Demographers have long recognized that fertility patterns observed in populations are driven by both behavioral and biological determinants. Yet, most studies of contemporary fertility focus solely on the behavioral dimension, treating biology as a residual determinant, if mentioned at all. In this dissertation, I address this shortcoming by drawing from a biosocial framework that allows for the integration of biological measures into explanatory models of socio-demographic outcomes. In particular, I consider the interplay between reproductive biology and behavior in three empirical chapters on childlessness and subfecundity in the United States. In doing so, I demonstrate that biology can create a frame in which individuals make decisions across their reproductive life course.

In the first empirical chapter, I focus on the process of remaining childless as a case study for examining how women's reproductive choices and behaviors take place against the backdrop of changing social structures, competing preferences, and age-related declines in fecundity. Specifically, I focus on an understudied, yet revealing dimension of why individuals remain childless-- stated fertility expectations over the life course. Using data from the National Longitudinal Survey of Youth-1979 cohort, I use sequence analysis and logistic regression models to identify and describe groups of permanently childless women who follow similar trajectories of stated fertility expectations. Results indicate that three dimensions of fertility expectation patterns-- the emergence of a childless expectation, the consistency of childless expectations, and the types of patterns that precede a childless expectation-- provide a more nuanced description of the composition of permanently childless women than the standard voluntary/involuntary framework. The results, moreover, reflect how women reevaluate expectations for the future when they encounter critical junctures in their life courses.

The second empirical chapter considers how women internalize personal subjective risks about their fecundity, or their biological capacity to reproduce. First, I propose a conceptual model of perceived fecundity that emphasizes the evaluation of perceptions as a process drawn from understudied contextual and individual-level influences. Second, I use data from the National Longitudinal Survey of Youth (NLSY) 1997 cohort to provide new, nationally representative estimates of perceived fecundity among young adult women aged 25-30. Finally, I draw from the proposed conceptual framework to examine predictors of and variation in perceived fecundity. Results show striking variation across a range of demographic, contextual, situational, and experiential factors and highlight the need for improved measures of fecundity perceptions and a deeper understanding of how women evaluate their reproductive potential.

The third empirical chapter focuses on an understudied process by which reproductive biology can influence behavior. Prior research suggests that one reason women do not use contraception is because they perceive themselves to be infertile or have difficulty becoming pregnant (i.e. subfecund) and thus, at low risk of pregnancy. These beliefs, however, may provide a false sense of protection from unintended pregnancy if they are not medically accurate. Using data from the National Longitudinal Survey of Youth-1997 cohort, a large, nationally representative survey of young adults, I demonstrate that women with low perceived fecundity have higher odds of non-use of contraception. Further, these results persist after controlling for either a medical diagnosis of infertility or absence of pregnancy following at least 6 months of unprotected sex, suggesting that fecundity risk perceptions often operate independently of experienced subfecundity.

This is a collection of three studies that, together, seek to characterize the strong structural factors that have shaped black and white American family life over the 20th Century. The first of these studies highlights the two main challenges to measuring long-term trends in family composition in the United States using existing data sources: the lack of information from earlier historical periods—what I call the temporal boundary problem—and the inability of official Census data to identify family members who do not co-reside—what I call the household boundary problem. In order to overcome these challenges, I motivate a microsimulation-based demographic estimation strategy that allows for detailed measurement of family composition, given observed rates of fertility and mortality.

The second study in this series uses this demographic estimation strategy to examine the distribution of kin over the course of life for select historical birth cohorts of black and white Americans over the 20th Century. In so doing, I characterize the distinct family life trajectories that have separated black and white Americans for most of our recent history: black Americans, compared to their white peers, are likely to have had fewer available kinship resources at most ages, leading to earlier orphanhood, lower lifetime probabilities of transitioning into most kinship relations, and shorter durations spent in them once there. Though these differences have diminished over time as the mortality and fertility rates of black and white Americans have converged, they have not disappeared entirely: I estimate that black minors born in the period 2000-2010 will experience 58% more deaths in the family by the time they reach adulthood compared to their white peers.

The final study examines the risk and prevalence of imprisonment within the full family networks of black and white Americans over the course of the “prison boom” (1985-1995). In brief, I estimate that the average black American born at the height of the prison boom experienced the imprisonment of a relative for the first time at age 7 and by age 65 will belong to a family in which more than 1 in 7 working-age relatives have ever been imprisoned. By contrast, the average white American who experiences the imprisonment of a relative will not do so until age 39 and by age 65 will belong to a family in which 1 in 20 working-age relatives have ever been imprisoned. Further simulation analyses suggest that future interventions to lower the national imprisonment rate have the potential to meaningfully reduce this race gap in family imprisonment for future generations of Americans.

In combination, these three studies represent the first extensive use of a demographic estimation strategy to characterize the consequences of the three largest structural drivers of family composition and dynamics among black and white Americans over the past century: changing patterns of fertility, mortality, and mass imprisonment. While black-white differences in family composition due to differences in fertility and mortality have been decreasing over time, the recent movement toward mass imprisonment has created a new source of divergence that is likely to be felt by affected families for generations to come. While certainly not as grave as death, the level of social stigma and exclusion that a prison record brings with it can become a kind of “social death” if affected family members are removed from the pool of potential family support. From this perspective, the race gap in kin availability has not changed much over the past century: black families still suffer much greater loss of members than white families. What has changed is the source of that loss. Those family members that death used to remove physically are now removed socially by the system of mass imprisonment.

We are far from a complete understanding of the social determinants of mortality in theUnited States. Despite the longstanding interest in racial and class-based inequalities in health and mortality in the United States, research is often hampered by data limitations. However, new advances in data linkage have allowed mortality researchers to construct ad- ministrative datasets with millions of mortality records and demographic covariates. The unprecedented scale and richness of these administrative datasets allow social scientists to make new discoveries into the contours of mortality disparities in the United States.

This dissertation is comprised of three studies of mortality using large-scale, linked U.S.Census and administrative death records. In my second chapter, I investigate the relationship between owning a home in early adulthood and life expectancy, demonstrating that owning a home in early adulthood has a causal effect on life expectancy. My second study assesses the predictability of individual-level longevity, demonstrating the challenges of predicting individual-level mortality. The final study of the dissertation examines the Black-White crossover, finding evidence that the crossover is not a data artifact and cannot be uncrossed using sociodemographic variables alone.

The development of mortality models is important in order to reconstruct historical processes, understand current patterns and predict future trends. Mortality models are particularly useful when the available data are sparse, unreliable or incomplete. Traditionally, mortality patterns in data-rich populations were used to observe mathematical or empirical regularities, which could be applied to data-sparse populations. However, as a wider variety of data have become available, the focus of model building has shifted to developing flexible models that perform well in a variety of contexts.

This dissertation introduces Bayesian methods of mortality estimation in three contexts where the available data are imperfect. The first paper develops a method to estimate subnational mortality in situations with small populations and highly-variable data. The second paper develops a unified modeling framework to estimate and project neonatal mortality in all countries worldwide, including those with limited and poor-quality data. The third paper introduces a new dataset to study mortality inequalities in the United States, and develops methods to deal with the truncated and censored mortality information that is available. In all three contexts, the modeling approaches combine strengths from traditional demographic models, which capture mortality regularities across age, with the flexibility of Bayesian frameworks, which allow for multiple data sources to be incorporated, information to be shared across time and space, and uncertainty to be assessed.

This dissertation embodies a series of four interrelated chapters, collectively aiming to uncover the role of state health policy (SHP) in population mortality outcomes across the US states and counties over the course of nearly 3 decades, from 1990 to 2018. Chapter 1 introduces necessary background literature on determinants of mortality, its causes and disparities across various sub-populations by socio-economic status, age, sex, place of residence, generational belonging, and health. Based on this evidence, in chapter 2 I propose a conceptual model with a particular focus on isolating assessed causal links between the key indicators, such as population age, health status, and state government ideology. I argue that these three concepts represent a minimal adjustment set to isolate the causal link between SHP and mortality. Additional variables broadly defining area-level socio-economic deprivation, income inequality, residential segregation are considered part of the conceptual model's expanded set that, in principle, ought to augment the precision of the target estimate in a quantitative analysis.

Chapter 3 follows with a detailed description of the method to generate health positivity index scores as well as an alternative, the rate of SHP innovation measure, from a selection of 110 groups of health-related policies, passed into law by states since 1970 through 2018. The main I score index of SHP health positivity takes advantage of the policies' cross-sectional and temporal cumulativity, its thematic relativity with respect to the global maximum in any year, and a double-weighting design to stem over-representation bias. I consider raw and transformed variations of the I score, its decade-lagged variant, and the indicator of the relative rate of new health policy passage. Preliminary testing using one- and two-way repeated measures ANOVA analyses demonstrates medium-to-large effect sizes for the associations between mortality and I score measures at the state level, with notable attenuation in the county group sample, whereas no such association exists for the alternative rate-of-health-policy-innovation indicator. Temporal dependence of the SHP-mortality association is established.

The aim of chapter 4 is to use spatial econometric modeling techniques to arrive at the best possible estimate of the state health policy effect on mortality, as measured by age-standardized death rate, and life expectancy estimates at birth and at age 65. Using mortality data from the United States Mortality Database (USMDB) chapter 4 thus builds upon the previous chapters as follows. First, a set of quantitative indicators representing key concepts presented in chapters 1 and 2 are introduced from secondary sources or are estimated directly on the basis of publicly available secondary data from the US Census Bureau and National Center for Health Statistics. Next, I provide an overview to popular spatial econometric methodology approaches using spatial panel fixed effects designs and submit these variable sets to modeling 3 sex-specific mortality outcomes in various specification combinations, each of which includes one of the measures of state health policy derived in chapter 3. I use evidence from the model diagnostics to arrive at the most optimal model type for each combination of outcome-SHP predictor. My findings indicate that state health policy represented by the two I score measures is the most effective in predicting state-level outcomes. It is estimated that conditional on the socio-economic, political, health and demographic factors SHP could explain up to 30\% of the longevity gap at birth and at age 65 for states on the extreme ends of the longevity spectrum. On the other hand, county group mortality is susceptible to the effects of the parent state health policy only to a limited degree, and is associated more with local socio-economic context. Approximately two-thirds of the SHP-mortality effect is direct and one-third is a by-product of spatial spillovers from the neighboring geography, whereas the reverse is true for the county groups.

The findings of this analysis produce several important contributions. First, the analysis highlights the utility of studying summary indicators of health policy at sub-national aggregate levels as these encompass valuable information that characterizes heterogeneous risk factors that impact the aggregate. Second, the effect of the policy action is inherently spatial, but depends greatly on the scale of aggregation. Longevity in smaller population sub-groups is more sensitive to its local and regional socio-economic and physical environments, with state policy effects percolating more slowly and in a reduced form. Lastly, I find that generalized synthetic measures of state health policy, as applied in the final part of the analysis, are best used in conjunction with summary measures of mortality at coarser geographic resolutions, such as states and regions. Future analyses involving county-level or finer resolution data would benefit from an application of thematic summary components of such measures.