This paper studies the relationship between pre-college skills and the gender gap in STEM majors. Using longitudinal data for the United States, I estimate a dynamic discrete choice model of initial and final major choices in which college students sort into majors based on observed characteristics and unobserved ability. More specifically, I distinguish observed test scores from latent ability. I find that math test scores significantly overstate gender gaps in math problem solving ability. Math problem solving ability strongly predicts STEM enrollment and completion for men and women. I further explore the importance of math self-efficacy, which captures students' beliefs about their ability to perform math-related tasks. Math self-efficacy raises both men's and women's probability of enrolling in a STEM major. Math self-efficacy also plays a critical role in explaining decisions to drop out of STEM majors for women, but not for men. The correlation between the two math ability components is higher for men than for women, indicating a relative shortfall of high-achieving women who are confident in their math ability. Lastly, I estimate the returns to STEM enrollment and completion and find large returns for high math ability women. These findings suggest that well-focused math self-efficacy interventions could boost women's STEM participation and graduation rates. Further, given the high returns to a STEM major for high math ability women, such interventions also could improve women's labor market outcomes.

US health surveys consistently report that men and those with higher socioeconomic status (SES) engage in more physical activity than women and lower SES counterparts, using questions that ask about physical activity during leisure time. However, social characteristics such as gender and SES shape understandings of and access to leisure-based physical activity as well as other domains where healthy activity is available – namely house work, care work, and paid work. Thus, the physical activity of US adults may look different when what counts as physical activity expands beyond leisure activity. The current study uses Amazon’s Mechanical Turk platform to conduct a 2-by-2-by-2 factorial experiment that crosses three types of physical activities: leisure, house or care work, and paid work. We find that physical activity questions that prime respondents – that is, ask respondents – to consider house/care work or paid work lead to increased minutes reported of physical activity compared to not priming for physical activity, while asking about leisure is no different from having no physical activity primed. The effect on reported physical activity of priming with house/care work is stronger for women than men, demonstrating support for gendered specialization of time spent in the house and care work domain. The effects on reported physical activity of priming with house/care work and paid work are stronger for those with less education compared to more education, consistent with socioeconomic divisions in access to physical activity in house/care work and employment. This study highlights the contingence of our understanding of the physical activity of US adults on both its measurement in surveys and the social forces which shape understanding of and access to physical activity.

We study the e↵ects of local partisanship in a model of electoral competition. Voters care about policy, but they also care about the identity of the party in power. These party preferences vary from person to person, but they are also correlated within each state or congressional district. As a result, most districts are biassed toward one party or the other (in popular parlance, most states are either ‘red’ or ‘blue’). We show that, under a large portion of the parameter space, electoral competition leads to maximization of the welfare of citizens of the ‘swing district,’ or ‘swing state,’ as the case may be: the one that is not biassed toward either party. The rest of the country is ignored. We show empirically that the US tari↵ structure is systematically biassed toward industries located in swing states, after controlling for other factors. Our best estimate is that the US political process treats a voter living in a non-swing state as being worth 70% as much as a voter in a swing state. This represents a policy bias orders of magnitude greater than the bias found in studies of protection for sale.

We examine the effect of interstate differences in income taxes on commuting times. Our theoretical model introduces a border into a model of an urban area and shows that differences in average tax rates distort commuting patterns, but the sign of the effect depends on whether taxes are residence‐based or employment‐based. Empirically, tax differentials have a large effect on commuting times for affluent households and mobile households. We show that commuting times are a sufficient statistic to measure the spatial welfare effects of tax policy. The model and empirical design can be used by economists to study other policy differences.

While traveling in a city you may pass through a variety of neighborhoods: poor, rich, old, new, dilapidated, or even delightful. How long have these neighborhoods been rich, or poor, or delightful? When and why did that character start? How long does that character persist? This thesis addresses these central questions in urban economics by compiling new data and developing new methods to link neighborhood-level conditions throughout history back to the time of settlement to explore the origins and persistence of variation in neighborhood character. Among the greatest challenges to exploring these questions is a lack of digitized information on neighborhood conditions over time. I overcome this obstacle by compiling the most comprehensive available geospatially reconstructed body of maps and archival documents available, detailing population, housing, and environmental characteristics within New York City and Boston beginning in the early 19th century. This allows the first opportunity to observe environmental conditions within neighborhoods since the formative years of settlement and assess the extent to which those historical conditions explain neighborhood character over time. The central finding is that the environmental desirability at the time of settlement continues to explain contemporary variation in income and housing prices in New York City and Boston; areas that were less desirable at the time of settlement remain relatively less desirable today. I examine several potential mechanisms underlying the persistence of neighborhood character including the entrenchment of ethnic enclaves, polluting manufactures, housing price controls, and even redevelopment patters, but none explain the observed persistence. Only neighborhoods near employment epicenters can break from their historical legacy. The first two chapters examine persistence of neighborhood character in New York City and Boston. The third chapter details the records I helped uncover, collect, and the methods I helped develop to geospatially compile into the Historical Urban Ecological (HUE) Data Set. HUE includes thousands of newly digitized municipal data sets from Baltimore, Boston, Brooklyn, Chicago, Cincinnati, Philadelphia, and New York City (starting with Manhattan Island, through the incorporation of each borough), from the earliest time the cities reported data through 1930. The tabular data primarily capture ward-level health and population characteristics, but also include the block-level sequence of water, sewer, and public transit infrastructure deployment where those data exist. The data also include geographic information system (GIS) ward boundary histories in each city for easy linkage of the tabular ward-level data to the correct geography for a given year. Finally, the HUE street centerline GIS data detail the early street paths within cities and allow accurate reconstruction and geolocation of historical data, including the residential and environmental exposure histories of the Union Army Veterans sample for which the data were designed. Beyond the new findings and data, this work additionally contributes a framework to approach related future research. The data sections explain how to efficiently build an accurate, versatile GIS foundation to link data from a variety of unrelated sources over time. I also explain how to use that geographic foundation to extract and assess information from historical maps and documents. Finally, my results demonstrate how to use the new geographically-linked data in a research framework. It is my hope that this thesis provides a guide for researchers to expand on this work.

We exploit plausibly exogenous changes in the value of reserves in Texas's giant oil fields to determine the impact of crime in Texas counties that have reserves. Texas provides an ideal setting for this research strategy. First, Texas has the largest number of giant oil fields. Second, Texas's giant oil fields possess the greatest remaining oil potential. Third, giant oil fields are dispersed throughout the state. We find that a 1% increase in the value of oil reserves increases murder by 0.16%, robbery by 0.55% and larceny by 0.18%. Using the estimated elasticities, an average increase in the value of oil reserves (26% increase in the value of reserves) results in a 4.15% increase in murder rates, 8% increase in robbery and 4.7% increase in larceny. These effects are not trivial. We explore potential mechanisms that could be driving this increase in crime and find that an increase in the value of local oil reserves improves the local economic conditions, increases the share of young males, and increases the share of individuals residing in group quarters (e.g. temporary worker housing) of its county with no effect on the local economic conditions, demographic changes or crime rates of adjacent counties.

On October 29, 1929 the US stock market crashed and opened the door for one of the most significant events in American history – The Great Depression. Savings were wiped out, banks closed, GDP dropped, and unemployment soared upward. At its peak, unemployment rose to nearly 25%, in 1933. GDP dropped by 29% between 1929 and 1933 (Swanson and Williams 1972). Given the Depression’s harsh economic times we would expect at least two demographic behaviors that would affect the demography of the US and student populations. First, we would expect geographic movement to different areas as individuals looked for work. However, extant research has indicated . . .

Background: While many researchers document the immediate and localized health effects of armed conflicts on combatants are well documented in the literature, less is known about the effects of armed conflict on individuals who have subsequently migrated elsewhere. Objective: This study aims to estimate associations between pre-migration armed conflict in Central America and post-migration health in the United States. Methods: We created a new dataset that combines information on armed conflicts in Central America and immigrant health in the United States. We used ordered probit regressions to estimate age-adjusted associations between pre-migration armed conflict and post-migration health. Findings: The study sample of Central American immigrants included 15,563 females and 16,236 males between the ages 15 and 69. The mean age was 37.2 years (standard deviation, 11.6 years) for females and 35.5 years (standard deviation, 11.2 years) for males. After adjusting for age, pre-migration armed conflict was associated with a 8.6 percentage point decrease in excellent health for females (95% confidence interval, 6.0 to 11.1), and a 7.3 percentage point decrease for males (95% confidence interval, 4.0 to 10.7). Each decade of pre-migration armed conflict was associated a 2.9-percentage point decrease in excellent health for females (95% confidence interval, 2.0 to 3.8) and a 1.6-percentage point decrease for males (95% confidence interval, 0.6 to 2.6). For those individuals exposed to armed conflict, each decade since the most recent armed conflict was associated with a 1.5 percentage point increase in excellent health for females (95% confidence interval, 0.4 to 2.5). For males, the average marginal effect of decades since last conflict was not statistically significant (95% confidence interval, –0.001 to 0.002). Conclusions: Pre-migration armed conflict in Central America is associated with decreases in excellent post-migration health in the United States. The effects of armed conflict are cumulative and fade over time for females.

A method is presented for estimating the “underlying” infant mortality rates for areas with small populations. It is described and illustrated in a case study that estimates infant mortality rates for 2,494 US counties that had less than 1,000 births in 1970. The method’s validity is tested using a synthetic population in the form of a simulated data set generated from a model life table infant mortality rate, representing Level 23 of the West Family Model Life Table for both sexes. The test indicates that the method is capable of producing estimates that represent underlying rates. Although some judgment is needed with the method, it has sufficient transparency that estimates can be replicated. The results support the argument that the method can produce reasonable estimates of underlying infant mortality rates for small populations subject to high levels of stochastic variation.

Word embeddings are a powerful machine-learning framework that represents each English word by a vector. The geometric relationship between these vectors captures meaningful semantic relationships between the corresponding words. In this paper, we develop a framework to demonstrate how the temporal dynamics of the embedding helps to quantify changes in stereotypes and attitudes toward women and ethnic minorities in the 20th and 21st centuries in the United States. We integrate word embeddings trained on 100 y of text data with the US Census to show that changes in the embedding track closely with demographic and occupation shifts over time. The embedding captures societal shifts—e.g., the women’s movement in the 1960s and Asian immigration into the United States—and also illuminates how specific adjectives and occupations became more closely associated with certain populations over time. Our framework for temporal analysis of word embedding opens up a fruitful intersection between machine learning and quantitative social science.

There is a growing trend of buying homes among the single population in the U.S. This trend has been referred to as “Going Solo” and is particularly evident among women who are the focus of our study. In this paper we investigate the hypothesis that homeownership probabilities can be affected by marriage market expectations and pessimistic marriage market expectations may raise home buying probabilities among never married singles. We focus solely on the sub population called the never married single females and our results provide evidence consistent with the above hypothesis. In particular, we find that up to a certain threshold, the probability of homeownership decreases when the marriage market prospect indicator improves and there is evidence of heterogeneity in this relationship across race, education level, age group and motherhood status

We estimate the effect of urban form on driving. We match the best available travel survey for the us to spatially disaggregated national maps that describe population density and demographics, sec- toral employment and land cover, among other things. To address infer- ence problems related to sorting and endogenous density, we develop an estimator that relies on assumption of imperfect mobility and exploit quasi-random variation in subterranean geology. The data suggest that increases in density cause small decreases in individual driving. Apply- ing our estimates to the observed distribution of density and driving in the us suggests that plausible densification policies cause decreases in aggregate driving that are small, both absolutely and relative to what might be expected from gas taxes or congestion charging.

The aim of this study is to analyse the fertility of immigrant women born in the Plurinational State of Bolivia, Paraguay and Peru, residing in the Metropolitan Area of Buenos Aires, and their contribution to the aggregate level of this component in 2001 and 2010. A comparative analysis is presented with respect to native-born female population and women in the countries of origin, based on cohort and period fertility indicators. The results show that immigrants from the countries studied present higher cumulative fertility, earlier reproductive structures and higher levels of period fertility than Argentine women. The relationship with the population of origin is, however, more complex. In this regard, some of the main hypotheses for the interaction between migration and fertility are examined. Finally, although the contribution of bordering immigrants to the birth rate is moderate, their impact on the level of total fertility is irrelevant and does not significantly change the aggregate results.

Despite a slight increase in the labor force participation rate of women age 1855 in the U.S. between 1990 and 2010, the labor force participation rate of military wives in this age cohort fell from 63 to 57%. The goal of this paper is twofold: to document and analyze the labor force participation of military wives between 1990 and 2010, using the U.S. Census and American Community Survey data, and to compare the relationship between migration and labor force participation for military and non-military wives. We find that the primary suspects to explain the widening gap are the repeated migration for military wives, and the deepening of the recession.

This paper reports language differences in poor/fair self-rated health (SRH) among adults from six Hispanic groups in the United States. Data are from the cross-sectional 1997–2013 National Health Interview Survey (NHIS). The total sample of Hispanic adults with valid information for the variables considered in the study (n = 156,374) included Mexican-Americans (Mex-Am; n = 43,628), Mexicans (n = 55,057), Puerto Ricans (n = 14,631), Cubans (n = 8,041), Dominicans (from Dominican Republican, n = 4,359) and Other Hispanics (n = 30,658). We compared percentage of the population that reported poor/fair SRH among Hispanic individuals by language of interview and across origins using bivariate tests of association. Multivariable logistic regression analysis was used to study the odds of reporting poor/fair SRH based on language among the overall population and each group. Among the six Hispanic origins Puerto Ricans (15.92%), Cubans (16.36%) and Dominicans (15.32%) reported poor/fair SRH at higher levels than the overall sample (12.32%). In the logistic regression model adjusting potential covariates, those interviewed in Spanish were at higher odds of reporting poor/fair SRH than those interviewed in English (OR = 1.47, p < 0.0001). In the stratified analysis, Mexican–Americans were the only group where language of interview did not affect the odds of reporting poor/fair SRH. There are differences by Hispanic origin for reporting poor/fair SRH, and also by language of interview. Achieving accurate measurement of health status among Hispanics is a concern for all researchers, in particular those who study differences in health status by race/ethnicity in the United States. Future, research should account for Hispanic background and language of interviews.

Hidden in the shadows of New Hampshire’s opioid epidemic are the children who live with their parents’ addiction every day. They fall behind in school as the trouble at home starts to dominate their lives, they make the 911 calls, they are shuttled about to live with relatives or in foster care, and they face an uncertain future when their parents can no longer care for them. In the United States, one in eight children under age 18, or about 8.7 million, live with at least one parent who has a substance use disorder.1 Although many of these children will not experience abuse or neglect, they are at increased risk for maltreatment and child welfare involvement compared with other children.2 Parents who seek treatment can recover, yet parents using opioids are often using other substances and confronting behavioral health issues that complicate recovery.

Science, Technology, Engineering, and Math (STEM) jobs are a key contributor to economic growth and national competitiveness. Yet STEM workers are perceived to be in short supply. This paper shows that the “STEM shortage” phenomenon is explained by technological change, which introduces new job tasks and makes old ones obsolete. We find that the initially high economic return to applied STEM degrees declines by more than 50 percent in the first decade of working life. This coincides with a rapid exit of college graduates from STEM occupations. Using detailed job vacancy data, we show that STEM jobs changed especially quickly over the last decade, leading to flatter age-earnings profiles as the skills of older cohorts became obsolete. Our findings highlight the importance of technology-specific skills in explaining life-cycle returns to education, and show that STEM jobs are the leading edge of technology diffusion in the labor market.

Background: The civil war between the indigenous Mayans and other Guatemalans lasted for 36 years, killed civilians, decimated villages, and resulted in many refugees. The Guatemalan Peace Agreement of 1996 aimed to alleviate the ongoing conflict. Studies of peace agreements more typically evaluate local political outcomes while neglecting global health outcomes.</p><p class="p2"><strong>Objective: </strong>Our research quantified associations between pre-migration exposure to the peace agreement in Guatemala and the post-migration health status of Guatemalan immigrants in the United States.</p><p class="p2"><strong>Methods: </strong>We used chi-square tests to compare the distribution of health status before and after peace. We used ordered probit regressions to estimate associations between peace in Guatemala and health in the United States, conditional on the observed distributions of age, age squared, age cubed, and linear time trends before and after peace.</p><p class="p2"><strong>Findings: </strong>The study sample included 4,115 female and 5,282 male Guatemalan immigrants between the ages of 15 and 85. The mean age was 38.8 years for females (standard deviation, 14.2) and 35.4 years for males (standard deviation, 12.6). Chi-square tests found statistically significant differences in the distribution of health status before and after the peace agreement, for females (P &lt; .001) and males (P &lt; .001). In unadjusted results, the peace agreement was associated with a 7.3 percentage point increase in excellent post-migration health for females (95% confidence interval, 4.9 to 9.8) and a 6.0 percentage point increase for males (95% confidence interval, 3.8 to 8.2). In adjusted results, we found that the peace agreement was associated with a 6.1 percentage point increase in excellent post-migration health for females (95% confidence interval, 0.8 to 11.4) and a 5.5-percentage point increase for males (95% confidence interval, 1.0 to 10.0).</p><p class="p2"><strong>Conclusions: </strong>The peace agreement in Guatemala was associated with statistically significant improvements in the health status of Guatemalan immigrants to the United States.

The economic wellbeing of the millennial generation, which entered its working-age years around the time of the 2007-09 recession, has received considerable attention from economists and the popular press. This chapter compares the socioeconomic and demographic characteristics of millennials with those of earlier generations and compares their income, saving, and consumption expenditures. Relative to members of earlier generations, millennials are more racially diverse, more educated, and more likely to have deferred marriage; these comparisons are continuations of longer-run trends in the population. Millennials are less well off than members of earlier generations when they were young, with lower earnings, fewer assets, and less wealth. For debt, millennials hold levels similar to those of Generation X and more than those of the baby boomers. Conditional on their age and other factors, millennials do not appear to have preferences for consumption that differ significantly from those of earlier generations.

Jeffersonian yeoman agricultural tradition represents the “pick yourself up by your bootstraps” narrative of United States identity with its emphasis on independence, hard-work, and strong family networks. This Jeffersonian narrative specifically focuses on the patriarchal authority of the white male farmer taming nature and the frontier, ignoring the importance and roles of women, children, and social networks on the farm. My dissertation uses farm diaries and the Census to address these invisible forms of labor largely ignored in the traditional narrative. Andrew Peterson’s diaries described family labor and neighbor labor exchanged with nearby families. While living in a frontier area, exchanged neighbor labor worked with the Peterson household through the 1860s until Andrew’s children were old enough to work in the fields. Neighborhood exchange of labor complemented a low worker to consumer ratio within the Peterson household, and was not simply a frontier or pre-capitalist form of bartering. Farm diaries better describe the work of these invisible groups than the Census, but Andrew still underreported women’s work due to traditional narrative biases. Gendered ideologies and census procedures emphasized norms of separate work spheres and reinforced the traditional agricultural narrative at the expense of these invisible groups. While most of the bias for women occurred in planning by Census officials, enumerator practices and biases resulted in inconsistent reporting for children. Biases such as month of enumeration and sex of the respondent were small but statistically significant for women and children. Other important socio-demographic variables for occupational responses included age, school attendance, marital status, and parental occupation. Availability of new complete count census data allows for measuring kin networks beyond the household. Kin propinquity declined in the United States from 30% in 1790 to 6% by 1940, which closely mirrored long-term declines in agriculture and intergenerational coresidence due to urbanization and industrialization. Kin propinquity was especially high in New England prior to 1840, and Appalachia and Utah after 1850. The convergence in kin propinquity rates for younger and elderly people between 1850 and 1940 were caused by declining fertility, declining mortality, and younger generations leaving the farm with better economic opportunities elsewhere.