Does personal and family history influence legislative behavior in democracies? Linking members of Congress to the census, we observe countries of birth for members, their parents, and their grandparents, allowing us to measure ancestry for the politicians in office when American immigration policy changed dramatically, from closing the border in the 1920s to reshaping admittance criteria in the 1960s. We find that legislators more proximate to the immigrant experience support more permissive immigration legislation. A regression discontinuity design analyzing close elections, which addresses selection bias and holds district composition constant, confirms our results. We then explore mechanisms, finding support for in-group identity in connecting family history with policymaking. Holding fixed family history, legislators with more visible indicators of immigration based on surnames are even more supportive of permissive immigration legislation. However, a common immigrant identity can break down along narrower ethnic lines when restrictive legislation targets specific countries. Our findings illustrate the important role of personal background in legislative behavior in democratic societies even on major and controversial topics like immigration and suggest lawmakers' views are informed by experiences transmitted from previous generations.

Nowadays, machine learning is playing a crucial role in harnessing the power of the massive amounts of data that we are currently producing every day in our digital world. With the booming demand for machine learning applications, it has been recognized that the number of knowledgeable data scientists can not scale with the growing data volumes and application needs in our digital world. In response to this demand, several automated machine learning (AutoML) techniques and frameworks have been developed to fill the gap of human expertise by automating the process of building machine learning pipelines. In this study, we present a comprehensive evaluation and comparison of the performance characteristics of six popular AutoML frameworks, namely, Auto-Weka, AutoSKlearn, TPOT, Recipe, ATM and SmartML across 100 data sets from established AutoML benchmark suites. Our experimental evaluation considers different aspects for its comparison including the performance impact of several design decisions including time budget, size of search space, meta-learning and ensemble construction. The results of our study reveal various interesting insights that can significantly guide and impact the design of AutoML frameworks.

As carbon emissions reach unprecedented levels and threats of climate change mount, countries all over the world are racing to implement climate change mitigation policies to limit their carbon footprint and minimize climate risk. This dissertation explores the macroeconomic effects of climate change mitigation policies, namely carbon taxes and green energy investments. The first chapter, co-authored with Tiago Cavalcanti and Cezar Santos, evaluates the aggregate and distributional effects of climate change mitigation policies using a multi-sector equilibrium model with intersectoral input–output linkages and worker heterogeneity calibrated to different countries. The introduction of carbon taxes leads to changes in relative prices and inputs reallocation, including labor. For the United States, reaching its original Paris Agreement pledge would imply at most a 0.6% drop in output. This impact is distributed asymmetrically across sectors and individuals. In the US, workers with a comparative advantage in dirty energy sectors who do not reallocate suffer a welfare loss 12 times higher than workers in non-dirty sectors, but constitute less than 1% of the labor force. The second chapter estimates the local multiplier of spending in green energy in the United States. I construct a novel state-level dataset, and isolate the exogenous variation in green energy spending by exploiting the institutional characteristics of the green budget allocation by the Department of Energy (DoE). I find that a $1 increase in green investment increases state-level output by $1.1 contemporaneously, and up to $4.2 within two years of implementation. These estimates are large in comparison to the findings of the literature on public infrastructure multiplier, or the multiplier of non-green investments by DoE. The third chapter builds on the second chapter and provides an empirical and a theoretical breakdown of the local green multiplier in the United States to understand why it is large. Empirically, I find that green energy spending has significant effects on state sectoral output, state-level employment, and state-level investment. Quantitatively, I contrast green and non-green multipliers by specifying an open economy New Keynesian model with public capital, where each US state is an open economy within a fiscal and monetary union. I calibrate the public capital to green and non-green energy using a transaction-level dataset on awards by the Department of Energy. Model-based counterfactual experiments suggest that 86% of the difference between the green and non-green multipliers is explained by the initial stock of public capital in each energy type. As green public capital is further away from the steady-state, the marginal productivity of investment is higher in the short-run, leading to higher multipliers relative to investment in non-green public capital.

This dissertation research consists of three essays on healthcare economics and policy analysis. Chapter 1 investigates and explains the failure of a proposition on limiting dialysis clinic profits in California in 2018. The proposition would have required dialysis clinics to issue refunds to patients or their payers for revenue that exceeds 115% of the direct cost of treatment. In this chapter, a conceptual framework of how voters weigh costs and benefits is developed and two different empirical approaches, simple OLS and Double Post LASSO, are employed to identify key determinants of the voting outcome. The empirical results suggest that counties with the presence of the giant dialysis chain, DaVita, are 5% less likely to support the proposition. Chapter 2 estimates the effect of retirement on weight change and prevalence to obesity among different groups of people using 13 waves of Health and Retirement survey data (RAND HRS) from 1992 to 2016. Retirees and working people are very different in their characteristics, especially, physical health. Retirees normally have poorer health outcomes compared to working people due to age and health conditions. Thus, possible selection bias and confounding bias need to be addressed when examining the causal effect of retirement on weight outcomes and prevalence to obesity. Fuzzy Regression Discontinuity Design (Fuzzy RDD) and Propensity Score Matching (PSM) approaches are employed in this chapter to overcome the bias issues. The results suggest that retirement does not have a significant effect on weight outcomes when controlling for selection and confounding biases. Chapter 3 studies the impact of the Affordable Care Act (ACA) on the labor market. The major healthcare reform sparks a debate on whether it could harm employment. The most controversial policy in the ACA is its employer mandate, which requires all employers with 50 or more full-time employees to provide their workers health insurance benefits. The new mandate was predicted to cause a decline in labor demand, hence, put workers into unemployment or involuntary part-time employment. The ACA’s Medicaid expansion was also predicted to cause a decline in labor demand. In this chapter, I use pre-ACA regional uninsured rate as approximation for the region’s degree of exposure the ACA and find that the ACA has positive impact on employment, specially for small businesses, without major obstruction to the country’s business structure.

This paper examined geographic patterns of changes in the density of Supplemental Nutrition Assistance Program (SNAP) applications at the zip code level in New Orleans, LA in the immediate aftermath of the COVID-19 pandemic (March–May 2020), compared to pre-pandemic times (March–May 2019). All zip codes analyzed experienced increases in SNAP application density, ranging from 25% to 360%. While disadvantaged zip codes had higher SNAP application densities at baseline, they experienced a comparatively lower increase across time. Results highlight the staggering need for food assistance as a result of the COVID-19 pandemic, including in areas with historically low need.

There are few concentrated studies on wage inequality across local labor markets at the city or metropolitan level. This paper studies the changes in wage inequality among 170 metropolitan areas by using micro-level data from the U.S. Census and American Community Survey from 1980 to 2019. We propose that shifts in the relative demand for “college-educated” or “college equivalent” workers have been persistent in both temporal and spatial dimensions; and that this persistence has contributed to the increase in wage inequality along with the rise in managerial employment. Using fixed-effects models, we find that on average, changes in managerial intensity between 1980 and 2019 accounts for 6.9% of the change in wage inequality across U.S. labor markets.

Despite the widely known preventive interventions, the dyad of acute respiratory infections (ARI) and diarrhoea remain among the top global causes of mortality in under– 5 years. Studies on child morbidity have enormously applied “traditional” statistical techniques that have limitations in handling high dimension data, which leads to the exclusion of some variables. Machine Learning (ML) models appear to perform better on high dimension data (dataset with the number of features p (usually correlated) larger than the number of observations N). Using Uganda’s 2006–2016 DHS pooled data on children aged 6–59 months, I applied ML techniques to identify rural-urban differentials in the predictors of child’s diarrhoea and ARI. I also used ML to identify other omitted variables in the current child morbidity frameworks. The predictors were grouped into four categories: child characteristics, maternal characteristics, household characteristics and immunisation. I used 90% of the datasets as a training sets (dataset used to fit (train) a prediction model), which were tested or validated (dataset (pseudo new) used for evaluating the performance of the model on a new dataset) on 10% and 30% datasets. The measure of prediction was based on a 10-fold cross-validation (resampling technique). The gradient-boosted machine (ML technique) was the best-selected model for the identification of the predictors of ARI (Accuracy: 100% -rural and 100%-urban) and diarrhoea (Accuracy: 70%-rural and 100%-urban). These factors relate to the household’s structure and composition, which is characterised by poor hygiene and sanitation and poor household environments that make children more suspectable of developing these diseases; maternal socio-economic factors such as education, occupation, and fertility (birth order); individual risk factors such as child age, birth weight and nutritional status; and protective interventions (immunisation). The study findings confirm the notion that ARI and diarrhoea risk factors overlap. The results highlight the need for a holistic approach with multisectoral emphasis in addressing the occurrence of ARI and diarrhoea among children. In particular, the results provide an insight into the importance of implementing interventions that are responsive to the unique structure and composition of the household. Finally, alongside traditional models, machine learning could be applied in generating research hypotheses and providing insight into the selection of key variables that should be considered in the model.

The action research project aimed to implement a counseling model to equip clients experiencing anxiety with a holistic health approach. Anxiety is often treated by prescribed anxiolytics, although many women seeking counseling at Madison Counseling have not been formally diagnosed with an anxiety disorder. The study's goal was to provide participants with physical, mental/emotional, spiritual, and relational tools that can aid the participants in the reduction or alleviation of anxiety. The proposed model incorporated empirically proven techniques that balance the women's self to reduce symptomology by providing a healthy alternative to dealing with daily anxiety and stressors, thus reducing the need for women to think they need anxiety medication. Cognitive Behavioral Techniques incorporated with Scripture and meditation were implemented to counter distorted and irrational thinking. Information obtained from this study influences the field by providing a holistic health approach encompassing all aspects of the person, including the body, spirit, and soul, that can be utilized in the Western church and Christ-based counseling as an alternative or supplement to medication. Data were collected by a questionnaire, Likert scales, and information obtained through self-reports and counseling sessions. All five participants showed improvement from pre and post Likert scales by learning and implementing the alternative skills acquired through the Rocket City Women's Study Guide and counseling.

The traditional approach to the quantitative study of segregation is to employ indices that are selected by "desirable properties". Here, we detail how information theory underpins entropy-based indices and demonstrate how desirable properties can be used to systematically construct models of segregation. The resulting models capture all indices which satisfy the selected properties and provide new insights, such as how the entropy index presumes a particular form of intergroup interactions and how the dissimilarity index depends on the regional composition. Additionally, our approach reveals that functions, rather than indices, tend to be necessary mathematical tools for a comprehensive quantification of segregation. We then proceed with exploratory considerations of two-group residential segregation, finding striking similarities in major U.S. cities, subtle segregation patterns that correlate with minority group diversity, and substantive reductions in segregation that may be overlooked with traditional approaches. Finally, we explore the promise of our approach for segregation forecasting.

More than 40 million immigrants live in the United States (U.S.), comprising about 13.7% of the population [1], and the U.S. Census Bureau projects that the foreign-born population will grow to around 20% by 2060 [2]. The DMV region has become a magnet for immigrants. Since rapid demographic growth amongst immigrants in the DMV is projected [3], it is crucial to understand immigrant residential and mode choices. The DMV metropolitan area has recently emerged as an immigrant gateway. Singer describes six types of immigrant gateways: Former, Continuous, Post-World War II, Emerging, Re-Emerging, and Pre-Emerging [4].

This paper examines the quality of quarterly records on work hours collected from employers in the State of Washington to administer the unemployment insurance (UI) system, specifically to determine eligibility for UI. We subject the administrative records to four “trials,” all of which suggest the records reliably measure paid hours of work. First, distributions of hours in the administrative records and Current Population Survey outgoing rotation groups (CPS) both suggest that 52–54% of workers work approximately 40 hours per week. Second, in the administrative records, quarter-to-quarter changes in the log of earnings are highly correlated with quarter-to-quarter changes in the log of paid hours. Third, annual changes in Washington's minimum wage rate (which is indexed) are clearly reflected in year-to-year changes in the distribution of paid hours in the administrative data. Fourth, Mincer-style wage rate and earnings regressions using the administrative data produce estimates similar to those found elsewhere in the literature.

In the 1960s, two landmark statutes-the Equal Pay and Civil Rights Acts-targeted the long-standing practice of employment discrimination against U.S. women. In their aftermath, the gender gap in median earnings among full-time, full-year workers remained stable for 15 years, leading many scholars to conclude the legislation was ineffectual. This paper revisits this conclusion using variation in legislative incidence across states and occupation-industry-state job classifications. We find that women's wages grew by 4-12 percent more on average in places or jobs where the legislation was more binding, with the effects concentrated among the lowest-wage employees. We find no evidence of short-term changes in employment but some suggestive evidence that firms reduced their hiring of women in the long-term.

The negative environmental, health, and social effects arising from U.S. military action in communities both domestically and abroad suggest that the military represents an understudied institutional source of environmental injustice. Moreover, scholars and activists have long argued that the state is an active or a tacit contributor to environmental inequality, thus providing an opportunity to link U.S. military activity with approaches to the state developed under critical environmental justice. We build on these literatures to ask: Does the presence of domestic military facilities significantly increase carcinogenic risks from air toxics? And do communities of color face additional military-associated carcinogenic risks? Multilevel analyses reveal that locales in closer proximity to a military facility and those exposed to greater military technological intensity, independent of each other, experience significantly higher carcinogenic risk from air toxics. We find that proximity to military facilities tends to intensify racial and ethnic environmental inequalities in exposure to airborne toxics, but in different ways for Latinx and Black populations. These results highlight the role of the state in perpetuating racial and environmental expendability as reflected in critical environmental justice and represent an important expansion of nationwide environmental justice studies on contributors to environmental inequality.

This dissertation examines how housing and location choice decisions contribute to spatial and social inequality. The first chapter studies the financial burdens of property taxes on homeowners. Exploiting a reform in Philadelphia that generated changes in property taxes without changing the provision of public goods and services, I measure how sensitive homeowners are to increases in their property tax bills. I find that a $100 increase in property taxes increases property tax delinquency by 3.9% after one year and 7.7% after two years. Home sales also increase by 4.1% after two years. There is no effect on house prices. Further, the financial burdens of property taxes vary considerably by owner race and occupancy status. White owners are more likely to recover from delinquency and sell their homes than Black and minority owners. Owners who live in their houses are also more likely to sell than landlords. The second chapter studies how the time spent commuting to work have evolved over the last four decades for White and Black commuters. In 1980, Black commuters spent 50.3 more minutes commuting per week than White commuters; by 2019, that difference declined to 22.4 minutes. Two factors account for the majority of this decline: Black workers are more likely to commute by transit, and Black workers make up a larger share of the population in cities with long average commutes. Increases in car commuting by Black workers account for nearly one quarter of the decline in the racialized difference in commute times between 1980 and 2019. Today, commute times have mostly converged (conditional on observables) for car commuters in small- and mid-sized cities. However, persistent differences in commute times still remain today in large, segregated, congested, and—especially—expensive cities, revealing the limits of cars in overcoming entrenched racialization of other factors of commuting

This article reviews evidence on the labor market performance of Hispanics in the United States, with a particular focus on the US-born segment of this population. After discussing critical issues that arise in the US data sources commonly used to study Hispanics, we document how Hispanics currently compare with other Americans in terms of education, earnings, and labor supply, and then we discuss long-term trends in these outcomes. Relative to non-Hispanic Whites, US-born Hispanics from most national origin groups possess sizeable deficits in earnings, which in large part reflect corresponding educational deficits. Over time, rates of high school completion by US-born Hispanics have almost converged to those of non-Hispanic Whites, but the large Hispanic deficits in college completion have instead widened. Finally, from the perspective of immigrant generations, Hispanics experience substantial improvements in education and earnings between first-generation immigrants and the second-generation consisting of the US-born children of immigrants. Continued progress beyond the second generation is obscured by measurement issues arising from high rates of Hispanic intermarriage and the fact that later-generation descendants of Hispanic immigrants often do not self-identify as Hispanic when they come from families with mixed ethnic origins.

This paper studies the cyclical behaviour of earnings risk and career changes. We document that the procyclical skewness of the earnings growth distribution arises mostly from the earnings changes of employer and occupation switchers. To uncover their relative importance in driving cyclical earnings changes and whether this arises from changes in the returns to mobility or mobility shocks, we propose a multi-sector business cycle model with on-the-job search and endogenous occupational mobility. Idiosyncratic occupational mobility is the main driver of cyclical earnings risk, mainly due to cyclical shifts in the returns to this mobility. This is the main reason why the sullying effects of recessions are long-lasting. These effects manifest themselves through a collapse of the job ladder and forgone lifetime earnings gains, especially for low-paid workers, and through large lifetime earnings losses among high-paid workers who experience forced occupational mobility and poor re-employment outcomes.

Growing interest in addressing health equity is fueling efforts to better understand the unique challenges faced by Asian American and Native Hawaiian/Pacific Islander (AANHPI) communities. A lack of language accessibility for AANHPI people who are not proficient in English, in particular, could restrict their access to health insurance and quality health care, especially for those who may have to navigate complicated systems to enroll in coverage. In this brief, we assess the extent of limited English proficiency (LEP) among AANHPI nonelderly adults and variation in LEP across AANHPI subgroups, using the most recently available reliable data from the American Community Survey (ACS). We find the following: In 2019, about 3 in 10 (30.8 percent) Asian American adults and 1 in 8 (12.1 percent) Native Hawaiian/Pacific Islander (NHPI) nonelderly adults had LEP, compared with 32.9 percent of Hispanic adults, 3.1 percent of Black adults, and 1.4 percent of white adults. An estimated 14.9 percent of Asian American adults lived in a household in which all members ages 14 and older reported having LEP. AANHPI adults with LEP were more likely than those proficient in English to be noncitizens and to have economic disadvantages such as lower incomes, lower levels of education, and higher uninsurance rates. Whereas almost all Hispanic adults with LEP reported speaking Spanish, the languages AANHPI adults with LEP speak were much more varied, making it more challenging to reach all members of this group with targeted language access interventions. Estimated LEP rates varied widely across subgroups of AANHPI adults; for instance, LEP rates were around 12 percent for NHPI adults, whereas more than 40 percent of Chinese, Bangladeshi, Vietnamese, Nepalese, and Burmese adults had LEP. Overall, AANHPI adults have LEP at rates nearly as high as Hispanic adults. While Spanish is frequently offered in health system settings and materials (e.g., the federal Marketplace for purchasing health insurance coverage or state Medicaid resources) as a language option for people who do not speak English, the diverse languages spoken by AANHPI adults are rarely available. These findings show the need for greater language accessibility for AANHPI adults with LEP in health care and other settings, especially as some pandemic-related health insurance coverage protections expire and the need for clear communication from state health insurance agencies to enrollees continues to grow.

The institutions of marriage and the family have undergone profound changes over recent decades in sub-Saharan Africa, following differentiated paths across and within countries. These changes, however, have not been systematically related to variation in fertility and its decline over time. We use Demographic and Health Survey data from 29 countries in sub-Saharan Africa to examine how nuptiality patterns have changed over the period 1991-2019, and how these changes are associated with changes in the total fertility rate and ideal family size. Using multi-level linear models, we find that our four marriage indicators are all significantly associated with the total fertility rate, but only the associations with polygyny and remarriage are robust to the inclusion of sub-national region fixed effects. Our results suggest that declines in the prevalence of remarriage and polygyny together may account for 17 percent of total fertility decline in the average sub-national region over the period of study. In addition to these results, we find a significant positive association between the prevalence of polygyny and ideal family size, but no association between ideal family size and divorce, remarriage, or the age at first marriage after including fixed effects.

Since the mid-1990s, deportations of noncitizens have risen significantly, from less than 51,000 in 1995 to almost 360,000 in 2019 (U.S. Department of Homeland Security 2020). This increase has been significantly influenced by the devolution of immigration enforcement, as federal immigration authorities have co-opted state- and local-law enforcement agencies into policing immigrants (Coleman 2007). Because the vast majority of recent deportees are Latinx1 men, Latinx families and communities have been devastated by the effects of this substantial increase in deportations (Golash-Boza and Hondagneu-Sotelo 2013), including the loss of much needed income and childcare, and a greater likelihood of psychological trauma and long-term health problems for children (Hacker et al. 2012:7)

Rationale: Minorities often bear the brunt of unequal enforcement of drug laws. In the U.S., Blacks have been disproportionately more likely to be arrested for cannabis possession than Whites despite a similar rate of cannabis use. Decriminalizing cannabis has been argued as a way to reduce racial disparity in cannabis possession arrests. To date, however, the empirical evidence to support this argument is almost non-existent. Objectives: To examine whether cannabis decriminalization was associated with reduced racial disparity in arrests for cannabis possession between Blacks and Whites in the U.S. Methods: Using FBI Uniform Crime Report data from 37 U.S. states, cannabis possession arrest rates were calculated separately for Blacks and Whites from 2000 to 2019. A difference-in-differences framework was used to estimate the association between cannabis decriminalization and racial disparity in cannabis possession arrest rates (Blacks/Whites ratio) among adults and youths. Results: Cannabis possession arrest rates declined over 70% among adults and over 40% among youths after the implementation of cannabis decriminalization in 11 states. Among adults, decriminalization was associated with a roughly 17% decrease in racial disparity in arrest rates between Blacks and Whites. Among youths, arrest rates declined among both Blacks and Whites but there was no evidence for a change in racial disparity between Blacks and Whites following decriminalization. Conclusions: Cannabis decriminalization was associated with substantially lower cannabis possession arrest rates among both adults and youths and among both Blacks and Whites. It reduced racial disparity between Blacks and Whites among adults but not youths. These findings suggested that cannabis decriminalization had its intended consequence of reducing arrests and may have potential to reduce racial disparity in arrests at least among adults.