This study compared the incidence rates of intimate partner homicide (IPH) in Massachusetts by place of birth and race/ethnicity. The analysis involved 340 IPH victim cases between 1994 and 2014. Victims were just under 40 years of age, on average, and most were female (85%), White (67%), and killed by stabbing (34.4%) or firearms (33%). The incidence of IPH victims ranged from 1.3 to 5.6 cases per million people per year between 1994 and 2014 (M = 2.4 per million). Foreign-born individuals had 1.9-fold higher IPH incidence rates of victims relative to U.S.-born individuals. The incidence of IPH-suicide victims was also significantly higher among foreign-born (M = 1.2 per million) relative to U.S.-born individuals (M = 0.4 per million). Furthermore, minority racial/ethnic groups had significantly high incidence rates of victims, with highest incidence of IPH among Blacks. In the full sample there was a 1.9% decline in the incidence of IPH victims per year, which was not statistically significant. The findings highlight the need for culturally specific prevention and intervention strategies to address risks of IPHs and IPH-suicides among diverse groups, particularly among groups most at-risk in Massachusetts such as foreign-born born individuals and racial and ethnic minorities.

Federated learning (FL), as a type of collaborative machine learning framework, is capable of preserving private data from mobile terminals (MTs) while training the data into useful models. Nevertheless, it is still possible for a curious server to infer private information from the shared models uploaded by MTs. To address this problem, we first make use of the concept of local differential privacy (LDP), and propose a user-level differential privacy (UDP) algorithm by adding artificial noise to the shared models before uploading them to servers. According to our analysis, the UDP framework can realize <formula><tex>$(\epsilon_{i}, \delta_{i})$</tex></formula>-LDP for the i-th MT with adjustable privacy protection levels by varying the variances of the artificial noise processes. We then derive a theoretical convergence upper-bound for the UDP algorithm. It reveals that there exists an optimal number of communication rounds to achieve the best learning performance. More importantly, we propose a communication rounds discounting (CRD) method, which can achieve a much better trade-off between the computational complexity of searching and the convergence performance compared with the heuristic search method. Extensive experiments indicate that our UDP algorithm using the proposed CRD method can effectively improve both the training efficiency and model quality for the given privacy protection levels.

This dissertation includes three studies that examine intra-year employment and work hours instability as it relates to children’s economic wellbeing and risk of child maltreatment, with particular attention to the responsiveness of government programs to precarious parental work. In the first paper, I use data from the monthly Current Population Survey to provide the first demographic analysis of month-to-month parental work-hours volatility. I find that higher instability in work hours overall is concentrated in the top and bottom tails of the wage distribution, with parents at the bottom end of the wage distribution experiencing the most volatile hours worked. Young parents and parents who did not graduate high school were consistently estimated to have more variability in their hours worked than older and more highly educated parents. Consistent patterns of racial disparities in work-hours instability did not emerge among the whole sample, but among unmarried parents, Black parents experienced greater work-hours instability than their white counterparts. In my second paper, I use data from the Survey of Income and Program Participation to investigate how intra-year caregiver work hours volatility is related to child poverty, measured through both the official and supplemental poverty measures. I further examine varying degrees of buffering effects of government programs on income declines resulting from work-hours volatility. The results suggest that greater work-hours volatility is related to a higher risk of childhood poverty. In-kind benefits are more effective in buffering household income declines resulting from unstable caregiver work hours. This is followed by tax systems and cash programs. I find that the effectiveness of the near-cash programs is particularly protective for the children of unpartnered single mothers and Black children. Hispanic children also benefit from the transfers’ compensating effects on instability in work hours and, thereby, earnings, but to a lesser degree. The third paper draws on administrative data from a sample of families at-risk for child protective services (CPS) involvement in Wisconsin to investigate the link between earnings instability and CPS involvement. Specifically, it examines whether adequate access to safety-net programs mitigates the likelihood of child welfare involvement when families encounter negative earnings shocks. I find evidence of a link between negative earnings shocks (losses of earnings) and a family’s subsequent CPS involvement. Findings suggest that unfavorable earnings instability is linked to greater CPS-involvement risk, particularly for child abuse (compared to child neglect). In addition, I find that accessing sufficient social benefits as supplemental income when negative earnings shocks occur serves to effectively buffer against the risk of child maltreatment, particularly among families with young children (ages 0–4).

This study examines racial and ethnic differences in self-rated health among rural residents and whether these differences can be explained by socio-demographic characteristics. We used data from the 2011–2017 National Health Interview Survey to assess differences in self-rated health by race and ethnicity among rural residents (living in non-metropolitan counties; n = 46,883). We used logistic regression analyses to estimate the odds of reporting fair/poor health after adjusting for individual socio-demographic characteristics. Non-Hispanic Black and American Indian rural residents reported worse self-rated health than their non-Hispanic White counterparts (25.8% and 20.8% reporting fair/poor health, respectively, vs. 14.8%; p < 0.001). After adjusting for socio-demographic characteristics, disparities remained for non-Hispanic Black rural residents (Adjusted Odds Ratio = 1.55; 95% CI 1.36, 1.76). This study suggests more attention is required to address inequities among rural people and to develop policies to address structural racism and improve the health of all rural residents.

The COVID-19 pandemic significantly affected data collection for the nation’s primary source of household-level labor force data, the Current Population Survey (CPS). In the first four months of the pandemic period (March- June 2020) the average month-over-month nonresponse rate increased by 58 percent, while the size of newly entering cohorts declined by 37 percent relative to the prior 15 months. Together, these factors reduced the overall sample size of the CPS by around 16 percent. We hypothesize that these changes, and significant associated shifts in the demographic composition of the sample, were caused by the cessation of in-person interviewing. Geographic variation in nonresponse over this period does not appear related to variation in COVID case rates across metro areas or states. Using this change in interview method as a natural experiment, we compare labor market outcomes of those who entered the survey before and after the start of the COVID pandemic and find that the change in how individuals were recruited into the survey affected estimates of unemployment and labor force participation. In an exercise generating a counterfactual group of “missing ”respondents, we estimate that, between April and August of 2020, the average unemployment rate was 0.5 to 0.7 percentage points higher, and the labor force participation rate was 0.4 to 0.8 percentage points lower than estimates using the actual sample of respondents. One implication of these results is that web-based surveys, which are increasingly relied on in empirical labor market studies, may fail to reach important subpopulations of the labor market and that reweighting is unlikely to address the selection on outcomes we document.

Salutogenic effects of living near aquatic areas (blue space) remain underexplored, particularly in non-coastal and non-urban areas. We evaluated associations of residential proximity to inland freshwater blue space with new onset type 2 diabetes (T2D) in central and northeast Pennsylvania, USA, using medical records to conduct a nested case-control study. T2D cases (n = 15,888) were identified from diabetes diagnoses, medication orders, and laboratory test results and frequency-matched on age, sex, and encounter year to diabetes-free controls (n = 79,435). We calculated distance from individual residences to the nearest lake, river, tributary, or large stream, and residence within the 100-year floodplain. Logistic regression models adjusted for community socioeconomic deprivation and other confounding variables and stratified by community type (townships [rural/suburban], boroughs [small towns], city census tracts). Compared to individuals living ≥ 1.25 miles from blue space, those within 0.25 miles had 8% and 17% higher odds of T2D onset in townships and boroughs, respectively. Among city residents, T2D odds were 38–39% higher for those living 0.25 to < 0.75 miles from blue space. Residing within the floodplain was associated with 16% and 14% higher T2D odds in townships and boroughs. A post-hoc analysis demonstrated patterns of lower residential property values with nearer distance to the region's predominant waterbody, suggesting unmeasured confounding by socioeconomic disadvantage. This may explain our unexpected findings of higher T2D odds with closer proximity to blue space. Our findings highlight the importance of historic and economic context and interrelated factors such as flood risk and lack of waterfront development in blue space research.

In the United States, little information is available about the number or characteristics of lesbian, gay, bisexual, and transgender (LGBT) immigrants. Using data from the Pew Research Center, the 2017 Gallup Daily Tracking survey, and the U.S. Census Bureau’s American Community Survey, this report estimates the number of LGBT-identified undocumented and documented adult immigrants and foreign-born residents who are part of a same-sex couple. It also describes some of their demographic characteristics. This report updates a 2013 Williams Institute report.1 LGBT foreign-born adults in the United States by documentation status Overall, there are an estimated 1,274,500 LGBT foreign-born adults in the U.S., including 289,700 (22.7%) who are undocumented and 984,800 (77.3%) who are documented. In the U.S., about 128,500 same-sex couples include at least one foreign-born partner/spouse. Approximately 94,400 of these couples are binational, meaning that one partner/spouse is a nativeborn U.S. citizen and the other is either a naturalized citizen or a non-citizen.

Research in the US on the social determinants of reduced physical functioning at older ages has typically not considered physical work conditions as contributors to disparities. We briefly describe a model of occupational stratification and segregation, review and synthesize the occupational health literature, and outline the physiological pathways through which physical work exposures may be tied to long-term declines in physical functioning. The literature suggests that posture, force, vibration, and repetition are the primary occupational risk factors implicated in the development of musculoskeletal disorders, through either acute injuries or longer-term wear and tear. Personal risk factors and environmental and structural work characteristics can modify this association. In the long-term, these musculoskeletal disorders can become chronic and ultimately lead to functional limitations and disabilities that interfere with one's quality of life and ability to remain independent. We then use data on occupational characteristics from the Occupational Information Network (O*NET) linked to the 2019 American Community Survey (ACS) to examine disparities among sociodemographic groups in exposure to these risk factors. Occupations with high levels of these physical demands are not limited to those traditionally thought of as manual or blue-collar jobs and include many positions in the service sector. We document a steep education gradient with less educated workers experiencing far greater physical demands at work than more educated workers. There are pronounced racial and ethnic differences in these exposures with Hispanic, Black, and Native American workers experiencing higher risks than White and Asian workers. Occupations with high exposures to these physical risk factors provide lower compensation and are less likely to provide employer-sponsored health insurance, making it more difficult for workers to address injuries or conditions that arise from their jobs. In sum, we argue that physical work exposures are likely an important pathway through which disparities in physical functioning arise.

Differentially private GANs have proven to be a promising approach for generating realistic synthetic data without compromising the privacy of individuals. Due to the privacy-protective noise introduced in the training, the convergence of GANs becomes even more elusive, which often leads to poor utility in the output generator at the end of training. We propose Private post-GAN boosting (Private PGB), a differentially private method that combines samples produced by the sequence of generators obtained during GAN training to create a high-quality synthetic dataset. To that end, our method leverages the Private Multiplicative Weights method (Hardt and Rothblum, 2010) to reweight generated samples. We evaluate Private PGB on two dimensional toy data, MNIST images, US Census data and a standard machine learning prediction task. Our experiments show that Private PGB improves upon a standard private GAN approach across a collection of quality measures. We also provide a non-private variant of PGB that improves the data quality of standard GAN training.

By age 77 a plurality of women in wealthy Western societies are widows. Comparing older (aged 70+) married women to widows in the American Time Use Survey 2003-18 and linking the data to the Current Population Survey allow inferring the short- and longer-term effects of an arguably exogenous shock—husband’s death—and measuring the paths of adjustment of time use to it. Widows differ from otherwise similar married women, especially from married women with working husbands, by cutting back on home production, mainly food preparation and housework, mostly by engaging in less of it each day, not doing it less frequently. French, Italian, German, and Dutch widows behave similarly. Widows are alone for 2/3 of the time they had spent with their spouses, with a small increase in time with friends and relatives shortly after becoming widowed. Evidence from the European countries shows that widows feel less time stress than married women but are also less satisfied with their lives. Following older women in 18 European countries before and after a partner’s death shows that widowhood reduces their feelings of time pressure. U.S. longitudinal data demonstrate that it increases feelings of depression. Most of the adjustment of time use in response to widowhood occurs within one year of the husband’s death; but feelings of reduced time pressure and of depression persist much longer.

The growth in the number of Latino-owned businesses is now outpacing the growth rate of the Latino population and the start-up growth of all other racial and ethnic groups in the United States. The sociological literature on ethnic entrepreneurship has tended to depict Latino-owned businesses as driven by necessity, operating in largely resourceconstrained environments, and focuses heavily on first generation immigrants. This dissertation explores modern ethnic entrepreneurship including its potential for economic return, the interplay between business strategies and ethnic identity, and the role of institutions in advancing minority entrepreneurship. I use quantitative and qualitative methods to understand what it is like to be an ethnic entrepreneur today including 101 interviews with entrepreneurs and institutional investors. Through census data, I find entrepreneurship to be an alternative and lucrative pathway among high-skilled Latinos. This is reinforced by a subset of interviews with microbusiness and scaled business owners that examines how they make sense of their ethnoracial identity in relation to their business. I find that ethnic strategies can yield benefits as a business strategy but choosing when and how to leverage an ethnic identity is largely reserved for entrepreneurs who have obtained higher education, the later generations, and those operating in professional industries. Interviews with investment professionals shed light on the institutional logics that systematically block minority-owned investment firms’ wealth-generating opportunities. Together, these findings aim to illuminate the experience of modern Latinx entrepreneurs seeking economic mobility outside of formal labor structures, with important implications for increasing diversity and capital allocation to minority-owned businesses.

The Stanford Education Data Archive (SEDA) is part of the Educational Opportunity Project at Stanford University (https:\\edopportunity.org), an initiative aimed at harnessing data to help scholars, policymakers, educators, and parents learn how to improve educational opportunities for all children. SEDA includes a range of detailed data on educational conditions, contexts, and outcomes in schools, school districts, counties, commuting zones, and metropolitan statistical areas across the United States. Available measures differ by aggregation; see Sections I.A. and I.B. for a complete list of files and data. By making the data files available to the public, we hope that anyone who is interested can obtain detailed information about U.S. schools, communities, and student success. We hope that researchers will use these data to generate evidence about what policies and contexts are most effective at increasing educational opportunity, and that such evidence will inform educational policy and practices. The construction of SEDA has been supported by grants from the Institute of Education Sciences, the Spencer Foundation, the William T. Grant Foundation, the Bill and Melinda Gates Foundation, the Overdeck Family Foundation, and by a visiting scholar fellowship from the Russell Sage Foundation. Some of the data used in constructing the SEDA files were provided by the National Center for Education Statistics (NCES). The findings and opinions expressed in the research and reported here are those of the authors alone; they do not represent the views of the U.S. Department of Education, NCES, or any of the aforementioned funding agencies.

The nineteenth and twentieth century saw two waves of state schooling laws. The first wave focused on children to age 14 and the second wave focused on high school. Using the full count 1940 census and a new coding of state laws, this paper provides new estimates of the effects of the first wave of laws. The analysis focuses on cohorts of prime working age between 1910 and 1940. IV estimates of returns to schooling range from 0.067 to 0.077. Quantile IV estimates show the returns were largest for the lowest quantiles, and were generally monotonically decreasing for higher quantiles.

This fact sheet estimates the characteristics and employment numbers of workers in essential industries in California over the period from May 2020 to June 2021. Accurate estimates of essential workers in California are hard to produce for a number of reasons, including significant lags in data collection and reporting, and the instability of the labor market during the last 18 months. We therefore use multiple data sources to produce a range of estimates of the number of workers in California from May 2020 to June 2021 who meet the following three conditions: (1) working in essential industries, (2) not working from home, and (3) working for at least 1,000 hours cumulatively. We refer to workers meeting these three characteristics as active on-site essential workers.

Hong Kong has long been a financial center in Asia and a major global city. Its unique colonial history, its laissez-faire economic policies in the last century, and its booming economic achievements today have attracted immigrants from all over the world. Ranging from high-skilled westerners engaging in the business of education to relatively less wealthy South Asians working in labor-intensive industries, these migrants with diverse socioeconomic backgrounds have contributed to the constant growth of immigrant entrepreneurship in Hong Kong. In this chapter, we dissect the 2016 Hong Kong By-census to document the diverse characteristics of immigrant entrepreneurs in Hong Kong. We will explore the sociodemographic backgrounds of immigrant entrepreneurs and the locational and industrial distribution of their businesses. Additionally, we will delineate the Chinese influence on immigrant entrepreneurs in Hong Kong after the Handover in 1997.

In an article in Perspectives on Psychological Science, Roberts et al. (2020) analyzed racial representation among publications and authors within three fields of psychology. This commentary points to two aspects of that article that may inhibit proper interpretation of the findings. First, Roberts et al. do not present population base rates in U.S. demographics when drawing inferences. Specifically, they interpret their bibliometric analysis as indicating an overrepresentation of White authors in social and developmental psychology with no consideration of base rates. I demonstrate that when base rates are considered, the data show equal representation in the 1980s, 1990s, and 2000s, and White under-representation in the 2010s in both subfields. They also report a correlation between non-White editorship, non-White authorship, and nonWhite participant recruitment, and then suggest that editorship causes an increase in authorship and participant recruitment. They do not consider that demographic change—an overall increase in the proportion of non-Whites in the U.S.—is an alternative explanation for this phenomenon. Lastly, they claim that race is an unpopular topic but a comparative PsycInfo analysis shows race may be one of the most popular topics in psychology. Thus, there are alternative ways to interpret their data.

We provide the first evidence that spatial variation in all-cause mortality risk is capitalized into US housing prices. Using a hedonic framework, we recover the annual implicit cost of a 0.1 percentage-point reduction in mortality risk among older Americans and find that this figure is both relatively low and decreasing in age, from $1,346 for a 67 year old to $246 for an 87 year old. These estimates are one-fifth of the size of comparable estimates found in the labor market, suggesting that the housing market provides an alternative, substantially cheaper channel to reducing mortality risk.

The Great Migration from the US South is a prominent theme in economic history research not only because it was a prime example of large scale internal migration, but also because it had far reaching ramifications for American economic, social, and political change. This article offers a concise review of the literature focused on questions of timing, selection, and migrants’ outcomes, and then offers a more speculative interpretation of how the Great Migration fostered the advancement of Civil Rights. It concludes by pointing out areas where further exploration would be valuable.

We show that a calibrated life cycle two-earner household model with endogenous labor supply can rationalize the extent of consumption insurance against shocks to male and female wages, as estimated empirically by Blundell, Pistaferri, and Saporta-Eksten (2016) in US data. In the model, 35 percent of male and 18 percent of female permanent wage shocks pass through to consumption, compared to the empirical estimates of 32 percent and 19 percent. Most of the consumption insurance against permanent male wage shocks is provided through the presence and labor supply response of the female earner. Abstracting from this private intrahousehold income insurance mechanism strongly biases upward the welfare losses from idiosyncratic wage risk as well as the desired extent of public insurance through progressive income taxation. Relative to the standard one-earner life cycle model, the optimal degree of tax progressivity is significantly lower and the welfare gains from implementing the optimal system are cut roughly in half.

In geographical literature, numerous studies have demonstrated the differences that arise if spatial autocorrelation (SAC) is incorporated into a conventional nonspatial modeling procedure, but little is known about when these differences might be magnified. This study addressed this query by conducting two sets of regression modeling for 561 variables representing housing prices, metropolitan industry, health, crime, education, and (un)employment across various parts of the United States: (1) nonspatial ordinary least squares (OLS) using a set of selected independent variables and (2) spatial regression incorporating spatial filters into the nonspatial OLS as additional independent variables. This incorporation generally improved the model outcomes through decreases in residual autocorrelation and Akaike’s information criterion (AIC). The degree of improvement correlated positively with the level of SAC inherent in the dependent variables. That is, strongly autocorrelated socioeconomic variables underwent greater decreases in residual autocorrelation and AIC than those variables with weaker SAC. The results imply that spatial modeling outcomes are sensitive to and potentially predictable by the level of SAC possessed by dependent variables. Therefore, the degree of SAC present in a socioeconomic variable can serve as a direct indicator of how much improvement a nonspatial OLS will experience if that SAC is properly accounted for.