We develop a measure of workers’ technology exposure that relies only on textual descriptions of patent documents and the tasks performed by workers in an occupation. Our measure appears to identify a combination of labor-saving innovations but also technologies that may require skills that incumbent workers lack. Using a panel of administrative data, we examine how subsequent worker earnings relate to workers’ technology exposure. We find that workers at both the bottom but also the top of the earnings distribution are displaced. Our interpretation is that low-paid workers are displaced as their tasks are automated while the highest-paid workers face lower earnings growth as some of their skills become obsolete. Our calibrated model fits these facts and emphasizes the importance of movements in skill quantities, not just skill prices, for the link between technology and inequality.

This dissertation was a compilation of three, distinct studies assessing effects of various exposures on ultimate cardiovascular outcomes. In Aim 1, we assessed the relative and absolute risk of statin-associated muscle symptoms (SAMS) by exposure to statins of varying intensity (none, moderate, or high) in RCTs with at least 1,000 enrolled patients with planned follow-up of at least 2 years (n=152,461 patients in 24 RCTs). Risk was significantly greater for high compared with moderate intensity statin therapy for any muscle problem (RR=1.04, 95% CI 1.00 to 1.07; I2=0%), myalgia (RR=1.04, 95% CI 1.00 to 1.08; I2=0%, number needed to harm (NNH)=173), attrition due to muscle problems (RR=1.37, 95% CI 1.09 to 1.73, I2=0%, NNH=218) and elevated CK (RR=4.69, 95% CI 2.50 to 8.80; I2=7%, NNH=527). Risk also was significantly higher for high intensity compared with placebo for any muscle problem (RR=1.05, 95% CI 1.01 to 1.09, I2=0%), myalgia (RR=1.13, 95% CI 1.05 to 1.23; I2=0%, NNH=182), attrition due to muscle problems (RR=1.55, 95% CI 1.15 to 2.08, I2=0%, NNH=187) and elevated CK (RR=5.37, 95% CI 2.48 to 11.61; I2=7%, NNH=589). Due to inconsistency of results across sensitivity analyses, estimates were inconclusive for rhabdomyolysis and CK. There were no significant differences in risk between moderate intensity therapy and placebo for all outcomes. In Aim 2, we assessed odds of VTE in a case-control study of women, free of VTE disease at baseline, aged 50-65 who had at least one year of continuous enrollment in the Optum database. Women were assessed for most recent exposure to hormone therapy (HT) – estrogen, progestogen, none, or both – and statin therapy (moderate intensity, high intensity, or none) after adjusting for confounders. Cases were matched, 1:10, by age +/- 2 years and index date of event (+/- 1 month). There were 20,359 cases matched 1:10 to 203,590 controls in the selected sample. Exposure to HT was defined as any prescription in the 60 days prior to index date after review of the literature and homogeneity estimates across exposure groups. Using the same methods, exposure to statins was defined as three or more prescriptions for statins in the last 90 days prior to index date. Odds of VTE were significantly higher for women taking any HT (OR=1.51, 95% CI: 1.43, 1.60) but lower for those exposed to any statins (OR=0.88, 95% CI: 0.84, 0.93). The effect of statin exposure increased in magnitude for those exposed to high intensity statins (OR=0.82, 95% CI: 0.74, 0.90; referent is no exposure) versus moderate intensity statins (OR=0.90, 95% CI: 0.86, 0.95). In Aim 3, we assessed feasibility and possible effectiveness of a novel disease management program implemented at the St. Vincent’s Free Clinic in Galveston. Post-hospitalization heart failure (HF) disease management represents an important area of focus in preventing morbidity, mortality, and excess healthcare costs. Uninsured patients were offered enrollment in the disease management clinic during or immediately following hospitalization for a primary HF diagnosis at University of Texas Medical Branch at Galveston from January 2021 - December 2021. The program included twice-weekly visits with multiple healthcare professionals, including nurses, physicians, occupational therapists, social workers, pharmacists, and counselors. Patients were scheduled for 16 visits (2 months of follow-up) post-hospitalization before returning to usual care. Patients who attended at least the introductory appointment and one follow-upappointment within 30 days of discharge were considered enrolled. Of 59 patients referred, 47 (80%) were enrolled. Just 4 patients (8.5%, 95% CI: 2.5%, 20.5%) were readmitted at 30 days, while 4 of 12 (33%, 95% CI: 13.6%, 61.2%) were readmitted at 30 days in those who did not enroll. Program participants were readmitted significantly less frequently than national readmission rate estimates (23%, p=0.02). Thus, each Aim represents a novel exploration of various interventions and associations with important, clinical outcomes. The methods and findings in this dissertation are foundational to my future career as a physician-scientist.

We estimate the effect of job loss on households' bank account ownership using a novel assembly of data: FDIC-sponsored biennial supplements to the Current Population Survey (CPS), linked to respondents' work history in surrounding months constructed from the basic monthly CPS. We leverage differences in the timing of unemployment spells across respondents to plausibly identify the effect of job loss. Our estimates indicate the effects of job loss are quite large in magnitude. For example, households that experienced a job loss in the months leading up to the FDIC survey are about 18 percentage points more likely to be unbanked than households that lost a job in the subsequent year. This effect is roughly three-quarters of the sample mean unbanked rate among the lower-income, renter households that we study. Job loss also leads to increased use of other transaction products and services that might substitute for a bank account, including prepaid cards, check cashing, and money orders.

End-to-end (E2E) artificial intelligence (AI) pipelines are composed of several stages including data preprocessing, data ingestion, defining and training the model, hyperparameter optimization, deployment, inference, postprocessing, followed by downstream analyses. To obtain efficient E2E workflow, it is required to optimize almost all the stages of pipeline. Intel ® Xeon ® processors come with large memory capacities, bundled with AI acceleration (e.g., Intel Deep Learning Boost), well suited to run multiple instances of training and inference pipelines in parallel and has low total cost of ownership (TCO). To showcase the performance on Xeon processors, we applied comprehensive optimization strategies coupled with software and hardware acceleration on variety of E2E pipelines in the areas of Computer Vision, NLP, Recommendation systems, etc. We were able to achieve a performance improvement, ranging from 1.8x to 81.7x across different E2E pipelines. In this paper, we will be highlighting the optimization strategies adopted by us to achieve this performance on Intel ® Xeon ® processors with a set of eight different E2E pipelines.

The main purpose of this study is to observe and analyze the various impacts mental health has on a person’s labor participation and whether these impacts are exacerbated among veterans. Many war veterans struggle with mental well-being and have diffculties fnding work or purpose in their lives when transitioning to the civilian world. Additionally, the steady decrease of mental well-being across the United States among the general populace has generated concerns regarding a loss of labor due to burnout and suicide, particularly among teens and young adults. This study discusses the impact cognitive diffculties have on labor participation through a weighted logit model for 450 households from 2001 to 2020. The result from this model demonstrates the adverse impact cognitive diffculties have on a person’s willingness to participate in the labor force, especially among older individuals, but are not a signifcant factor in determining a veteran’s motivation to work.

Detailed and accurate labor statistics are fundamental to support social policies that aim to improve the match between labor supply and demand, and support the creation of jobs. Despite overwhelming evidence that labor activities are distributed unevenly across space, detailed statistics on the geographical distribution of labor and work are not readily available. To fill this gap, we demonstrated an approach to create fine-scale gridded occupation maps by means of downscaling district-level labor statistics, informed by remote sensing and other spatial information. We applied a super-learner algorithm that combined the results of different machine learning models to predict the shares of six major occupation categories and the labor force participation rate at a resolution of 30 arc seconds (~1x1 km) in Vietnam. The results were subsequently combined with gridded information on the working-age population to produce maps of the number of workers per occupation. The super learners outperformed (n = 6) or had similar (n = 1) accuracy in comparison to best-performing single machine learning algorithms. A comparison with an independent high-resolution wealth index showed that the shares of the four low-skilled occupation categories (91% of the labor force), were able to explain between 28% and 43% of the spatial variation in wealth in Vietnam, pointing at a strong spatial relationship between work, income and wealth. The proposed approach can also be applied to produce maps of other (labor) statistics, which are only available at aggregated levels.

Research on mixed heritage people has often focused on their reported race(s) and what these self-reports may reveal about their racial leanings. In the U.S., people are also asked: “What is this person’s ancestry or ethnic origin?” as a census fill-in-the-blank question. Existing research has analyzed the census data on race and ancestry and has uncovered meanings about ethnicity among White Americans, but little is known about the importance of ancestry in the case of mixed heritage people. In this paper, we draw on our interviews with 68 American-Indian-White, Black-White, and Asian-White mixed heritage people in the U.S. Given that many mixed heritage people’s connections with their disparate ancestries can be hindered by generational distance, lack of cultural contact and exposure, and social rejection, what motivates mixed heritage people to report their disparate ancestries, and how do their ancestry claims relate to their racial identities? How may the differential histories and racialization of groups in the U.S. shape mixed heritage people’s ability and inclination to assert either their White or minority ancestries? We found that mixed heritage individuals who were motivated to assert their ancestry claims did so for two main reasons: First, by claiming a specific ancestry (or ancestries), participants wished to assert a more individualistic sense of self than was typically allowed, given their racial treatment based upon their racial appearance; this could be especially meaningful if those individuals felt a mismatch between their racial assignment by others and their sense of self. Second, a claim to a specific ancestry was a way for individuals to forge connections with family, relatives, or an ancestry group that had not existed before. Overall, while most of our mixed heritage participants reported details of their European ancestries, it was their Black, American Indian, or Asian ancestries that were deemed to be most salient and/or meaningful to who they were.

Social scientists have frequently sought to understand the distinct effects of age, period, and cohort, but disaggregation of the three dimensions is difficult because cohort = period − age. We argue that this technical difficulty reflects a disconnection between how the cohort effect is conceptualized and how it is modeled in the traditional age-period-cohort framework. We propose a new method, called the age-period-cohort-interaction (APC-I) model, that is qualitatively different from previous methods in that it represents Ryder’s theoretical account about the conditions under which cohort differentiation may arise. This APC-I model does not require problematic statistical assumptions and the interpretation is straightforward. It quantifies inter-cohort deviations from the age and period main effects and also permits hypothesis testing about intra-cohort life-course dynamics. We demonstrate how this new model can be used to examine age, period, and cohort patterns in women’s labor force participation.

Importance Racial and ethnic disparities in delayed medical care for reasons that are not directly associated with the cost of care remain understudied. Objective To describe trends in racial and ethnic disparities in barriers to timely medical care among adults during a recent 20-year period. Design, Setting, and Participants This was a serial cross-sectional study of 590 603 noninstitutionalized adults in the US using data from the National Health Interview Survey from 1999 to 2018. Data analyses were performed from December 2021 through August 2022. Exposures Self-reported race, ethnicity, household income, and sex. Main Outcomes and Measures Temporal trends in disparities regarding 5 specific barriers to timely medical care: inability to get through by telephone, no appointment available soon enough, long waiting times, inconvenient office or clinic hours, and lack of transportation. Results The study cohort comprised 590 603 adult respondents (mean [SE] age, 46.00 [0.07] years; 329 638 [51.9%] female; 27 447 [4.7%] Asian, 83 929 [11.8%] Black, 98 692 [13.8%] Hispanic/Latino, and 380 535 [69.7%] White). In 1999, the proportion of each race and ethnicity group reporting any of the 5 barriers to timely medical care was 7.3% among the Asian group; 6.9%, Black; 7.9%, Hispanic/Latino; and 7.0%, White (P > .05 for each difference compared with White individuals). From 1999 to 2018, this proportion increased across all 4 race and ethnicity groups (by 5.7, 8.0, 8.1, and 5.9 percentage points [pp] among Asian, Black, Hispanic/Latino, and White individuals, respectively; P < .001 for each), slightly increasing the disparities between groups. In 2018, compared with White individuals, the proportion reporting any barrier was 2.1 and 3.1 pp higher among Black and Hispanic/Latino individuals (P = .03 and P = .001, respectively). There was no significant difference in prevalence between Asian and White individuals. There was a significant increase in the difference in prevalence between Black individuals and White individuals who reported delaying care because of long waiting times at the clinic or medical office and because of a lack of transportation (1.5 pp and 1.8 pp; P = .03 and P = .01, respectively). In addition, the difference in prevalence between Hispanic/Latino and White individuals who reported delaying care because of long waiting times increased significantly (2.6 pp; P < .001). Conclusions and Relevance The findings of this serial cross-sectional study of data from the National Health Interview Survey suggest that barriers to timely medical care in the US increased for all population groups from 1999 to 2018, with associated increases in disparities among race and ethnicity groups. Interventions beyond those currently implemented are needed to improve access to medical care and to eliminate disparities among race and ethnicity groups.

The objective of this study was to compare health outcomes by sexual orientation identity and cohabiting partnership status (defined as whether heterosexual, lesbian, and bisexual women were non-partnered, partnered in a cohabiting same-sex relationship, or partnered in a cohabiting different-sex relationship). Methods: We used data on heterosexual (n = 95,289) and sexual minority (n = 2,600) women aged 18 years and older from the 2013–2018 National Health Interview Survey. We estimated adjusted odds ratios (AORs) with 95% confidence intervals (CIs) from multivariable logistic regression models comparing health outcomes by sexual orientation identity and cohabiting partnership status while controlling for sociodemographic characteristics. Results: Compared with heterosexual women in a different-sex relationship, nonpartnered women generally reported worse health outcomes regardless of sexual orientation. Lesbian women with a same-sex partner were more likely to report poor/fair health (AOR, 1.61; 95% CI, 1.09–2.37), current cigarette smoking (AOR, 1.48; 95% CI, 1.14–1.94), and binge drinking (AOR, 1.63; 95% CI, 1.19–2.23) compared with heterosexual women with a different-sex partner. Bisexual women with a different-sex partner were more likely to report poor/fair health (AOR, 1.91; 95% CI, 1.23–2.97), severe psychological distress (AOR, 2.86; 95% CI, 1.78–4.59), current cigarette smoking (AOR, 1.38; 95% CI, 1.01–1.88), and binge drinking (AOR, 1.66; 95% CI, 1.18–2.32) compared with heterosexual women with a different-sex partner. Conclusion: More research is needed to identify the processes in which heterosexual and sexual minority women partner and cohabitate with members of the same or different sex—and whether or how this influences their health. Meanwhile, health care providers should be mindful that families are diverse, and acknowledging this diversity could be a first step toward achieving health equity for all women regardless of sexual orientation.

IMPORTANCE Identifying modifiable risk factors that are associated with dementia burden across racial and ethnic groups in the population can yield insights into the potential effectiveness of interventions in preventing dementia and reducing disparities. OBJECTIVE To calculate the population attributable fraction (PAF) of dementia associated with 12 established modifiable risk factors for all US adults, as well as separately by race and ethnicity. DESIGN, SETTING, AND PARTICIPANTS This cross-sectional study used survey data from nationally representative samples of US adults. PAFs were calculated using relative risks and prevalence estimates for 12 risk factors. Relative risks were taken from meta-analyses, as reported in a 2020 systematic review. Prevalence estimates for risk factors were derived from nationally representative cross-sectional survey data collected between 2011 and 2018. Combined PAFs were adjusted for risk factor communality using weights derived from the Atherosclerosis Risk in Communities (ARIC) study (1987-2018). Analyses were conducted May through October 2021. EXPOSURES Low education, hearing loss, traumatic brain injury, hypertension, excessive alcohol consumption, obesity, smoking, depression, social isolation, physical inactivity, diabetes, and air pollution. MAIN OUTCOMES AND MEASURES PAF for each dementia risk factor, a combined PAF, and the decrease in the number of prevalent dementia cases in 2020 that would be expected given a 15% proportional decrease in each exposure. RESULTS Among all US adults, an estimated 41.0% (95% CI, 22.7%-55.9%) of dementia cases were attributable to 12 risk factors. A 15% proportional decrease in each risk factor would reduce dementia prevalence in the population by an estimated 7.3% (95% CI, 3.7%-10.9%). The estimated PAF was greater for Black and Hispanic than it was for White and Asian individuals. The greatest attributable fraction of dementia cases was observed for hypertension (PAF, 20.2%; 95% CI, 6.3%-34.4%), obesity (PAF, 20.9%; 95% CI, 13.0%-28.8%), and physical inactivity (PAF, 20.1%; 95% CI, 9.1%-29.6%). These factors were also highest within each racial and ethnic group, although the proportions varied. CONCLUSIONS AND RELEVANCE A large fraction of dementia cases in the US were associated with potentially modifiable risk factors, especially for Black and Hispanic individuals. Targeting and reducing these risk factors may curb the projected rise in dementia cases over the next several decades.

Climate change is a global phenomenon with markedly local manifestations. Accordingly, territories differ in terms of exposure to climate events, of capacity to adopt climate mitigation policies and of the welfare effects that these deep transformations entail. The paper brings together these threads with an empirical study of the relationship between green technology development and income inequality in US Metropolitan Areas over the period 2005-2015. We find a positive association between local patenting capacity and growing income gaps to the detriment of the least affluent. Further, higher patenting propensity in early-stage technologies has a stronger association with income inequality, whereas such a relationship dissipates at later stages of the life cycle.

The Upjohn Institute was asked by the Kalamazoo County Board of Commissioners to complete a housing plan for Kalamazoo County. This plan includes a housing needs assessment, market demands, results from the county-wide housing survey, goals and objectives, and strategies to move forward. Community and committee engagement helped guide the direction and focus.

Gender equality is a fundamental human right, considered essential for the creation of a just society and a key ingredient of the United Nations Sustainable Development Goals. A key determinant of gender inequality in terms of the gender wage gap and other dimensions are the attitudes and stereotypes associated with women (Blau and Kahn (2017), Fern´andez (2007), Klasen (2002)). Here, we investigate the evolution of gender norms for 160 years in the US. Socioeconomists have posited two fundamental and widely debated theories on the evolution of cultural norms across time. One school argues that cultural norms (including gender norms) should converge across time (Marx (1867), Bell (1976), Lerner (1958), Inglehart (1997)) as economies become more advanced and globalised and technological progress allows for easier information sharing. The other school states that cultural traits are highly persistent, passed down from generation to generation and will remain divergent across regions (Weber (1904), Huntington (1993), Huntington (1996), DiMaggio (1994)). To test these theories quantitatively, researchers need data on attitudes measured at appropriate levels of granularity, at a high frequency and over a long time series. However, no such data currently exists. Here we propose an unsupervised machine learning methodology to measure attitudes at the document level. We apply this methodology to 193 million pages of local newspaper text to produce localised attitudes towards women on four different dimensions: career vs family, attitudes towards abortion, attitudes towards feminism/suffrage, and violence against women. We establish five novel facts on the evolution of attitudes across time with these measures. (i) Attitudes are less persistent than the existing literature hypothesises. (ii) The persistence of attitudes varies considerably across different regions and dimensions. (iii) Attitudes exhibit cyclical patterns. (iv) Regional variation in attitudes decreases considerably over time and has fallen between 64% to 79%. (v) A decrease in transport costs that allows for easier information sharing is associated with a homogenisation of the norms. Our paper calls for more research on whether similar patterns exist across countries and the causal factors that make culture converge.

This study evaluates how health care employers’ staffing decisions and labor demand respond to increased Medicaid eligibility using a database of over 14 million job vacancies for health care workers. The results indicate that increased Medicaid eligibility leads to health care employers posting more job vacancies and hiring additional health care workers. Health care employers that typically hire lower-skilled workers appear to be the most responsive to Medicaid expansions. These results imply that there are providers that respond to Medicaid patients’ needs but also highlight issues arising from Medicaid’s low reimbursement rates.

This study examines the long-term link between in-utero cigarette taxes and adult prenatal smoking. We use U.S. birth certificate records to demonstrate that exposure to higher in-utero cigarette taxes (over 1965-2001) reduces later-life adult pre-pregnancy and prenatal smoking. We also show that higher in-utero cigarette taxes have long-lasting effects on adult health, and intergenerational consequences for infant health. Finally, we demonstrate that larger in-utero tax responsiveness correlates with smaller contemporary cigarette tax responsiveness, suggesting that higher in-utero taxes may alter the composition of remaining smokers and contribute to reductions in contemporary cigarette tax responsiveness.

Report submitted to the Minnesota ERASE Campaign (End Rental Arrears and Stop Evictions). ERASE is an effort to ensure that the historic aid enacted by Congress reaches the lowestincome and most marginalized renters it is intended to help. We are working to eliminate rental indebtedness caused by the pandemic, prevent evictions, and create support for longterm policy changes to end housing instability and homelessness. Co-convenors of the campaign include ACER (African Career Education Resource); Housing Justice Center, and MHP (Minnesota Housing Partnership).

We analyze heterogeneous effects of new digital technologies on individual-level wage and employment dynamics in the United States from 2011-2018. To this end, we employ four digital technology measures from recent literature: computerization probabilities of occupations, occupational impacts of artificial intelligence, and the suitability of tasks for machine learning and their within-occupation variance. Based on CPS and ASEC panel data, the results indicate that labor-displacing digital technologies are associated with slower wage growth and higher probabilities of switching one's occupation and becoming non-employed. In contrast, labor-reinstating digital technologies improve individual labor market outcomes. Workers with high levels of formal education are most affected by the new generation of digital technologies.

The relationship between employee quality and firm outcomes is extensively explored in the literature. Prior research focuses on examining the quality of top managers in a limited number of industries or using smaller datasets. This study fills this literature gap using demographic characteristics from Metropolitan Statistical Areas, where firm headquarters are located as a proxy for employee quality. The findings show that firms with high employee quality in terms of employee competency and diversity exhibit higher firm value. The results show that ensuring the availability of an educated and diverse workforce provides long-lasting firm value.

The Growth Management Act in Washington allows subarea plans as an optional element, where the subarea plan is consistent with the Lewis County Comprehensive Plan (RCW 36.70A.080). A starting point for any comprehensive planning effort is to establish the factual basis from which the plan is developed. This existing conditions report, completed in April 2022, forms the factual basis for the Packwood Subarea Plan.