Privacy leakage is an important issue for machine learning. Existing privacy-preserving approaches with differential privacy usually allow the server to fully control users’ information. This may be problematic since the server itself may be untrusted, leading to serious privacy leakage. Besides, existing approaches need to choose a fixed number of iterations, so that the total privacy budget is finite. Fine-tuning is a common practice, which needs a lot of opportunities to try. However it is not allowed in the actual environment, because multiple access to user data will bring serious privacy risks. In this paper, we aim to address the problem of achieving privacy preserving with distributed differential privacy. In this scenario, different from the traditional need to disclose some local data to the centralized server, each participant keeps the data locally, to achieve better privacy protection effect. We propose a novel algorithm for privacy-preserving training with adjustable iteration steps by sampling techniques. The validity of the algorithm is verified by theoretical analysis and experimental evaluation on real-world datasets

Theories and evidence suggest that older women may experience unique discrimination for being both old and female in the workplace. To provide a remedy for this type of discrimination – known as intersectional discrimination – legal scholars argue that age and sex discrimination laws must be used jointly and acknowledge intersectional discrimination (age-plus-sex or sex-plus-age discrimination) as a separate cause of action. Nonetheless, in general, courts have declined to do so even though older women are protected under both age and sex discrimination laws. This raises a concern that age discrimination laws may be ineffective, or less effective in protecting older women. I test this implication by estimating the differential effect of age discrimination laws on labor market outcomes between older women and older men. My findings show that age discrimination laws did far less to improve labor market outcomes for older women than for older men. These results may explain the persistent findings of discrimination against older women in the existing literature and support the legal scholars’ argument that older women’s intersectional discrimination must be recognized as a separate cause of action.

Foreign-born doctors are an integral part of the U.S. labour market for doctors. About 29% of physicians, 24% of dentists, 16% of optometrists and 11% of podiatrists are not born in the United States. This study investigates whether, on average, there is a difference between the earnings of foreign-born doctors and their native-born counterparts, all else equal. The American Community Survey Data from 2006 to 2017 and two measures of doctors’ earnings: wage (and salary) income and total income (includes wages/salaries, business, investment income, etc.) are used. The results indicate that all else constant, all foreign-born doctors earn significantly (p < 0.01) less wage than native-born doctors for the first 5 years of their stay in the U.S. The wage gap decreases with their length of stay in the U.S. After 10 years of stay, foreign-born male doctors start to earn significantly more than their native-born counterparts (p < 0.05). However, when total income is considered, all foreign-born doctors earn significantly (p < 0.01) less than their native-born counterparts for up to 20 years of stay in the U.S. The magnitude of the total income gap is larger than that of the wage gap. All results remain robust even after ability is controlled for.

Women continue to face discrimination in the workplace and unequal burdens in non-market work and reproductive health access. This dissertation studies how women make decisions about work and family life in the current policy environment. Chapter 1: Does the introduction of paid family leave in the United States increase fertility? Fertility in California increased by 2.5 percent relative to the rest of the country following the 2004 implementation of a statewide paid family leave. This increase is primarily among higher order (2nd or higher birth parity) births to mothers in their 30s. Chapter 2: Do gender norms influence housework distribution? Immigrants from source countries with more progressive gender norms share housework and childcare more equally between men and women once they immigrate to the U.S., though men primarily spend more time in childcare rather than housework. Immigrants from source countries with less gender equality allocate housework and childcare more traditionally, with women doing a significantly larger share of the housework and childcare. These effects persist into the second generation for men, particularly fathers, but not for women. Chapter 3: Does the type of parent involvement in abortion law differentially impact sexual behavior among minors? Notification laws require abortion providers to notify, via phone or email, a minor’s guardians prior to providing the procedure. Consent laws require a notified signature from a guardian to obtain an abortion. I find a statistically significant 4-7 percent increase in pill use at last intercourse for sexually active females in response to a notification law, and a 5 percent decrease in intercourse in the last 3 months following the implementation of a consent law. Neither effect is statistically distinguishable from the effect of the alternate law, suggesting that using one indicator for either type of law is not obscuring individual effects.

It has long been understood by archaeologists that while cemeteries are built by the living to serve the dead, burial grounds can also serve as significant cultural spaces utilized by and integral to the cultural traditions of the living. The study of cemetery sites is therefore critical to the understanding of many aspects of a given culture. Archaeologists often analyze the cemetery sites in a larger region through the lens of a “deathscape,” a macro-scale analytical tool similar to the anthropological concept of a landscape, but which instead focuses on the various cultural processes associated with death: from illness and dying to mortuary behavior, burial, and memorialization. New York City—including the five boroughs of the Bronx, Brooklyn, Manhattan, Queens, and Staten Island—has been a center of commerce since its establishment as a colonial outpost in the 17th century and its urban development has outpaced many other major American cities. The city has been the site of hundreds upon hundreds of burial places, some of which have remained perpetually preserved and others that have been obliterated and redeveloped with or without the removal of the human remains lying within. While previous attempts have been made to document the cemeteries of New York City from historical or genealogical perspectives, to date, a comprehensive archaeological analysis of the city’s cemeteries has not been completed. This study was therefore completed to better understand the use and reuse of burial space in New York City and to identify patterns in how that portion of its deathscape made up of cemeteries was formed, reshaped, and maintained over time leading to the burial landscape seen in the modern city in the present day. Following a period of intensive documentary research, 527 burial sites were identified within New York City as part of this study of the city’s deathscape. Each cemetery was mapped within a Geographic Information System (GIS) and entered into a database and classified according to a series of characteristics (e.g., location, type, dates of use, current status, etc.). This study examines the cemeteries within the data set in order to better understand several critical aspects of the New York City deathscape, including its initial formation within the context of New York City’s historic period occupation, development, and its increasingly stringent municipal regulation of burial space. The burial sites included in the data set were then compared and contrasted to identify the patterns that governed (and continue to govern) the establishment and use of cemeteries in New York City and the patterns that lead to the selective preservation or obliteration of certain cemeteries and/or the relocation of human remains to new burial sites as the deathscape evolved. This analysis concludes that cemetery obliteration and the removal and relocation of remains were heavily influenced by three factors: institutionalized colonial power structures that continued to govern land use and access to space throughout the post-colonial period and into the present; the intensity of urban development and population growth in New York City; and changing connections between kin networks and social groups to specific spaces and places over time. Finally, this study includes a summary of the growth of archaeology as a profession in New York City and a synthesis of historical and modern archaeological investigations in the region that have contributed to archaeologists’ knowledge of the deathscape. The GIS-based maps documenting the past and present locations of burial grounds are intended to be used as sensitivity maps that can be utilized by archaeologists to protect sites known to be sensitive for human remains from disturbance during future development. Despite the large number of burial sites identified during this study, it is likely that many more remain undiscovered, and there is therefore great potential for the continued analysis of the New York City deathscape in the future as new burial sites are identified and documented.

How does immigration affect incomes in the countries migrants go to, and how do rising incomes shape emigration from the countries they leave? The answers depend on whether people who migrate have higher or lower productivity than people who do not migrate. Theory on this subject has long exceeded evidence. We present estimates of emigrant selection on both observed and unobserved determinants of income, from across the developing world. We use nationally representative survey data on 7,013 people making active, costly preparations to emigrate from 99 developing countries during 2010–2015. We model the relationship between these measures of selection and the income elasticity of migration. In low-income countries, people actively preparing to emigrate have 30 percent higher incomes than others overall, 14 percent higher incomes explained by observable traits such as schooling, and 12 percent higher incomes explained by unobservable traits. Within low-income countries the income elasticity of emigration demand is 0.23. The world’s poor collectively treat migration not as an inferior good, but as a normal good. Any negative effect of higher income on emigration within subpopulations can reverse in the aggregate, because the composition of subpopulations shifts as incomes rise—an instance of Simpson’s paradox.

This dissertation presents three independent essays in applied econometrics and macroeconomics. Chapter 1 studies the research question: How do firms’ productivity and markups respond to energy price shocks? Productivity and markups are two concepts closely related to the production process and welfare in economics. In general terms, these variables help us to understand: (i) how much a firm can produce given the scarcity of resources; and (ii) the gap between the marginal costs of producing goods and the final price that is charged to consumers. I exploit variations in costs due to, for instance, energy price shocks, to understand more about the behavior of these two variables. I proceed to answer this question as follows: First, I state two potential issues in the commonly used proxy-variable technique when employed as an intermediate step to recover markups: circularity and violation to the monotonicity/scalar unobservable assumption(s). I then present a novel structural estimator that overcomes these issues. Second, taking advantage of a panel of Chilean manufacturing plants, I study the relationship between energy prices, markups, and productivity using an instrumental variables research design. I explore a natural experiment, the 2004 Argentine crisis, as a potential source of exogenous variations for electricity prices. I also complement the research design by instrumenting for average variable cost constructing shift-share type instruments from variations in energy prices. Results suggest that a relevant channel through which firms adjust energy shocks is flexible markups. Specifically, my estimates suggest that a 10 percent average variable cost energy cost-shock increase leads to a roughly 3 percent decrease in markups. However, productivity remains unchanged. Chapter 2, focuses on the intersection of macroeconomics and labor economics and studies the following research question: What does hiring discrimination against black Americans imply for unemployment, job finding, and separation rates over the business cycle? Taking as given the empirical evidence suggesting discrimination against black Americans during the hiring process, I research the labor market consequences of such behavior. Because the question involves outcomes that we cannot directly measure or observe in the data, I rely on a the oretical model. I use the model to conduct a counterfactual experiment: whether the labor outcomes of black people would have been different had they had a relatively higher hiring probability. In 2010, Peter Diamond, Dale Mortensen, and Christopher Pissarides won the Nobel Prize in Economics “for their analysis of markets with search frictions.” They contributed to the development of a model that represents the workhorse in macroeconomics for understanding labor market outcomes. Based on this standard search-and-matching model, I introduce a novel modification that allows the model to generate heterogeneity among agents: the combination of endogenous separations with an urn-ball matching function. After calibrating the model to match the U.S. economy’s aggregate labor market statistics, I find that discrimination in the early stages of the hiring process leads to adverse labor outcomes over the business cycle: lower job finding probabilities and higher unemployment volatility for discriminated groups. Conversely, the model does not predict significant differences in separation rates. Finally, chapter 3 lies at the intersection of applied econometrics and the economics of crime. Here I propose an econometric approach to address two significant methodological challenges that emerge when studying aggregated crime figures: (i) allocation of weights, and (ii) underreporting. Thus, the chapter presents a way to aggregate crime variables into one non-linear index for which each relative importance or weight assigned to the different types of crime is determined from the relationship with potential covariates that may affect overall crime. Furthermore, the index deals with potential underreporting issues that characterize police crime datasets by implementing a zero-inefficiency stochastic frontier modeling approach. To illustrate the practical implementation of the crime index, I take advantage of two different datasets: (i) a yearly province-level panel of crime in Canada, during the period 2000-2010; and (ii) a detailed database of police-recorded crimes in Bogot´a, Colombia, over 2010-2018. The former allows researchers to access relevant socioeconomic information, while the latter is representative of a developing country in Latin America with serious crime problems. I find that the estimates in the econometric specification are a good match with previous results in the literature. Furthermore, the crime index provides a good unified global mapping of the evolution of different crime categories over time and locations in these two countries.

This paper analyzes the relationship between oil and gas development and in-migration of workers into boomtowns, taking into account their human capital. Using zero-inflated negative binomial estimation methodology, we find that shale development has differing scale and demand shock impacts on U.S. interregional migration flows. The results demonstrate the heterogeneity of migration responses to shale developments with a disproportionately higher positive effect for medium-high human capital workers with technical degrees or trainings common in the energy industry. Furthermore, labor demand shocks from oil and gas development have a modest association with migration flows, which is contrary to the assumption that natural resource boom is a considerable attraction for migrants. This study highlights the types of human capital gained by oil and gas development areas characterized as being rural and sparsely populated, which can have important implications for the long-run growth and economic resilience of the boomtowns.

The relative importance of racial and class inequality in incarceration in the United States has recently become the subject of much debate. In this paper, we seek to give this debate a stronger empirical foundation. First, we update previous research on racial and class inequality in people’s likelihood of being imprisoned. Then we examine racial and class inequality in people’s risk of having a family member imprisoned or living in a high-imprisonment neighborhood. We find that racial inequality in prison admissions has fallen in the twenty-first century, while class inequality has surged. However, in recent years, Black people with high levels of education and income were more likely than White people with low levels of education and income to experience the imprisonment of a family member or to live in a neighborhood with a high imprisonment rate. These seemingly contradictory conclusions can be reconciled by the fact that class boundaries among Black people are more permeable than they are among White people. Imprisonment in the United States is increasingly reserved for the poor. But because Black people are disproportionately connected to the poor through their families and neighborhoods, racial inequality exceeds class inequality in people’s indirect experiences with imprisonment.

Ban the Box (BTB) policies mandate deferred access to criminal history until later in the hiring process. However, these policies chiefly target public employers. The study is the first to focus on the primary goal of BTB reform, by measuring the impact of BTB policies on the probability of public employment for those with convictions. To execute the analyses, the study uses data from the National Longitudinal Survey of Youth 1997 Cohort (2005–2015) and difference‐in‐difference (DD) estimation. The study finds that BTB policies raise the probability of public employment for those with convictions by about 30% on average. Some scholars argue that BTB policies encourage statistical discrimination against young low‐skilled minority males. The study employs triple‐difference (DDD) estimation to test for statistical discrimination, but uncovers no evidence to support the hypothesis. (JEL J15, J71, J78, K4).

So now we get to the math! Just kidding. Mostly. In this chapter, you are going to learn more about bivariate analysis, or analyzing the relationship between two variables. I don’t expect you to finish this chapter and be able to execute everything you just read about—instead, the big goal here is for you to be able to understand what bivariate analysis is, what kinds of analyses are available, and how you can use them in your research. Take a deep breath, and let’s look at some numbers!

We show that changes in birth order during the baby boom can explain a substantial share of the stagnation and recovery in educational attainment among cohorts born between 1946 and 1974. Combining birth order effects estimated using the Health and Retirement Survey and birth order data from Vital Statistics, we estimate that changes in birth order can explain more than 20 percent of the decline in white male college completion rates among the 1946–1960 cohorts, and more than one third of the rebound among the 1960–1974 cohorts. We also revisit the role of cohort size, finding smaller effects than previously reported.

The economic impacts of the COVID-19 pandemic have been dramatic. Analysts are still working to understand how the pandemic has uniquely impacted different sectors of the economy and different parts of the workforce. In Minnesota, as in other states, the employment impacts have sharply diverged by gender, race, and ethnicity. Through analysis of state-level occupational survey data, state unemployment data, and interviews with community service organizations and unions, this report provides a clearer picture of which Minnesota workers have been most impacted by COVID-19, how they have been impacted, and what state and community economic supports they have relied upon during the dislocations caused by the pandemic. Our analysis reveals clear patterns by which women, men, and different racial and ethnic groups in Minnesota have been distinctly impacted by workplace closures and modifications as a result of COVID-19. The data show a dual vulnerability for women on the whole, and one faced disproportionately by women of color: these workers are concentrated in the essential workforce with high risk of virus transmission and have been particularly vulnerable to layoffs. Other demographic groups, especially white men, are more likely to work in low-risk essential positions or have been able to work from home. This report breaks down these trends in greater depth and suggests why these patterns exist. Moreover, our interviews with community service organization and union leaders provide insight into what has and has not worked to support the most vulnerable workers in the COVID-19 context. Together, these data support a set of policy recommendations that can strengthen our pandemic responses now and in the future.

The diversity of GIS application patterns leads to the demand for multilevel GIS data declassification. For example, Publicly used data must be declassified to hide confidential spatial information. The reversion process is not a common data permutation like the conventional encryption method does. The reverted data should also keep the general geospatial features. Furthermore, when facing different levels of confidentiality, different levels of reversion were needed. In this paper, A declassification and reversion method with controllable accuracy is realized using geometric algebra (GA). The geographical field is expressed as a GA object and the unified representation of the field is further realized. By introducing the rotor operator and perturbation matrix, the declassification methods are proposed for geographic field data, which can progressively revert the features of the field. A geometric algebraic declassification operator is also constructed to realize the unification operations of field features and spatial coordinate. By exploring the space error and space structure characterization of the results, a quantitative performance evaluation is provided. Experiments have shown that the method can carry out effective precision control and has good randomness and a high degree of freedom characteristics. The experimental data show a correlation coefficient of 0.945, 0.923 and 0.725 for the longitude-oriented field data during the low level, medium level and high level declassification, respectively. The algorithm characteristics meet the application needs of geographic field data in data disclosure, secure transmission, encapsulation storage, and other aspects.

Although the US population is aging, the population living in nursing homes has fallen. The decline is the largest among low income older adults. This report explores two possible demographic drivers of this decline: (1) increasing racial and ethnic diversity and (2) increasing life expectancy among men. Using decennial census and American Community Survey data from and nonlinear regression decomposition techniques, I estimate the share of the overall decline in institutional residency attributable to these demographic trends. Additionally, I explore which living arrangements have risen as institutional residence fell and discuss implications for the Supplemental Security Income (SSI) program. I am able to explain 99 percent of the decline in institutional residency, of which changes in racial and ethnic diversity alone explain 19 percent. Medicaid Home and Community Based Care waiver programs alone explain approximately 60 percent. As nursing home residency fell, assisted living rose but not by enough to fully offset the decline in institutional residence and very unequally by race. Co-residence with persons other than a spouse and unmarried partnerships both grew dramatically. Findings indicate increases in community-residence may increase SSI payments and rising rates of co-residence may lead to more complex benefit determinations and greater administrative cost.

In this paper, I document that workers in larger cities have significantly more occupational options than workers in smaller ones. They are able to form better occupational matches and earn higher wages. I also note differences in the occupational reallocation patterns across cities. I develop a dynamic model of occupational choice that microfounds agglomeration economies and captures the empirical patterns. The calibration of the model suggests that better occupational match quality accounts for approximately 35% of the observed wage premium and a third of the greater inequality in larger cities.

This paper combines cross-sectional and longitudinal labor income data to present a comparison between anonymous and non-anonymous growth incidence curves in the United States during the past 50 years. If anonymous growth incidence tend to be upward sloping because of increasing inequality during that period, the same is not true of non-anonymous curves. The latter prove to be flat or non-significantly down- ward sloping, suggesting some neutrality of growth when initial income positions are accounted for. This is true when using either panel data or synthetic panels based on CPS data and one-parameter functional representations of income mobility. Flat non- anonymous curves are observed even in periods of increasing cross-sectional income inequality. Differences between anonymous and non-anonymous curves thus matter for the interpretation of inequality changes, social welfare and policy.

Americans now spend the majority of their food budgets eating outside the home (Economic Research Service 2016). This trend began in the previous century, a reflection of an economy built on longer work hours, multiple jobs, and increasingly precarious work conditions. Eating out is so widespread that restaurant workers make up nearly 10 percent of the private sector workforce, totaling more than fourteen million jobs. In fact, the restaurant industry is growing on pace to surpass manufacturing as the fourth­largest employer by 2020. The industry is resilient and was one o( the few to grow through the Great Recession, quickly bouncing back from a short employment dip. However, employment growth has not meant greater prosperity for workers. Restaurant workers live in poverty at more than twice the rate of the rest of the workforce; they are a plurality of minimum wage workers and more than half of workers earning below minimum wage' (Bureau of Labor Statistics 2018). A segment of the restaurant workforce is subject to a subminimum wage. More than one-third of all restaurant workers live in states where the hourly subminimum wage for tipped workers is only $2.13, and nearly three-quarters Live in states where the subminimum wage falls below the federal minimum of $7 .25. As a result, five of the ten lowest-paying occupations in the country are in the restaurant industry.

Arkansas was a demographic frontier after the U.S. Civil War. Despite marked agricultural land deforestation and development after the 1870s, it remained agrarian well into the twentieth century. We fuse life course and racial state frameworks to examine Black and White women’s settlement in Arkansas over the post-Civil War period (1880-1910). A racial state empowers residents and enacts policies based on race rather than equal citizenship rights. We highlight three institutional domains shaped by racial state policies: productive economies (subsistence, mixed commercialism, and plantation production); stratification on an agricultural ladder (from sharecropping to forms of tenancy to farm ownership); and rules of raced (and gendered) social control . We examine women’s settlement patterns and related outcomes in an institutional context at different life course stages using mixed methods: women’s oral histories and Census data analysis. We find that by 1880 White women and families, less attracted by forces of marketization, had largely migrated to subsistence and mixed commercial subregions. Black women and families, generally desiring to rise on the agricultural ladder to farm ownership, largely migrated to the rich lands found in plantation production counties. Black women in Arkansas could rise but, by 1910, new racial state (Jim Crow) policies more severely limited travel, material resources, and education for tenant farm families, predominantly Black, in the plantation subregion. Commensurate with this, Black women in the plantation subregion had experienced less status mobility on the agricultural ladder, with reduced living standards, by later life.