Seminar Series

Analyses of the Health and Retirement Study (HRS) between 1992 and 2014 compare the relationship between different levels and forms of debt and heart attack risk trajectories across four cohorts. Although all cohorts experienced growing household debt, including the increase of both secured and unsecured debt, they nevertheless encountered different economic opportunity structures and crises at sensitive times in their life courses, with implications for heart attack risk trajectories. Results from frailty hazards models reveal that unsecured debt is associated with increased risk of heart attack across all cohorts. Higher levels of housing debt, however, predict higher rates of heart attack among only the earlier cohorts. Heart attack risk trajectories for Baby Boomers with high levels of housing debt are lower than those of same-aged peers with no housing debt. Thus, the relationship between debt and heart attack varies by level and form of debt across cohorts but distinguishes Baby Boomer cohorts based on their diverse exposures to volatile housing market conditions over the sensitive household formation period of the life course.

The COVID19 pandemic has exposed and amplified entrenched socioeconomic inequalities and health disparities among underserved communities. Increased testing and upcoming vaccinations are two important strategies to manage COVID19. Mounting evidence indicates that underserved communities are less likely to actively participate in mass testing and upcoming immunization due to inadequate information, logistics and issues surrounding fear, stigma and trust.  As we employ rigorous testing and tracing, and prepare for a likely mass vaccination, it is critical that we understand and seek solutions to historical distrust of medical, public health, and vaccination programs in underserved minority communities. North Carolina Central University has created an Advanced Center for COVID19 Related Disparities (ACCORD) (www.nccu.edu/accord) to address this public health emergency and conduct multidisciplinary research to study the public health and economic impact of COVID19 on underserved communities of NC. ACCORD is supported by the North Carolina Policy Collaboratory at the University of North Carolina at Chapel Hill with funding from the North Carolina Coronavirus Relief Fund established and appropriated by the North Carolina General Assembly.” 

In March 2020, U.S. public health and government officials began recommending physical distancing behaviors to slow the spread of COVID-19. However, a growing line of research underscores that socio-environmental factors that limit the ability to physically distance may contribute to disparities in the impact of COVID-19. For example, overcrowded housing may increase the transmission of infectious diseases despite county- or state-level physical distancing procedures. In this paper, we examine the relationship between the percentage of overcrowded households and COVID-19 deaths across U.S. counties. We find that percentage of overcrowded households is a significant predictor of cumulative COVID-19 deaths and that the relationship between overcrowding and COVID-19 deaths changes overtime. Furthermore, we find evidence that poverty exacerbates the relationship between overcrowding and COVID-19 deaths. Our findings underscore that disadvantaged areas may be more vulnerable to the effects of COVID-19 and that this vulnerability may lead to growing disparities over time.

We estimate racial differences in longevity using records from cohorts of Union Army veterans. Since veterans received pensions based on proof of disability at medical exams, estimates of the causal effect of income on mortality may be biased, as sicker veterans received larger pensions. To circumvent endogeneity bias, we propose an exogenous source of variation in pension income: the judgment of the doctors who certified disability. We find that doctors appeared to discriminate against black veterans. The discrimination we observe is acute-- we would not observe any racial mortality differences had physicians not been racially biased in determining pension awards. The effect of income on health was indeed large enough to close the black-white mortality gap in the period. Our work emphasizes that the large effects of physicians' attitudes on racial differentials in health, which persist today amongst both veterans and the civilian population, were equally prominent in the past.

Life course sociology and criminology provide compelling evidence that the distinction between “criminal” and “non-criminal” is largely a matter of time. Yet crime discourse and policy are still deeply rooted in the notion of criminality as an immutable characteristic. This talk contrasts the fluidity in criminal behavior with the growing stickiness of public labels, drawing from demographic analysis of changes in the population bearing such records and the spillover effects on U.S. politics (including the 2020 election), health care, and labor market institutions. I conclude by considering policy interventions tailored to specific life course stages (e.g., raise-the-age) and institutions (e.g., ban-the-box, compassionate release) amidst the call for more systemic change.

Recent qualitative research argues that Asian Americans’ educational attainments are not predicated on their parent’s education, diverging from status attainment theory. Using data from two nationally representative studies, the analysis reveals extremely high levels of offspring education and no association with parents’ education among Chinese, Indian, Korean, and Vietnamese immigrants. High adolescent educational expectations and parental pressure regardless of parental education partially account for the lack of association. In turn, this education pattern translates into high levels of income for this population. In contrast, Whites, Blacks, Mexican Americans, and later generation Asian Americans’ education and income patterns are generally consistent with status attainment theory. These results demonstrate that educational attainment among certain Asian American populations diverges from classic stratification models and indicate the need for more detailed explorations to further contextualize these patterns. 

The Great Smoky Mountains Study (GSMS) is a longitudinal, population-based community survey of children and adolescents in North Carolina. The study is part of a collaborative effort between Duke University and the North Carolina State Division of Developmental Disabilities, Mental Health and Substance Abuse Services. The collaborative study started in 1992 and continued until 2015. Important goals of the study were to estimate the number of youth with emotional and behavioral disorders; investigate the persistence of those disorders over time; examine the need for, and use of, services for emotional and behavioral disorders; and identify possible risk factors for developing emotional and behavioral disorders. Drawing from 11 counties in western North Carolina, the screening sample consisted of 4,500 children: 1,500 each aged 9, 11, and 13 years at baseline. The study included both urban and rural sectors, and all the agencies that provide child mental health services in the area. This region is also home to a fairly large American Indian population, and 349 of the youth in the study are enrolled members of the Eastern Band of the Cherokee Nation. These youths represent a population that has been under-represented in mental health research across the country. The GSMS has provided policy-relevant information in the areas of: 1) need for mental health services, 2) risks for emotional and behavioral disorders, 3) outcomes of serious emotional disorders, 4) use of mental health services across sectors and 5) effectiveness of mental health services among cohorts.

Individuals do not respond uniformly to treatments, events, or interventions. Social scientists routinely partition samples into subgroups to explore how the effects of treatments vary by covariates like race, gender, and socioeconomic status. In so doing, analysts determine the key subpopulations based on theoretical priors. Data-driven discoveries are also routine, yet the analyses by which social scientists typically go about them are problematic and seldom move us beyond our expectations, and biases, to explore new meaningful subgroups. Emerging machine learning methods allow researchers to explore sources of variation that they may not have previously considered, or envisaged. In this paper, we use causal trees to recursively partition the sample and uncover sources of treatment effect heterogeneity. We use honest estimation, splitting the sample into a training sample to grow the tree and an estimation sample to estimate leaf-specific effects. Assessing a central topic in the social inequality literature, college effects on wages, we compare what we learn from conventional approaches for exploring variation in effects to causal trees. Given our use of observational data, we use leaf-specific matching and sensitivity analyses to address confounding and offer interpretations of effects based on observed and unobserved heterogeneity. We encourage researchers to follow similar practices in their work on variation in effects.

Hispanic children account for more than one third of children living in poverty in the U.S. Yet, eligible low-income Hispanic families appear to have lower rates of participation in most public programs designed to reduce poverty. In 2009, for example, Hispanics made up 24 percent of the EITC-eligible population, but only 15 percent of those receiving the federal or state EITC were Hispanics. In recent estimates from the 2018 American Community Survey, only 2 out of 3 native-born Hispanic child households living at the federal poverty level reported receiving SNAP. This presentation will share findings from two related endeavors to understand uptake among Hispanic families and how the design of public policies might be affecting uptake. First, drawing on data from the 2014 panel of the Survey of Income and Program Participation, estimates of EITC uptake from 2013-15 by race and ethnicity are derived adjusting for detailed characteristics of families with state- and year-fixed effects, and with a variety of new variables collected from primary sources to capture EITC marketing and outreach, availability of tax sites, and generosity of related state programs. Second, descriptive analyses of new collated information on state-level policies regarding eligibility determination, documentation requirements, and practices affecting the user-application experience for TANF will be presented in the 13 states that are host to over 80 percent of low income Hispanic children. This is complemented by new primary data collected in 2020 from TANF state and local administrators, and front line workers, in several of these same states regarding their perspectives about serving Hispanic families.

Expressing the significance of COVID-19 in a relatable metric is important because public awareness is critical to participation, on which mitigating policies depend. Mortality indicators are among the most salient measures of the impact of COVID-19. Following well-established practices in demography, several CoViD-19 mortality indicators can be derived from the cumulative number of CoViD-19 deaths. The first indicator is an occurrence-exposure rate comparable to the Crude Death Rate. Unstandardized, it may not be appropriate for comparisons between populations that have very different age compositions, but it allows for a direct comparison between CoViD-19 and all causes mortality over periods of any length. The second measure is an indirectly standardized rate which appears to perform quite like a directly standardized rate but without requiring a breakdown of CoViD-19 deaths by age and sex. While age-standardized death rates have excellent properties for tracking the pandemic, those are expressed in underwhelming metrics: deaths per 1,000 or fraction thereof. With extant life tables, reductions in 2020 life expectancies can be estimated. Declines in life expectancies are intuitive indicators, but they are unsuitable for fine-grained tracking of a fast-moving epidemic because their estimation requires an assumption of unchanged future mortality. To avoid making any assumption about future mortality, I introduce a Mean Unfulfilled Lifespan (MUL), defined as the average difference between the actual and otherwise expected ages at death in a recent death cohort. For fine-grained tracking of the pandemic across small areas or over short periods of time, MUL values can be quickly approximated. To illustrate I estimate that using a seven-day rolling window, the MUL peaked at 7.32 years in Lombardy, 8.96 years in Madrid, and 8.93 years in New York, but reached 12.86 years for the entire month of April in Guayas (Ecuador).