Pesticides are linked to negative health outcomes, but a causal relationship is difficult to establish due to nonrandom pesticide exposure. I use a peculiar ecological phenomenon, the mass emergence of cicadas in 13 and 17-year cycles across the eastern half of the US, to estimate the short and long-term impacts of pesticides. With a triple-difference setup that leverages the fact that cicadas only damage tree crops and not agricultural row crops, I show that insecticide use increases with cicada emergence in places with high apple production. Exposed cohorts experience higher infant mortality and adverse health impacts, followed by lower test scores and higher dropout rates. I exploit geo-spatial sources of variation and find evidence for pesticide exposure through a water channel. Moderate levels of environmental pollution, not just extreme exposure, can affect human health and development. The study design, which encompasses the entire chemical era of US agriculture since 1950, provides insights into the regulation of pesticides in the US and globally

Timely communication of warnings is essential to protection of lives and properties during tornado out-breaks. Both official and personal channels of communication prove to have considerable impact on the overall outcome. In this study, an agent-based model is developed to simulate warning’s reception–dissemination process in which a person is exposed to, receives, and sends information while interacting with others. The model is applied to an EF5 tornado (EF indicates enhanced Fujita scale) that struck Moore, Oklahoma, in 2013. The parameters are calibrated using publicly available data or a poststorm telephone survey or were derived from literature reviews, expert judgement, and sensitivity analysis. The result shows a reasonable agreement between modeled and observed reception rates for older and younger adults and for different channels, with errors of less than 20 percentage points. Similar agreement is also seen for the average numbers of warning sources. The subsequent simulation indicates that, in the absence of tornado sirens, the overall reception rates for younger and older adults would drop from the baseline by 17 and 6 percentage points, respectively. Concurrently, there is a large decline in the number of warning sources. When a persons’ social network is enlarged, the reception rate for older adults improves from 77% to 80%, whereas for younger adults it stays unchanged. The impact of increased connectivity is more pronounced when people are not watching television or a tornado siren is not available.

Big data, the “new oil” of the modern data science era, has attracted much attention in the GIScience community. However, we have ignored the role of code in enabling the big data revolution in this modern gold rush. Instead, what attention code has received has focused on computational efficiency and scalability issues. In contrast, we have missed the opportunities that the more transformative aspects of code afford as ways to organize our science. These “big code” practices hold the potential for addressing some ill effects of big data that have been rightly criticized, such as algorithmic bias, lack of representation, gatekeeping, and issues of power imbalances in our communities. In this article, I consider areas where lessons from the open source community can help us evolve a more inclusive, generative, and expansive GIScience. These concern best practices for codes of conduct, data pipelines and reproducibility, refactoring our attribution and reward systems, and a reinvention of our pedagogy.

We analyze the relationships of three different types of patented technologies, namely artificial intelligence, software and industrial robots, with individual-level wage changes in the United States from 2011 to 2021. The aim of the study is to investigate if the availability of AI technologies is associated with increases or decreases in individual workers’ wages and how this association compares to previous innovations related to software and industrial robots. Our analysis is based on available indicators extracted from the text of patents to measure the exposure of occupations to these three types of technologies. We combine data on individual wages for the United States with the new technology measures and regress individual annual wage changes on these measures controlling for a variety of other factors. Our results indicate that innovations in software and industrial robots are associated with wage decreases, possibly indicating a large displacement effect of these technologies on human labor. On the contrary, for innovations in AI, we find wage increases, which may indicate that productivity effects and effects coming from the creation of new human tasks are larger than displacement effects of AI. AI exposure is associated with positive wage changes in services, whereas exposure to robots is associated with negative wage changes in manufacturing. The relationship of the AI exposure measure with wage increases has become stronger in 2016–2021 in comparison to the 5 years before.

This study examined the spatiotemporally varying association between mental illness and substance use (MISU) mortality and unemployment, which was rarely explored in previous studies. We employed a spatiotemporal Bayesian hierarchical model to examine how the association varies over space and time after controlling eight covariates on county-level socioeconomic status. MISU mortality rates in 3108 counties in the contiguous U.S. between 2001 and 2014, a period including two economic recessions, were analyzed. The results showed that the association between MISU mortality and unemployment was spatially and temporally varying. Three compelling patterns were found: (1) the persistent high association in the rural Appalachian; (2) the rising high association in the Midwest; and (3) nearly all counties showed a weaker association during the Great Recession period than before and after. We provided possible explanations for such patterns as regional contexts including the insufficiency of healthcare facilities and relative deprivation. Our study demonstrated that the association between MISU mortality and unemployment is spatiotemporally different across counties in the U.S., and the paradoxical “benefit” of the economic recession to the effect of unemployment rates on MISU mortality rates. The results also reiterate the need for locally-focused policies for mitigating the impact of unemployment on mortalities.

Objective: This study seeks to investigate how social, political, and economic factors shaped demand for membership in three major federated fraternal orders in the United States during the late 19th and early 20th centuries: the Independent Order of Odd Fellows, the Knights of Pythias, and the Improved Order of Red Men. Methods: Error-correction, time-series-cross-sectional models are estimated using state-level data, with the first differences in total membership, new initiations, and suspensions serving as dependent variables. Results: We find evidence that railroads, urbanization, immigration , bank panics, and presidential election years all had significant effects on membership; those effects, however, varied by fraternal order. Conclusion: The demand for American fraternalism was affected by civil society, politics, and the economy in ways that scholars have not previously studied.

Purpose: COVID-19 mortality rates are higher in rural versus urban areas in the United States, threatening to exacerbate the existing rural mortality penalty. To save lives and facilitate economic recovery, we must achieve widespread vaccination coverage. This study compared adult COVID-19 vaccination rates across the US rural-urban continuum and across different types of rural counties. Methods: We retrieved vaccination rates as of August 11, 2021, for adults aged 18+ for the 2,869 counties for which data were available from the CDC. We merged these with county-level data on demographic and socioeconomic composition, health care infrastructure, 2020 Trump vote share, and USDA labor market type. We then used regression models to examine predictors of COVID-19 vaccination rates across the USDA's 9-category rural-urban continuum codes and separately within rural counties by labor market type. Findings: As of August 11, 45.8% of adults in rural counties had been fully vaccinated, compared to 59.8% in urban counties. In unadjusted regression models, average rates declined monotonically with increasing rurality. Lower rural rates are explained by a combination of lower educational attainment and higher Trump vote share. Within rural counties, rates are lowest in farming and mining-dependent counties and highest in recreation-dependent counties, with differences explained by a combination of educational attainment, health care infrastructure, and Trump vote share. Conclusion: Lower vaccination rates in rural areas is concerning given higher rural COVID-19 mortality rates and recent surges in cases. At this point, mandates may be the most effective strategy for increasing vaccination rates.

Background: Middle Eastern (ME) immigrants are one of the fastest-growing groups in the US. Although ME countries have a high burden of atherosclerotic cardiovascular disease (ASCVD), the cardiovascular health status among ME immigrants in the US has not been studied in detail. This study aims to characterize the cardiovascular health status (CVD risk factors and ASCVD burden) among ME immigrants in the US. Methods: We used 2012–2018 data from the National Health Interview Survey, a US nationally representative survey. ME origin, CVD risk factors, and ASCVD status were self-reported. We compared these to US-born non-Hispanic white (NHW) individuals in the US. Results: Among 139,778 adults included, 886 (representing 1.3 million individuals, mean age 46.8) were of ME origin, and 138,892 were US-born NHWs (representing 150 million US adults, mean age 49.3). ME participants were more likely to have higher education, lower income and be uninsured. The age-adjusted prevalence of hypertension (22.4% vs 27.4%) and obesity (21.4% vs 31.4%) were significantly lower in ME vs NHW participants, respectively. There were no significant differences between the groups in the age-adjusted prevalence of ASCVD, diabetes, hyperlipidemia, and smoking. Only insufficient physical activity was higher among ME individuals. ME immigrants living in the US for 10 years or more reported higher age-adjusted prevalence of hypertension, hyperlipidemia, and ASCVD. Conclusions: ME immigrants in the US have lower odds of hypertension and obesity, and of having a suboptimal CRF profile compared to US-born NHWs. Further studies are needed to determine whether these findings are related to lower risk, selection of a healthier ME subgroup in NHIS, or possible under-detection of cardiovascular risk factors in ME immigrants living in the US.

In the first chapter, I explore the role of gender discrimination in children’s investments by parents. I specifically examine whether parents in urban China devote more household resources, specifically education, health, and total expenditures, to girls than to boys. The empirical literature has extensively examined trends in the sex ratio at birth and the effect of gender on the extensive margin of fertility. Much less of the existing literature has explored the impact of gender on the intensive margin of parental inputs mainly because of lack of individual child-level data, especially in urban China. This is an important area for economic research to help disentangle child gender bias and provide insight on the well-being of Chinese children. To answer my research question, I use unique data, Chinese Child Twin Survey (CCTS) that includes family expenditure information for individual children within the family. Estimating the causal effect of child gender on parental investments requires boys and girls to live in families with similar observable and unobservable characteristics. This assumption may be violated in China for two reasons.

This dissertation represents a study of how machine learning can be incorporated into existing econometric causal techniques, with explorations both in the costs and benefits of making that choice. The first chapter explores a simulated instrumental variables setting to evaluate the ease of incorporating unmodified machine learning techniques into the ”first stage“ problem. The first stage of two-stage least squares (2SLS) is a prediction problem—suggesting gains from utilizing ML in 2SLS’s first stage. However, little guidance exists on when ML helps 2SLS—or when it hurts. We investigate the implications of inserting ML into 2SLS, decomposing the bias into three informative components. Mechanically, ML-in-2SLS procedures face issues common to prediction and causal-inference settings—and their interaction. Through simulation, we show linear ML methods (e.g.post-Lasso) work “well,” while nonlinear methods (e.g.random forests, neural nets) generate substantial bias in second-stage estimates—some exceeding the bias of endogenous OLS. This work was performed in conjunction with professors Edward Rubin and Glen Waddell. The chapter author wrote simulation code, excepting the substantial portions used for table creation and to iterate over differing methods, to evaluate and run the methods tested in this chapter, and we iv designed the DGP function based on those found in Belloni, Chen, Chernozhukov, and Hansen (2012).

Introduction: This report examines key trends in commuting among the employed population in New York City between 1990 and 2019. Methods: This report uses the American Community Survey PUMS (Public Use Microdata Series) data for all years released by the Census Bureau and reorganized for public use by the Minnesota Population Center, University of Minnesota, IPUMSusa, (https://usa.ipums.org/usa/index.shtml). See Public Use Microdata Series Steven Ruggles, J. Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek. Integrated Public Use Microdata Series: Version 5.0 [Machine-readable database]. Minneapolis: University of Minnesota, 2021. Discussion: Between 1990 and 2019, most of the employed population of New York City (around 75%) spent between 10 and 60 minutes commuting to work. During this period, workers in New York City also experienced a general increase in travel time to work: the proportion of workers who reported travel times of less than 10 minutes dropped from 16.4% in 1990 to 14.6% in 2019, while the proportion of workers who reported travel times of 60 minutes of more increased from 7.8% in 1990 to 10.9% in 2019. Public transit was the most popular means of transportation among the New York City working population. From 1990 to 2019, almost half of the New York City working population relied on public transit to work, and the share gradually increased from 51.8% in 1990 to 55.6% in 2019. The second most popular means of transportation was by a private vehicle (including auto, truck, van, Taxicab, and motorcycle), but the proportion of people using them dropped from 34.3% in 1990 to 27.5% in 2019. Around 11% of New York City’s working population chose to bike or walk to work, and the proportion remained relatively stable from 1990 (10.9%) to 2019 (11.6%). Working at home gradually gained popularity, with its share of the population rising from 2.5% in 1990 to 4.7% in 2019. These data are pre-COVID-19. Other trends by sex, income, marital status, and poverty status are further analyzed in the report.

This piece challenges two “enabling myths” concerning African American male employment. The myths are-1) that the unemployment rate for African Americans in general, and African American males in particular, has “always” been multiples of the white male unemployment rate, and 2) that African American males have “always” participated in the labor force at a lower rate than white males. These “always” myths enable a belief that nothing can be done to address the current disparities, either because the disparities are baked into the “cultural deficiencies” of African American males as a “vestige of slavery”, or more insidiously that they are a result of a “biological deficiency” that cannot be overcome. These “enabling myths” enable political inaction. I demonstrate that the myths are just thatmyths-and that the disparities are a result of the post-emancipation actions of whites enabled by white labor market power under increased labor market competition during the Great Depression. The outcomes have been sustained in the decades since the Great Depression through causal mechanisms powered by American social structure, not by African American cultural deficiency and certainly not African American biological deficiency.

Objectives: Later adult work attachments and exits are in flux, suggesting the need for understanding both the range of contemporary population-level pathways of work and nonwork and variations by overlapping social locations. We document patterned continuity and change in monthly work attachments and analyze the intersecting effects of age, gender, education, and race/ethnicity. Methods: We capitalize on massive microlevel 16-month panel data from the Current Population Survey from 2008 through 2016 to empirically identify patterned pathways of monthly states: working full-time, long hours, part-time; being self-employed or unemployed; not working because of a disability, due to family care or other reasons, or because one defines oneself as retired. Results: Analyses of 346,488 American women and men aged 50-75 years reveal patterned elasticity in the timing and nature of work attachments in the form of six distinctive pathways. Our intersectional analyses illustrate divergences and disparities: advantages for educated White men, disadvantages for low-educated Black men and women through their early 60s, and intersecting effects of gender, education, and race/ethnicity during the later work course across age groups. We find convergence across social markers by the 70s. Discussion: This research highlights the importance of intersectional analysis, recasting the gendered work course in later adulthood into a framework of even greater complexities within mutually shaping categories of race/ethnicity, class, and age. Older Americans experience patterned, uneven pathways around work and nonwork. We recommend additional scholarship on the dynamics of constrained and disparate choices unfolding across multiple intersecting social locations.

Background: It is unknown whether racial/ethnic disparities exist in surgical utilization for children. The aim, therefore, was to evaluate the odds of surgery among children in the US by race/ethnicity to test the hypothesis that minority children have less surgery. Methods: Cross-sectional data were analyzed on children 0–18 years old from the 1999 to 2018 National Health Interview Survey, a large, nationally representative survey. The primary outcome was odds of surgery in the prior 12 months for non Latino African-American, Asian, and Latino children, compared with non Latino White children, after adjustment for relevant covariates. The National Surgical Quality Improvement Program Pediatric Dataset was used to analyze the odds of emergent/urgent surgery by race/ethnicity. Results: Data for 219,098 children were analyzed, of whom 10,644 (4.9%) received surgery. After adjustment for relevant covariates, African-American (AOR, 0.54; 95% CI, 0.50–0.59), Asian (AOR, 0.39; 95% CI, 0.33–0.46), and Latino (AOR, 0.62; 95% CI, 0.57–0.67) children had lower odds of surgery than White children. Latino children were more likely to require emergent or urgent surgery (AOR, 1.71; 95% CI, 1.68–1.74). Conclusions: Latino, African-American, and Asian children have significantly lower adjusted odds of having surgery than White children in America, and Latino children were more likely to have emergent or urgent surgery. These racial/ethnic differences in surgery may reflect disparities in healthcare access which should be addressed through further research, ongoing monitoring, targeted interventions, and quality-improvement efforts. Level of evidence: II. Type of study: Prognosis study.

In February 2020, the U.S. government began to implement a new Public Charge rule that greatly expands the definition of “public charge” when determining admissibility for legal permanent residency (LPR). The rule seeks to determine not only whether applicants used public benefits in the past, but also whether they are likely to use them in the future. However, predicting future use based on characteristics measured at the time of application, such as English language proficiency and income, is difficult. We evaluate the risk of being deemed inadmissible as well as the likelihood of using public assistance by regional group. Using a sample of recently arrived LPRs in the 2013–2017 American Community Survey, we find that Mexicans/Central Americans face disproportionate risk of being deemed a public charge despite their relatively low public assistance use. This increased risk would likely alter the composition of newly admitted LPRs with relatively fewer Mexican/Central American LPRs.

A growing literature demonstrates that birth order affects educational attainment, but the impact of public policy on sibling inequality remains largely unknown. Using linked historical Census data and a family fixed effects model, we examine the impact of birth order for U.S. boys born during the late 1800s and early 1900s, a period of increased public investment in education. Consistent with evidence from recent cohorts in the U.S. and Western Europe, we find that men's educational attainment declines with birth order. Later-born boys obtain 0.2-0.6 years (3-7 percent) fewer years of education than their firstborn brothers. Among whites, later-born boys also have lower earnings and occupation scores. Next, exploiting variation in compulsory schooling laws across states and time, we show that laws requiring eight or more years of schooling substantially compress birth order gaps in educational attainment between white brothers born outside the South.

Introduction: This report analyzes different demographic cross-sections for cost-burdened households at various times over the study period (2000, 2010, and 2017). Methods: The metro areas include the Public Use Micro Areas (PUMAs) associated with following counties for New York (Rockland, Orange, Westchester, Putnam, Duchess, Nassau, Suffolk, Bronx, Kings, New York, Queens, and Richmond), New Jersey, (Passaic, Bergen, Hudson, Essex, Union, and Middlesex), and Connecticut (Fairfield). Since counties are not identified in public-use microdata from 1950 onward and PUMAs change over time, we used consistent PUMA boundaries from 2000 to 2010 (https://usa.ipums.org/usa-action/variables/CPUMA0010#description_section). For more on this see a discussion here https://forum.ipums.org/t/i-can-see-couple-of-distinct-countyfips-whereas-the-rest-of-them-are-under-0-countyfips-for-minnesota/1585/4 and here https://forum.ipums.org/t/1990-or-2000-puma-equivalency-files/2842. Census microdata are used for both analyses; they include data from the Decennial 2000 5% sample, the 2006-2010 5-Year ACS estimates, and the 2013-2017 ACS 5-Year estimates, hereafter referred to as 2000, 2010, and 2017, respectively. The cross-sectional analysis reports summary statistics across the entire metro region with the individual household head as the unit of analysis. The spatial analysis collapses these observations into Public Use Micro Areas (PUMAs) and, using appropriate household weights, reports descriptive statistics for each geographic area. Data comes from Steven Ruggles, Sarah Flood, Ronald Goeken, Josiah Grover, Erin Meyer, Jose Pacas, and Matthew Sobek. Integrated Public Use Microdata Series: Version 9.0. Minneapolis: University of Minnesota, 2019. http://doi.org/10.18128/D010.V9.0. Discussion: All race and ethnic groups saw increasing levels of rent burden across the study period, with Latino households reporting the highest rates of rent burden compared to every other racial and ethnic group at every point over the study period. Latino households have consistently seen higher rates of rent burden than non-Hispanic whites, non-Hispanic blacks, and non-Hispanic Asians—going from 40.3% in 2000 to 59.2% in 2017. While all Latino subgroups saw increases in rent burden over the study period, Mexican and Dominican-headed households saw the largest increases, moving from the lowest levels in 2000 to the highest levels in 2017. Homeowners saw increases in mortgage burden between 2000 and 2010 but decreases between 2010 and 2017. Non-Hispanic black and Latino populations saw the highest levels of mortgage burden at every point over the study period. While the Latino population reported the highest prevalence of mortgage burden in 2000 and 2010 (53.6% and 44.6% respectively), non-Hispanic black households experienced the highest rate in 2017 (45.2%). Asians and non-Hispanic white households saw the lowest rates among all four groups.

Genetic testing is increasingly used in clinical practice to assist with the diagnosis of genetic diseases and/or provide information about disease risk, and genetic counseling supports patient understanding of test results before and/or after genetic testing. Therefore, access to genetic testing and counseling is important for patient care. Health insurance coverage is a major determinant of access to health care in the United States. Uninsured individuals are less likely to have a regular source of health care than their insured counterparts. Different health insurance types and benefits also influence access to health care. Data on the association of health insurance and uptake of genetic testing and/or counseling for cancer risk are limited. Using data from the National Health Interview Survey, we examined the uptake of genetic testing and/or counseling for colorectal cancer (CRC) risk by health insurance type. We found that only a small proportion of individuals undergo genetic testing and/or counseling for CRC risk (0.8%), even among subgroups of individuals at risk due to family or personal history (3.7%). Insured individuals were more likely to undergo genetic testing and/or counseling for CRC risk, particularly those with Military and Private insurance plans, after adjusting for various demographic, socioeconomic, and health risk covariates. Further investigations are warranted to examine potential disparities in access and health inequities.

Domestic violence is a serious under-reported crime in the United States, especially among immigrant women. The Violence Against Women Act (VAWA) was an attempt to partially address this problem by allowing battered immigrants to petition for legal status without relying on the sponsorship of an abusive U.S. citizen or legal permanent resident spouse. The tougher immigration policy climate may have made immigrant women more vulnerable to domestic violence, as well as more reluctant to report domestic violence to law enforcement. Sanctuary policies, which limit local law enforcement's cooperation with federal immigration authorities, may counteract these effects. After exploiting the temporal and geographic variation in the adoption of interior immigration enforcement and sanctuary policies, we can successfully identify the impact of sanctuary policies, which help boost the rate of VAWA self-petitions. Additionally, we provide suggestive evidence of the channel through which this impact is likely taking place - namely through victims' increased willingness to report cases and leave their abusers. Understanding survivors' responses to immigration policy is crucial given growing police mistrust and immigrants' vulnerability to crime.

The COVID-19 pandemic necessitated a large-scale shift toward working from home, leading to a sea change in the level of remote work likely to continue after the pandemic. Previous research has quantified the importance of home location relative to work location (for example, Kneebone and Holmes, 2015), but to date, little work has been done to show the potential impact of an unwinding of that relationship. This analysis quantifies how many renter households could potentially take advantage of teleworking to buy a home. These renter households at the “tipping point” of homeownership are identified using income, industries, and occupations from the 2018 American Community Survey (ACS) rates of teleworking potential from a Bureau of Labor Statistics (BLS) analysis of the American Time Use Survey 2017–2018 (ATUS) and the Zillow Home Value Index (ZHVI) for the lower third of home values in the for-sale market. This analysis finds that 1.92 million U.S. renter households are on the telework tipping point for homeownership. The highest rates of tipping point households are found in expensive west coast markets. For example, more than one-fourth (25.2 percent) of renter households in San Jose could theoretically afford to buy a home in a less pricey locale if they were able to take advantage of more permanent work-from-home policies. In a few metropolitan areas with an extra-expensive principal city, a sizable share of renter households would have an additional incentive to move to the suburbs—up to 10.4 percent of renter households in the city of San Francisco. Nationwide, Asian renter households have the highest share at the tipping point (9.0 percent), followed by Latinx (5.0 percent), White (4.1 percent), and Black (3.7 percent) renter households. This finding means that the Asian homeownership rate is most likely to have observable increases due to telework. Across metropolitan areas, Black renter households are typically more likely to be at the tipping point (29.0 percent more likely than other racial groups), and Latinx renter households are far less likely (26.2 percent less likely than other racial groups).