This dissertation estimates voting behavior by race and ethnicity in Georgia’s 2020 presidential election, one of two long-time Republican states ‘flipped’ by Democratic candidate Joe Biden. The purpose is to evaluate media claims that demographic change and minority voter mobilization clinched the victory. Developing empirical evidence elevates public discourse by challenging the public misconception that the political dividends of race/ethnic diversity flow uniquely to the Democratic party. Parties, in turn, can be more responsive to their constituencies. State-level research can guide policies on election security, redistricting, and minority rights. Two methodological approaches are implemented that have not previously been directly compared. The first is the aggregation technique of ecological inference, which is the current standard used for Voting Rights Act litigation concerning racially polarized voting. The second is the disaggregation technique of multilevel regression and poststratification, an increasingly popular approach for producing statistical estimates for sociodemographic subgroups at lower-level geographies. The findings challenge the conventional wisdom that a combination of increased race/ethnic diversity and grassroots voter mobilization activity secured Democratic candidate Joe Biden’s electoral triumph. The empirical evidence supports the hypothesis that the defection of majority White voters to the Democratic party tipped Georgia’s partisan balance. In addition, White voters with post-secondary educational attainment of a four-year degree or greater displayed the largest shift. The results align with previous research on the secular realignment of college-educated White voters with the Democratic party. Future research directions are discussed in the context of the role political demography could play in promoting high-quality information related to race/ethnic change in the electorate.

The rise of remote work across the United States has corresponded with an increase in a new type of local economic development strategy: remote worker attraction incentive programs. The first remote worker attraction incentive program was implemented in 2018 by the state of Vermont, followed quickly by the city of Tulsa, Oklahoma. In 2020, the Covid-19 pandemic changed the nature of work for many Americans, and the use of these attraction programs escalated quickly. Now, dozens of communities across the country have established remote worker incentive programs, seeking to benefit from the disentanglement of office space and work enabled by both new norms and new technologies. This brief conducts a preliminary national analysis of remote work attraction programs by identifying trends in their adoption, structure, and programmatic goals. The rise of this tool, with its focus on individual community investment rather than traditional models of employer attraction or workforce development, has the potential to reshape economic development policy at the local level. We chose to investigate 26 programs covering 36 counties and municipalities across the United States. To better understand the impact of remote worker incentive programs at the local level, we excluded those programs run at the state level. We collected population, age, employment, and income data on participating geographies to explore what may motivate communities to use this attraction tool.

Sub-Saharan Africa (SSA) accounts for around 40% of projected global births over the next 80 years. To investigate the roots of persistently high fertility rates across the region, I assemble micro data from 192 Demographic and Health Surveys covering 66 low-and-middle-income countries and document three key facts. First, women's fertility ideals and intentions are, on average, substantially higher in SSA than other low-and-middle-income regions. This gap is particularly large among poorer households: the socioeconomic gradient in desired fertility is twice as steep (more negative) on the sub-continent. Second, poorer women are also significantly less likely to work for a wage in SSA, where there exists a robust negative relationship between female wage work prevalence and desired fertility across provinces. Third, exploiting within-SSA variation across 25 countries, I find that increases in female salaried employment opportunities at the province level are associated with a flattening of this gradient over time, conditional on a rich set of covariates. These findings provide suggestive evidence that the nature of SSA's occupational change process may be an important contributor to the region's distinct fertility transition.

Prior literature on the economic impact of immigration has largely ignored changes to the composition of labor demand. In contrast, this paper uses a comprehensive collection of survey and administrative data to show that heterogeneous establishment entry and exit drive immigrantinduced job creation and a rightward shift of the productivity distribution in U.S. local industries. High-productivity establishments are more likely to enter and less likely to exit in high immigration environments, whereas low-productivity establishments are more likely to exit. These dynamics result in productivity growth. A general equilibrium model proposes a mechanism that ties immigrant workers to high-productivity firms and shows how accounting for changes to the employer distribution can yield substantially larger estimates of immigrantgenerated economic surplus than canonical models of labor demand.

Racially mixed children are a rapidly expanding segment of American families, signaling the ongoing blurring of racial boundaries. Most of what is known about multiraciality is drawn from analyses of two-parent families even as marriage became decreasingly tied to childbearing. The current study tracked the prevalence and racial composition of multiracial families where parents are married and unmarried from 1980 until 2018 using data from the U.S. Census and American Community Survey. We find that multiracial families are increasingly common amongst married and unmarried parents with the greatest growth occurring among White unmarried mothers, nearly 15% of whom have multiracial children as of 2018. Additionally, we find that Asian-White and Hispanic-White children are more likely to live in married families while Black-White and dual minority children are disproportionately represented amongst single-parent families. Ultimately, capturing the complexity of racialized contexts where multiracial children are found, as well as how the prevalence of these contexts has changed over time, requires accounting for family structure differences.

Many data applications have certain invariant constraints due to practical needs. Data curators who employ differential privacy need to respect such constraints on the sanitized data product as a primary utility requirement. Invariants challenge the formulation, implementation and interpretation of privacy guarantees. We propose subspace differential privacy, to honestly characterize the dependence of the sanitized output on confidential aspects of the data. We discuss two design frameworks that convert well-known differentially private mechanisms, such as the Gaussian and the Laplace mechanisms, to subspace differentially private ones that respect the invariants specified by the curator. For linear queries, we discuss the design of near optimal mechanisms that minimize the mean squared error. Subspace differentially private mechanisms rid the need for post-processing due to invariants, preserve transparency and statistical intelligibility of the output, and can be suitable for distributed implementation. We showcase the proposed mechanisms on the 2020 Census Disclosure Avoidance demonstration data, and a spatio-temporal dataset of mobile access point connections on a large university campus.

I analyze how Chinese import competition impacts community college enrollments in US commuting zones by using the variation of import exposure across the United States. After controlling for broader changes in the US economy, the findings show that there is fairly large and positive effect on enrollments for men between the years 1990 and 2007. The results for female enrollments are small or negative. However, these results for both men and women are imprecisely estimated. The findings of this paper resemble previous findings in the literature, namely Greenland and Lopresti (2016), Ferriere et. al (2021), and Lee (2021) which demonstrate increases in high school graduation rates and college enrollments. At the same time, due to the demographics of community colleges, these results expand insight on the human capital adjustment of lower income individuals caused by import competition (Kane & Rouse, 1995).

We study the problem of differentially private synthetic data generation for hierarchical datasets in which individual data points are grouped together (e.g., people within households). In particular, to measure the similarity between the synthetic dataset and the underlying private one, we frame our objective under the problem of private query release, generating a synthetic dataset that preserves answers for some collection of queries (i.e., statistics like mean aggregate counts). However, while the application of private synthetic data to the problem of query release has been well studied, such research is restricted to non-hierarchical data domains, raising the initial question-what queries are important when considering data of this form? Moreover, it has not yet been established how one can generate synthetic data at both the group and individual-level while capturing such statistics. In light of these challenges, we first formalize the problem of hierarchical query release, in which the goal is to release a collection of statistics for some hierarchical dataset. Specifically, we provide a general set of statistical queries that captures relationships between attributes at both the group and individual-level. Subsequently, we introduce private synthetic data algorithms for hierarchical query release and evaluate them on hierarchical datasets derived from the American Community Survey and Allegheny Family Screening Tool data. Finally, we look to the American Community Survey, whose inherent hierarchical structure gives rise to another set of domain-specific queries that we run experiments with.

This dissertation investigates the consequences of university-driven development in Philadelphia, especially for the African American communities that surround the University of Pennsylvania, Temple University, and Drexel University. It uses the theoretical contributions of W.E.B. Du Bois and David Harvey to conceptualize Philadelphia’s high rate of low-income homeownership as a product of the struggle of black workers and communities for democracy and the Right to the City. Thirty-three qualitative interviews with long-time residents, political activists, university administrators, and community institutions were conducted. Quantitative analysis including logistic regression analysis of Home Mortgage Disclosure Act (HMDA) data comparing outcomes in gentrifying and non-gentrifying neighborhoods and spatial K-cluster analysis were also conducted. Results show that university-driven development is leading to the conversion of single-family homes into apartment buildings and multifamily rentals, and a vision of the city in which developers, city officials, and university administrators wish to (in the words of one interviewee) “bring Manhattan to Philadelphia”. For homeowners, density is a shorthand for social, economic, and political displacement of the black working class and the disappearance of affordable homeownership opportunities. Density and affordable housing—and an ideology of urbanism—as conceptualized by city planners, university officials, developers, and new residents, clash with communities’ definitions of what the urban fabric of Philadelphia should be, as well as what truly affordable housing looks like. Furthermore, the influx of a student and professional population and its definition of progressivism has led to the political displacement of constituencies that have been shaped by black liberation movements. Resistance to universitydriven development, whether it is the movement against the building of Temple’s Stadium, or the drive to “save-zone” neighborhoods by rezoning them from mixed residential to single family, are led by black homeowners to preserve homeownership and black electorates. They are rooted in the historic struggles of the black worker in Philadelphia. I conclude with a discussion of the context of decreasing rates of homeownership in the country as a threat to a truly democratic society.

While the relationship between automobile ownership and the built environment is well established, less is known about how household relocations—specifically, moves between urban and suburban geographies—affect the likelihood of owning an automobile. Using the Panel Study of Income Dynamics and a refined neighborhood typology, I examine the relationship between inter-geography moves and transitions into and out of carlessness. Results suggest that among low-income households, urban-to-suburban movers have an increased likelihood of becoming car owners; those moving in the “opposite” direction—from suburban to urban neighborhoods—show a high propensity to transition into carlessness. Patterns among higher-income households, while similar, are more pronounced. In particular, higher-income carless households that make urban-to-suburban moves are far more likely to become car owners than their low-income counterparts. This highlights the ease with which higher-income households adjust their car ownership levels to suit their post-move neighborhoods. Higher-income suburban-to-urban movers are also more likely to transition into carlessness than low-income households. Importantly, however, only households at the bottom end of the “higher income” distribution have an increased propensity to become carless; suburban-to-urban movers with more financial resources maintain vehicle ownership rates similar to households that remain in the suburbs.

Marital behavior has changed dramatically in the U.S. (FP-21-24), with variation across sociodemographic characteristics such as education and race-ethnicity (FP-21-12). Using data from the American Community Survey (ACS), this profile examines the share of never-married men aged 35-39 by race-ethnicity and educational attainment. We focus on those aged 35-39 because this age bracket is above the median age of first marriage (the age in which at least 50% of men were married) in 2019 (FP-21-12) and captures most first marital experiences. White, Black, and Asian refers to those who are non-Hispanic and report a single race in the ACS, and Hispanic refers to those who report their ethnicity as Hispanic, regardless of race. Those with a two-year degree are included in the “some college” category.

I study the effect of higher education on lifecycle earnings and focus on three ways in which education may help drive earnings: the effect of higher education on one’s skills, the effect on one’s ability to accumulate new skills in the future, and the relationship between education and one’s likelihood of unemployment. In the first chapter, I examine whether a rising college earnings premium has contributed to increased earnings inequality over the past 50 years. I answer this with a model of college linked to a Ben-Porath human capital problem. The model is estimated from data on college attainment rates and lifecycle income profiles, which I document by education level. I find that lifetime earnings inequality within cohorts has risen significantly since the 1970s but little of this is directly attributable to the college premium. Higher tuition is also not particularly important. Increasing earnings risk seems to be the main driver, with higher college return dispersion a contributing factor. In the second chapter, I look at how unemployment risk affects the earnings distribution and the trajectory of lifetime earnings. I use a lifecycle human capital model where workers accrue both general and career-specific skills and face unpredictable transitory unemployment shocks. Through unemployment’s effect on career-specific human capital, I find that unemployment risk explains a seeming contradiction where college graduates’ earnings are characterized by high lifecycle dispersion but low year-to-year variation relative to those with less education. As in prior work, I find that variation in total lifetime earnings is mostly driven by differences among people before they enter adulthood, but unemployment risk significantly affects the mean. Moreover, my framework also explains recent findings on the distribution of earnings changes, particularly the notable skew and leptokurtosis.

Since the end of the Cold War, the health of the national nuclear enterprise workforce has been a matter of abiding concern to senior U.S. officials. The two government agencies with principal responsibility for this workforce—the U.S. Department of Defense (DoD) and the U.S. Department of Energy's National Nuclear Security Administration (NNSA)—have had to contend with adverse demographic trends, recruitment and retention challenges, and intense competition for specific skills and expertise, especially in scientific and engineering fields, to maintain a workforce with the capabilities and experience needed for nuclear-related duties. This report summarizes the results of a quick-turn, 90-day assessment of the health of the national nuclear enterprise workforce, focusing on federal personnel working in acquisition and scientific, technical, engineering, and math occupations. The study team used a mixed methods approach that relied primarily on extant data to consider workforce health in terms of workforce planning, recruiting and hiring, employee engagement and development, leader development, and morale and retention. The report features findings about enterprise strengths, such as promising practices that are candidates for broader use, and those about factors that challenge workforce health, such as evolving demand for more and different talent in light of simultaneous modernization and sustainment needs. The study team also offers recommendations to bolster the health of the nuclear enterprise workforce, both now and over the next decade.

INTRODUCTION This study sought to investigate the trend and factors associated with DHT (Digital Health Tools) utilization among individuals with SPD (serious psychological distress) in the US. METHODS Data were drawn from the 2013 to 2017 National Health Interview Survey. Descriptive statistics and multivariable logistic regression were employed to assess the use of DHT among individuals aged ≥18 years with SPD. RESULTS A total of 6110 adults reported SPD and of these, 15.6% reported at least one technology-based interaction with the health system. During the 5-year period, the proportion of individuals with SPD who utilized any DHT to interact with the healthcare system doubled from 10.0% in 2013 to 21.3% in 2017 (p<0.001). In multivariable models, several sociodemographic factors predicted DHT use. CONCLUSIONS The use of DHT among individuals with SPD in the US increased between 2013 and 2017. However, sociodemographic disparities in DHT use among this population exist.

Existing research establishes connections between legacies of slavery and racial economic inequality (Curtis and O’Connell 2017; O’Connell 2012) but has yet to examine connections between legacies of slavery, gentrification, and related state policy. By conducting OLS regression and spatial autoregression using data from the 1860 U.S. Census, 2000 U.S. Census, and 2012 American Community Survey, I explore the relationship between legacies of slavery and contemporary gentrification at the county-level. This study demonstrates evidence of a meaningful association between legacies of slavery and income by race, and between legacies of slavery and the likelihood a place will be made more vulnerable to gentrification by targeted government policy. This bridges the literatures on legacies of slavery and gentrification by proposing and finding measured support for their association yet leaves questions remaining about the total impacts of government policies conceptually associated with gentrification.

In the U.S., Black adults consistently have higher allostatic load – an indicator of physiological dysregulation – than White adults. Education is considered a likely mechanism given racial differences in attainment, but evidence is mixed. This may be due, in part, to data limitations that have made it difficult for scholars to account for the structurally rooted systemic racism that shaped the U.S. education system and led to large racial inequities in school term length and school attendance among older adults who grew up in the Jim Crow South. Our study addresses this limitation by linking historical data on Black and White segregated school systems in the U.S. South from 1919 to 1954 to the Health and Retirement Study (HRS) to determine if a new measure of educational attainment that accounts for structural racism that led to differences in the number of school days attended by Black and White students across years and states better explains Black-White inequities in allostatic load among older adults who attended school during Jim Crow. We restrict our sample to HRS respondents racialized as White or Black, who resided in the South when they were school-aged, completed primary/secondary school between 1919 and 1954, and provided a measure of allostatic load (n = 1932). We find that our new measure of schooling – duration in school – reduced the Black-White inequity in allostatic load more so than self-reported years of schooling whether we measured allostatic load continuously (34% vs 16%) or categorically (45% vs 20%). Our findings highlight the importance of identifying and using historically informed measures of schooling that account for structurally rooted systemic racism when trying to understand how education shapes the health of individuals racialized as Black in the United States.

Objectives Although education is a key determinant of cognitive function, its role in determining Black–White disparities in cognitive function is unclear. This may be due, in part, to data limitations that have made it difficult to account for systemic educational inequities in the Jim Crow South experienced by older cohorts, including differences in the number of days Black students attended school compared to their White counterparts or Black peers in better-funded southern states. We determine if accounting for differential rates of school attendance across race, years, and states in the Jim Crow South better illuminates Black–White disparities in trajectories of cognitive function. Methods We linked historical state-level data on school attendance from the 1919/1920 to 1953/1954 Biennial Surveys of Education to the Health and Retirement Study, a nationally representative, longitudinal study of U.S. adults older than age 50. We restricted our sample to Black and White older adults who attended school in the Jim Crow South and began primary school in/after 1919/1920 and completed primary/secondary school by 1953/1954 (n = 4,343). We used linear mixed models to estimate trajectories of total cognitive function, episodic memory, and working memory. Results Self-reported years of schooling explained 28%–33% of the Black–White disparity in level of cognitive function, episodic memory, and working memory. Duration of school, a measure that accounted for differential rates of school attendance, explained 41%–55% of the Black–White disparity in these outcomes. Discussion Our study highlights the importance of using a more refined measure of schooling for understanding the education–cognitive health relationship.

In this dissertation, I study three topics ranging from political economics to the economics of crime. In the first chapter, “Using Rain for Electoral Gain: Evidence from FEMA’s Public Assistance Program”, I study distortions in public spending— following large natural disasters—caused by elected officials at the state level in the United States (U.S.). I construct a novel county-level dataset that merges fine-grain physical measures of large destructive storms, satellite data on existing infrastructure, demographic information from Census, and administrative data that tracks multiple types of relief spending from several federal programs. To organize my empirical analysis I utilize a model of political competition between parties, that provides clear testable predictions. I find robust empirical evidence that political parties target public spending to counties with higher historic turnout relative to other counties within the same electoral district. To assuage fears that my results are driven by some unobserved bias or simply by chance, I propose two credible instrumental variables, explicitly model—and control for—the selection process that determines a counties eligibility for relief aid, and perform multiple placebo tests. In the second chapter “It’s Always Sunny in Politics”—co-authored with Carolina Concha-Arriagada—we study how election day weather impacts voter choice. Specifically we study an understudied dimension of weather—sunshine. Using novel daily weather measurements from satellites, linked to county-level U.S. Presidential electoral returns from 1948-2016, we document how sunshine affects the decision making of voters. We find that election-day exposure to sunshine increases support for the Democratic party on average. Additionally, we show that, contrary to prior findings that do not control for sunshine, precipitation has no detectable impact on partisan support, but universally depresses turnout. To rationalize our results we propose a mechanism whereby sunshine modulates voter mood which causes a change in voter choice, while precipitation only impacts turnout through increasing the cost of voting. We then build a theoretical model, which features this mechanism, and generates additional tests that we take to our data. Our results suggest that uninformative weather on election day, specifically sunshine, has detectable electoral impacts that teach us about voter choice. In the third chapter, “Decomposing Racial Disparities in Incarceration: Evidence from a Southern U.S. Jurisdiction” I utilize high quality data developed from an almost two year relationship with a prosecutor’s office located in a medium-sized jurisdiction in the southern United States. These data were carefully curated by a team of data scientists and engineers with institutional knowledge from prosecutors within the office. We first documents large raw racial disparities in incarceration and then decompose this raw disparity into systemic and direct components. We find that the systemic component vastly outweighs the direct component, implying that upstream racial differences drive the raw disparity.

The past few decades have seen major shifts in childbearing behavior. The overall share of women who have ever had a birth has declined, the mean age at first birth has increased, and the share of women with three or more children has decreased (FP-22-19, FP-20-06, FP-20-04). These changes affect the number of years, as well as the ages, women spend parenting minor children. This profile uses data from the 1980 and 2020 June Fertility Supplement of the Current Population Survey (CPS) to estimate changes in the share of women who are in their prime parenting years and, for 2020, we focus on educational differences in the primary parenting years. We define prime parenting years as the age group at which over 75% of women were parenting a child aged 0 to 17. We estimate the ages spent parenting based on women’s age at entry into motherhood and women’s age at which her youngest current biological, adopted or stepchild will turn 18, essentially allowing us to capture the age women started parenting and can expect to complete parenting minor children. This is then applied to singular ages to estimate the share of women at each age who were parenting a minor child. The range of parenting years depends not only on the age at which women first have a child but also on the number and spacing of children. The figures below are color coded by quartile such that the darkest color represents a share of 75% or more at that age and the lightest color represents a share below 25% at that age.

Introduction: Prepared (ready-to-eat) foods are sold in >90% of U.S. supermarkets, but little is known about their nutritional quality. This study examined trends in purchases of supermarket prepared foods and compared their nutritional profile with that of supermarket packaged foods and restaurant foods. Methods: Nutrition data were obtained on prepared foods sold from 2015 to 2019 in 2 supermarket chains (∼1,200 stores). One chain (193 stores) provided transaction-level sales data from 2015 to 2017. Analyses (conducted in 2021–2022) examined trends in the number of different prepared foods offered by the chains and trends in purchases of calories, total sugar, saturated fat, and sodium from prepared foods. Calorie and nutrient densities (i.e., per 100 g of food) and prevalence of being high in calories or nutrients (on the basis of Chilean standards) were analyzed among supermarket prepared foods, supermarket packaged foods, and restaurant foods consumed in the National Health and Nutrition Examination Surveys 2015–2018. Results: The number of different prepared foods offered at supermarket chains increased from 1,930 in 2015 to 4,113 in 2019. Calories per transaction purchased from supermarket prepared foods increased by 1.0 calories/month (95% CI=0.8, 1.1), a ∼3% annual increase, with similar trends for other nutrients. At supermarkets, >90% of prepared bakery and deli items and 61% of prepared entrees/sides were high in calories or another nutrient of concern, similar to supermarket packaged foods and restaurant foods. Conclusions: Supply of and demand for supermarket prepared foods have grown substantially over time. These trends are concerning given these foods’ overall poor nutritional quality.