 18 – 22 year old women – the Relationship Dynamics and Social Life (RDSL) project. The
RDSL data allows us to build on prior research, while also overcoming some of its limitations, which included
being unable to explore the role of rapidly changing relationship characteristics as predictors of subsequent
dissolution. The data includes weekly time-varying measures of intimate relationship characteristics, which we
will use to predict subsequent dissolution. Further, because of the frequent interviews with a short period of
retrospective reporting (approximately one week), we can include very brief intimate relationships and/or
other types of intimate relationships that young people might not remember (or choose to forget) when they
are asked to recall their experiences over the past several years.
First, we propose to estimate discrete-time logit hazard models that take advantage of up to 165 weekly
interviews per woman, to further our understanding of how affiliation/intimacy, unequal decision-making,
conflict/intimate partner violence, concurrent sexual partners, partner socioeconomic status, shared/unshared
birth history, and churning (break-ups/reconciliations) contribute to the risk of dissolution. We will estimate
models of the time to first break-up for each relationship, and subsequently estimate models for higher-order
break-ups among relationships that reconciled. We will incorporate random effects into our multilevel hazard
models to account for an individual’s underlying propensity to be in a relationship that dissolves quickly, and
to account for the nested structure of the data (the data include approximately three relationships per woman).
Second, we propose to examine race and socioeconomic (SES) differences in these patterns. Using
graphing techniques, tests of formal mediation, and Oaxaca-Blinder decomposition models, we will explore
whether there are race and SES differences in overall rates of relationship dissolution, and if so, why.
This is the ideal dataset (RDSL), team (Barber, Kusunoki, Gatny), and approach (logistic regression
hazard models, decomposition, mediation analyses) to answer these research questions, which have been
unexplored due to data limitations.