Objectives: Epidemiological patterns for lung cancer among never smokers (LCNS) are largely unknown, even though LCNS cases comprise 15% of lung cancers. Past studies were based on epidemiologic or health system cohorts, and not fully representative of the underlying population. The objective was to analyze rates (and trends) of LCNS by sex, age group, and race and ethnicity based on all-inclusive truly population-based sources. Materials and methods: Individual-level data from 2014 to 2018 on smoking status among microscopically-confirmed lung cancer cases from Florida's cancer registry were combined with population denominators adjusted with NHIS data on smoking prevalence to compute population-based LCNS incidence rates and rate ratios. Incidence rates and proportional mortality were ranked against other cancers. Joinpoint regression analyses examined trends. Results: Proportions of LCNS ranged from 9% among White men to 83% among Chinese women. Overall, LCNS was 13% (IRR 1.13, 95%CI 1.08-1.17) more common among men than women, but variation occurred by age group, with female rates exceeding male in younger ages. Age-adjusted rates per 100,000 were highest among Asian/Pacific Islander (API) men and women (15.3 and 13.5, respectively) and Black populations (14.6, 12.9), intermediate for White (13.2, 11.8) and lowest among the Hispanic population (12.1, 10.6). Among API women, LCNS was the second leading cause of cancer death, surpassed only by breast cancer. LCNS trends were stable over time. Conclusion: LCNS is the 11th most frequently occurring cancer in men and 8th in women. LCNS differences by race/ethnicity are small, within a 15% range of the White population's rates. Surprisingly, API men and women have the highest LCNS rates and proportional mortality. As smoking prevalence decreases in the US, LCNS cases will inevitably increase, warranting inquiry into risk factors across the lifespan.

The study of longitudinal multimorbidity trajectories (MOTRs) in older adults is a growing field of study. Many previously used multimorbidity measurement methods have been cross-sectional in nature, and more studies utilizing longitudinal methods are needed. This population-based study aims to (1) investigate the construction of MOTRs using semi-parametric group based trajectory models (GBTM) employing multiple combinations of comorbidity indices and index score calculation methods, and stratified by sex, (2) generate non-parametric unsupervised machine learning multimorbidity clustering trajectories and compare them to similar GBTM models, and (3) using supervised machine learning classification algorithms, determine the influence of earlylife conditions compared to unsupervised machine learning multimorbidity clustering trajectories on mortality. We found that the least-chronic disease (Escaper) and highest-chronic disease (Frail) MOTRs consistently manifested with all combinations of model types and comorbidity index score calculation methods. Additionally, we found that unsupervised machine learning cluster trajectories had higher performance than GBTM models and may be candidates for implementation in a clinical setting. Futhermore, multimorbidity trajectory membership was more influential in the prediction of mortality than early-life conditions. Although early-life conditions have some effect on the risk of mortality, current disease burden is an appropriate area of focus when the goal is to decrease the risk of mortality in an older adult population. As far as we know, development of longitudinal unsupervised machine learning clustering of chronic disease in older adults using common comorbidity index scores and development of supervised machine learning predictive classification models using MOTRs as features has not been done. This dissertation has contributed to the literature by helping to understand how MOTRs are built, how multimorbidity manifests over time, and how MOTRs influence mortality in an older adult population. Clinicians may use the information and techniques developed here to identify older adult patients on an unhealthy disease trajectory with the goal of a possible intervention to shift the patient to a healthier trajectory.

Aims: While nationally representative alcohol surveys are a mainstay of public health monitoring, they underestimate consumption at the population level. This paper demonstrates how to adjust individual-level survey data using aggregated alcohol per capita (APC) data for improved individual- and population-level consumption estimates. Design and methods: For the period 1984-2020 data on self-reported alcohol consumption in the past 30 days were taken from the Behavioral Risk Factor Surveillance System (BRFSS) involving participants (18+ years) in the US. Monthly abstainers were re-allocated into lifetime abstainers, former drinkers and 12-month drinkers using the 2005 National Alcohol Survey data.  To correct for under-coverage of alcohol use, we triangulated APC and survey data by upshifting quantity (average grams/ day) and frequency (drinking days/week) of alcohol use based on national and state-level alcohol per capita data. Findings The corrections described above resulted in better correspondence between survey and APC data. Following our procedure, national estimates of alcohol quantity increased from 45% to 77% of APC estimates. Both quantity and frequency of alcohol use were upshifted; by upshifting to 90% of APC, we were able to fit trends and distributions in APC patterns for individual states and the US. Conclusions: An individual-level dataset which more accurately reflects the alcohol use of US citizens was achieved. This dataset will be invaluable as a research tool and for the planning and evaluation of alcohol control policies for the US. The methodology described can also be used to adjust individual-level alcohol survey data in other geographical settings.

Despite the various health and social benefits walkable environments are expected to confer, initial studies indicate that older adults in the United States are primarily moving away, rather than toward, urban centers. Yet, amid this broad demographic movement, is there any indication that older adults are more likely to be found in more walkable neighborhoods or that the growth in older adult populations (both percentage change and change in proportion) is positively related to neighborhood walkability? To address these questions, the spatial relationship between older adult population and neighborhood walkability (assessed using Walk Score®) and other amenities was examined for about 37,000 urban census tracts situated within U.S. metropolitan areas with populations of 500,000 or more. The results of the modeling procedure suggest a positive association between the population of older adults in 2015–2019 and neighborhood walkability. Crucially, however, the change in older adults aged 65+ between 2010 and 2015–2019 indicates a trend toward less walkable areas. The relationship between older adults and neighborhood walkability varied among “pre-retirement” (50–64) and “retirement” (65+) age groups and by the size and degree of urban sprawl of the encompassing metropolitan area.

Objectives/Hypothesis: Patients with cleft lip and/or palate (CLP) are at increased risk of malnutrition. Acute and chronic malnutrition have been associated with elevated risk of postsurgical wound complications, adding morbidity and cost to patients and their families. To study the association between demographic factors, including insurance type, race, and median neighborhood income (MNI), and malnutrition in patients with CLP. Study Design: Retrospective cohort study. Methods: Retrospective review was performed in patients undergoing their first cleft-related surgery at a large tertiary pediatric hospital from 2006 to 2018. Demographic data, weight and height at surgery, type of insurance, race, and primary residential address were collected. Geocoded information on MNI was generated using patient address. World Health Organization Z-scores for weight-forage (WFA) and height-forage (HFA) were used as proxies for acute and chronic malnutrition, respectively. Linear regression models were generated to analyze the relationship of insurance type, race, and MNI on WFA and HFA Z-scores. Results: About 313 patients met inclusion criteria. Increasing MNI predicted increasing WFA Z-score (0.05 increase in WFA per $1,000 increase, P = .047) as well as HFA Z-score (0.09 increase in HFA per $1,000 increase, P = .011). The effect of MNI was not independently modified by race for either WFA (P = .841) nor HFA (P = .404). Race and insurance type did not predict WFA or HFA. Conclusions: Lower MNI is a significant independent risk factor for acute and chronic malnutrition in children with CLP. Combined with previous investigation linking malnutrition to surgical outcomes in this population, this offers a target area for intervention to improve patient outcomes.

This article conducts a benefit-cost analysis of a child allowance. Through a systematic literature review of the highest quality evidence on the causal effects of cash and near-cash transfers, this article produces core estimates on the benefits and costs per child and per adult of increasing household income by $1000, which can be used for any cash or near-cash program that increases household income. We then apply these estimates to three child allowance proposals, with the main proposal converting the $2000 Child Tax Credit in the federal income tax code into a fully refundable and more generous child allowance of $3600 per child ages 0-5 and $3000 per child ages 6-17, as enacted for 1 year in the American Rescue Plan. Aggregate costs and benefits are estimated via micro-simulation. Our estimates indicate that making the $2000 Child Tax Credit fully refundable and increasing benefits to $3000/$3600 would cost $97 billion per year and generate social benefits of $929 billion per year. Sensitivity analyses indicate that the results are robust to alternative assumptions and that each of the three child allowance proposals produces a very strong to an extraordinarily strong return for the U.S. population.

We build a consistent measure of male and female work for the US for the period 1880-2019-encompassing intensive and extensive margins-by combining data from the US Census and several early sources. The resulting measure of hours, including paid work as well as unpaid work in family businesses, displays an asym-metric U-shape for women, with a modest decline up to mid-20th century and a sustained rise afterwards. For men, hours fall throughout the sample period. We empirically and theoretically relate these trends to the process of structural transformation , and namely the reallocation of labour across agriculture, manufacturing and services, and the marketization of home production. We propose a multisector model of the economy with uneven productivity growth, income effects, and consumption complementarity across sectoral outputs. At early stages of development, declining agriculture leads to rising services (both in the market and the home) and leisure, implying a fall in market work for both genders. At later stages of development , structural transformation reallocates labor from manufacturing into services, and a large service economy implies an important marketization process, progressively reallocating work from home to market services. Given gender comparative advantages, the first channel is more relevant for men, implying a decrease in male hours, and the second channel is more relevant for women, implying an increase in female hours.

Introduction: We investigate the association neighborhood cohesion, as source of social support, has with psychological distress among white, Black, and Latinx lesbian, gay, and bisexual (LGB) individuals, compared to heterosexual individuals in the United States. Method: We estimate zero-order multinomial logistic regression models to assess the likelihood of moderate and severe psychological distress among respondents. Result: In the models accounting for neighborhood cohesion and all other covariates, white, Black, and Latinx lesbian, gay, and bisexual individuals are more likely to meet the criteria for moderate and severe psychological distress than non-LGB people. Conclusion: Neighborhood cohesion has differing impact on psychological distress outcomes by racial/ethnic-sexual orientation groups, but in general provides a greater magnitude of protection against moderate psychological distress for non-LGB groups and a greater magnitude of protection against severe psychological distress for LGB groups.

Purpose: This study examined whether socioeconomic inequalities account for Black/White disparities in: (a) the prevalence of potential risk factors for overdose among adults using cocaine; and (b) national mortality rates for cocaine-involved overdose. Methods: Data from 2162 Non-Hispanic (NH) Black or White adults (26 +) who reported past-year cocaine use in the 2015–2019 National Survey of Drug Use and Health were analyzed to obtain predicted probabilities of potential overdose risk factors by race and sex, using marginal effects via regression analyses, adjusting for age and socioeconomic indicators. Next, National Center for Health Statistics data (for 47,184 NH Black or White adults [26 +] who died of cocaine-involved overdose between 2015 and 2019) were used to calculate cocaine-involved overdose mortality rates by race and sex across age and educational levels. Results: Several potential overdose vulnerabilities were disproportionately observed among NH Black adults who reported past-year cocaine use: poor/fair overall health; cocaine use disorder; more days of cocaine use yearly; hypertension (for women); and arrests (for men). Adjusting for age and socioeconomic indicators attenuated or eliminated many of these racial differences, although predicted days of cocaine use per year (for men) and cocaine use disorder (for women) remained higher in NH Black than White adults. Cocaine-involved overdose mortality rates were highest in the lowest educational strata of both races; nonetheless, Black/White disparities were observed even at the highest level of education, especially for adults ages 50 +. Conclusion: Age and socioeconomic characteristics may account for some, yet not all, of Black/White disparities in vulnerability to cocaine-involved overdose.

While labor unions bargain for more equality among their members and in the general society, little is known about their own compensation practices. Using newly assembled administrative data covering union NHQs for the period 1959- 2016 and almost all U.S. labor unions workers over the period 2000-2016, we show that unions do “as they preach”. They pay wages that are on average almost 20% higher than in comparable U.S. private firms, but much more equally distributed: Gini coefficients are 20% smaller among unions and the share of total earnings ac- cruing to the top 1% of wage earners is twice smaller. We argue that such a low level of inequality, especially at the top, is puzzling because union leaders do have substantial margins to set their own pay. We show that the low level of inequality observed among union employees cannot be accounted for by market-type expla- nations, such as a low dispersion of skills among them, a lower average size in union than non-union firms, or fewer hierarchical levels in these firms. Rather, we provide evidence of the existence of a social norm against high pay in the union sector, which results in low inequality. Our results shed new light on how pay norms and institutions can affect real pay, even in a declining sector where firms have strong incentives to perform well in order to survive.

The discourse of economic development through culture and its applications have increasingly received more attention in geographic academia. However, there has been little insight into how the breakthroughs and paradigms of more successful experiments could be sensitively and carefully used for the benefit of less experienced areas. This paper presents an attempt to rectify this, by proposing the use of tangible data from the United States, a country with extensive experience in cultural development and governance. It presents a piece of that research, in the form of a trial methodology for assessing significant clusters of cultural development and identifying their causes. After briefly overviewing the development of the theory of cultural development and defining some basic terms-artists and their definition for quantitative research, (creative) clusters, and creative cities-, a methodology will be proposed and showcased. It will depend on exploratory spatial analysis and the concept of the "artistic dividend", a method of more directly measuring artists' contributions to their local economies by counting their numbers and aggregating their income. Data will be taken from the American Community Survey for its thematic and spatial detail, with visual artists being used as an example category. The decennial evolution of clusters will also be inspected and displayed. Finally, the methodology's further applications, possible evolutions (use of further literature review and regression methods for discovering factors), and distilled focus (improvements by qualitative methods) will be assessed, for its implementation in the final thesis.

Background: Community socioeconomic deprivation (CSD) may be related to higher oil and natural gas development (OGD) exposure. We tested for distributive and benefit-sharing environmental injustice in Pennsylvania's Marcellus Shale by examining (1) whether OGD and waste disposal occurred disproportionately in more deprived communities and (2) discordance between the location of land leased for OGD and where oil and gas rights owners resided. Materials and Methods: Analyses took place at the county subdivision level and considered OGD wells, waste disposal, and land lease agreement locations from 2005 to 2019. Using 2005-2009 American Community Survey data, we created a CSD index relevant to community vulnerability in suburban/rural areas. Results: In adjusted regression models accounting for spatial dependence, we observed no association between the CSD index and conventional or unconventional drilled well presence. However, a higher CSD index was linearly associated with odds of a subdivision having an OGD waste disposal site and receiving a larger volume of waste. A higher percentage of oil and gas rights owners lived in the same county subdivision as leased land when the community was least versus most deprived (66% vs. 56% in same county subdivision), suggesting that individuals in more deprived communities were less likely to financially benefit from OGD exposure. Discussion and Conclusions: We observed distributive environmental injustice with respect to well waste disposal and benefit-sharing environmental injustice related to oil and rights owner's residential locations across Pennsylvania's Marcellus Shale. These results add evidence of a disparity between exposure and benefits resulting from OGD.

Functional linear regression gets its popularity as a statistical tool to study the relationship between function-valued response and exogenous explanatory variables. However, in practice, it is hard to expect that the explanatory variables of interest are perfectly exogenous, due to, for example, the presence of omitted variables and measurement error. Despite its empirical relevance, it was not until recently that this issue of endogeneity was studied in the literature on functional regression, and the development in this direction does not seem to sufficiently meet practitioners' needs; for example, this issue has been discussed with paying particular attention on consistent estimation and thus the distributional properties of the proposed estimators still remain to be further explored. To fill this gap, this paper proposes new consistent FPCA-based instrumental variable estimators and develops their asymptotic properties in detail. We also provide a novel test for examining if various characteristics of the response variable depend on the explanatory variable in our model. Simulation experiments under a wide range of settings show that the proposed estimators and test perform considerably well. We apply our methodology to estimate the impact of immigration on native wages.

The growth of many industries, neighborhoods, and cities differ with some growing slowly and others experiencing rapid change-i.e., rushes. To explain these differences and illuminate the mechanisms of growth, we develop a model centered on a new trade-off between time-varying fundamentals and time-invariant-but rank-dependent-opportunities. Early growth depends on the opportunities new entities provide, whether from accumulating entrepreneurship human capital in firms, real estate in neighborhoods , or land in cities. Our model can explain the existence of rushes and their size. We provide suggestive empirical evidence on industry growth, neighborhood change, and city growth consistent with the model predictions.

During the past decade, several states have adopted statewide educational attainment goals. Kentucky’s 60x30 goal seeks to credential 60% of the adult population in the state by 2030, bringing Kentucky closer to the national average. Using a difference-in-differences empirical strategy I compare Kentucky to a control group of states from 2010-2019 in order to understand statewide gains in varying levels of postsecondary achievement, and in what subsamples of the population the largest gains have been made. The results indicate that Kentucky has been successful in increasing educational attainment compared to states who have not adopted an educational attainment goal, and that these increases are mostly seen at lower levels of postsecondary education. The results of this analysis also indicate greater equity in postsecondary credential gains for “other race” residents at all attainment levels, and Hispanic residents at the Bachelor’s or higher level in Kentucky relative to the control group of states. Relative to the control group, setting the 60x30 goal did not have an effect on closing the racial attainment gap for Black residents in Kentucky.

This paper shows that the ancestry composition shaped by century-long immigration to the US can explain the current structure of global supply chain networks. Using an instrumental variable strategy, combined with a novel dataset that links firm-to-firm global supply chain information with a US establishment database and historical migration data, we find that the co-ethnic networks formed by immigration have a positive causal impact on global supply chain relationships between foreign countries and US counties. Such a positive impact not only exists in conventional supplier-customer relationships but also extends to strategic partnerships and trade in services. Examining the causal mechanisms, we find that the positive impact is stronger for counties in which more credit-constrained firms are located and that such a stronger effect becomes even more pronounced for foreign firms located in countries with weak contract enforcement. Collectively, the results suggest that co-ethnic networks serve as social collateral to overcome credit constraints and facilitate global supply chain formation.

Objective: To examine the association of social determinants of health (SDOH) on prevalence of stroke in non-elderly adults (<65 years of age). Methods: We used the National Health Interview Survey (2013–2017) database. The study population was stratified into younger (<45 years of age) and middle age (45 to 64 years of age) adults. For each individual, an SDOH aggregate score was calculated representing the cumulative number of individual unfavorable SDOH (present vs absent), identified from 39 subcomponents across five domains (economic stability, neighborhood, community and social context, food, education, and health care system access) and divided into quartiles (quartile 1, most favorable; quartile 4, most unfavorable). Multivariable models tested the association between SDOH score quartiles and stroke. Results: The age-adjusted prevalence of stroke was 1.4% in the study population (n=123,631; 58.2% (n=71,956) in patients <45 years of age). Young adults reported approximately 20% of all strokes. Participants with stroke had unfavorable responses to 36 of 39 SDOH; nearly half (48%) of all strokes were reported by participants in the highest SDOH score quartile. A stepwise increase in age-adjusted stroke prevalence was observed across increasing quartiles of SDOH (first, 0.6%; second, 0.9%; third, 1.4%; and fourth, 2.9%). After accounting for demographics and cardiovascular disease risk factors, participants in the fourth vs first quartile had higher odds of stroke (odds ratio, 2.78; 95% CI, 2.25 to 3.45). Conclusion: Nearly half of all non-elderly individuals with stroke have an unfavorable SDOH profile. Standardized assessment of SDOH risk burden may inform targeted strategies to mitigate disparities in stroke burden and outcomes in this population.

The facts, not the fiction, of America’s immigration experience. Immigration is one of the most fraught, and possibly most misunderstood, topics in American social discourse—yet, in most cases, the things we believe about immigration are based largely on myth, not facts. Using the tools of modern data analysis and ten years of pioneering research, new evidence is provided about the past and present of the American Dream, debunking myths fostered by political opportunism and sentimentalized in family histories, and draw counterintuitive conclusions, including...

In the first part of this work, we propose a novel efficient sampling method for measurement-constrained data. Under “measurement constraints,” responses are expensive to measure and initially unavailable on most of records in the dataset, but the covariates are available for the entire dataset. Our goal is to sample a relatively small portion of the dataset where the expensive responses will be measured and the resultant sampling estimator is statistically efficient. Measurement constraints require the sampling probabilities can only depend on a very small set of the responses. A sampling procedure that uses responses at most only on a small pilot sample will be called “response-free.” We propose a response-free sampling procedure (OSUMC) for generalized linear models (GLMs). Using the A-optimality criterion, i.e., the trace of the asymptotic variance, the resultant estimator is statistically efficient within a class of sampling estimators. We establish the unconditional asymptotic distribution of a general class of response-free sampling estimators. This result is novel compared with the existing conditional results obtained by conditioning on both covariates and responses. Under our unconditional framework, the subsamples are no longer independent and new martingale techniques are developed for our asymptotic theory. We further derive the A-optimal response-free sampling distribution. Since this distribution depends on population level quantities, we propose the Optimal Sampling Under Measurement Constraints (OSUMC) algorithm to approximate the theoretical optimal sampling. Finally, we conduct an intensive empirical study to demonstrate the advantages of OSUMC algorithm over existing methods in both statistical and computational perspectives. We find that OSUMC’s performance is comparable to that of sampling algorithms that use complete responses. This shows that, provided an efficient algorithm such as OSUMC is used, there is little or no loss in accuracy due to the unavailability of responses because of measurement constraints. In the second part of this work, we develop uniform inference methods for the conditional mode based on quantile regression. Specifically, we propose to estimate the conditional mode by minimizing the derivative of the estimated conditional quantile function defined by smoothing the linear quantile regression estimator, and develop two bootstrap methods, a novel pivotal bootstrap and the nonparametric bootstrap, for our conditional mode estimator. Building on high-dimensional Gaussian approximation techniques, we establish the validity of simultaneous confidence rectangles constructed from the two bootstrap methods for the conditional mode. We also extend the preceding analysis to the case where the dimension of the covariate vector is increasing with the sample size. Finally, we conduct simulation experiments and a real data analysis using U.S. wage data to demonstrate the finite sample performance of our inference method. The supplemental materials include the wage dataset, R codes and an appendix containing proofs of the main results, additional simulation results, discussion of model misspecification and quantile crossing, and additional details of the numerical implementation. In the third part of this work, we develop a multi-round aggregated one-step estimator and a scalable bootstrap method for distributed sparse least absolute deviation (LAD) regression with high-dimensional covariates. The proposed one-step estimator is based on multi-round distributed quantile regression and linear regression estimators. We derive convergence rates and sparsity properties of the new multi-round estimators and show that our multi-round one-step estimator requires less restrictive sample complexity than the one-shot aggregation for valid inference. We also develop a novel pivotal bootstrap for simultaneous inference that is scalable to the distributed setting. Building on high-dimensional Gaussian approximation techniques, we establish the validity of simultaneous confidence rectangles constructed from the pivotal bootstrap. Finally, we conduct numerical experiments using the simulated data, which demonstrate encouraging performance of our methods.

Poverty scholarship in the United States is increasingly reliant upon the Supplemental Poverty Measure (SPM) as opposed to the Official Poverty Measure of the United States for research and policy analysis. However, the SPM still faces several critiques from scholars focused on poverty in nonmetropolitan areas. Key among these critiques is the geographic adjustment for cost of living employed in the SPM, which is based solely upon median rental costs and pools together all nonmetropolitan counties within each state. Here, we evaluate the current geographic adjustment of the SPM using both microdata and aggregate data from the American Community Survey for 2014–2018. By comparing housing costs, tenure, and commuting, we determine that median rent is likely an appropriate basis for geographic adjustment. However, by demonstrating the wide variability between median rents of nonmetropolitan counties within the same state, we show that the current operationalization of this geographic adjustment could be improved through the use of more-specific categories such as metropolitan adjacency or Rural Urban Continuum Codes.