The purpose of this study is to comprehensively examine the association between alcohol consumption behavior and health and mortality risks in the United States. The three interconnected aims of this dissertation broaden the relationship between alcohol intake levels, health, and mortality risks to newer perspectives: 1) the 1984 Minimum Legal Drinking Age Act, cohort variation in alcohol use and mortality, 2) mortality among working-age women, and 3) adult health outcomes. Data from the National Health Interview Survey-Linked Mortality Files (2001 – 2015) and the National Longitudinal Study of Adolescent to Adult Health (1995 – 2019) are used to explore these associations. Cox proportional hazard models, discrete-time hazard models, vector generalized linear models, and generalized linear models are used to estimate the impact of alcohol consumption on health and mortality risks. A major conclusion is that the 1984 MLDA significantly decreased moderate to heavy alcohol drinking among cohorts aged 16 – 20 after 1988 when all the states fully enacted the policy. Furthermore, participation in the labor force serves as a protective mechanism against increased mortality risks for women via alcohol consumption levels. Finally, physical and mental health outcomes in adulthood (age 32 – 42) are significantly affected by lifetime alcohol use disorder. The empirical findings reiterate the importance of studying alcohol as a major threat to adult health and mortality outcomes in the United States.

The nineteenth and twentieth century saw two waves of state schooling laws. The first wave focused on children to age 14 and the second wave focused on high school. Using the full count 1940 census and a new coding of state laws, this paper provides new estimates of the effects of the first wave of laws. The analysis focuses on cohorts of prime working age between 1910 and 1940. IV estimates of returns to schooling range from 0.067 to 0.077. Quantile IV estimates show the returns were largest for the lowest quantiles, and were generally monotonically decreasing for higher quantiles.

The Maryland Health Benefit Exchange, which operates Maryland's state-based health insurance marketplace, Maryland Health Connection, analyzed the Maryland uninsured population by geography, demographic characteristics, and eligibility for government programs before and after the job losses resulting from the COVID-19 pandemic. This report details the methodology and results of this effort along with the construction of an interactive web dashboard that will enable other researchers, outreach, and marketing to effectively identify and target those most at risk of being uninsured and those experiencing new uninsurance due to the pandemic.

Multi-temporal and spatially explicit population data are vital in many fields, such as demography, urban planning, disaster prevention, economics, and environmental modeling. Population data used in these studies are typically aggregated at census enumeration units, which are too coarse for many applications. Accurate population downscaling methods are needed to obtain population data at finer spatial resolutions. We use a novel settlement-related database, Built-Up Property Records (BUPR) from the Historical Settlement Data Compilation for the United States (HISDAC-US) to downscale population from census tracts to block groups. The BUPR dataset provides the number of built-up property records for each 250-m grid at 5-year temporal resolution from 1810 to 2015 for most contiguous United States (CONUS). The ability of BUPR to downscale population from census tracts to block groups for four states, representing a range of population densities, is evaluated here by comparing against other commonly-used ancillary datasets. The BUPR-based method outperforms all other methods in all but one state with highly-incomplete BUPR. A more detailed accuracy assessment is performed by dividing each state into low, medium, and high population density categories. The BUPR method produces more accurate downscaled population estimates for low and medium categories, though its performance deteriorates in the high density category due to its relatively coarse spatial resolution. BUPR-based dasymetric mapping is subsequently applied to the CONUS and found to generalize well beyond the four comparison states with high downscaling accuracy. The long-term record of the HISDAC-US dataset enables the potential construction of fine-grained population data back to 1810.

We document the intergenerational mobility of black and white American men from 1880 through 2000 by building new historical datasets for the late nineteenth and early twentieth century and combining them with modern data to cover the middle and late twentieth century. We find large disparities in mobility, with white children having far better chances of escaping the bottom of the distribution than black children in every generation. This mobility gap was more important in proximately determining each generation’s racial gap than was the initial gap in parents’ economic status.

This dissertation explores techniques for automating program analysis, with a focus on validating and securely executing differentially private programs. Differential privacy allows analysts to study general patterns among individuals, while providing strong protections against identity leakage. To automatically check differential privacy for programs, we develop Fuzzi: a three-level logic for differential privacy. Fuzzi’s lowest level is a general-purpose logic; its middle level is apRHL, a program logic for mechanical construction of differential privacy proofs; and its top level is a novel sensitivity logic for tracking sensitivity bounds, a fundamental building block of differential privacy. Some differentially private algorithms have sophisticated proofs that cannot be derived by a compositional typechecking process. To detect incorrect implementations for these algorithms, we develop DPCheck for testing differential privacy automatically. Adapting a well-known “pointwise” proof technique for differential privacy, DPCheck observes runtime program behaviors, and derives formulas that constrain potential privacy proofs. Once we are convinced that a program is differentially private, we often still have to trust that the machine executing the program does not misbehave and leak sensitive results. For analytics at scale, computation is often delegated to networked computers that may become compromised. To securely run differentially private analytics at scale, we develop Orchard, a system that can answer many differentially private queries over data distributed among millions of user devices. Orchard leverages cryptographic primitives to employ untrusted computers, while preventing untrusted computers from observing sensitive results.

In a 2014 paper, Garthwaite, Gross, and Notowidigdo examined the employment impact of the 2005 TennCare contraction. We extend their approach in several directions. First, we use consistent Conley-Taber estimation. Second, we transform their estimates to make them comparable to previous work; the transformed effects have large confidence intervals. We estimate their models using several larger data sets in an attempt to get more precise estimates but find that the results can be quite different. We consider two modifications to account for a major disruption to coverage in 2002, and one of these reduces the differences in the results.

Disability is an important characteristic to consider in survey research. However, people with disabilities are a hard-to-reach population. Internet survey methods offer tremendous potential to expand researchers’ ability to reach and learn about people with disabilities. The goal of this study is to examine potential bias when using nonprobability Internet samples to investigate demographics and socioeconomic outcomes of people with disabilities. We compare the findings based on a national employment and disability survey instrument fielded to four samples: (1) a random-digit dial (RDD) sample, (2) a prescreened sample from a nonprobability Internet access panel, for which screening was based on the presence of 139 previously reported health conditions, (3) an unscreened sample from another nonprobability Internet access panel (without previously prescreened health conditions), and (4) a mixed nonprobability self-recruited (river and snowball) sample. Each sample was weighted on four demographic variables (gender, age, race/ethnicity, and region) using benchmarks from the American Community Survey (ACS). Three dichotomous outcome variables of interest (level of education, household income, and current employment status) were contrasted with weighted population estimates from the ACS. Results showed that the sample resulting from the RDD and all three nonprobability Internet samples differed significantly from ACS population estimates on all three outcome variables. Reweighting to include type of functional disability did not significantly reduce dissimilarities with ACS for any of the four samples. Nonprobability Internet survey methods offer relatively low-cost, easy-to-use avenues for disability-related research. Yet, researchers must proceed with caution to reduce or avoid known sources of bias in both the methodology and the interpretation of results.

We test the effectiveness of a link-tracing sampling approach, Network Sampling with Memory (NSM) to recruit samples of rare immigrant populations with an application among Chinese immigrants in the Raleigh-Durham Area of North Carolina. NSM uses the population network revealed by data from the survey to improve the efficiency of link-tracing sampling, and has been shown to substantially reduce design effects in simulated sampling. Our goals are: (1) to show that it is possible to recruit a probability sample of a locally rare immigrant group using NSM and achieve high response rates; (2) to demonstrate feasibility of collection and benefits of new forms of network data that transcend kinship networks in existing surveys and can address unresolved questions about the role of social networks in migration decisions, the maintenance of transnationalism, and the process of social incorporation; (3) to test the accuracy of the NSM approach to recruit immigrant samples by comparison with the American Community Survey (ACS). Our results indicate feasibility, high performance, cost-effectiveness and accuracy of the NSM approach to sample immigrants for studies of local immigrant communities. This approach can also be extended to recruit multi-site samples of immigrants at origin and destination.

Despite ample supply of COVID-19 vaccines, the proportion of fully vaccinated individuals remains suboptimal across much of the US. Rapid vaccination of additional people will prevent new infections among both the unvaccinated and the vaccinated, thus saving lives. With the rapid rollout of vaccination efforts this year, the internet has become a dominant source of information about COVID-19 vaccines, their safety and efficacy, and their availability. We sought to evaluate whether trends in internet searches related to COVID-19 vaccination — as reflected by Google’s Vaccine Search Insights (VSI) index — could be used as a marker of population-level interest in receiving a vaccination. We found that between January and August of 2021: 1) Google’s weekly VSI index was associated with the number of new vaccinations administered in the subsequent three weeks, and 2) the average VSI index in earlier months was strongly correlated (up to r=0.89) with vaccination rates many months later. Given these results, we illustrate an approach by which data on search interest may be combined with other available data to inform local public health outreach and vaccination efforts. These results suggest that the VSI index may be useful as a leading indicator of population-level interest in or intent to obtain a COVID-19 vaccine, especially early in the vaccine deployment efforts. These results may be relevant to current efforts to administer COVID-19 vaccines to unvaccinated individuals, to newly eligible children, and to those eligible to receive a booster shot. More broadly, these results highlight the opportunities for anonymized and aggregated internet search data, available in near real-time, to inform the response to public health emergencies. After reaching a minimum in June 2021, coronavirus disease 2019 (COVID-19) cases and hospitalizations rose rapidly throughout the summer across the United States as the more contagious Delta variant became the dominant strain of SARS-CoV-2, the virus that causes COVID-19. The vast majority of new hospitalizations across the US in mid-to-late 2021 have been occurring among unvaccinated individuals. 1,2 Despite ample supply of COVID-19 vaccines, the proportion of fully vaccinated individuals remains below recommended levels across much of the US. Specifically, as of October 20, 2021, the US Centers for Disease Control and Prevention (CDC) reports that only 57% of Americans (67% of the population ≥ 12 years of age) have been fully vaccinated (https://covid.cdc.gov/covid-data-tracker). Rapid vaccination of additional people will prevent new infections among both the unvaccinated and the vaccinated, reduce the severity of infections among the vaccinated, and thus, save lives. With the rapid roll out of vaccination efforts this year, the internet has become a dominant source of information (and misinformation) about COVID-19 vaccines, their safety and efficacy, and their availability. Prior studies have shown that internet search patterns based on anonymized and aggregated data can be used to predict the occurrence of Lyme disease and outbreaks of influenza; to nowcast COVID-19 cases, hospitalizations, and deaths; and to identify food establishments that would benefit from food safety inspections to limit the further spread of food borne illness. Internet search patterns may similarly provide novel insights that could be used to inform public health efforts to increase vaccination uptake, but this hypothesis has not been examined in detail. Internet searches related to COVID-19 vaccines began rising in January 2021 and then rose further starting in March. 8 Early evidence suggests that internet search activity (aggregated to the state level) is associated with higher rates of vaccination in that state. Google recently began publishing the COVID-19 Vaccination Search Insights (VSI) index, a publicly available dataset showing trends in Google searches related to COVID-19 vaccination from January 2021 through the present. We sought to evaluate whether patterns in internet searches across locations and over time could be used as a marker of population-level interest in receiving a vaccination and, if so, to explore how this information might be used by public health officials to identify geographic areas with particularly high amenability towards vaccination despite low uptake.

Paraeducators perform multiple roles in U.S. classrooms, including among others preparing classroom activities, working with students individually and in small groups, supporting individualized programming for students with disabilities, managing classroom behavior, and engaging with parents and communities. Yet, little research provides insights into this key group of educators. This study combines an analysis of national administrative data to describe the paraeducator labor market with a systematic review of collective bargaining agreements and other job-defining documents in ten case-study districts. We find a large and expanding labor market of paraeducators, far more diverse along ethnic and racial lines than certified teachers but with far lower wages, fewer performance incentives, less professional development, and fewer opportunities for advancement within the profession.

There’s been no escaping the COVID-19 pandemic, with its toll of hospitalizations, layoffs, and quarantines. Every place in America has been affected, often in drastic ways, as the coronavirus hit home and laid bare—like an X-ray—an array of underlying economic and social challenges wherever it arrived. And so it has been in Indiana. While it has managed, by some measures, one of the stronger recoveries from the initial crisis among states, the Hoosier State has also contended with major dislocations and challenges. Not only did COVID-19 interrupt several years of relatively decent growth prior to the pandemic shock, but the pandemic and its impacts have intensified an array of concerns about the underlying health and resilience of the state’s economy, ranging from its technological competitiveness (region by region) to its adaptability to the pay of its jobs.

The US fertility transition in the nineteenth century is unusual. Not only did it start from a very high fertility level and very early in the nation's development, but it also took place long before the nation's mortality transition, industrialization, and urbanization. This paper assembles new county-level, household-level, and individual-level data, including new complete-count IPUMS microdata databases of the 1830-1880 censuses, to evaluate different theories for the nineteenth-century American fertility transition. We construct cross-sectional models of net fertility for currently-married white couples in census years 1830-1880 and test the results with a subset of couples linked between 1850-1860, 1860-1870, and 1870-1880 censuses. We find evidence of marital fertility control consistent with hypotheses as early as 1830. The results indicate support for several different but complimentary theories of the early US fertility decline, including the land availability, conventional structuralist, ideational, child demand/quality-quantity tradeoff, and life cycle savings theories.

This article uses baseline data from an observational study to estimate the determinants of racial and gender disparities in obesity. Samples of low-income workers in Minneapolis and Raleigh reveal that respondents in Minneapolis have lower body mass indices (BMIs) than respondents in Raleigh. There are large, statistically significant race and gender effects in estimates of BMI that explain most of the disparity between the two cities. Accounting for intersectionality-the joint impacts of being Black and a woman-reveals that almost all the BMI gaps between Black women in Minneapolis and Raleigh can be explained by age and education differences.

Purpose: Burden of disease (BoD) assessments typically rely on national-level incidence rates for the health outcomes of interest. The impact of using a constant national-level incidence rate, versus a more granular spatially varying rate, remains unknown and understudied in the literature. There has been an increasing number of publications estimating the BoD of childhood asthma attributable to air pollution, as emerging evidence demonstrates that traffic-related air pollution (TRAP) leads to onset of the disease. In this study, we estimated the burden of incident childhood asthma cases which may be attributable to nitrogen dioxide (NO2), a criteria pollutant and a good marker of TRAP, in the contiguous United States. We used both a national-level and newly generated state-specific asthma incidence rates and compared results from the two approaches. Methods: We estimated incident childhood asthma cases which may be attributable to NO2 using standard BoD assessment methods. We combined child (<18 years) counts with 2010 NO2 exposures at the census block level, concentration-response function, and state-specific asthma incidence rates. NO2 concentrations were obtained from a previously validated land-use regression model. We sourced the concentration-response function from a meta-analysis on TRAP and risk of childhood asthma. We estimated incidence rates using raw data collected in the 2006–2010 Behavioral Risk Factor Surveillance System and Asthma Call-back Surveys. We stratified the estimated BoD by urban versus rural status and by median household income, explored trends in BoD across 48 states and the District of Columbia, and compared our results with a published BoD analysis which used a constant national-level incidence rate across all states. Results: The overall mean (min–max) NO2 concentration(s) was 13.2 (1.5–58.3) ug/m3 and was highest in urbanized areas. The estimated national aggregate asthma incidence rate was 11.6 per 1000 at-risk children and ranged from 4.3 (Montana) to 17.7 (District of Columbia) per 1000 at-risk children. The 17 states that did not have data to estimate an incidence rate were assigned the national aggregate asthma incidence rate. Using the state-specific incidence rates, we estimated a total of 134,166 (95% confidence interval: 75,177–193,327) childhood asthma incident cases attributable to NO2, accounting for 17.6% of all childhood asthma incident cases. Using the national-level incidence rate, we estimated a total of 141,931 (95% confidence interval: 119,222–163,505) incident cases attributable to NO2, accounting for 17.9% of all childhood asthma incident cases. Using the state-specific incidence rates therefore reduced the attributable number of cases by 7,765 (5.5% relative reduction), compared with estimates using the national-level incidence rate. Across states, the change in the attributable number of cases ranged from −64.1% (Montana) to +33.8% (Texas). California had the largest absolute decrease (−6,190) in attributable cases, whereas Texas had the largest increase (+3,615). Stratifying by socioeconomic status and urban versus rural status produced new trends compared with the previously published BoD analysis showing high heterogeneity across the states. Conclusions: We estimated new state-specific asthma incidence rates for the contiguous United States. Using state-specific incidence rates versus a constant national incidence rate resulted in a small change in the NO2 attributable BoD at the national level, but had a more prominent impact at the state level.

Rationale: Health equity is a significant concern of public health, yet a comprehensive assessment of health equity in the United States over time is lacking. While one might presume that overall health will improve with rising living standards, no such presumption is warranted for health equity, which may decline even as average health improves. Objectives: To assess trends in national and state-level health equity in mortality for people up to age 25, ages 25–64 and aged 65 and older. Methods: A health equity metric was calculated as the weighted mean life expectancy relative to a benchmark level, defined as the life expectancy of the most socially-privileged subpopulation (white, non-Latinx males with a college education or higher). We analyzed 114,558,346 death records from the National Center for Health Statistics, from January 1, 1969 to December 31, 2019 to estimate health equity annually at the national and state-level. Using ICD-9/ICD-10 classification codes, inequities in health were decomposed by major causes of death. Results: From 1969 to 2019, health equity in the United States improved (+0.36 points annually [95% CI 0.31–0.41]), albeit at a slower rate over the last two decades (+0.08 points annually [95% CI 0.03–0.14] from 2000 to 2019, compared to +0.57 points annually from 1969 to 2000 [95% CI 0.50–0.65]). Health equity among those under 25 improved substantially (+0.82 points annually [95% CI 0.75–0.89]) but remained flat for adults 25–64 (−0.01 points annually [95% CI -0.03-0.003]) For those over 65, health equity displayed a downward trend (−0.08 points annually [95% CI -0.09 to −0.07]). Gains in equity from reduced unintentional injuries and homicides have been largely offset by rising mortality attributable to drug overdoses. Conclusions: The US is failing to advance health equity, especially for adults. Keeping policy-makers accountable to a summary measure of health equity may help coordinate efforts at improving population health.

Federal tax data, derived from individuals’ and businesses’ tax and information returns, are invaluable resources for research on a range of topics. That research improves our understanding of individuals’ and firms’ responses to economic incentives. However, full access to these data is available only to select government agencies, to a very limited number of researchers working in collaboration with analysts in those agencies, or through highly selective programs within the Internal Revenue Service Statistics of Income Division. In addition, the existing process of manually vetting each statistical release for disclosure risks is labor intensive and imperfect because it relies on subjective human review. As part of larger project to implement an automated validation server, we conduct an extensive feasibility study on several differentially private methods for releasing tabular statistics, mean and quantile statistics, and regression analyses with cross-sectional data. We provide a discussion on which methods we tested and which methods could not be implemented in practice. We then evaluate the selected differentially private methods based on their impact on tax public policy decisions and several other utility metrics. From our findings, we outline the outstanding challenges and future work.

We construct a time series of college attendance patterns for the United States and document a reversal: family background was a better predictor of college attendance before World War II, but academic ability was afterward. We construct a model of college choice that explains this reversal. The model's central mechanism is that an exogenous surge of college attendance leads better colleges to be oversubscribed, institute selective admissions, and raise their quality relative to their peers, as in Hoxby (2009). Rising quality at better colleges attracts high-ability students, while falling quality at the remaining colleges dissuades low-ability students, generating the reversal.

A key challenge for research on many questions in the social sciences is that it is difficult to link records in a way that allows investigators to observe people at different points in their life or across generations. In this paper, we contribute to recent efforts to create these links with a new approach that relies on millions of record links created by individual contributors to a large, public, wiki-style family tree. We use these “true” links both to inform the decisions one needs to make when using automated methods to link records and as a training data set for use in a supervised machine learning approach. We describe our procedure and illustrate its potential by linking individuals across the 100% samples of the US censuses from 1900, 1910, and 1920. When linking adjacent censuses, we obtain an overall match rate of 62-65 percent (for over 88.9 million matches), with a false positive rate that is around 6-7 percent and with links that are similar to the population along observable characteristics. Thus, our method allows us to link records with a combination of a high match rate, precision, and representativeness that is beyond the current frontier. Finally, we demonstrate the potential of the data by estimating the degree of intergenerational transmission of literacy between father-son and mother-daughter pairs.

This paper analyzes how increased trade exposure affects students' choice of STEM major. I first present a simple model to illustrate how trade exposure impacts students' utility functions through their self-beliefs about labor outcomes and then use assorted data to show that import competition positively affects the choice of STEM major. The associated effects are 1.05 and 0.72 percentage point increases in the probability of choosing STEM majors in the first and second parts of college, respectively. As for labor market outcomes, my results suggest that a rise in import competition leads to a pronounced negative effect on weekly wages, employment status, and full-time employment across STEM and non-STEM occupations from the late 1990s through the 2000s. STEM occupations, however, are less negatively impacted by import competition, which helps explain why a rise in import exposure increases the probability of students choosing STEM majors.