CURRENT PROJECTS include only the principal investigator and the project title. COMPLETED PROJECTS have titles linked to the project abstract.

CURRENT PROJECTS

  • Anders T Van Sandt, Exploring the Impact of Comprehensive Health Care Access on Health Outcomes Across the U.S.
  • Anita Pena, Displacement, Neighborhood Change, and Residential Migration Patterns:  Causes and Consequences
  • Ashley H Hirai, Sub-State Analysis of National Survey of Children's Health (NSCH) Oversamples for State and Local Public Health Planning & Assessment
  • Benjamin Gilbert, Economic Data Aggregation Bias: Empirical Evidence from the Energy Sector
  • Blair Darney, !SALE!: Social determinants of health and adolescent pregnancy in US Latinas
  • Daniel Simon, Death by Despair? Individual and Contextual Predictors of Suicide Mortality Risk
  • Franz Fuchs, Health Insurance Payer Mix Map of Wyoming
  • James M Flynn, Do Sugar-Sweetened Beverage Taxes Work?
  • Jared Carbone, How Do Households Relocate in Response to the Changes in Moving Costs?
  • Jeronimo R Carballo, Analyzing the Impact of Firms' Trade Activities on Labor Market Outcomes. A Matched Employee-Employer Perspective
  • Joe Chestnut, Estimating Commuting Accessibility by Income and Departure Time
  • Joshua Black, Analysis of Drug Mentions with Involvement of Buprenorphine
  • Joshua Black, Analysis of Drug Induced Mortality Data for Opioid Analgesics
  • Joshua Black, Classifying individuals who progress from nonfatal, opioid-related hospital encounters to fatal overdose
  • Joshua Clapp, The Social and Economic Consequences of Violent Victimization in the NCVS
  • Justin Denney, Hearing Quality, Social Resources, and Mortality among U.S. Adults
  • Kassandra M McLean, Hyper-Local Economic Ecosystems
  • Kim Truong-Vu, Gender and Racial/Ethnic Differences in the Timing of Initiating the HPV Vaccine: The Case of Southeast Asian Americans
  • Lori Hunter, Environmental Extremes and Rural Health
  • Myron P Gutmann, Trends, Transitions, and Well-Being in Small-Town America
  • Patrick Krueger, Small Businesses and Employer-Sponsored Health Insurance
  • Richard K Mansfield, Estimating Two-Sided Assignment Models Using LEHD Data
  • Ryan C Lewis, The Welfare Implications of Corporate Policy
  • Sarah Small, State Variation in Intimate Partner Violence Rates, Reporting, and Healthcare Utilization
  • Shaowei Wan, Identifying Geographic Areas of Unmet Needs for Hospice Care among Persons with Cancer: A National Perspective
  • Sophia R Newcomer, Population-level assessment of early childhood vaccination timeliness, parental vaccine hesitancy, and immunization schedule adherence in the United States, including rural-urban disparities
  • Soumak Basumallik, Did ACA Reduce Disparities in Preventive Healthcare between Hispanics and Non-Hispanic Whites?
  • William B Allshouse, Oil and Gas Siting, Housing Choices, and Environmental Justice
  • Wolfgang Keller, Offshoring and Innovation: How Large are the Benefits of Co-Location?

COMPLETED PROJECTS

SNAP comprises 70% of total U.S. expenditures on food and nutrition programs, 20% of all U.S. spending on safety net programs, and provides benefits to 12% of the U.S. population making it the key safety net program today.2 However, because there has typically not been variation in benefit amounts or eligibility, which is typically used to estimate the causal effects of safety net programs, very little is known about the effects of SNAP. I plan to take advantage of the only recent variation of this type--the loss of SNAP eligibility for immigrant families as a result of welfare reform and then the haphazard restoration of eligibility across states and over time in the following years--to estimate the effects of SNAP on children's health. In order to analyze how these eligibility changes affected children's health I need a large dataset that allows me to observe state of residence and birth, country of birth (foreign or not), andthe  number of years foreign-born have been in the U.S., because the eligibility rules depend on whether the family is a recent immigrant or not. The NHIS has a large sample size, this necessary demographic information and detailed health and health care utilization information. The NHIS will allow me to have more power and examine more health outcomes than I can with existing public use data. Unfortunately the public use NHIS data do not have state of residence, state of birth, or a detailed measure of number of years immigrants have lived in the U.S., which is the information necessary to take advantage of the policy variation, so I need access to the restricted use version of the data 

The Spatial Sciences Census Research Node (SSCRN) will foster a connection between the spatial and the survey sciences. This bridge will yield both immediate and long‐term benefits for the estimation, dissemination, and usability of the small area statistics produced by the Census Bureau. Small area statistics describe the character of the population within small geographic zones, such as census tracts. Small area estimates from the American Community Survey (ACS) are imprecise. Historically, very little attention has been paid to the geographic distribution of populations within these small areas. The basic research of SSCRN will increase knowledge about the organization of the American population within small geographic areas for the purpose of improving small area estimates. Apart from the impractical solution of increasing the sample size, the only way to reduce the uncertainty of survey estimates is to utilize ancillary information about the population. SSCRN will exploit new forms of geographic information and recent advances in spatial statistics to make small area estimates more accurate. In addition to improving small area population estimates, an improved understanding of the geographic micro‐structure of the American population is of broad scientific interest and may expand knowledge about socio‐spatial processes like segregation and neighborhood effects. SSCRN will bridge the spatial and the survey sciences through basic research that addresses the current needs of the Census Bureau and yields tangible products for researchers. SSCRN will develop and disseminate training materials, software, and research for a broad community of Census data users and producers.

SSCRN will convene a series of interdisciplinary meetings designed to identify research frontiers at the intersection of the spatial and the survey sciences. The meetings will be used to design and refine a novel curriculum and workshop series aimed at developing a cohort of scientists that are capable of establishing and elucidating links between the spatial and survey sciences. SSCRN will develop software tools that will enhance the usability of ACS small area estimates by allowing users to intelligently combine tracts to reduce uncertainty in variables of interest. SSCRN will educate and train a large group of students, post docs, and scientists through meetings, structured mentoring, and workshops. The impact of these activities will be multiplied through design and dissemination of a broadly accessible model curriculum and training workshop. Finally, SSCRN will have a broad impact by releasing a novel form of spatial microdata that will improve small area estimation and enable researchers to explore the spatial structure of the American population. This activity is supported by the NSF‐Census Research Network funding opportunity 

The past several decades have seen an increase in immigration policy and immigration enforcement. These policies have been increasingly conducted by states and localities, rather than the federal government, as well through piecemeal federal immigration policy changes. These changes have resulted in a wide variety of policies being implemented across the country that affect immigrants, and in this project we will examine the consequences of these different policies on immigrants’ health and the spillover effects of this onto natives’ health. In order to analyze how these policy changes affected immigrants’ and natives’ health, we need a large data set that allows us to observe state of residence, county of residence, and country of birth, as well as the number of years foreign-born immigrants have been in the U.S., because the policies will likely have heterogeneous effects depending on whether individuals are recent immigrants or not. The NHIS has a large sample size, the necessary demographic information, and detailed health, health insurance, and health care utilization information. The NHIS will allow us to have more power and examine more relevant outcomes than with public use data. Unfortunately, the public use NHIS data do not have state of residence or county of residence, which is the information necessary to take advantage of the policy variation, so we need access to the restricted use version of the data. 

Economic theory suggests that trading firms should be particularly dependent on external and internal financing due to issues such as long shipment lags, currency and regulatory risks, and high initial fixed costs to set up a distribution network abroad. This implies that firms with tighter credit constraints should be significantly less likely to export to a given destination (extensive margin) and have significantly lower export shares (intensive margin) than comparable, but financially unconstrained, firms. However, existing strategies fail to document how financial constraints determine trade shares because they lack firm-level microdata. By making use of linked Economic Census, Quarterly Financial Report, LFTTD, and Compustat data, we can more directly measure firm-level constraints and determine how biased standard techniques for assessing this relationship at the industry-level have been. Results will better characterize the true contribution of financing constraints on trade shares in the US economy. 

In this project, we estimate the association between residential coethnic concentration, socioeconomic status, and neighborhood turnout, and individual level adult health outcomes and risk factors according to race/ethnicity/nativity, focusing on four main groups: foreign-born Mexican immigrants, US-born Mexican-Americans, US-born Non-Hispanic Blacks, and US-born Non-Hispanic Whites. Prior work on residential area “effects” has only examined particular cities with high concentrations of (especially Mexican Americans). We use restricted-use data from the continuous National Health and Nutrition Examination Survey (NHANES) 1999-2010 survey cycles linked to 1970,1980, 1990, 2000, and 2010 Census as well as to 2005-2009 ACS data, at Census Tract levels using the Longitudinal Tract Database put together by the Spatial Structures in the Social Sciences Center at Brown University. We also use more limited Block Group data from the 2000 Census and 2005-2009 ACS, downloaded directly from American Fact Finder. These data have already been merged and used at the Hyattsville, MD NCHS RDC. We are requesting approval for Census-RDC use to transfer the data when the Rocky Mountain RDC open in Boulder, CO, in early 2017. 

Extreme heat is the leading cause of annual weather-related deaths in the United States and is expected to increase in duration, intensity, and frequency with climate change. It is well-known that extreme heat can have disparate effects on different geographic areas due to varying levels of acclimatization and vulnerability in each community. Therefore, some heat-related deaths may be prevented with proper resources to educate the public about extreme heat specific to their community. In this project, we first aim to find the unique relationship between temperature and mortality in every county within the state of Virginia using NCHS’s National Vitals Statistics System data through the RMRD a nd then secondly, to use the exposure-response relationships found in the first step to predict the number of future heat-related deaths associated with daily weather forecasts in real-time. This prediction of heat-related mortality will be made available online through an interactive web map which shows the projected death counts (not any of the original and sensitive health data used in the analysis). The use of an interactive web application can provide quick and effective information about a resident’s community and allow for better understanding of the severity of an impending extreme heat event. We will model the historical temperature-mortality relationship using standard time-series methods adjusted for important confounding variables such as seasonality, day of week, and long-term trend. The daily temperature data will be derived from NOAA’s weather stations and daily non-accidental deaths from the National Vitals Statistics System (NVSS) data set for the period 1989-2015. The daily temperature data is publicly available online through NOAA’s National Centers for Environmental Information; however, the death data (containing exact date of death, county of residence, and cause of death) requires restricted data from the National Center for Health Statistics (NCHS)’s National Vitals Statistics System data set. After piloting the project for Virginia counties, our goal is to then expand the analysis to all counties in the United States. At this stage, we are just requesting the data for the state of Virginia and will submit an additional request when we get to that stage of the analysis 

 Heterogeneity in Health Returns to Medical Spending" style="regular"]This project estimates the health returns to medical spending and examines whether these returns vary with age, gender, and pre-existing health conditions. We start by studying how medical spending affects an individual’s mortality rate, physical health, and mental health by running instrumental variable regressions of health outcomes on medical spending. This enables us to quantify the average marginal health returns to medical spending in the population. Next, we test for age-specific marginal returns to medical spending by estimating the same relationships on different age groups in the data, and test for gender-specific returns by estimating the relationships on men and women separately. Lastly, we examine whether the returns to medical spending vary with pre-existing health conditions by comparing the marginal returns to medical spending for people that have been diagnosed with angina, arthritis, asthma, cancer, coronary heart disease, high cholesterol, diabetes, emphysema, high blood pressure, heart attack, other heart diseases, or stroke. 

 This analysis will provide needed public health surveillance on mortality involving stimulant drug substances. Little mortality data are provided in the literature for specific drug substances beyond the level of International Statistical Classification of Diseases and Related Health Problems, 10th Edition (ICD-10) coding. This analysis will calculate age-adjusted rates and trends over time for several prescription stimulant substances. A set of illicit stimulants, opioids, and nonsteroidal anti-inflammatory drugs (NSAIDs) are also examined to provide benchmark comparisons for substances expected to be commonly and uncommonly involved in death. Finally, polysubstance mortality will also be examined.

 Polysubstance use has higher risks of mortality, and multiple substances are frequently reported on death certificates. Due to the huge plethora of drugs that could be involved in mortality, polysubstance analysis can be overwhelming. However, statistical strategies are available for such high dimensional problems. Multivariate analyses can be useful in reducing the dimensionality and can provide insight into how variables cluster together according to latent variables. MCA has been used to investigate facial injuries in violent crime, trauma patterns in blast injuries, and malaria infection patterns. This proposal will analyze polysubstance drug mortality from the perspective of a multidimensional analysis. The primary strength of this approach is that many more drug substances can be analyzed concurrently than in traditional polysubstance analyses, which typically rely on ICD-10 codes and univariate proportions. The purpose of the analysis is to determine what patterns (if any) exist in which drugs are concomitantly present in fatal overdoses. Further, the analysis will show what characteristics are associated with the different dimensions of polysubstance overdose, which could be valuable to broader drug safety initiatives. A multivariate analysis is proposed to provide needed public health information on mortality involving multiple drug substances. Mortality data are incomplete in the literature for if and how polysubstance use of drug substances contributes to mortality. Multiple correspondence analysis (MCA) will examine how drug substances cluster together along latent variables; clustering of drugs within drug classes will also be examined. This analysis will also examine associations between demographic and other characteristics among drug substance clusters in mortality data. The primary objective of this analysis is to provide public health surveillance on mortality involving the clustering of drug substances. This study is descriptive, and no formal hypotheses are pre-defined. Death certificate information from Drug Mentions with Involvement (DMI) database will be analyzed to characterize how specific drug substances are involved at the same time.

The primary objective of this analysis is to provide public health surveillance of buprenorphine overdose mortality. This study is descriptive, and no formal hypotheses are pre-defined. Death certificate information from the Research Data Center (RDC) will be analyzed to calculate counts and proportions of deaths associated with buprenorphine and selected comparator opioid analgesics. Age-specific, geography-specific, and polysubstance fatalities will be characterized (with counts and proportions). Additionally, sensitivity analyses will be conducted to assess the impact the drug overdose definition and reporting specificity on rates. Counts and proportions will be obtained from the RDC. After data are obtained from the RDC, counts will be used to calculate population-adjusted and drug utilization-adjusted rates of mortality for buprenorphine and comparator substances. Calculations of rates will be at the national and state level, and therefore individual-level linkages to denominator data (population and drug utilization) within the RDC are not necessary.

The immigrant health advantage suggests that, despite significant socioeconomic disadvantage, immigrants report better health than U.S.-born counterparts. The immigrant health advantage has been shown in Hispanic-origin immigrant populations in the United States, with little focus on other racial/ethnic groups. In a previous study, I examined this immigrant health advantage as it pertained to overweight, obesity, hypertension, and diabetes in African-origin black immigrants relative to U.S.-born non-Hispanic blacks. Additionally, to investigate health deterioration associated with duration, I compared the health of African-origin black immigrants to non-Hispanic white and Mexican immigrants. I analyzed data on adults aged 18-85+ in the 2002-2017 National Health Interview Survey using multinomial (overweight; obesity) and binomial (hypertension; diabetes) logistic regressions. Findings provide support for the immigrant health advantage and suggest that, relative to U.S.-born non-Hispanic blacks, African-origin black immigrants are at lower odds for obesity (p<0.001), hypertension (p<0.001), and diabetes (p<0.001), regardless of duration. Further, I find that African-origin black immigrants report similar health status and health deterioration when compared to non-Hispanic white and Mexican immigrants. These findings provide new insights into the health of African-origin black immigrants, a rapidly growing proportion of the immigrant population of the United States. This study only answers a portion of questions. Africans, and black immigrants as a whole, represent a largely heterogeneous population. As such, it is essential that researchers disaggregate by country of origin to better understand differential health outcomes. I seek to address this by identifying country of origin groups using NHIS and NHANES. Additionally, to better understand the influence of period and cohort effects, I seek to analyze year of arrival data in the same two datasets. Country of origin and year of arrival are restricted and only available through the Research Data Center.

This project aims to provide a more comprehensive understanding of the change in U.S. income inequality and labor market dynamics by characterizing the nation’s income distribution using an innovative set of distributional forms, namely a mixture model with a finite number of components. We use data from the Census Bureau’s Current Population Survey Annual Social and Economic Supplement (CPS ASEC) to construct fine-grained, highly populated frequency histograms that allow us to estimate a multi-component statistical model of the income distribution, which better capture the three generative processes that prior researchers have speculated shape the income distribution. Findings will identify the crucial characteristics of each of these processes and inform empirically-based theorizations of different modalities of income appropriation. This will ultimately cast light on the occurrence and consequences of unemployment or partial engagement with labor markets, segmentation and stratification in labor markets, and the relationship between functional and individual income.

Although U.S. early life mortality rates are magnitudes lower than later life mortality rates and have continued to decline (Murphy et al. 2013), they are unacceptably high, particularly for some population subgroups. Nonetheless, social demographic and epidemiological research on early life mortality, especially beyond infancy, has been scarce over the past several decades. This is most likely because research attention has focused on other stages of the life course given that deaths are highly concentrated at older ages, and because there are very few large, nationally representative U.S. data sets that facilitate research on early life mortality. However, U.S. infants, children, adolescents, and young adults are growing up in a context of widening socioeconomic inequality and rapidly changing family structures. Overall, such social and economic changes may differentially affect early life mortality risks, with particularly harmful consequences for the most vulnerable population subgroups. But very little recent research has examined early life mortality disparities and trends in the context of these broad social and economic changes. We plan to use the recently-released National Health Interview Survey Linked Mortality Files together with: multivariate linear regression, multivariate logistic regression, multivariate hazard models, and multivariate Poisson analyses to examine patterns and trends in early life mortality within the United States. 

The Spatial Sciences Census Research Node (SSCRN) will foster a connection between the spatial and the survey sciences. This bridge will yield both immediate and long‐term benefits for the estimation, dissemination, and usability of the small area statistics produced by the Census Bureau. Small area statistics describe the character of the population within small geographic zones, such as census tracts. Small area estimates from the American Community Survey (ACS) are imprecise. Historically, very little attention has been paid to the geographic distribution of populations within these small areas. The basic research of SSCRN will increase knowledge about the organization of the American population within small geographic areas for the purpose of improving small area estimates. Apart from the impractical solution of increasing the sample size, the only way to reduce the uncertainty of survey estimates is to utilize ancillary information about the population. SSCRN will exploit new forms of geographic information and recent advances in spatial statistics to make small area estimates more accurate. In addition to improving small area population estimates, an improved understanding of the geographic micro‐structure of the American population is of broad scientific interest and may expand knowledge about socio‐spatial processes like segregation and neighborhood effects. SSCRN will bridge the spatial and the survey sciences through basic research that addresses the current needs of the Census Bureau and yields tangible products for researchers. SSCRN will develop and disseminate training materials, software, and research for a broad community of Census data users and producers.

SSCRN will convene a series of interdisciplinary meetings designed to identify research frontiers at the intersection of the spatial and the survey sciences. The meetings will be used to design and refine a novel curriculum and workshop series aimed at developing a cohort of scientists that are capable of establishing and elucidating links between the spatial and survey sciences. SSCRN will develop software tools that will enhance the usability of ACS small area estimates by allowing users to intelligently combine tracts to reduce uncertainty in variables of interest. SSCRN will educate and train a large group of students, post docs, and scientists through meetings, structured mentoring, and workshops. The impact of these activities will be multiplied through design and dissemination of a broadly accessible model curriculum and training workshop. Finally, SSCRN will have a broad impact by releasing a novel form of spatial microdata that will improve small area estimation and enable researchers to explore the spatial structure of the American population. This activity is supported by the NSF‐Census Research Network funding opportunity

The main paper is titled "Multinationals, Markets and Mark-Ups" and is co-authored with Wolfgang Keller of the University of Colorado. This paper contributes to a literature that measures the evolution of market power by documenting how the mark-ups charged by firms have changed over time and across industries. While it is difficult to summarize the literature succinctly, it is widely understood that the market shares of the largest firms have risen and that average mark-ups have risen most in those industries that have concentrated the most. Much of the evidence has been piecemeal, however, and sorting out causality has been controversial. We develop a technique for measuring market power that relies on relatively weak assumptions and only requires data that is readily available in the foreign investment surveys. We then use the BEA's affiliate level data to measure the strength of competition across countries, industries, and time. This data has many strengths over those used in other studies because we can follow the same firms in multiple markets allowing us to control for firm-year fixed effects. We find that the mark-ups being charged by the affiliates of U.S. multinationals have risen substantially over the period 1999-2014 and that mark-ups have increased the most in less developed countries.

This project seeks to quantify the relationship between offshoring activities and the rate of innovation of U.S. firms. While the phenomenon of offshoring is not new, it has become central to the expansion strategies of U.S. firms only over the past fifteen years. At the same time, empirical work that documents its effect on innovation is virtually non-existent. While it is well-recognized that the shift of production capacity overseas affects the number of jobs in the U.S., it is not clear to what extent offshoring affects the rate of innovation of U.S. firms, which will impact future domestic job growth. To analyze this question, we will compare the innovation rates of firms that offshore with those that do not, using China’s accession to the World Trade Organization (WTO) in 2001 as an external shock that generates a quasi-random sample of firms. The key challenge is to ensure that firms who do offshore are not too different from firms that do not offshore in terms of their determinants of innovation, which will require an appropriate comparison group that only Census micro-level data can provide.

Offshoring can have opposing effects on innovation depending on the extent and type of offshoring taking place. It can positively affect the firm’s innovation rate by either lowering production costs, allowing the firm to pool more resources towards innovation (“productivity effect on innovation”) or by enabling firms to tap into foreign pools of skilled labor. At the same time, the shifting of production capacities overseas may leak out U.S. technological knowledge to foreign competitors and lead to reduced knowledge spillovers within the firm. Building upon two previous studies that point to evidence of R&D spillovers to domestic firms from foreign-owned production (Keller and Yeaple 2009) and the potential knowledge costs from separating production facilities and firm headquarters (Keller and Yeaple 2013), this empirical study will attempt to disentangle these opposing effects and quantify the influence of offshoring on different measures of innovation, including R&D expenditures, patenting and trademarks.

The data source for our treatment-and-control group approach will be the Census of Manufacturing (CMF), the Annual Survey of Manufacturing (ASM), along with the Longitudinal Firm Trade Transaction Database (LFTTD). For our measures of innovation, we will rely primarily on R&D expenditure data from the Business R&D and Innovation Survey (BRDIS) and Survey of Industrial R&D (SIRD). This project will benefit the Census along several dimensions. In addition to providing a better understanding of plant characteristics and how they relate to offshoring, it will inform the Census Bureau on the international scope of R&D for many firms and improve the accuracy of the BRDIS by comparing locational outcomes and activities of innovation.