Structural racism manifests through systematic neighborhood disin-vestment in local infrastructure where racial and ethnic minorities re-side,1leading to a disproportionate burden of environmental pollutionand associated respiratory health hazards.2A former federal agency,the Home Owners’Loan Corporation (HOLC, established in 1933),codified neighborhood disinvestment across the United States. Usinga grading system, HOLC appraisers assessed perceived mortgage se-curity risk of a neighborhood’s residents defaulting on their homeloans.3Riskier grades were consistently assigned to neighborhoodswith high concentrations of racial and ethnic minorities, whomfinan-cial institutions perceived as“undesirable”recipients of home loans.3The legacy of this system is often referred to as“redlining,”atermbased on the red color assigned to the highest-risk neighborhoods onmaps used by agencies and organizations involved in real estate lawsand lending. We hypothesized that neighborhoods with historically(c. 1933) higher-risk HOLC grades have worse contemporary respira-tory health than neighborhoods with lower-risk HOLC grades.Current residents of historically disinvested neighborhoods mayexperience an increased prevalence of asthma through past land-use patterns that led to worse environmental quality, such as lessgreen space or higher air pollution.

Despite a robust literature base that has explored links between household crop diversity and children’s dietary diversity, evidence continues to yield mixed results regarding the significance and efficacy of crop diversity to improve childhood dietary outcomes. Given wide variance in the association between agrobiodiversity and dietary diversity across contexts, identifying factors that influence the direction and significance of the relationship can help inform targeted and appropriate development policies and interventions to maximize beneficial impact on livelihoods. Household characteristics such as wealth and distance to markets may provide important insight into the linkages between crop diversity and dietary diversity, given that low-resource households are often more reliant on localized production (including their own) for nutritional needs as compared to wealthier households with the resources to purchase more diverse foods from markets. This study examines the associations between crop diversity and dietary diversity among farm households at different levels of wealth in 10 sub-Saharan African countries. Drawing on the Integrated Public Use Microdata Series (IPUMS)-Demographic and Health Surveys (DHS) system, we show that the significance and direction of the association between crop diversity (as proxied by the Simpsons Diversity Index (SDI)) and children’s dietary diversity (as measured by the Household Dietary Diversity Score (HDDS)) varies by wealth quintile across all study countries. We find that the significance and direction of the relationship between crop diversity and dietary diversity depends on the socioeconomic status of a household: in richer households, crop diversity has a negative effect on dietary diversity and in poorer households, there is no significant effect. Further, findings demonstrate the importance of considering contextual factors, such as wealth and distance to markets, in assessing how the presence of crop diversity may improve nutritional outcomes. This study indicates the need to better understand the factors that impact the relationship between agricultural diversity and dietary diversity in order to inform agricultural development strategies.

Innovation is a crucial determinant of long-run economic growth in advanced economies. This dissertation explores the economic and social determinants of the production and diffusion of innovation in the context of Europe and the United States in the late nineteenth and early twentieth century.The first chapter (jointly authored with Gaia Dossi) documents how out-migration impacts innovation in the country of origin of migrants. During the Age of Mass Migration, nearly fourmillion English migrants settled in the US. We construct a novel individual-level dataset link-ing English immigrants in the US to the UK census and complement it with the newly digitizeduniverse of UK patents. Using a new shift-share instrument for bilateral migration flows and atriple-differences design, we document a positive, significant, and persistent effect of exposureto US technology through migrant ties on the direction of innovation in Britain in 1870–1940.The individual-level analysis suggests that physical return migration is not the main factor un-derlying this “return innovation” effect. Instead, we find that migration ties generate infor-mation flows that facilitate the cross-border diffusion of novel knowledge. Furthermore, ourfindings suggest that market integration fostered by migration linkages is a crucial driver ofinformation flows.The second chapter (jointly authored with Lorenzo Spadavecchia) interprets out-migrationthrough the lenses of directed technical change and adoption theory. We study the impact ofimmigration restriction policies on technology adoption in countries sending migrants. Be-tween 1920 and 1921, the number of Italian immigrants to the United States dropped by 85%after Congress passed the Emergency Quota Act, a severely restrictive immigration law. In adifference-in-differences setting, we exploit variation in exposure across Italian districts to thismassive restriction against human mobility. Using novel individual-level data on Italian immi-grants to the US and newly digitized historical censuses, we show that this policy substantially hampered technology adoption and capital investment. We interpret this as evidence of di-rected technical adoption: an increase in the labor supply dampens the incentive for firms toadopt labor-saving technologies. To validate this mechanism, we show that more exposed dis-tricts display a sizable increase in overall population and employment in manufacturing. Weiii provide evidence that “missing migrants,” whose migration was inhibited by the Act, drive thisresult.The third chapter (jointly authored with Enrico Berkes, Gaia Dossi, and Mara P. Squiccia-rini) investigates how societies respond to adversity. After a negative shock, separate strandsof research document either an increase in religiosity or a boost in innovation efforts. In thispaper, we show that both reactions can occur at the same time, driven by different individuals within the society. The setting of our study is 1918–1919 influenza pandemic in the United States.To measure religiosity, we construct a novel indicator based on the naming patterns of new-borns. We measure innovation through the universe of granted patents. Exploiting plausibly exogenous county-level variation in exposure to the pandemic, we provide evidence that more-affected counties become both more religious and more innovative. Looking within counties,we uncover heterogeneous responses: individuals from more religious backgrounds further embrace religion, while those from less religious backgrounds become more likely to choose a scientific occupation. Facing adversity widens the distance in religiosity between science-oriented individuals and the rest of the population, and it increases the polarization of religious beliefs.

he Center for the Study of Asian American Health (CSAAH) within the NYU Grossman School of Medicine at NYU Langone Health and the Coalition of Asian American Children and Families (CACF) have created a manual to promote best practices for collecting, analyzing, and reporting Asian American health data. We recommend anyone who is working with quantitative data that pertains to Asian Americans to review this guide in order to promote an accurate and nuanced characterization of Asian American health.Data Collection The current national standards for collecting racial/ethnic data were issued by the Office of Man-agement and Budget’s 1997 Directive 15 though leaders in civil rights and academia have voiced concerns over their definition and application. CSAAH, CACF and the New York Academy of Med-icine (NYAM) developed a race/ethnicity form for collecting disaggregated race/ethnicity infor-mation in the New York metropolitan area which was informed by three main sources: 1) national policy recommendations by the State Data Collaborative;1 2) American Community Survey 2019 data; and 3) feedback from community focus groups and trusted subject matter experts. Our Revised Race/Ethnicity Form asks participants to identify as one or more of seven aggregate race/ethnicity groups – American Indian/Alaska Native (AI/AN), Asian, Black or African American, Hispanic/Latino, Middle Eastern or North African (MENA), Native Hawaiian or Pacific Islander (NH/PI), or White – and then asks participants to identify as one or more of 70+ detailed race/ethnicity categories. The form can be adapted in format and to include different detailed categories based on project or research focus and needs. When conducting primary data collection among Asian American populations, there are several factors to consider regarding the study population of focus and survey design. These factors include but are not limited to the study population: age, income, immigration history, education level, and English proficiency. Data from the U.S. Census Bureau like the Decennial Census and American Community Survey can be used for information on the demographic makeup of the study population in a specific geographic region or neighborhoods where the population resides for targeted recruitment efforts. Data Analysis When constructing racial/ethnic group variables for analysis we recommend creating: 1) A mutually exclusive aggregate race/ethnicity variable where individuals who identify as more than one aggregate racial/ethnic identities are labeled ‘Multiracial’ (AI/AN, Asian, Black, Hispanic/Latino, MENA, NH/PI, White, Multiracial, Other); 2) A mutually exclusive detailed race/ethnicity variable where individuals who identify as more than one detailed racial/ethnic identifies are labeled ‘Mixed’ (e.g., Chinese, Asian Indian, Filipino, Japanese, Asian Mixed, Detailed Other, etc.); 3) If appropriate, for Asian populations, a mutually exclusive Asian regional race/ethnicity variable (e.g., East Asian, South Asian, Southeast Asian, Central Asian); 4) If appropriate, a non-mutually exclusive race/ethnicity variable for each unique aggregate and detailed race/ethnicity category; and 5) If appropriate, further granular variables for Multiracial/Mixed individuals. When conducting secondary data analysis, it is important to consider how Asian American race/ethnicity data has been collected in the original dataset. Race/ethnicity data may be missing, unknown or other. Detailed race/ethnicity data may be limited or nonexistent. Additionally, many datasets are not always designed to be inclusive of non-English speakers, and sampling bias may result from higher rates of English proficient Asian Americans who tend to have more favorable socioeconomic indicators.2 Data quality and therefore representation should be qualified in Methods, Discussion and Limitations sections. Sensitivity analyses should be conducted for more precise evaluation of the heterogeneity of the Asian American sample. Careful consideration of who was included or not included in the data being analyzed will help researchers evaluate biases in data collection and recruitment and the generalizability of the analyses to Asian American populations. Partial missingness of data using complete case analysis, imputation and advanced imputation and total missingness using standardization, simulation, and sensitivity analyses should be performed. In the case studies, we examine the impact of undercounting in minority populations. The U.S. Census has a history of undercounting immigrant and minority populations, which can impact the accuracy of public health statistics. The 2020 U.S. Census may have had a greater undercount due to a potential citizenship question and the COVID-19 pandemic.3,4 Correcting for the undercount requires specific birth, death, and migration data, making it challenging for most researchers. In the case of New York City (NYC) COVID-19 vaccination statistics, the implications of an undercount for Asian American and NH/PI populations were examined. Collection, Analysis, and Reporting of Asian American Health Data5Data Reporting When reporting Asian American data, descriptions should include all eligibility and exclusion criteria, including explicit details about the disaggregated Asian racial/ethnic group(s) sampled, the dates of the study period, the type of study design, and the number of individuals surveyed or enrolled. Having this information can help researchers better determine next steps for research and resource allocation by contextualizing the population sample.The demographic details of the study population play a crucial role in preventing harmful stereotypes and generalizing data. These details should include racial/ethnic information, income, education, English language proficiency, language spoken at home, and nativity. It is important to be transparent about any groupings, such as aggregating into an “Other” category or combining smaller Asian race/ethnicity categories into regional groupings. When describing multiracial or Asian mixed populations, it is useful to include details of the combinations of race/ethnicities. Reporting race and ethnicity to meet OMB guidelines is required for federal agencies as well as other state and local organizations that model their reporting guidelines on the OMB’s 1997 directive. The Agency for Healthcare Research and Quality (AHRQ) provides a guideline for roll-up of detailed race/ethnicity categories while the Institute of Medicine (IOM) provides further granular guidance. The process of rolling up detailed race/ethnicity categories into aggregate categories has gray areas and can lead to mismatches between aggregate categories and self-identification of race/ethnicity. This can result in individuals who identify as one aggregate category but are categorized differently based on their detailed race/ethnicity. It is recommended that participants’ reported aggregate race/ethnicity should always reflect their selected aggregate category and not be changed. In cases where only detailed race/ethnicity information is provided, researchers should detail their assignment rules for ambiguous categories. The representation of Asian American experiences in data can be distorted by three racialized stereotypes - the model minority, healthy immigrant effect, and perpetual foreigner. These stereotypes can be a result of implicit biases in researchers – embedded in their own lived experiences – and should be considered when reporting Asian American data.5 Recommendations for considerations when discussing results from Asian American data: 1) Discuss who is represented in the data to provide context to research findings and avoid creating generalizations about the Asian American population. 2) Be cautious when discussing Asian multiracial and mixed data as these categories are diverse. 3) Hypothesize the potential impact of missing data in the analysis. 4) Consider complementing statistics from large databases with community-based research to uncover health problems that may be hidden. 5) Seek and include community feedback to enhance trust and cultural resonance. Collection, Analysis, and Reporting of Asian American Health Data66) If Asian data is not available, discuss the implications and provide recommendations for additional research that includes Asian Americans. This manual provides recommendations for public health practitioners, allied health professionals, health equity researchers, and data managers to employ in their own research with immigrant and minority groups in the U.S. Understanding and addressing the nuances of data collection, analysis, and reporting is the cornerstone to accurate representation of groups in data. Suggested Citation: Chin MK, Yusuf Y, Wyatt LC, Ðoàn LN, Russo RG, Kader F, Feng L, Fu L, Kwon, SC, and Yi SS. Collection, Analysis, and Reporting of Asian American Health Data. E-published by Center for the Study of Asian American Health at NYU Langone; 2023. Available from: https://aanhpihealth.org/resource/asian-american-manual-202

This paper relies on estimates from a microsimulation model to analyze the distributional impacts of the preferential tax treatment of owner-occupied housing by race and income. Compared to other homeowner tax advantages, which are more equally distributed than tax burdens, the mortgage interest deduction is more unequally distributed than tax burdens. Although current tax policy is more progressive than a tenure-neutral tax policy that treats homeowners and renters the same, a tenure-neutral tax policy yields larger reductions in after-tax income inequality. Among median-income households, white households receive a larger share of homeownership tax benefits than Black households, relative to each racial group’s share of the overall tax burden. Residential location explains the largest portion of the racial gap in the imputed rent exclusion, while household income explains the largest portion of the racial gap in the mortgage interest deduction and capital gains tax exclusion.

Objectives. To assess the relationship between poverty, delayed care, unaffordable care, and functional limitations among individuals with chronic obstructive pulmonary disease (COPD). Methods. Using the National Health Interview Survey data, we selected respondents with COPD, aged 40 years and older. The predictor variables were poverty and measures of delayed and unaffordable care. The outcome variable was functional limitations. We performed a survey-weighted multivariate logistic regression analysis, adjusting for sociodemographic characteristics. Results. Respondents classified as poor had three times the odds of functional limitations compared with those classified as not poor. Respondents who reported having measures of delayed care or unaffordable care had two to nine times and two to four times the adjusted odds of functional limitations compared with those who did not report such measures of delayed and unaffordable care, respectively. Conclusions. Poverty and delayed and unaffordable care are associated with functional limitations among individuals with COPD.

We estimate a life-cycle model of consumption, human capital investment, and labor supply and use it to analyze changes in Social Security rules. The most important aspect of our paper is human capital towards the end of the life cycle which responds to changes in the rules. Retirement arises endogenously as part of the labor supply decision. The model allows for both an endogenous wage process through human capital investment (which is typically assumed exogenous in the retirement literature), an endogenous retirement decision (which is typically assumed exogenous in the human capital literature), and accounts for the Social Security system. We estimate the model using indirect inference to match the life-cycle profiles of employment and measured wages from the SIPP data. The model replicates the main features of the data—in particular the large increase in measured wages and small increase in labor supply at the beginning of the life cycle as well as the small decrease in measured wages but large decrease in labor supply at the end of the life cycle. We use the model to estimate the effects of various changes to tax and Social Security policies and show that allowing for human capital accumulation is critical.

The ruling in Plessy v. Ferguson changed the United States from a color-focused society to a race- focused society and stripped the privileged formal status from mulattos. Black identity was flattened to what we know today: a collective identity with rampant color stratification. While research on colorism has expanded, we need to understand how contemporary colorism evolved from the systems of the past. I use Census microdata from 1870 and 1920 and regression analysis to examine colorism before and after the one-drop-rule. I reveal that colorism worsened after the one- drop-rule, and mulattos tried even harder to preserve their distinct identity.

The authors examine work patterns in the United States from 1973 to 2018, with the novel focus on days per week, using intermittent CPS samples and one ATUS sample. Among full-time workers, the incidence of four-day workweeks tripled, adding 7 million four-day workers. Similar growth occurred in the Netherlands, Germany, and South Korea. The rise was not due to changes in demographics or industrial structure. Such schedules are more common among those who are 1) less educated, 2) younger, 3) white non-Hispanic, 4) men, 5) natives, 6) people with young children, 7) non-unionized, 8) police and firefighters, 9) health care workers, and 10) restaurant workers. Based on an equilibrium model, the authors show that the increase in four-day workweeks results more from workers’ preferences and/or their daily fixed costs of working than production costs. The wage penalty for a four-day workweek is greater when such work is more prevalent, and the penalty has diminished over time.

This study examines social media users' preferences for the use of platform-wide moderation in comparison to user-controlled, personalized moderation tools to regulate three categories of norm-violating content - hate speech, sexually explicit content, and violent content. Via a nationally representative survey of 984 US adults, we explore the influence of third-person effects and support for freedom of expression on this choice. We find that perceived negative effects on others negatively predict while free speech support positively predicts a preference for having personal moderation settings over platform-directed moderation for regulating each speech category. Our findings show that platform governance initiatives need to account for both actual and perceived media effects of norm-violating speech categories to increase user satisfaction. Our analysis suggests that users do not view personal moderation tools as an infringement on others' free speech but as a means to assert greater agency over their social media feeds.

We study the adoption of remote work within cities and its effect on city structure and welfare. We develop a dynamic model of a city in which workers can decide to work in the central business district (CBD) or partly at home. Working in the CBD allows them to interact with other commuters, which enhances their productivity through a standard production externality, but entails commuting costs. Switching between modes of labor delivery is costly, and workers face idiosyncratic preference shocks for remote work. We characterize the parameter set in which the city exhibits multiple stationary equilibria. Within this set, a coordination mechanism can lead to stationary equilibria in which most workers commute or most of them work partially from home. In these cases, large shocks in the number of commuters , like the recent lockdowns and self-isolation generated by the COVID-19 pandemic, can result in dynamic paths that make cities converge to a stationary equilibrium with large fractions of remote workers. Using cell-phone-based mobility data for the U.S., we document that although most cities experienced similar reductions in commuting during the pandemic, commuting shares in the 20 largest cities have stabilized at commuting levels that are only about 60% of pre-pandemic levels. In contrast, the smallest 500 cities have returned to pre-pandemic commuting levels. House price panel data by city show consistent changes in house price CBD-distance gradients. We structurally estimate the model for 274 U.S. cities and show that cities that have stabilized at a large fraction of remote work are much more likely to have parameters that result in multiple stationary equilibria. Our results imply welfare losses in these cities that average 2.7%.

We evaluate progress in the War on Poverty as President Johnson defined it, which established a 20 percent baseline poverty rate and adopted an absolute standard. While the official poverty rate fell from 19.5 percent in 1963 to 10.5 percent in 2019, our absolute Full-income Poverty Measure, which uses a fuller income measure and updates thresholds only for inflation, fell from 19.5 to 1.6 percent. However, we also show that relative poverty reductions have been modest. Additionally, government dependence increased over this time, with the share of working-age adults receiving under half their income from market sources more than doubling.

Federal standards that assign counties to Core Based Statistical Areas (CBSAs) and designate tracts with a Rural Urban Commuting Area (RUCA) code rely on journey to work data from the American Community Survey (ACS). Because the ACS is based on a relatively small sample of the population commuting flows are represented as point estimates, frequently with high margins of error. In this paper we examine the impact of uncertainty in commuting flows on these two critical designations. We find that, for the vast majority of counties (85%) and tracts (87%), the official designation remains consistent and that, in over 1,000 replications of the data, 98% of county assignments and 97% of tract assignments were consistent with the official delineations. While these results are reassuring, a small number of counties and tracts do experience assignments that are different from the official delineation at very high rates. We also test the official delineations against an alternative data source, the Longitudinal Employer- Household Dynamics Survey’s Origin–Destination Employment Statistics (LEHD-LODES). We find the LODES data result in designations that are largely consistent with those from the ACS but we lack a clear way to choose between the data products in places where they differ. Overall, our findings suggest that the current delineation method for assigning counties to CBSAs and tracts to RUCA codes is sufficiently robust to uncertainty to continue going forward. Nevertheless, the presence of significant uncertainty for some observations suggests the need for continued consideration of uncertainty in the context of future data releases.

We demonstrate that regression models can be estimated by working independently in a row-wise fashion. We document a simple procedure which allows for a wide class of econometric estimators to be implemented cumulatively, where, in the limit, estimators can be produced without ever storing more than a single line of data in a computer's memory. This result is useful in understanding the mechanics of many common regression models. These procedures can be used to speed up the computation of estimates computed via OLS, IV, Ridge regression, LASSO, Elastic Net, and Non-linear models including probit and logit, with all common modes of inference. This has implications for estimation and inference with 'big data', where memory constraints may imply that working with all data at once is particularly costly. We additionally show that even with moderately sized datasets, this method can reduce computation time compared with traditional estimation routines.

The prevalence of school-based healthcare has increased markedly over the past decade. We study a modern mode of school-based healthcare, telemedicine, that offers the potential to reach places and populations with historically low access to such care. School-based telemedicine clinics (SBTCs) provide students with access to healthcare during the regular school day through private videoconferencing with a healthcare provider. We exploit variation over time in SBTC openings across schools in three rural districts in North Carolina. We find that school-level SBTC access reduces the likelihood that a student is chronically absent by 2.5 percentage points (29 percent) and reduces the number of days absent by about 0.8 days (10 percent). Relatedly, access to an SBTC increases the likelihood of math and reading test-taking by between 1.8-2.0 percentage points (about 2 percent). Heterogeneity analyses suggest that these effects are driven by male students. Finally, we see suggestive evidence that SBTC access reduces violent or weapons-related disciplinary infractions among students but has little influence on other forms of misbehavior.

What is the relationship between gender segregation in higher education and gender segregation in the labor market? Using Fossett's (2017) difference-of-means method for calculating segregation indices and data from the American Community Survey, we show that approximately 36% of occupational segregation among college-educated workers is associated with gender segregation across 173 fields of study, and roughly 64% reflects gender segregation within fields. A decomposition analysis shows that fields contribute to occupational segregation mainly through endowment effects (men's and women's uneven distribution across fields) than through the coefficient effects (gender differences in the likelihood of entering a male-dominated occupation from the same field). Endowment effects are highest in fields strongly linked to the labor market, suggesting that educational segregation among fields in which graduates tend to enter a limited set of occupations is particularly consequential for occupational segregation. Within-field occupational segregation is higher among heavily male-dominated fields than other fields, but it does not vary systematically by fields' STEM status or field-occupation linkage strength. Assuming the relationship between field segregation and occupational segregation is at least partly causal, these results imply that integrating higher education (e.g., by increasing women's representation in STEM majors) will reduce but not eliminate gender segregation in labor markets.

Food-energy-water (FEW) systems are increasingly vulnerable to shocks. Repeated floods, worsening droughts, sudden tariffs, and disease outbreaks all underscore the importance of strengthening production systems during a time of rapid global change. However, the laws, regulations, and incentive programs that govern these sectors were often developed in isolation, creating fragmented and lagged responses to previous crises, ineffective governance of FEW security, and unintended effects even when achieving policy goals. Here, we examine the Mississippi River Basin in the Midwest US to illustrate how policies designed to address one challenge had other unanticipated consequences. We argue for a long view of the future that honors the interconnectedness of FEW sectors with ecosystems (FEWE); values non-provisioning ecosystem services; and prioritizes incentives that improve FEW production, farm profitability, and ecosystem health. Now is the time for reassessment of how well FEWE provide security to all humans and the environment, and to support integrated policies that avoid unintended future consequences.

Objective Tracking the US smoking cessation rate over time is of great interest to tobacco control researchers and policymakers since smoking cessation behaviors have a major effect on the public's health. A couple of recent studies have employed dynamic models to estimate the US cessation rate through observed smoking prevalence. However, none of those studies has provided recent annual estimates of the cessation rate by age group. Methods We employed a Kalman lter approach to investigate the annual evolution of age-group-specic cessation rates, unknown parameters of a mathematical model of smoking prevalence, during the 2009-2018 period using data from the National Health Interview Survey. We focused on cessation rates in the 24-44, 45-64 and 65 + age groups. Results The ndings show that cessation rates follow a consistent u-shaped curve over time with respect to age (i.e., higher among the 25-44 and 65 + age groups, and lower among 45-64-year-olds). Over the course of the study, the cessation rates in the 25-44 and 65 + age groups remained nearly unchanged around 4.5% and 5.6%, respectively. However, the rate in the 45-64 age group exhibited a substantial increase of 70%, from 2.5% in 2009 to 4.2% in 2017. The estimated cessation rates in all three age groups tended to converge to the weighted average cessation rate over time. Conclusions The Kalman lter approach offers a real-time estimation of cessation rates that would be helpful for monitoring smoking cessation behavior, of interest in general but also for tobacco control policymakers.

In network services, there are often conflicts among privacy protection, trust, and service quality, which greatly reduces the regulatory effectiveness of networks. Using the basic method of information theory, this paper considers the optimization mechanism of trust and the balance between privacy and service utility from the perspective of quantifying information flow. First, we build a trust model based on multiple factors, such as direct trust, service trust, and trust cloud network. The weight of trust factors is determined by category diversity and information entropy theory. Second, based on statistical knowledge and information entropy, we propose a new privacy measurement model with hierarchical weights. Third, we establish a trust privacy service relationship model to realize the trade-off between privacy, service, and trust in networks. The demander can choose to establish trust, protect privacy, or obtain high-quality services according to personal preferences and needs. Fourth, we propose a privacy trust mapping relationship model and a privacy disclosure selection model. Finally, we design different experimental schemes. These simulation results show that our research can not only provide better quality of service but also better protect privacy and help interactive parties in networks establish a strong trust relationship. This research will improve the effectiveness of networks and can be applied to class II vaccine multi-agent collaborative supervision based on technology.

We study the problem of online prediction, in which at each time step t ∈ {1, 2, · · · T }, an individual xt arrives, whose label we must predict. Each individual is associated with various groups, defined based on their features such as age, sex, race etc., which may intersect. Our goal is to make predictions that have regret guarantees not just overall but also simultaneously on each sub-sequence comprised of the members of any single group. Previous work [Blum and Lykouris, 2019] provides attractive regret guarantees for these problems; however, these are computationally intractable on large model classes (e.g., the set of all linear models, as used in linear regression). We show that a simple modification of the sleeping-experts-based approach of Blum and Lykouris [2019] yields an efficient reduction to the well-understood problem of obtaining diminishing external regret absent group considerations. Our approach gives similar regret guarantees compared to Blum and Lykouris [2019]; however, we run in time linear in the number of groups, and are oracle-efficient in the hypothesis class. This in particular implies that our algorithm is efficient whenever the number of groups is polynomially bounded and the external-regret problem can be solved efficiently, an improvement on Blum and Lykouris [2019]'s stronger condition that the model class must be small. Our approach can handle online linear regression and online combinatorial optimization problems like online shortest paths. Beyond providing theoretical regret bounds, we evaluate this algorithm with an extensive set of experiments on synthetic data and on two real data sets-Medical costs and the Adult income dataset, both instantiated with intersecting groups defined in terms of race, sex, and other demographic characteristics. We find that uniformly across groups, our algorithm gives substantial error improvements compared to running a standard online linear regression algorithm with no groupwise regret guarantees.