Center for Economic Studies Research Report: 2019 Research and Methodology Directorate
Issued June 2020 MISSION The Center for Economic Studies partners with stakeholders within and outside the U.S. Census Bureau to improve measures of the economy and people of the United States through research and the development of innovative data products.
HISTORY The Center for Economic Studies (CES) was established in 1982 on a foundation laid by a generation of visionaries both inside and outside the Census Bureau. CES’s early mission was to house data- bases on businesses, link them cross-sectionally and longitudinally, conduct economic research with them, and make them available to researchers.
Pioneering CES staff and visiting academic esearchersr began fulfill- ing that vision. Using these new data, their analyses sparked a revo- lution of empirical work in the economics of industrial organization.
Researcher access to these restricted-access data grew with the establishment of secure research data centers, the first of which was opened by CES in Boston in 1994. Today, there are such facilities located at dozens of universities and research organizations across the country.
In time, CES expanded its focus from data and research on busi- nesses to also include workers and households. Today, CES staff carry out empirical research on a wide array of subjects, leading to important discoveries in economics and other social sciences, improvements in existing Census Bureau surveys and data products, enhanced research databases, and new statistics and information products for public use.
ACKNOWLEDGMENTS Randy Becker coordinated the production of this report and wrote, compiled, or edited its various parts. John Voorheis and Nikolas Pharris-Ciurej authored Chapter 2. Andrew Foote and Lee Tucker authored Chapter 3. Other CES staff contributed updates used throughout.
Linda Chen and Faye Brock of the Public Information Office provided publication management, graphics design and composi- tion, and editorial review for print and electronic media. The Census Bureau’s Administrative and Customer Services Division provided printing management.
DISCLAIMER Research summaries in this report have not undergone the review accorded Census Bureau publications, and no endorsement should be inferred. Any opinions and conclusions expressed herein are those of the author(s) and do not necessarily represent the views of the Census Bureau or other organizations. All results have been reviewed to ensure that no confidential information is disclosed. Center for Economic Studies
Research Report: 2019 Issued June 2020 Research and Methodology Directorate
U.S. Department of Commerce Wilbur Ross, Secretary Karen Dunn Kelley, Deputy Secretary U.S. CENSUS BUREAU Steven D. Dillingham, Director SUGGESTED CITATION U.S. Census Bureau, Center for Economic Studies Research Report: 2019, U.S. Government Printing Office, Washington, DC, 2020.
U.S. CENSUS BUREAU Steven D. Dillingham, Director Ron S. Jarmin, Deputy Director and Chief Operating Officer John M. Abowd, Associate Director for Research and Methodology John L. Eltinge, Assistant Director for Research and Methodology Lucia S. Foster, Chief, Center for Economic Studies CONTENTS
A Message From the Chief Economist ...... 1
Chapters
1. 2019 News ...... 3
2. Improving Census Bureau Demographic Surveys Using Administrative Records ...... 13
3. Transitions Into the Labor Force: Using the Census Bureau’s National Jobs Frame to Measure College Graduates’ and Veterans’ Outcomes ...... 21
Appendixes
1. Overview of the Center for Economic Studies ...... 29
2. Publications and Working Papers by Center for Economic Studies Staff: 2019 ...... 31
3 Abstracts of Center for Economic Studies Working Papers by Census Bureau Staff: 2019 ...... 37
4. Center for Economic Studies Working Papers: 2019 . . . . . 51
5. Longitudinal Employer-Household Dynamics (LEHD) Partners ...... 55
6. Center for Economic Studies Organizational Chart (February 2020) ...... 59 This page is intentionally blank. A MESSAGE FROM THE CHIEF ECONOMIST
Collaborative teamwork enables the Center further from for Economic Studies (CES) to conduct researchers pathbreaking research and to develop new and program- innovative data products. Much of the work mers. Taking highlighted in this annual report represents the subject of many different types of teams. Following our Chapter 3 as overview of activities at CES in Chapter 1, an example, Chapters 2 and 3 illustrate how these part- some of the nerships lead to improved measures of our popularity nation’s economy and people and new data of PSEO and products. Chapter 2 provides an example of VEO stems partnerships across the U.S. Census Bureau from the user-friendly tools developed to with programmatic areas in support of demo- showcase the data. The development team, led graphic surveys including the National Survey by Matt Graham with Jody Hoon-Star, created of College Graduates, the Current Population tools that allow the data to be used in an intui- Survey, and the American Community Survey. tive fashion. The tool is hosted on our Web Chapter 3 discusses partnerships with exter- site managed by Heath Hayward and kept up nal stakeholders and experts in developing to date by Chaoling Zheng. The data infra- two new data products, the Post-Secondary structure for these products is supported by Employment Outcomes (PSEO) and Veterans the Local Employment Dynamics partnership Employment Outcomes (VEO). maintained through the efforts of Keith Bailey, Kimberly Jones, and Claudia Perez. A large While the chapters highlight the importance part of this project requires the acquisition of of partnerships between CES and other external datasets for which the team works groups, it is good to take a moment to reflect with the Economic Reimbursable Division. on the importance of collaborative teamwork Finally, all of the research and data products within CES. Since these are annual research released must go through the Census Bureau’s reports, it is not surprising that we focus on disclosure avoidance reviews for which we rely the work of researchers, but all CES staff on our work with the Disclosure Review Board. contribute to our research and development activities—some directly and some indi- Supporting a research office also includes rectly. For example, in addition to the many researchers who take on the tasks of editing researchers working on the Longitudinal our working paper series (Christopher Goetz), Business Database (LBD) and its public prod- coordinating our reviews (Scott Ohlmacher), uct, the Business Dynamics Statistics (BDS), running our seminar series (Emek Basker and for many years the main programmer for the Danielle Sandler), providing conference alerts LBD/BDS was Ronald Davis whose code is (Bitsy Perlman), and managing our mentor- still used throughout. ing program (Randy Becker). It also requires staff who handle our budget and agreements For large-scale projects that involve creating (Cheryl Grim, Donna Myers, Towana Nix, and new data products, the team expands even (Continued)
U.S. Census Bureau Center for Economic Studies Research Report: 2019 1 A MESSAGE FROM THE CHIEF ECONOMIST—Con.
Sierra Noland) and administrative staff (Dawn and eligibility for and participation in the Anderson, Rebecca Turner, and Deborah Supplemental Nutrition Assistance Program Wright). Appendix 6 of this report provides (SNAP) and Special Supplemental Nutrition the complete listing of CES staff. Assistance Program for Women, Infants, and Children (WIC). Our work to provide mortality We will continue research and development statistics to qualified researchers on approved activities to improve our existing data prod- projects continues as usual. ucts over the coming year. We are working to produce an enhanced version of the BDS. Thank you to everyone who contributed to The team working on business technology our annual report. Randy Becker compiled adoption continues to develop modules for and edited all of the material. Editorial review use in the Annual Business Survey. Teams also was performed by Faye Brock, and design continue their work on measures of pro- services and cover art production by Linda ductivity dispersion; business expectations Chen, both of the Public Information Office. and uncertainty; nonemployer statistics by Other contributors are acknowledged on the demographics; income and mobility statistics; inside cover.
Lucia S. Foster, Ph.D. Chief Economist and Chief of the Center for Economic Studies
2 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Chapter 1. 2019 News
THE CENTER FOR (or forthcoming) as journal public-use data products and ECONOMIC STUDIES articles or book chapters (see our various analysis and visual- EXPANDS ITS AREAS OF Appendix 2). Some of these ization tools that are discussed EXPERTISE newly published journal articles next. are highlighted on the next This year was our first full two pages. The accompany- RELEASES OF PUBLIC-USE year with an increased focus ing figure shows that CES staff DATA on research and development research is being published in CES continued to maintain and activities. In October 2018, many of the top peer-reviewed update its public-use data prod- research staff previously with journals in economics including ucts in 2019, including Business the Center for Administrative the American Economic Review, Dynamics Statistics, Business Records Research and Quarterly Journal of Economics, Formation Statistics, Quarterly Applications (CARRA) joined and Review of Economics and Workforce Indicators, LEHD the Center for Economic Studies Statistics. Origin-Destination Employment (CES), extending our expertise For more information about our Statistics, OnTheMap, OnTheMap in businesses and workers to researchers and our research, for Emergency Management, include people and households. including access to papers in Job-to-Job Flows, Post- This is also the first annual report our working paper series (which Secondary Employment to not discuss the activities of continues to include working Outcomes, and the Opportunity the Federal Statistical Research papers by FSRDC researchers), Atlas. In addition, 2019 saw the Data Center (FSRDC) network, visit our newly redesigned Web launch of a new, experimental which was founded by CES a site
U.S. Census Bureau Center for Economic Studies Research Report: 2019 3 NOTABLE 2019 PUBLICATIONS BY CES STAFF
“Do Institutions Determine Economic Geography? Evidence from the Concentration of Foreign Suppliers”
Fariha Kamal and Asha Sundaram Journal of Urban Economics Volume 110, March 2019, pp. 89–101
Do institutions shape the geographic concentra- product-country-specific measure of supplier tion of industrial activity? We explore this ques- concentration for U.S. importers. Results show tion in an international trade setting by examining that U.S. importers source in a more spatially the relationship between country-level institu- concentrated manner from countries with weaker tions and patterns of spatial concentration of contract enforcement. We find support for the global sourcing. A priori, weak institutions could idea that, where formal contract enforcement be associated with either dispersed or concen- is weak, local supplier networks compensate by trated sourcing. We exploit location and transac- sharing information to facilitate matching and tion data on imports by U.S. firms and adapt the transactions. Ellison and Glaeser (1997) index to construct a
“Local Labor Demand and Program Participation Dynamics: Evidence from New York SNAP Administrative Records”
Erik Scherpf and Benjamin Cerf Journal of Policy Analysis and Management Volume 38, Issue 2, Spring 2019, pp. 394–425
This study estimates the effect of local labor entrants and program rules brought on by the demand on the likelihood that Supplemental Great Recession, we find that fluctuations in labor Nutrition Assistance Program (SNAP) beneficia- demand in industries with high shares of SNAP ries are able to transition out of the program. Our participants—especially food service and retail— data include SNAP administrative records from change the likelihood of exiting the program. New York (2007 to 2012), linked at the person- Notably, estimated industry effects vary across level to the 2010 Census and linked at the county- race and parental status, with Black participants month-level to industry-specific, labor market being most sensitive to changes in local labor conditions. We find that local labor markets market conditions and mothers benefiting less matter for the length of time spent on SNAP, but from growth in local labor demand than fathers there is substantial heterogeneity in estimated and nonparents. We confirm that our results are effects across local industries. Using Bartik-style not driven by endogenous intercounty mobility instruments to isolate the effect of labor demand or New York City labor markets and are robust to and controlling for the changing composition of multiple specifications.
“Modeling Endogenous Mobility in Earnings Determination”
John M. Abowd, Kevin L. McKinney, and Ian M. Schmutte Journal of Business and Economic Statistics Volume 37, Issue 3, pp. 405–418
We evaluate the bias from endogenous job longitudinally linked employer-employee data mobility in fixed-effects estimates of worker- from the Longitudinal Employer-Household and firm-specific earnings heterogeneity using Dynamics infrastructure file system of the
4 Center for Economic Studies Research Report: 2019 U.S. Census Bureau U.S. Census Bureau. First, we propose two new explained by individual heterogeneity and residual diagnostic tests of the assumption that reduces the proportion due to employer and mobility is exogenous to unmodeled determi- match effects. To assess external validity, we nants of earnings. Both tests reject exogenous match our estimates of the wage components mobility. We relax exogenous mobility by mod- to out-of-sample estimates of revenue per eling the matched data as an evolving bipartite worker. The mobility-bias-corrected estimates graph using a Bayesian latent-type framework. attribute much more of the variation in revenue Our results suggest that allowing endogenous per worker to variation in match quality and mobility increases the variation in earnings worker quality than the uncorrected estimates.
“Race Matters: Income Shares, Income Inequality, and Income Mobility for All U.S. Races”
Randall Akee, Maggie R. Jones, and Sonya R. Porter Demography Volume 56, Issue 3, June 2019, pp. 999–1021
Using unique linked data, we examine income the lowest-income groups: Blacks, American inequality and mobility across racial and ethnic Indians, and Hispanics have lower within-group groups in the United States. Our data encompass inequality and immobility. On the other hand, the universe of income tax filers in the United low-income groups are also highly immobile States for the period 2000–2014, matched in terms of overall, rather than within-group, with individual-level race and ethnicity infor- mobility. These same groups also have a higher mation from multiple censuses and American probability of experiencing downward mobility Community Survey data. We document both compared with Whites and Asians. We also find income inequality and mobility trends over the that within-group income inequality increased period. We find significant stratification in terms for all groups between 2000 and 2014, and the of average incomes by racial/ethnic group and increase was especially large for Whites. The pic- distinct differences in within-group income ture that emerges from our analysis is of a rigid inequality. The groups with the highest incomes— income structure, with mainly Whites and Asians Whites and Asians—also have the highest levels positioned at the top and Blacks, American of within-group inequality and the lowest levels Indians, and Hispanics confined to the bottom. of within-group mobility. The reverse is true for
“What Drives Differences in Management Practices?”
Nicholas Bloom, Erik Brynjolfsson, Lucia Foster, Ron Jarmin, Megha Patnaik, Itay Saporta-Eksten, and John Van Reenen American Economic Review Volume 109, Issue 5, May 2019, pp. 1648–1683
Partnering with the U.S. Census Bureau, we and development, information and communica- implement a new survey of "structured" man- tion technologies, or human capital. We find agement practices in two waves of 35,000 man- evidence of two key drivers to improve manage- ufacturing plants in 2010 and 2015. We find an ment. The business environment, as measured enormous dispersion of management practices by right-to-work laws, boosts incentive manage- across plants, with 40 percent of this variation ment practices. Learning spillovers, as measured across plants within the same firm. Management by the arrival of large "Million Dollar Plants" in practices account for more than 20 percent of the country, increase the management scores of the variation in productivity, a similar, or greater, incumbents. percentage as that accounted for by research
U.S. Census Bureau Center for Economic Studies Research Report: 2019 5 In October 2018, the U.S. Census third quarter of 2004. With the The DiSP is the product of the Bureau released the 2016 release in October 2019, the expertise of BLS staff in measur- Business Dynamics Statistics series now extends to the third ing productivity and that of CES (BDS), which provides annual quarter of 2019. staff in working with business- statistics from 1976 to 2016 on level microdata. The accompa- For further details on the BFS establishment openings and nying text box illustrates these and to access the latest data, closings, firm startups and shut- data in use. visit
6 Center for Economic Studies Research Report: 2019 U.S. Census Bureau WHAT DRIVES PRODUCTIVITY GROWTH? Why does one manufacturer generate more revenue per Productivity Dierences Between More and Less total hours worked by its Productive Plants Within an Industry employees than another (Di erences shown are in revenue per total hours worked) manufacturer in the same industry? Take the computer equipment industry. The figure here shows that the more productive plants in Overall mean for that industry generate about manufacturing 400 percent more real rev- ( ) enue per total hours worked
than its less productive Shoes Cement Computers plants. Other industries, such Source US Bureau of Labor Statistics and US Census Bureau Dispersion as shoes and cement, have Statistics on Productivity smaller differences in produc- tivity but those differences are still large. In fact, within business conditions within plants that are getting smaller. the average manufacturing an industry. For example, if it These two drivers are likely industry in 2015, the latest year becomes harder for businesses connected since plants that for which data are available, to adjust their labor force in are more productive are more plants exhibit enormous differ- reaction to a change in busi- likely to grow. The DiSP can ences in real revenue per total ness climate, we may see a rise help us understand how large hours worked. More productive in dispersion. We may also see the gap is between more and plants, at the 75th percentile a rise in dispersion when there less productive plants in an of the productivity distribu- is a lot of experimentation and industry—painting a more tion, generate approximately innovation within an industry. complete picture for each 150 percent more real revenue These two examples highlight industry. The measured pro- per hour than less produc- that high or low dispersion is ductivity gap provides infor- tive plants that are at the 25th not necessarily either good or mation about the potential percentile. bad, but it is useful to know for improvement. If there is a whether an industry has high large gap between plants in an What drives these differences or low dispersion. It is a way industry, the less productive is an open question and one to find out information about plants have room to become that can be explored using the many factors that can lead more productive. They can the Dispersion Statistics on to productivity gaps between make changes such as real- Productivity (DiSP)—a new, businesses within the same locating resources, innovat- experimental data product industry. ing, or adopting new business recently released by the practices. U.S. Census Bureau and the There are essentially two driv- Bureau of Labor Statistics. ers of industry-level productiv- To download DiSP data and ity growth: (1) individual plants for methodological details and The size of the productivity become more productive, and research related to them, visit dispersion within an industry (2) more productive plants
U.S. Census Bureau Center for Economic Studies Research Report: 2019 7 This year’s releases update the analysis of the impacts of young/ of jobs by industry, worker age, base geography to TIGER 2018. old firms or small/large firms in earnings, and other worker char- They also implement an auto- relation to commuting patterns acteristics. To provide users with mated mechanism to identify and worker characteristics. Both the latest information on rapidly single-quarter firm underreport- LODES and OnTheMap can be changing events, OTMEM auto- ing issues and impute miss- used to answer a variety of ques- matically incorporates real-time ing worker earnings microdata tions on the spatial, economic, data updates from the National records. Imputations are applied and demographic aspects of Weather Service, Departments of to the historical data series as workplaces and home-to-work Interior and Agriculture, and the well as current data. flows. Federal Emergency Management Agency. See Chapter 2 of our CES staff continue to main- In August, version 6.7 of 2013 annual report for a more tain and improve the LEHD OnTheMap was released, adding detailed overview of OTMEM. Origin-Destination Employment an additional 2 years of LODES Statistics (LODES) and the data, extending availability from The latest release updates the OnTheMap application. LODES 2002 through 2017. This release American Community Survey is a partially synthetic data set also updates the base geogra- data to the 2013–2017 5-year that describes the geographic phy to TIGER 2018. estimates and updates the patterns of jobs by their employ- underlying LODES data to 2017. OnTheMap can be accessed at ment locations and residential OTMEM can be accessed at
8 Center for Economic Studies Research Report: 2019 U.S. Census Bureau sex, age, education, race, and Wisconsin–Madison were added, In 2018, in collaboration ethnicity. This year’s releases joining the University of Texas with researchers at Harvard incorporate data to the first system and public institutions University and Brown University, quarter of 2019. in Colorado. Newly introduced the Census Bureau launched tabulations include the destina- the Opportunity Atlas, a new These J2J data files and asso- tion industry and geography interactive tool providing ciated documentation are of employed graduates. In access to highly localized data available for download at November, the PSEO Explorer on social mobility. Using ano-
This year also saw the fur- Many graduate students use restricted-use U.S. Census Bureau ther expansion and develop- microdata in the Federal Statistical Research Data Centers for ment of the experimental their Ph.D. dissertation research, and many of these doctoral Post-Secondary Employment candidates are eligible to apply to the Center for Economic Outcomes (PSEO) statistics and Studies (CES) Dissertation Mentorship Program. Program visualization tool. First intro- participants are assigned one or more CES researchers as duced in 2018, PSEO provides mentors, who advise students on the use of Census Bureau earnings and employment out- microdata. Students are also given the opportunity to visit CES comes of post-secondary gradu- to meet with our research staff and present research in prog- ates, by institution, degree field, ress. This year, CES accepted two new participants into the and degree level for 1, 5, and program and, at the close of 2019, mentored 46 students from 10 years after graduation. This 25 different universities and a variety of different disciplines year, the University of Michigan– since the program began in 2008. Ann Arbor and University of
U.S. Census Bureau Center for Economic Studies Research Report: 2019 9 STATISTICAL AGENCIES • The Role of Recruiting • Job-to-Job Flows and COLLABORATE ON Intensity on Vacancy Yields: the Consequences of Job RESEARCH WORKSHOPS Evidence From a Large-Scale Separations Merge of Job Postings and • Opportunity Atlas BLS-CENSUS Research Survey Data Workshop The workshop was a success • Maternal Labor Dynamics: thanks to the researchers from On June 17, BLS and the Census Participation, Earnings, and both agencies who participated Bureau cohosted their ninth Employer Changes and especially to Kristin Sandusky annual workshop featuring • Labor Market Concentration, and Jim Spletzer (Census Bureau) empirical research by econo- Earnings Inequality, and and Sabrina Pabilonia and mists from both agencies. These Earnings Mobility Elizabeth Handwerker (BLS) who annual workshops are intended • Megafirms and Monopsonists: organized the workshop. Planning to encourage and nurture col- Not the Same Employers, Not for the tenth annual BLS-Census laboration between researchers the Same Workers Research Workshop is currently at BLS and the Census Bureau. underway. • New Experimental State-level William Beach, commissioner Labor Productivity Measures of BLS and Steven Dillingham, LED PARTNERSHIP • Analysis of Revisions to director of the Census Bureau WORKSHOP Aggregate Labor Productivity provided welcoming remarks. Measures The 2019 Local Employment This year’s workshop consisted Dynamics (LED) Partnership • The Gender Gap in of three themed sessions with Workshop was held at the Entrepreneurship two papers each—one from each Census Bureau on September agency—with discussants from • Migration From Sub- 4 and 5. Now in its twentieth the other agency. In addition, a National Administrative Data: year, this workshop has been a poster session of eight papers Problems and Solutions With key component in strengthen- was held. Workshop papers an Application to Higher ing the voluntary partnership included: Education between state data agencies and • Automation, Labor Share, • Nonemployer Statistics by the Census Bureau, leveraging and Productivity: Plant- Demographics (NES-D): A existing data in the development Level Evidence From Blended-data Approach of new sources of economic and U.S. Manufacturing to Demographic Business demographic information for Statistics policymakers and data users. • Improving Estimates of Hours The workshop brings together Worked for U.S. Productivity • Driving the Gig Economy key stakeholders including state Measurement
Steven Dillingham, director of the Census Bureau, and William Beach, commissioner of BLS, offer welcoming remarks at the ninth annual BLS-Census Research Workshop.
10 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Labor Market Information direc- tors, data analysts and data providers at state and federal agencies, nonprofit organiza- tions, businesses, and other users of Longitudinal Employer- Household Dynamics (LEHD) data products. They discuss the latest product enhancements, discover how their peers are using the data, and learn about the research that will shape future improvements.
Topics addressed by presenta- tions, panel discussions, and roundtable sessions at this year’s John Friedman of Brown University gave the keynote address at the 2019 workshop included the manu- LED Partnership Workshop. facturing sector, transportation planning, various other state and offered training sessions on the examine income inequality and local uses of LEHD data, and J2J, OnTheMap, and the new social mobility. PSEO. John Friedman, professor making the most of published Presentations and materials from of economics and international LEHD data. CES staff also dis- the 2019 workshop (and those affairs and public policy at Brown cussed newly available public-use from previous years) can be University, gave the workshop’s data and tools, products under found at
THE LED WEBINAR SERIES The U.S. Census Bureau and the Local • What Causes Labor Turnovers to Vary Employment Dynamics (LED) Partnership, in (Kristin McCue, CES) collaboration with the Council for Community • What May Be Driving Growth in the “Gig and Economic Research, hosts an ongoing Economy?” (Kristin Sandusky, CES) Webinar series focusing on uses of Longitudinal • D.C.'s Startup Scene, Part II: Opportunity Employer-Household Dynamics (LEHD) data. In Costs (Shirin Arslan, D.C. Policy Center) 2019, the following Webinars were held: • Recent Updates to LODES and OnTheMap • Housing and the Tech Boom: Using LEHD (Matthew Graham, CES) and Zillow Data to Understand Housing • Recent Updates to Job-to-Job Flows Market Impacts (Aaron Terrazas, Zillow) Explorer: Job Hopping Across Cities (Heath • Older People Working Longer, Earning More Hayward, CES) (James Spletzer, CES) To view recordings of these and earlier • Job Growth and Spatial Mismatch Between Webinars, visit
U.S. Census Bureau Center for Economic Studies Research Report: 2019 11 CES STAFF RECEIVE differential privacy to protect develop new products using RECOGNITION the underlying data. blended data. Their work led to innovative statistical products In March, Andrew Foote received At the same ceremony, Sonya state agencies use to improve a Bronze Medal Award for his Porter, Mark Leach, Benjamin and increase efficiencies in pub- work in developing and launch- Cerf, Brad Foster, Rachel lic assistance programs. ing PSEO, a set of experimen- Shattuck, and other team tal statistics on the earnings members received a Bronze The Bronze Medal Award for and employment of graduates Medal Award for their efforts to Superior Federal Service is the of particular postsecondary acquire Supplemental Nutrition highest honorary recognition institutions. These are just the Assistance Program and Women, given by the Census Bureau. second set of statistics released Infants, and Children program by the Census Bureau to utilize administrative records and
The State Data Acquisition and Product Development Team created innovative statistical products that help state agencies improve their public assistance programs.
12 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Chapter 2. Improving Census Bureau Demographic Surveys Using Administrative Records John Voorheis and Nikolas Pharris-Ciurej, Center for Economic Studies
The U.S. Census Bureau con- The Demographic Research weighting, and data production. ducts dozens of household Area in the Center for Economic Here we will discuss our work surveys covering a wide variety Studies (CES) has been at the on several surveys, includ- of topics, including housing, forefront of efforts to utilize ing the National Survey of education, health, and expen- administrative records, in coor- College Graduates, the Current ditures. These surveys provide dination with survey produc- Population Survey (CPS), and the backbone for measuring the tion teams and subject-matter the American Community people and households of the experts in the Census Bureau’s Survey (ACS). United States, providing vital Demographic Directorate.1 This information to policymakers, chapter provides an overview of USING ADMINISTRATIVE researchers, and other stake- some of our efforts to improve RECORDS IN SAMPLING holders. Conducting these survey measurement. In general, Most household surveys con- surveys, however, has become our approach involves: (1) work- ducted by the Census Bureau increasingly challenging as ing with survey operations staff use the Census Bureau’s Master response rates have declined, to identify areas where admin- Address File (MAF) as their costs have increased, and wor- istrative records could be used, sampling frame. However, many ries about measurement error (2) performing basic research on surveys have either a nar- have compounded. administrative records’ fitness rower target universe than all for use, (3) prototyping potential While no singular approach can U.S. households or a need to uses for administrative records, fully reverse these trends, the oversample certain sociode- and (4) providing support for use of administrative records mographic subgroups. In these survey operations staff to incor- and third-party data in the sur- cases, it is possible to use porate administrative records, vey production process is a very administrative records to flag first experimentally and then as promising avenue for improv- households that are in these tar- ongoing production activities. ing survey measurement and get groups and thereby improve reducing respondent burden and In this chapter, we will describe sampling efficiency. Our work on survey costs. Indeed, the use of several case studies illustrating the National Survey of Children’s administrative records is a prior- this approach. Cases fall into one Health (NSCH) is one case study ity not just for convenience, but of two categories. In one cat- of this approach. because Title 13 of the U.S. Code egory, administrative records are The NSCH is a survey fielded mandates their maximum use used to enhance precollection by the Census Bureau and where possible. Here, adminis- activities through better sam- sponsored by the Maternal trative records generally refer to pling, and here we will highlight and Child Health Bureau of the microdata records contained in our work on the National Survey Health Resources and Services files collected and maintained of Children’s Health. In the other Administration, a unit in the by government agencies and category, administrative records Department of Health and commercial entities for the are used to enhance postcollec- Human Services. The NSCH is purpose of administering pro- tion activities through improve- designed to provide nationally grams and providing services. ments in editing, imputation, and state-representative data Data acquired from commercial on the physical and emotional entities specifically are often 1 Some of the work described in health of children under the age referred to as “third party” data. this chapter began in the Center for Administrative Records Research and of 18. Until 2015, the NSCH was Applications, which merged with the Center for Economic Studies in October conducted as a random-digit 2018.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 13 dial phone survey, conducted they are often undercounted in The MAF-ARF is another com- about every 4 years by the traditional household surveys or posite administrative records National Center for Health censuses, requiring additional file, produced annually, which Statistics. Since 2015, the Census efforts to accurately capture combines information on Bureau has been the data col- them in survey or census data addresses (MAFIDs) from lector, and the survey has been collections. multiple sources, including the substantially redesigned. Most National Change of Address In order to improve the effi- notably, after 2015 the NSCH data set from the U.S. Postal ciency of the NSCH sampling became an address-based Service, IRS Form 1040s and IRS process, we use administrative survey, conducted annually, and Form 1099 and W-2 information records to assign addresses consolidating questions from returns, Medicare enrollment to different strata that can be the related National Survey of data, and multiple data sets sampled at different rates based Children with Special Health from the Department of Housing on the likelihood that a child Care Needs. and Urban Development. After is in the household. Of course, linking the Numident to the Since 2015, the NSCH has been the first and most powerful CHCK and to the MAF-ARF, we conducted in two parts. First, way of identifying households flag all addresses that have at sampled addresses are sent likely to have children is to use least one child and assign them a short “screener” survey to administrative records contain- to stratum 1. We additionally determine if there is a child in ing information on the location retain information on the count the household. Then, households of children, which we do by of children at each address, by that return these screeners and combining three different data age bins that is then used for have a child present are sent a sources. designing oversampling rates. more in-depth topical survey First, we use the Numident— module that can be either done Although stratum 1 is a very an administrative records file online or returned via a paper powerful tool for identifying sourced from the Social Security form. households with children, it is Administration that provides not perfect, nor does it com- The difficulty in conducting information on the birthplace, pletely identify all children. the NSCH as an address-based birthdate, and other demo- Specifically, in 2018, only about survey is mainly due to its graphic characteristics for all 80 percent of ACS households target population. The NSCH is individuals with social security flagged as stratum 1 actually had designed to cover only individu- numbers—to identify all chil- a child under 18 years, and only als under the age of 18, how- dren under the age of 18 at the about 70 percent of all house- ever only about a quarter of all start of survey data collection. holds with children in the ACS households nationwide have Second, we use a composite were in stratum 1.2 Therefore, a child. Because of this, sim- administrative records data set relying on stratum 1 alone would ply sampling from the Census called the Census Household be less than ideal for coverage Bureau’s master list of all unique Composition Key (CHCK), which purposes, and randomly sam- addresses in the United States combines family relationship pling from nonstratum 1 house- (i.e., the MAF) is very ineffi- information from the Numident, holds would present the same cient—requiring the mailing of IRS Form 1040 data, and the cost concerns that necessitated many screeners relative to the decennial census, to identify the creation of stratum 1 in the desired number of completed the parents of each child in the first place. To bridge this gap, surveys. The relative scarcity Numident. Finally, we link each we have developed a modeling of households with children is child (and their parents) to the and optimization approach to further complicated by the fact Master Address File Auxiliary that households with children Reference File (MAF-ARF) to 2 The Census Bureau Disclosure have higher-than-average rates obtain their address in the year Review Board has cleared the statistics in this paper for public dissemination, of residential mobility. In addi- of data collection. DRB approval numbers CBDRB-FY2020- tion, very young children are CES010-008, CBDRB-FY2020- CES010-001, and CBDRB-FY18-433, a hard-to-count population; respectively.
14 Center for Economic Studies Research Report: 2019 U.S. Census Bureau allow for efficient sampling of are any adults between the of the child-present score for the nonstratum 1 households ages of 20 and 50, whether each state that makes this true while maintaining adequate there are opposite-sex coresi- and apply these cutoff values coverage. Specifically, we divide dents in the household between to the entire MAF. After con- the nonstratum 1 households the ages of 20 and 50, hous- structing the stratum flags, we into two substrata (2a and 2b) ing tenure, number of children then pass them to the NSCH so that stratum 2b contains in the census block group, survey operations team, who fewer than 5 percent of all chil- and the number of IRS Form then design (high) sampling dren. Sampling can then pro- 1040 child exemptions in the rates from stratum 1 and (lower) ceed using only stratum 1 and household and in the 9-digit sampling rates from stratum 2a, 2a, which, together, contain 95 ZIP Code. We then train this while ignoring stratum 2b. percent of all children. model separately for each state This sampling strategy has been on the most recently available To do this, we developed a quite productive, resulting in a ACS microdata, augmented statistical model that leverages higher proportion of successful with information from adminis- data from previously mentioned screeners—and hence a lower trative records and third-party administrative records (includ- number of overall screeners data. We then use this model to ing the MAF-ARF and IRS Form needed to reach the desired calculate a “child-present” score 1040s), third-party data from sample size—than otherwise for all addresses on the MAF. several commercial vendors, would be the case. Additionally, and aggregate neighborhood Our goal is to create a stratum as newer administrative records (census block group) informa- 2a that is as small as possible and survey data have become tion from the ACS to predict with the constraint that 95 available, it has been possible to whether a household has one percent of children in each state test the coverage of the stratum or more children. This model are covered by stratum 1 and 2a flags, to verify that the model- considers whether there are (or that no more than 5 percent ing and optimization step (using women of childbearing age in of children are in stratum 2b). the slightly older data) is not the household, whether there To do so, we identify the value systematically missing some
Figure 2-1. Coverage Rates of the 2018 NSCH Stratum Flags
Source: U.S. Census Bureau, 2018 American Community Survey and the 2018 National Survey of Children’s Health Stratum Flags.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 15 children. For example, Figure number of surveys. Even surveys occupation, employment, and 2-1 summarizes the results of with a broader target universe social safety net programs to one such audit, comparing the than the NSCH may benefit, as name a few. An excellent exam- coverage of the 2018 NSCH stra- even nationally representative ple of this work is our efforts tum flags to the 2018 ACS. Of surveys often need to overs- to link the National Survey of households with children at valid ample certain subpopulations— College Graduates (NSCG) and (deliverable) addresses in the a task that can be improved administrative records contained ACS, in most states, more than through the use of appropriate in the Longitudinal Employer- 95 percent were in stratum 1 or administrative records. Household Dynamics (LEHD) 2a, indicating that the stratum files, which gives us the ability flags are performing in line with USING ADMINISTRATIVE to compare NSCG responses expectations. RECORDS IN DATA to questions about earnings, VALIDATION employment, and employer char- Going forward, there are two acteristics, such as industry and key areas for further refine- Besides the use of administra- number of employees, to values ments to our methodology. tive records to improve survey reported by employers to their First, postcollection analysis has sampling, CES staff have also states’ unemployment insurance suggested that, although the applied administrative records system. Our comparison of earn- stratum flags have good cover- to the task of improving surveys ings across the earnings distribu- age overall, they appear to cover post data collection. One post- tion reveals a striking pattern, very young children at lower collection activity includes data as seen in Figure 2-2. We find rates than older children. Efforts quality and validation exercises. that individuals with low admin- to either improve coverage In particular, we have linked istrative records earnings report or adjust for this undercover- administrative records data to relatively higher earnings on the age are the subject of ongoing various surveys to understand survey, while individuals at the research. A second, somewhat how the two data sources align. top of the income distribution related interest involves incor- This can be a first step toward report substantially lower earn- porating an even larger array possibly removing questions ings on the survey. of administrative records and from a survey, for some or all third-party data sources into respondents, or using the linked Related validation and alignment our process, such as informa- administrative records values work is currently underway that tion from state-administered in editing or imputation proce- examines income measurement social safety net programs dures. To take one example, we in the ACS. Using an expanded like Supplemental Nutrition have collaborated with other set of tax information made Assistance Program, as well as Census Bureau staff on research accessible through a joint sta- newly available housing data on examining the feasibility of tistical project with the Internal home values, mortgages, and replacing housing questions on Revenue Service (IRS), CES housing characteristics. the ACS and American Housing researchers, in collaboration with Survey with available third-party Census Bureau colleagues in While we have focused here on data. the Demographic and Decennial just one survey, the essence of directorates, are investigat- this approach—flagging house- So far, most of our efforts in ing how responses to the ACS holds of interest by combining this area have focused on data income questions (specifically information on the characteris- validation work, examining a self-employment, investment tics of individuals with admin- wide variety of topical areas income, earnings, and retirement istrative records on individual on numerous surveys including income) compare with detailed residences—offers a flexible demographic measures, edu- information from tax records. way to improve sampling for a cational attainment, income,
16 Center for Economic Studies Research Report: 2019 U.S. Census Bureau that will be representative of the Figure 2-2. population of interest, such as Comparison of Earnings Reported in the NSCG and the noninstitutionalized popula- Administrative Records Earnings From LEHD tion in the case of the CPS or the population under the age of 18 in the previously discussed NSCH. One important input into the final weights used by most surveys is a factor to account for nonresponse, which is con- structed by splitting all sampled cases into cells based on observ- able characteristics.
In most Census Bureau house- hold surveys sampled from the MAF, the number of observ- able household characteristics is limited and, therefore, the nonresponse cells are not as detailed as they could be, poten- tially resulting in uncorrected nonresponse bias. However, Source: U.S. Census Bureau, 2010 National Survey of College Graduates and by integrating administra- 2011 Longitudinal Employer-Household Dynamics Snapshot. tive records data, it is possible to have much more informa- tion about responding and To date, such comparisons there may be tax administration nonresponding households, have not been possible, and we benefits, including better under- allowing for more detailed expect that this work will yield standing the nonfiler population, nonresponse cells. We are cur- important insights into measure- as well as the aggregate report- rently working with our Census ment error and improvements in ing behavior of self-employed Bureau and Bureau of Labor measurement on both sides of individuals. Statistics colleagues on research the comparison. This is particu- to incorporate administrative larly the case for surveys like USING ADMINISTRATIVE records, such as income from the ACS, which tend to have RECORDS IN WEIGHTING IRS Form 1040s as well as the high item nonresponse rates for previously described MAF-ARF Another postcollection use questions about income. And and third-party housing data, of administrative records and because ACS data are used to into the nonresponse adjustment third-party data is to improve allocate more than 675 billion factors for both the Consumer survey weights to better address dollars in state and federal funds Expenditure Survey and the CPS. nonresponse bias. Sample sur- annually, improving measure- We expect this research will veys conducted by the Census ment in key survey items, such be broadly applicable to other Bureau construct survey weights as income, can have large real- household surveys. in order to produce estimates world impacts. Additionally,
U.S. Census Bureau Center for Economic Studies Research Report: 2019 17 USING ADMINISTRATIVE information available in admin- just administrative tax records RECORDS TO CREATE NEW istrative records, albeit on a (these can only measure college STATISTICS more limited set of variables. attendance but not comple- One such example is a project tion). The result has been new A final area of focus for CES we are conducting with the ACS insights into patterns of educa- is the production of entirely Office that employs data from tional attainment. new statistics by blending data ACS respondents from 2006 from household surveys, the Specifically, Figure 2-3 shows to 2018 and the universe of IRS decennial census, third-party the high school completion, Form 1040 data to understand data, and various administra- college attendance, and col- trends in educational attain- tive records sources, thereby lege completion rates for 24- to ment. In particular, we have expanding the scope of 26-year-olds whose parents linked data from 24- to 26-year- information available in any were in the bottom versus top old ACS respondents to their one source without increas- tercile of the income distribu- parents’ tax return information ing respondent burden or tion when they were in high when they were in high school survey length. One way to school. We see that high school (i.e., when aged 16 to 18). This construct such blended data completion and college atten- use of blended ACS and IRS products is to combine the dance have both increased for data improves both on what rich cross-sectional informa- low-income children, reducing can be done with just survey tion in a household survey income gaps, but the gap in data (relatively few 25-year- with the extensive longitudinal college completion has actually olds live at home) and with
Figure 2-3. Trends in Educational Attainment by Family Income in the ACS
Low Middle High High school completion College attendance College completion (24−26−year−olds) (24−26−year−old high school completers) (24−26−year−old college attenders)
.04 .04 .04
.02 .02 .02
0 0 0
−.02 −.02 −.02
−.04 −.04 −.04 1983 1986 1989 1983 1986 1989 1983 1986 1989 Year of birth Year of birth Year of birth
Source: U.S. Census Bureau, 2006-2017 American Community Survey, 1998-2015 IRS Form 1040 data, and Census Numident.
18 Center for Economic Studies Research Report: 2019 U.S. Census Bureau widened slightly. Additional work are concerns about misreporting cycle and lead to less costly data examines how these income on surveys. collection, better data quality, gradients in educational attain- and/or richer data, all without A second limitation of admin- ment vary by sociodemographic increasing respondent burden or istrative tax data is that it and neighborhood character- even decreasing such burden. only provides information for istics. This work complements resource-sharing groups as Despite our successes, there is both existing data products defined by tax law. That is, still much research and planning from the National Center for an IRS tax unit, i.e., everyone that must occur before we are Education Statistics, as well as appearing on a single IRS Form able to exploit the full potential other Census Bureau blended 1040, can differ from usual of these innovative administra- data products, such as the household or family concepts. tive records-based approaches. Opportunity Atlas, which was For example, two cohabiting We look forward to continuing highlighted in last year’s annual unmarried partners will not share our work with collaborators report. a tax return, but they would be within the Census Bureau and Another type of blended data in the same survey household across the federal government, product combines survey con- or family. By combining the as we incorporate administrative tent and administrative records best features of these two data records into survey operations at a single point in time. An sources, it is possible to con- and create blended data prod- example of this that we are struct a conceptually attractive ucts that provide new, timely, currently working on uses ACS blended measure of household valuable information to the and federal income tax data pretax posttransfer income. With American public. to produce new subnational these new blended income data, The Census Bureau is tasked statistics on the distribution we have begun to construct with providing data on the of income. One limitation of new statistics on the distribution nation’s people and economy, administrative tax data is that of income for states and other and administrative records have it only provides information on subnational geographies. proven to be an increasingly income that is relevant to the tax important way for us to fulfill authorities. On the other hand, CONCLUSION this mission. As the Census household surveys, such as the Throughout this chapter, we Bureau continues to explore the ACS, provide broader definitions have highlighted various efforts full potential of administrative of income—including in-kind and to develop new and innovative records, CES will continue to be cash transfers—and also allow ways to incorporate administra- at the forefront of these efforts. for the study of household or tive records into survey mea- family income, although there surement. These efforts focus on all phases of the survey life
U.S. Census Bureau Center for Economic Studies Research Report: 2019 19 This page is intentionally blank. Chapter 3. Transitions Into the Labor Force: Using the Census Bureau’s National Jobs Frame to Measure College Graduates’ and Veterans’ Outcomes Andrew Foote and Lee Tucker, Center for Economic Studies Through the Local Employment a number of research stud- information on long-run earnings Dynamics (LED) partner- ies have shown that the initial and employment outcomes. In ship with states, the District entry conditions of workers particular, the Post-Secondary of Columbia, and U.S. territo- have long-run impacts on earn- Employment Outcomes (PSEO) ries, the U.S. Census Bureau ings outcomes for both college contains data on the career has developed a national jobs graduates and veterans (e.g., outcomes for graduates of post- frame and data products includ- Oreopolous et al., 2012 and Zou secondary institutions, while the ing the Quarterly Workforce 2018). Furthermore, some of Veterans Employment Outcomes Indicators, OnTheMap, and this research has examined how (VEO) contains earnings out- Job-to-Job Flows. Moreover, career trajectories differ based comes for veterans who have the unique features of this data on an individual’s field of study left the armed forces. These data infrastructure, based on worker- or training background (Altonji products are possible through level employment and earnings et al., 2016). Until recently, new partnerships with univer- data from state unemployment Census Bureau public-use data sity and military stakeholders, insurance records, have allowed products provided no informa- respectively, expanding on the researchers, both within the tion on these long-run out- Census Bureau’s long history Census Bureau and through the comes. Moreover, because the of fostering productive part- Federal Research Statistical Data demographic data contained in nerships. Moreover, through Center network, to study numer- administrative records obtained the novel application of formal ous important economic ques- through the LED partnership are differential privacy techniques, tions that might otherwise be limited, policymakers (including each product is able to provide unanswerable. While other large our LED partners) have been detailed information of interest data sets permit researchers to unable to analyze and under- to our partners, policymakers, construct limited snapshots of stand the career trajectories and the general public, while labor market outcomes for par- of the groups of individuals of also upholding our legal obliga- ticular subpopulations of inter- importance to them. tions to protect individuals’ pri- est, few provide the longitudinal vacy. In what follows, we provide Two new public-use data prod- information needed to analyze an overview of the history, goals, ucts developed by Center job transitions or career trajecto- and techniques associated with for Economic Studies (CES) ries, while being comprehensive these new data products. researchers fill this void of enough to examine those key subpopulations.
Because of its unique advan- The PSEO and VEO provide critical information to better tages, these data have long understand labor market outcomes. These experimental data been used to analyze short-term products resulted from new partnerships the Census Bureau labor market trends. For exam- has developed with colleges, universities, and the U.S. military. ple, Job-to-Job Flows allows Importantly, the PSEO and VEO directly inform household users to easily construct detailed decision making about career planning and educational invest- analyses of job transitions. Yet, ments. Moreover, these products are based entirely on existing there is abundant evidence that information, requiring no additional burden on individuals. We earnings upon entry into the look forward to scaling these products to cover all higher edu- labor market, and subsequent cational institutions and all branches of the armed forces. long-run career trajectories, are Ron Jarmin, Deputy Director of the Census Bureau important as well. In particular,
U.S. Census Bureau Center for Economic Studies Research Report: 2019 21 DEVELOPING THE PSEO data. Their input also enabled us Here earnings for the 25th, 50th, to choose the most consistent and 75th percentiles of each cell Until recently, data on the measure of earnings, the appro- as well as higher-level aggrega- employment outcomes of gradu- priate levels of aggregation, and tions are reported. These data ates of postsecondary institu- the acceptable level of “noise” to are useful for assessing the initial tions have had limitations. A infuse into estimates (to pro- earnings of graduates, compar- number of states have matched tect privacy). Our partners also ing different types of degrees, data on workers in their indi- helped us develop the original and examining the earnings tra- vidual state (from their state’s PSEO data visualization tool. jectory of degrees over time. unemployment insurance pro- gram) to data on graduates from The UT System has been one of The second set of PSEO statis- their universities. While this pro- the main public advocates for tics are the employment flows vides useful information, it misses PSEO and has generated much table, which show census divi- graduates who move to a differ- publicity and visibility for this use- sion and industry sector of a ent state—an issue that grows ful product. Their advocacy and graduate’s main job at 1, 5, and worse over a longer time horizon. support for the broader Census 10 years after graduation. These Schools now have earnings out- Bureau effort helped expand this tabulations allow users to see comes at the program level (i.e., partnership to other states. where graduates work imme- degree level and major) through diately following graduation as the Department of Education’s THE PSEO IN ACTION well as their employment in the College Scorecard. However, the middle of their careers. The PSEO consists of two sets of College Scorecard data are lim- tables about graduates. The first As noted above, an advantage ited to graduates who received set provides statistics on earn- of the PSEO over similar data Title IV financial aid as students ings by postsecondary institu- is its ability to compare longer- and only include earnings for the tion, degree level, degree field, run earnings outcomes across first year after graduation, which and graduation cohort at 1, 5, fields of study. Figure 3-1 shows typically do not reflect long-run and 10 years after graduation. the median earnings at 1 and 10 earnings outcomes.
The PSEO addresses the need Figure 3-1. for both national earnings and Initial Earnings and 10-Year Earnings Growth for employment outcomes data and 2001–2003 University of Texas at Austin Graduates by a longer time horizon. Though Degree Field there are many data sources (Earnings are in 2016 dollars) containing college graduate out- comes, PSEO is the only source PETROLEUM ENG. 200 that contains all graduates of a postsecondary institution. The BIOLOGY BIOCHEMISTRY PSEO also calculates consistent LIBERAL ARTS & SCI., GEN. STUDIES & HUMANITIES
150 PHYSICS earnings outcomes across majors PHILOSOPHY BUSINESS, GENERAL MICROBIOLOGY & IMMUNOLOGY and institutions, enabling direct FINANCE ACCOUNTING comparisons. COMMUNICATION & MEDIA STUDIES 100 ECONOMICS PSYCHOLOGY The PSEO began as a partner- RADIO HISTORY JOURNALISM MATH MECHANICAL ENG. ship with the University of Texas ENGLISH HEALTH & PHYS. ED. ELECTRICAL ENG. (UT) System. The Census Bureau 50 ANTHROPOLOGY COMPUTER & INFO. SCI. ARCHITECTURE benefited from the expertise SOCIAL WORK REGISTERED NURSING of UT staff who had previously MULTI/INTERDISCIPLINARY STUDIES developed a similar set of mea- 0 20,000 40,000 60,000 80,000 sures and who could help us to Percentage gain in median earnings at 10 years Median earnings 1 year after graduation understand how students and Source: U.S. Census Bureau, Post-Secondary Employment Outcomes. administrators interact with such
22 Center for Economic Studies Research Report: 2019 U.S. Census Bureau years after graduation for select growth—on par with social work. There are some large differences graduates from the University Thus, when making decisions across these schools. In particu- of Texas at Austin. While gradu- about borrowing against future lar, graduates from the University ates in a number of fields earn income, information on long-run of Texas at Austin are much around $40,000 in their first earnings is much more relevant more likely to stay in Texas (80.4 year, subsequent earnings than first-year earnings. percent), while University of growth varies from less than 25 Michigan Ann Arbor graduates For state policymakers, a pri- percent to almost 200 percent. are much more likely to leave mary interest is the retention For example, biology and bio- Michigan (58.6 percent). Besides of university graduates within chemistry majors in their tenth out-of-state mobility rates, there the state, which depends on the year earn over 150 percent more are also significant differences in match between employment than what they earned in their regional destinations. For exam- opportunities and training. The first year, while the earnings of ple, while both schools are in the PSEO employment flows data social work and anthropology Midwest, University of Michigan sheds light on the job mobility of graduates grew less than 50 Ann Arbor graduates are almost graduates over time. percent over the same period. twice as likely to be employed in Meanwhile, registered nursing Figure 3-2 shows where gradu- the Pacific Division as University graduates have very high median ates work 1 year after graduating of Wisconsin Madison graduates, earnings (over $60,000) in their from the four flagship institu- while they have similar rates of first year, but very low earnings tions currently in the PSEO data. remaining in the Midwest.
Figure 3-2. Where Graduates With a Bachelor’s Degree From Four Flagship Institutions Work by Region (In percent) University of Colorado Boulder University of Texas at Austin
0.9 0.4
2.0 0.9 4.4 4.1 12.9 2.6 6.0 3.9 1.5 2.1 65.6
3.8 2.8 0.7 0.7 1.1 3.2 80.4
University of Wisconsin-Madison University of Michigan
0.7 1.7 54.9 9.3 2.4 41.4 5.7 15.0 5.2 10.0 2.2 2.3 15.0 15.4
4.1 7.4 0.8 1.3
2.1 3.3
Source: U.S. Census Bureau, Post-Secondary Employment Outcomes.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 23 The PSEO employment flows data also sheds light on the Figure 3-3. industry of employment and Percentage of Graduates by Industry how those employment pat- Business management marketing and related support services terns differ by field of study and Social sciences institution. Figure 3-3 compares Communication journalism and related programs the relative flow of graduates into different industry sectors for three of the most popular Retail bachelor degree fields: business, social sciences, and communica- Information tions. Not surprisingly, a large proportion of business graduates Finance/ become employed in the finance insurance and insurance sector, while com- paratively few go into educa- tion. Meanwhile, social science Professional majors are much more likely to be employed in the health and Education education sectors than in other sectors of the economy. These examples illustrate the PSEO’s Health utility for understanding the transition of college graduates Source: U.S. Census Bureau, Post-Secondary Employment Outcomes. into the labor market, filling a void that had existed in available data. veterans is not only of interest enlisted veterans who completed To learn more about the PSEO to the military, but also to other their initial term of service and and to download data, visit policymakers and to individuals then separated from active duty
24 Center for Economic Studies Research Report: 2019 U.S. Census Bureau and earnings dispersion of and earnings by using short, documentation are available at veterans based on their military repeated cohorts. For example,
Figure 3-4. Median Annual Earnings for Army Veterans 1, 5, and 10 Years After Discharge by Occupation
1 year after discharge 5 years after discharge 10 years after discharge 2016 dollars $
$
$
$
$
$
$
$
$ Infantry, Unmanned Operational Supply Automotive, Food service, general vehicle intelligence administration general general system operators
Source: U.S. Census Bureau, Veterans Employment Outcomes, 2000–2007 cohort.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 25 Figure 3-5. Dispersion in Annual Earnings for Army Veterans 5 Years After Discharge by Occupation th percentile th percentile th percentile dollars $
$
$
$
$
$
$ Infantry Unmanned Operational Supply Automotive Food service general vehicle intelligence administration general general system operators
Source: U.S. Census Bureau, Veterans Employment Outcomes, 2000–2007 cohort.
Figure 3-6. Median Annual Earnings for Army Veterans 1 Year After Discharge by Educational Attainment at Enlistment
GED High school diploma Some college or higher dollars $
$
$
$
$
$
$
$ - - - - - - - -
Source: U.S. Census Bureau, Veterans Employment Outcomes.
26 Center for Economic Studies Research Report: 2019 U.S. Census Bureau We are also using PSEO as a We are also investigating branches of the armed forces. pilot for incorporating Internal opportunities to expand the As we continue our research Revenue Service earnings data VEO and analyzing its potential and development efforts, we into the LED data infrastruc- for research that improves our expect new partnerships that ture, as a secondary source understanding of labor market will yield similarly valuable data of earnings and employment outcomes for military veterans, products. information. including veterans from other
PSEO AND VEO: THE PRIVACY CHALLENGES AND SOLUTIONS The rise of administrative data sources and of public-use data that might otherwise have their use by statistical agencies, as well as data been impossible without sacrificing individual partnerships, have led to a proliferation of new privacy. and valuable information on the population The specific issue with both PSEO and VEO and the economy. However, it has also created is that the partners who provide us the data many new challenges for statistical agencies, have access to the complete frame of individu- which have a legal obligation to ensure indi- als—graduates and veterans, respectively—that vidual privacy. Increasingly, statistical agencies is used to produce our statistics. In the case must face the reality that outside entities have of PSEO, our partners also observe earnings access to large portions of the data underlying for the subset of graduates employed within their public-use outputs. When partnering with the state. So, in the worst case, where just an outside data provider, such as in the cases of one graduate is employed outside the state, PSEO and VEO, this is obviously explicitly true. the earnings of an individual graduate could The Census Bureau must always ensure that the be inferred from a single aggregate statistic statistics it publishes are robust to any outside such as median earnings. Meanwhile, detailed attempts to combine those statistics with exter- tabulations by, say, college major are especially nal data to identify the existence or attributes difficult to protect using traditional statistical of any given individual. Many of the traditional disclosure limitation methods because out- privacy protection measures employed by side data users can create implicit samples statistical agencies are not well-suited to this that threaten the privacy of individuals in the particular challenge because they were devel- microdata. Yet these tabulations are also of oped for survey-based statistics, where only particular interest to our partners and to the the statistical agency possesses the underlying general public. microdata. One solution is to add sufficient “noise” to the To address these new privacy challenges, the underlying statistic so that it is more difficult Census Bureau has become a leader in the to identify any one individual, no matter what application of “differential privacy” techniques. information the end user has. However, any At their core, these techniques allow the Census introduction of noise comes with an inevitable Bureau to quantify the worst-case loss to trade-off in the accuracy of the information individual-level privacy associated with each that is released. The differential privacy tech- statistic that it releases and therefore formally niques applied to both the PSEO and VEO (in a mathematical sense) guarantee that the have allowed us to quantify this trade-off and risk of privacy loss is appropriately limited. therefore maximize our ability to release useful While PSEO and VEO are not the first Census statistics without creating unacceptable risks to Bureau data products to employ differential individual privacy. privacy methods, they provide a clear demon- stration of how the innovative use of differential See page 28 privacy enables the production and release
U.S. Census Bureau Center for Economic Studies Research Report: 2019 27 Continue from page 27... The VEO also utilizes the above code and built on the innovations from the PSEO by address- A major contribution of both the PSEO and ing another significant challenge. Differential VEO is that they provide earnings percentiles privacy methods, such as the histogram for many subsets of college graduates and method, incur a privacy loss for each tabula- veterans, respectively. Earnings percentiles tion that is produced, while the overall loss to are typically difficult to publish because true privacy is not permitted to exceed a threshold percentiles correspond to specific individu- “budget” beyond which the privacy loss is con- als. Previous methods for releasing percentiles sidered unacceptable. The VEO team wished to using differential privacy produced estimates provide statistics on earning percentiles across that were too noisy to be useful in this setting. several demographic and economic dimensions In a key methodological innovation, the PSEO such as occupation, industry, and geography. team first used public information about the This meant that, in addition to limiting the over- earnings distribution to construct a histogram all loss to privacy, the project team needed to of earnings and then added geometric noise balance the trade-off between privacy loss and to the counts associated with each bin in the statistical accuracy for each individual table and histogram. This approach has two advantages efficiently allocate its privacy budget across over previous methods. First, the use of public the numerous desired tables. To accomplish information permitted more accurate ranges on this, the team first developed new metrics to the possible values for the earnings percentile. quantify the accuracy of privacy-protected data Second, the method injects noise across the that could be compared across tables. Then, entire histogram, allowing the release of multi- in keeping with the research of Abowd and ple percentiles while only incurring one instance Schmutte (2019), the team built an optimization of privacy loss. See Foote et al. (2019) for more method that determined which combination of detail and for comparisons to other method- individual table budgets would ensure desired ologies. The code is posted on GitHub for use minimum accuracy for each table without by other statistical agencies. Given the dearth exceeding the overall privacy loss budget and of off-the-shelf implementations of differential without having to omit detailed tabulations that privacy, we hope that providing these resources might be of benefit to users. Besides the VEO, is helpful to others seeking to implement formal this method is useful for any similar product privacy protections. that involves multiple tables.
REFERENCES Abowd, John M., and Ian M. Borgschulte, Mark, and Paco (PSEO),” Journal of Privacy Schmutte, “An Economic Martorell, “Paying to Avoid and Confidentiality, 9(2), 2019. Analysis of Privacy Protection Recession: Using Reenlistment Oreopoulos, Philip, Till von and Statistical Accuracy as to Estimate the Cost of Wachter, and Andrew Heisz, Social Choices," American Unemployment,” American “The Short- and Long-Term Economic Review, 109(1):171- Economic Journal: Applied Career Effects of Graduating 202, 2019. Economics, 10(3):101-127, 2018. in a Recession,” American Altonji, Joseph G., Lisa B. Kahn, Foote, Andrew David, Ashwin Economic Journal: Applied and Jamin D. Speer, “Cashier Machanavajjhala, and Kevin Economics, 4(1):1-29, 2012. or Consultant? Entry Labor McKinney, “Releasing Earnings Zou, Ben, “The Local Economic Market Conditions, Field of Distributions using Differential Impacts of Military Personnel,” Study, and Career Success,” Privacy: Disclosure Avoidance Journal of Labor Economics, Journal of Labor Economics, System for Post-Secondary 36(3):589-621, 2018. 34(S1):S361-S401, 2016. Employment Outcomes
28 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Appendix 1. OVERVIEW OF THE CENTER FOR ECONOMIC STUDIES
The Center for Economic Studies PRODUCTS • OnTheMap. Online mapping (CES) partners with stakeholders and reporting application CES uses microdata from exist- within and outside the U.S. Census showing where people work ing censuses and surveys, and Bureau to improve measures of and workers live, with infor- from administrative sources, to the economy and people of the mation on worker and busi- create innovative public-use data United States through research ness characteristics. and the development of innova- products, including: tive data products. • OnTheMap for Emergency • Business Dynamics Statistics Management. Intuitive Web- (BDS). Tabulations of estab- based interface for accessing RESEARCH lishment openings and U.S. population and workforce closings, firm startups and CES research staff use confi- statistics, in real time, for shutdowns, and job creation dential microdata from Census areas being affected by natu- and destruction with unique Bureau censuses and surveys ral disasters. of businesses and households, information on firm age and linked employer-employee data, firm size. • Opportunity Atlas. Interactive and administrative records mapping tool showing mea- • Business Formation Statistics from federal and state agencies sures of social mobility for (BFS). Quarterly statistics on and other sources to carry out every Census tract in the business applications and for- empirical research that leads to: United States. mations including projections • Discoveries in economics and for recent and future quarters. • Post-Secondary Employment other social sciences not pos- Outcomes (PSEO). Statistics • Dispersion Statistics on sible using publicly available on the earnings and employ- Productivity (DiSP). Annual data. ment outcomes for college statistics on within-industry graduates by institution, • Improvements in existing dispersion of productivity for degree field, and degree level. Census Bureau surveys and the manufacturing sector. data products. • Quarterly Workforce • Job-to-Job Flows (J2J). Indicators (QWI). Workforce • Enhancements to micro-level Statistics on worker realloca- statistics, including employ- data sets for researcher-use tion, including job change, ment, earnings, job creation, in the Federal Statistical hires and separations from and turnover, by demography, Research Data Center and to nonemployment, and geography, and industry for (FSRDC) network. characteristics of origin and each state. destination jobs. • New statistics and information • Synthetic Longitudinal products for public use. • National Longitudinal Business Database (SynLBD). Mortality Study (NLMS). Experimental synthetic Research findings are dissemi- Database for studying the microdata on all U.S. estab- nated through publications (see effects of demographic and lishments including employ- Appendix 2), CES working papers socioeconomic characteristics ment, payroll, and age. (see Appendixes 3 and 4), confer- on differential in mortality ences, seminars, workshops, and rates. this annual report.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 29 HISTORY and research organizations • Those with whom we are col- across the country. In addition to laborating on joint research CES was established in 1982 to restricted-use data on businesses and development including: house databases on businesses, and households from the Census link them cross-sectionally and Bureau, the FSRDCs now also pro- ○ Fellow statistical agen- longitudinally, conduct economic vide secure access to restricted- cies including the Bureau research with them, and make use data from other federal of Economic Analysis, them available to researchers. statistical agencies. As of 2018, Bureau of Justice Statistics, In his 1991 Nobel Prize lecture, the FSRDCs are administered by Bureau of Labor Statistics, economist Ronald Coase noted, the newly established Center for Economic Research Service, “We can also hope to learn much Enterprise Dissemination. and National Center for more in the future from the stud- Science and Engineering ies of the activities of firms which With time, CES’ focus evolved Statistics. have recently been initiated by the from a near-exclusive focus on Center for Economic Studies of the manufacturing sector to ○ Other federal agencies the Bureau of the Census of the include nonmanufacturing sectors including the Agency United States.” and data on workers and house- for Healthcare Research holds. In 2008, the Longitudinal and Quality, the Food Elaborating on these thoughts Employer-Household Dynamics and Nutrition Service, in a letter sent to CES following (LEHD) program joined CES and the Small Business a visit there in June 1993, Coase from the Census Bureau’s Administration. states: Demographic Directorate, and in 2018, researchers from the former ○ Academic institutions “It must be a matter of pride for Center for Administrative Records including Brown University, all in the Bureau of the Census Research and Applications Harvard University, to have a unit which, through its (CARRA) joined CES. University of California— research activities, is playing such Irvine, University of a valuable role in increasing our Today, CES is comprised of several Maryland, and University of understanding of the working of dozen researchers with doctorates North Carolina at Chapel our economic system. Of course, in economics, sociology, demog- Hill. no individual or institution can do raphy, public policy, statistics, and everything. The Center will have history and with research that is ○ Other research organiza- to depend on research conducted even more diverse. tions including the Institute elsewhere (particularly in universi- for Research on Innovation ties) . . . to develop a more com- and Science, NORC at the PARTNERSHIPS plete and more accurate picture University of Chicago, and of the structure of the economy. CES relies on many partners the RAND Corporation. For this reason I greatly welcome within and outside the Census • The members of the Local the initiative of the Bureau of the Bureau, including: Employment Dynamics Census in establishing an office Partnership and other LEHD of the Center in Boston . . . and I • Census Bureau divisions that partners (see Appendix 5), hope, after assessing your expe- collect, process, and produce who provide data critical to rience in Boston, that it will be the business and household a number of our public-use found desirable to establish simi- microdata at the heart of our data products, including J2J, lar offices in other places.” research and that provide us their expert knowledge of the OnTheMap, PSEO, and the Indeed, CES opened the first methodologies underlying QWI. research data center in Boston in those surveys and censuses. 1994 and continued to grow the network over the next quarter • Our colleagues in other cen- century. Today, there are FSRDCs ters within the Research and located at dozens of universities Methodology Directorate.
30 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Appendix 2. PUBLICATIONS AND WORKING PAPERS BY CENTER FOR ECONOMIC STUDIES STAFF: 2019
PUBLICATIONS Akee, Randall, Maggie R. Jones, Bloom, Nicholas, Erik Sonya R. Porter, and Emilia Brynjolfsson, Lucia Foster, Abowd, John M., Kevin L. Simeonova, “Hispanic and Ron Jarmin, Megha Patnaik, McKinney, and Ian M. Asian Earnings Inequality Itay Saporta-Eksten, and John Schmutte, “Modeling and the Role of Labor Market Van Reenen, “What Drives Endogenous Mobility in Entrants and Immigrants,” Differences in Management Earnings Determination,” AEA Papers and Proceedings, Practices?” American Journal of Business and forthcoming. Economic Review, 2019, Economic Statistics, 2019, 109:1648–1683. 37:405–418. Altekruse, Sean F., Candace M. Cosgrove, William C. Altekruse, Bronchetti, Erin T., Garret Abraham, Katharine G., John Richard A. Jenkins, and Carlos Christensen, and Hilary W. Haltiwanger, Kristin Sandusky, Blanco, “Socioeconomic Hoynes, “Local Food Prices, and James R. Spletzer, “The Risk Factors for Fatal Opioid SNAP Purchasing Power, Consequences of Long Term Overdoses in the United and Child Health,” Journal Unemployment: Evidence States: Findings from the of Health Economics, 2019, from Linked Survey and Mortality Disparities in 68:102231. Administrative Data,” ILR American Communities Study Review, 2019, 72:266–299. Brown, J. David, John S. Earle, (MDAC),” PLoS One, 2020, Mee Jung Kim, and Kyung Abraham, Katharine, John 15(1):e0227966. Min Lee, “Large Start-ups, Haltiwanger, Kristin Sandusky, Azoulay, Pierre, Benjamin Jones, Job Creation, and Founder and James Spletzer, J. Daniel Kim, Javier Miranda, Characteristics,” Industrial “Measuring the Gig Economy: “Age and High-Growth and Corporate Change, 2019, Current Knowledge and Open Entrepreneurship,” American 28:1637–1672. Issues,” In Measuring and Economic Review: Insights, Accounting for Innovation in Brown, J. David, John S. 2020, 2:65–82. the 21st Century, edited by Earle, Mee Jung Kim, and Carol Corrado, Javier Miranda, Behrendt, Carolyn E., Kyung Min Lee, “Immigrant Jonathan Haskel, and Daniel Candace M. Cosgrove, Entrepreneurs and Innovation Sichel, University of Chicago Norman J. Johnson, and in the U.S. High-Tech Sector,” Press, forthcoming. Sean F. Altekruse, “Are In The Roles of Immigrants Associations between and Foreign Students in Abraham, Katharine G., John Psychosocial Stressors Science, Innovation, and Haltiwanger, Kristin Sandusky, and Incident Lung Entrepreneurship, edited and James Spletzer, “The Rise Cancer Attributable to by Ina Ganguli, Shulamit of the Gig Economy: Fact or Smoking?” PLoS One, 2019, Kahn, and Megan MacGarvie, Fiction?” AEA Papers and 14(6):e0218439. University of Chicago Press, Proceedings, 2019, 109:357– 2020. 361. Birke, David, Garret Christensen, Rebecca Littman, Edward Brown, J. David, Misty L. Akee, Randall, Maggie R. Jones, Miguel, Elizabeth Levy Heggeness, Suzanne M. and Sonya R. Porter, “Race Paluck, Nicholas Swanson, Dorinski, Lawrence Warren, Matters: Income Shares, and Zenan Wang, “Research and Moises Yi, “Predicting the Income Inequality, and Transparency is on the Rise in Effect of Adding a Citizenship Income Mobility for All Economics,” AEA Papers and Question to the 2020 U.S. Races,” Demography, Proceedings, forthcoming. Census,” Demography, 2019, 2019, 56:999–1021. 56:1173–1194.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 31 Chetty, Raj, Nathaniel Hendren, Christensen, Garret, and Edward Dreisigmeyer, David, “A Quasi- Maggie R. Jones, and Miguel, “Improving Research Isometric Embedding Sonya R. Porter, “Race and Transparency in the Social Algorithm,” Pattern Economic Opportunity Sciences: Registration, Recognition and Image in the United States: Pre-registration, and Multiple Analysis, 2019, 29:230–233. An Intergenerational Testing Adjustments,” In Dreisigmeyer, David W., Perspective,” Quarterly Research Integrity in the “Whitney’s Theorem, Journal of Economics, Behavioral Sciences, edited by Triangular Sets, and forthcoming. Lee Jussim, Sean T. Stevens, Probabilistic Descent on and Jon A. Krosnick, Oxford Christensen, Garret, Allan Dafoe, Manifolds,” Journal of University Press, forthcoming. Edward Miguel, Don A. Moore, Optimization Theory and and Andrew K. Rose, “A Study Cougle, Jesse R., Katherine A. Applications, 2019, 182:935– of the Impact of Data Sharing McDermott, Jahn K. Hakes, 946. on Article Citations using and Keanan J. Joyner, Fairlie, Robert W., Javier Journal Policies as a Natural “Personality Disorders and Miranda and Nikolas Zolas, Experiment,” PLoS One, 2019, Social Support in Cannabis “Measuring Job Creation, 14(12):e0225883. Dependence: A Comparison Growth, and Survival among with Alcohol Dependence,” Christensen, Garret, Jeremy the Universe of Start-ups Journal of Affected Disorders, Freese, and Edward Miguel, in the United States Using 2020, 265:26–31. Transparent and Reproducible a Combined Start-up Panel Social Science Research: Dilworth, Richardson, and Todd Dataset,” ILR Review, 2019, How to Do Open Science, Gardner, “White Flight,” In The 72:1262–1277. University of California Press, Wiley-Blackwell Encyclopedia Fernandez Leticia, and 2019. of Urban and Regional Norman J. Johnson, Studies, edited by Anthony Christensen, Garret, and Edward “Demographics and the Orum, Wiley-Blackwell, 2019. Miguel, “Transparency Epidemiological Risk Factors and Reproducibility: Dinlersoz, Emin, Nathan for Dementia in Hispanic/ Conceptualizing the Goldschlag, Amanda Myers, Latino Populations,” In Caring Problem,” In The Production and Nikolas Zolas, “The for Latinxs with Dementia in a of Knowledge: Enhancing Anatomy of Trademarking by Globalized World: Behavioral Progress in Social Science, Firms in the United States,” In and Psychosocial Treatments, edited by Colin Elman, John Measuring and Accounting for edited by Hector Y. Adames Gerring, and James Mahoney, Innovation in the 21st Century, and Yvette N. Tazeau, Cambridge University Press, edited by Carol Corrado, Springer, 2020. 2020. Javier Miranda, Jonathan Foote, Andrew, and Michel Haskel, and Daniel Sichel, Christensen, Garret, and Edward Grosz, “The Effect of Local University of Chicago Press, Miguel, “Transparency and Labor Market Downturns on forthcoming. Reproducibility: Potential Postsecondary Enrollment Solutions,” In The Production Dinlersoz, Emin M., Henry R. and Program Choice,” of Knowledge: Enhancing Hyatt, Hubert P. Janicki, Education Finance and Policy, Progress in Social Science, “Who Works for Whom? forthcoming. edited by Colin Elman, John Worker Sorting in a Model Gerring, and James Mahoney, of Entrepreneurship with Cambridge University Press, Heterogenous Labor 2020. Markets,” Review of Economic Dynamics, 2019, 34:244–266.
32 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Foote, Andrew, Michel Grosz, Fraser, Gary E., Candace M. Hosgood III, H. Dean, Candace and Stephanie Rennane, Cosgrove, Andrew D. Cosgrove, Madelyn Klugman, “The Effect of Lower Mashchak, Michael J. Orlich, Thoman Rohan, Sean Transaction Costs on and Sean F. Altekruse, Altekruse, “Lung Cancer Social Security Disability “Lower Rates of Cancer and Mortality among Smokers Insurance Application Rates All-Cause Mortality in an and Never-Smokers in the and Participation,” Journal Adventist Cohort Compared United States,” Epidemiology, of Policy Analysis and with a US Census Population,” forthcoming. Management, 2019, 38:99–123. Cancer, 2020, 126:1102–1111. Hungerman, Daniel M., Kevin Foote, Andrew, Michel Grosz, Goldschlag, Nathan, Ron Jarmin, Rinz, and Jay Frymark, and Ann Stevens, “Locate Julia Lane, and Nikolas Zolas, “Beyond the Classroom: Your Nearest Exit: Mass “The Link between University The Implications of School Layoffs and Local Labor R&D, Human Capital and Vouchers for Church Market Response,” ILR Business Startups,” In Finances,” Review of Review, 2019, 72:101–126. Measuring and Accounting for Economics and Statistics, Innovation in the 21st Century, 2019, 101:588–601. Foster, Lucia, Cheryl Grim, John edited by Carol Corrado, Haltiwanger, and Zoltan Wolf, Hyatt, Henry, “Co-working Javier Miranda, Jonathan “Innovation, Productivity Couples and the Similar Jobs Haskel, and Daniel Sichel, Dispersion, and Productivity of Dual-Earner Households,” University of Chicago Press, Growth,” In Measuring and Monthly Labor Review, forthcoming. Accounting for Innovation in November 2019. the 21st Century, edited by Goldschlag, Nathan, Julia Lane, Hyatt, Henry, and Tucker Carol Corrado, Javier Miranda, Bruce A. Weinberg, and McElroy, “Labor Reallocation, Jonathan Haskel, and Daniel Nikolas Zolas, “Proximity Employment, and Earnings: Sichel, University of Chicago and Economic Activity: An A Vector Autoregression Press, forthcoming. Analysis of Vendor-University Approach,” Labour, 2019, Transactions,” Journal of Foster, Lucia, Cheryl Grim, and 33:463–487. Regional Science, 2019, Nikolas Zolas, “A Portrait of 59:163–182. Ikudo, Akina, Julia I. Lane, U.S. Firms that Invest in R&D,” Joseph Staudt, and Economics of Innovation Goldschlag, Nathan, Travis J. Bruce A. Weinberg, and New Technology, 2020, Lybbert and Nikolas J. Zolas, “Occupational Classifications: 29:89–111. “Tracking the Technological A Machine Learning Composition of Industries Foster, Lucia, Julia Manzella, Approach,” Journal of with Algorithmic Patent Erika McEntarfer, and Economic and Social Concordances,” Economics Danielle H. Sandler, Measurement, 2019, 44:57–87. of Innovation and New “Employment and Earnings Technology, forthcoming. Jee, Eunjung, Joya Misra, for Federal Government and Marta Murray-Close, Economists: Empirical Goldschlag, Nathan, and “Motherhood Penalties in the Evidence by Gender and Javier Miranda, “Business U.S., 1986-2014,” Journal of Race,” AEA Papers and Dynamics Statistics of High Marriage and Family, 2019, Proceedings, forthcoming. Tech Industries,” Journal of 81:434–449. Economics & Management Strategy, 2020, 29:3–30. Kamal, Fariha, Mary E. Lovely, and Devashish Mitra, “Trade Liberalisation and Labour Shares in China,” The World Economy, 2019, 42:3588–3618.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 33 Kamal, Fariha, and Asha Scherpf, Erik, and Benjamin WORKING PAPERS Sundaram, “Do Institutions Cerf, “Local Labor Demand Abowd, John M., Joelle Determine Economic and Program Participation Abramowitz, Margaret C. Geography? Evidence from Dynamics: Evidence Levenstein, Kristin McCue, the Concentration of Foreign from New York SNAP Dhiren Patki, Trivellore Suppliers,” Journal of Urban Administrative Records,” Raghunathan, Ann M. Economics, 2019, 110:89–101. Journal of Policy Analysis and Rodgers, Matthew D. Shapiro, Management, 2019, 38:394– Luque, Adela, and Maggie R. and Nada Wasi, “Optimal 425. Jones, “Differences in Self- Probabilistic Record Linkage: Employment Duration by Year Schulkind, Lisa, and Danielle H. Best Practice for Linking of Entry & Pre-Entry Wage- Sandler, “The Timing of Employers in Survey and Sector Attachment,” Journal Teenage Births: Estimating Administrative Data,” Center of Labor Research, 2019, the Effect on High School for Economic Studies 40:24–57. Graduation and Later Life Working Paper 19-08, 2019. Outcomes,” Demography, Miranda, Javier, and Nikolas Akcigit, Ufuk, Emin Dinlersoz, 2019, 56:345–365. Zolas, “Measuring the Impact Jeremy Greenwood, and of Household Innovation Schuster, Steven Sprick, Matthew Veronika Penciakova, Using Non Employer Jaremski, Elisabeth Ruth “Synergizing Ventures,” NBER Administrative Data,” In Perlman, “An Empirical Working Paper No. 26196, Measuring and Accounting for History of the United States 2019. Innovation in the 21st Century, Postal Savings System,” Social Akee, Randall, and Maggie R. edited by Carol Corrado, Science History, forthcoming. Jones, “Immigrants’ Earnings Javier Miranda, Jonathan Staudt, Joseph, “Mandating Growth and Return Migration Haskel, and Daniel Sichel, Accesss: Assessing the from the U.S.: Examining their University of Chicago Press, NIH’s Public Access Policy,” Determinants using Linked forthcoming. Economic Policy, forthcoming. Survey and Administrative Monti, Holly, Martha Stinson, Data,” Center for Economic Staudt, Joseph, Yifang Wei, and Lori Zehr, “How Long Studies Working Paper 19- Lisa Singh, Shawn D. Klimek, Do Early Career Decisions 10; NBER Working Paper No. J. Bradford Jensen, and Follow Women? The Impact 25639, 2019. Andrew L. Baer, “Automating of Employer History on Response Evaluation for Akee, Randall, and Maggie R. the Gender Wage Gap,” Franchising Questions on Jones, “Foreign vs. Journal of Labor Research, the 2017 Economic Census,” U.S. Graduate Degrees: forthcoming. In Big Data for 21st Century The Impact on Earnings Noon, James M., Leticia E. Economic Statistics, edited Assimilation and Return Fernandez, and Sonya R. by Katharine G. Abraham, Migration for the Foreign Porter, “Response Error and Ron S. Jarmin, Brian Moyer, Born,” Center for Economic the Medicaid Undercount and Matthew D. Shapiro, Studies Working Paper 19-17, in the Current Population University of Chicago Press, 2019. Survey,” Health Services forthcoming. Basker, Emek, Randy A. Research, 2019, 54:34–43. Swanson, Nicholas, Garret Becker, Lucia Foster, T. Kirk Sanzenbacher, Geoffrey T., Christensen, Rebecca Littman, White, and Alice Zawacki, Anthony Webb, Candace M. David Birke, Edward Miguel, “Addressing Data Gaps: Four Cosgrove, and Natalia Orlova, Elizabeth Levy Paluck, and New Lines of Inquiry in the “Rising Inequality in Life Zenan Wang, “Research 2017 Economic Census,” Expectancy by Socioeconomic Transparency is on the Rise in Center for Economic Studies Status,” North American Economics,” AEA Papers and Working Paper 19-28, 2019. Actuarial Journal, forthcoming. Proceedings, forthcoming.
34 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Brown, J. David, John S. Jones, Maggie R., and James P. Luque, Adela, Michaela Dillon, Earle, Mee Jung Kim, and Ziliak, “The Antipoverty Julia Manzella, James Noon, Kyung Min Lee, “Immigrant Impact of the EITC: New Kevin Rinz, and Victoria Entrepreneurs and Innovation Estimates from Survey and Udalova, “Nonemployer in the U.S. High-Tech Sector,” Administrative Tax Records,” Statistics by Demographics Center for Economic Studies Center for Economic Studies (NES-D): Exploring Working Paper 19-06; NBER Working Paper 19-14, 2019. Longitudinal Consistency Working Paper No. 25565, and Sub-national Estimates,” Kamal, Fariha, and Asha 2019. Center for Economic Studies Sundaram, “Do Institutions Working Paper 19-34, 2019. Brown, J. David, Misty L. Determine Economic Heggeness, Suzanne M. Geography? Evidence Luque, Adela, and Maggie R. Dorinski, Lawrence Warren, from the Concentration of Jones, “Gender Differences and Moises Yi, “Predicting the Foreign Suppliers,” Center for in Self-employment Duration: Effect of Adding a Citizenship Economic Studies Working the Case of Opportunity and Question to the 2020 Census,” Paper 19-05, 2019. Necessity Entrepreneurs,” Center for Economic Studies Center for Economic Studies Krizan, C.J., “Statistics Working Paper 19-18, 2019. Working Paper 19-24, 2019. on the Small Business Choi, Joonkyu, Nathan Administration’s Scale-Up Manzella, Julia, Evan Totty, Goldschlag, John Haltiwanger, America Program,” Center for and Gary Benedetto, “A and J. Daniel Kim, “Founding Economic Studies Working Task-based Approach to Teams and Startup Paper 19-11, 2019. Constructing Occupational Performance,” Center for Categories with Implications Kurmann, André, and Erika Economic Studies Working for Empirical Research in McEntarfer, “Downward Paper 19-32, 2019. Labor Economics,” Center for Nominal Wage Rigidity in the Economic Studies Working Foote, Andrew, Ashwin United States: New Evidence Paper 19-27, 2019. Machanavajjhala, and from Worker-Firm Linked Kevin McKinney, “Releasing Data,” Center for Economic Marinoni, Astrid, and John Earnings Distributions using Studies Working Paper 19-07, Voorheis, “Who Gains Differential Privacy: Disclosure 2019. from Creative Destruction? Avoidance System for Post Evidence from High-Quality Luque, Adela, Renuka Bhaskar, Secondary Employment Entrepreneurship in the James Noon, Kevin Rinz, Outcomes (PSEO),” Center for United States,” Center for and Victoria Udalova, Economic Studies Working Economic Studies Working “Nonemployer Statistics by Paper 19-13, 2019. Paper 19-29, 2019. Demographics (NES-D): Using Handley, Kyle, Fariha Kamal, Administrative and Census Miller, Sarah, Sean Altekruse, and Ryan Monarch, “Rising Records Data in Business Norman Johnson, and Import Tariffs, Falling Export Statistics,” Center for Laura R. Wherry, “Medicaid Growth: When Modern Economic Studies Working and Mortality: New Evidence Supply Chains Meet Old- Paper 19-01, 2019. from Linked Survey and Style Protectionism,” NBER Administrative Data,” NBER Working Paper No. 26611, Working Paper No. 26081, 2019. 2019.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 35 Murray-Close, Marta, and Schuster, Steven Sprick, Matthew Staudt, Joseph, Yifang Wei, Misty L. Heggeness, “Manning Jaremski, and Elisabeth Lisa Singh, Shawn Klimek, Up and Womaning Down: Ruth Perlman, “An Empirical J. Bradford Jensen, and How Husbands and Wives History of the United States Andrew L. Baer, “Automating Report Earnings When She Postal Savings System,” NBER Response Evaluation for Earns More,” Federal Reserve Working Paper No. 25812, Franchising Questions on Bank of Minneapolis, Institute 2019. the 2017 Economic Census,” Working Paper 28, 2019. Center for Economic Studies Shattuck, Rachel, “High Labor Working Paper 19-20; NBER Rinz, Kevin, “Did Timing Matter? Force Attachment, but Few Working Paper No. 25818, Life Cycle Differences in Social Ties? Life-Course 2019. Effects of Exposure to the Predictors of Women’s Great Recession,” Center for Receipt of Childcare Zawacki, Alice M., Jessica P. Economic Studies Working Subsidies,” Center for Vistnes, and Thomas C. Paper 19-25, 2019. Economic Studies Working Buchmueller, “Why Are Paper 19-26, 2019. Employer-sponsored Health Sandler, Danielle, and Nichole Insurance Premiums Higher in Szembrot, “Maternal Labor the Public Sector than in the Dynamics: Participation, Private Sector?” Center for Earnings, and Employer Economic Studies Working Changes,” Center for Paper 19-03, 2019. Economic Studies Working Paper 19-33, 2019.
36 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Appendix 3. ABSTRACTS OF CENTER FOR ECONOMIC STUDIES WORKING PAPERS BY CENSUS BUREAU STAFF: 2019
Center for Economic Studies Working Paper 19-01 NONEMPLOYER STATISTICS BY DEMOGRAPHICS (NES-D): USING ADMINISTRATIVE AND CENSUS RECORDS DATA IN BUSINESS STATISTICS Adela Luque Renuka Bhaskar James Noon Kevin Rinz Victoria Udalova
January 2019
The quinquennial Survey of Business Owners, or Department of Veteran Affairs will provide admin- SBO, provided the only comprehensive source of istrative records data on veteran status. information in the United States on employer and nonemployer businesses by the sex, race, ethnic- The use of blended data in this manner will make ity, and veteran status of the business owners. The possible the production of NES-D, an annual series annual Nonemployer Statistics series (NES) pro- that will become the only source of detailed and vides establishment counts and receipts for non- comprehensive statistics on the scope, nature, employers but contains no demographic informa- and activities of U.S. businesses with no paid tion on the business owners. With the transition of employment by the demographic characteristics the employer component of the SBO to the Annual of the business owner. Using the 2015 vintage of Business Survey, the Nonemployer Statistics by nonemployers, initial results indicate that demo- Demographics series, or NES-D, represents the graphic information is available for the overwhelm- continuation of demographics estimates for non- ing majority of the universe of nonemployers. employer businesses. NES-D will leverage exist- For instance, information on sex, age, place of ing administrative and census records to assign birth, and citizenship status is available for over demographic characteristics to the universe of 95 percent of the 24 million nonemployers, while approximately 24 million nonemployer businesses race and Hispanic origin are available for about (as of 2015). Demographic characteristics include 90 percent of them. These results exclude owners key demographics measured by the SBO (sex, of C-corporations, which represent only 2 percent race, Hispanic origin, and veteran status) as well of nonemployer firms. Among other things, future as other demographics (age, place of birth, and work will entail imputation of missing demograph- citizenship status) collected but not imputed by ics information (including that of C-corporations), the SBO if missing. A spectrum of administrative testing the longitudinal consistency of the esti- and census data sources will provide the nonem- mates, and expanding the set of characteristics ployer universe and demographics information. beyond the demographics mentioned above. Specifically, the nonemployer universe originates Without added respondent burden and at lower in the Business Register; the Census Numident will imputation rates and costs, NES-D will meet the provide sex, age, place of birth, and citizenship needs of stakeholders, as well as the economy as a status; race and Hispanic origin information will be whole, by providing reliable estimates at a higher obtained from multiple years of the decennial cen- frequency (annual vs. every 5 years) and with a sus and the American Community Survey; and the more timely dissemination schedule than the SBO.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 37 Center for Economic Studies Working Paper 19-03 WHY ARE EMPLOYER-SPONSORED HEALTH INSURANCE PREMIUMS HIGHER IN THE PUBLIC SECTOR THAN IN THE PRIVATE SECTOR? Alice M. Zawacki Jessica P. Vistnes Thomas C. Buchmueller
February 2019
In this article, we examine the factors explaining difference in public and private sector premiums differences in public and private sector health in 2000, by 2014, public premiums had exceeded insurance premiums for enrollees with single private premiums by 14 to 19 percent. We find coverage. We use data from the 2000 and 2014 that differences in plan characteristics played a Medical Expenditure Panel Survey-Insurance substantial role in explaining premium differences Component, along with decomposition methods, in 2014, but they were not the only or even the to explore the relative explanatory importance most important factor. Differences in worker age, of plan features and benefit generosity, such as gender, marital status, and educational attainment deductibles and other forms of cost sharing, basic were also important factors, as was workforce employee characteristics (e.g., age, gender, and unionization. education), and unionization. While there was little
Center for Economic Studies Working Paper 19-05 DO INSTITUTIONS DETERMINE ECONOMIC GEOGRAPHY? EVIDENCE FROM THE CONCENTRATION OF FOREIGN SUPPLIERS Fariha Kamal Asha Sundaram
February 2019
Do institutions shape the geographic concentra- product-country-specific measure of supplier tion of industrial activity? We explore this ques- concentration for U.S. importers. Results show tion in an international trade setting by examining that U.S. importers source in a more spatially the relationship between country-level institutions concentrated manner from countries with weaker and patterns of spatial concentration of global contract enforcement. We find support for the sourcing. A priori, weak institutions could be idea that, where formal contract enforcement associated with either dispersed or concentrated is weak, local supplier networks compensate by sourcing. We exploit location and transaction sharing information to facilitate matching and data on imports by U.S. firms and adapt the transactions. Ellison and Glaeser (1997) index to construct a
38 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-06 IMMIGRANT ENTREPRENEURS AND INNOVATION IN THE U.S. HIGH-TECH SECTOR J. David Brown John S. Earle Mee Jung Kim Kyung Min Lee
February 2019
We estimate differences in innovation behavior every level of the entrepreneur’s education. The between foreign- versus U.S.-born entrepreneurs size of the estimated immigrant-native differences in high-tech industries. Our data come from the in product and process innovation activities rises Annual Survey of Entrepreneurs, a random sam- with detailed controls for demographic and human ple of firms with detailed information on owner capital characteristics but falls for research, devel- characteristics and innovation activities. We find opment, and patenting. Controlling for finance, uniformly higher rates of innovation in immigrant- motivations, and industry reduces all coefficients, owned firms for 15 of 16 different innovation but for most measures and specifications, immi- measures; the only exception is for copyright/ grants are estimated to have a sizable advantage trademark. The immigrant advantage holds for in innovation. older firms as well as for recent start-ups and for
Center for Economic Studies Working Paper 19-07 DOWNWARD NOMINAL WAGE RIGIDITY IN THE UNITED STATES: NEW EVIDENCE FROM WORKER-FIRM LINKED DATA André Kurmann Erika McEntarfer
February 2019
This paper examines the extent and conse- than hourly wage rates due to systematic varia- quences of Downward Nominal Wage Rigidity tions in hours worked. The results are consistent (DNWR) using administrative worker-firm linked with concurrent findings in the literature that data from the Longitudinal Employer Household reductions in base pay are exceedingly rare but Dynamics (LEHD) program for a large representa- that firms use different forms of nonbase pay and tive U.S. state. Prior to the Great Recession, only variations in hours worked to flexibilize labor cost. 7–8 percent of job stayers are paid the same nom- We then exploit the worker-firm link of the LEHD inal hourly wage rate as 1 year earlier—substan- and find that during the Great Recession, firms tially less than previously found in survey-based with indicators of DNWR reduced employment data—and about 20 percent of job stayers experi- by about 1.2 percent more per year. This negative ence a wage cut. During the Great Recession, the effect is driven by significantly lower hiring rates incidence of wage cuts increases to 30 percent, and persists into the recovery. Our results suggest followed by a large rise in the proportion of wage that despite the relatively large incidence of wage freezes to 16 percent as the economy recovers. cuts in the aggregate, DNWR has sizable alloca- Total earnings of job stayers exhibit even fewer tive consequences. zero changes and a larger incidence of reductions
U.S. Census Bureau Center for Economic Studies Research Report: 2019 39 Center for Economic Studies Working Paper 19-08 OPTIMAL PROBABILISTIC RECORD LINKAGE: BEST PRACTICE FOR LINKING EMPLOYERS IN SURVEY AND ADMINISTRATIVE DATA John M. Abowd Joelle Abramowitz Margaret C. Levenstein Kristin McCue Dhiren Patki Trivellore Raghunathan Ann M. Rodgers Matthew D. Shapiro Nada Wasi
March 2019
This paper illustrates an application of record Study (HRS) with employers and establishments linkage between a household-level survey and an in the Census Bureau’s Business Register to cre- establishment-level frame in the absence of unique ate a new data source, which we call the CenHRS. identifiers. Linkage between frames in this setting Multiple imputation is used to propagate uncer- is challenging because the distribution of employ- tainty from the linkage step into subsequent analy- ment across firms is highly asymmetric. To address ses of the linked data. The linked data reveal new these difficulties, this paper uses a supervised, evidence that survey respondents’ misreporting machine-learning model to probabilistically link and selective nonresponse about employer charac- survey respondents in the Health and Retirement teristics are systematically correlated with wages.
Center for Economic Studies Working Paper 19-09 WHY THE ECONOMICS PROFESSION MUST ACTIVELY PARTICIPATE IN THE PRIVACY PROTECTION DEBATE John M. Abowd Ian M. Schmutte William N. Sexton Lars Vilhuber
March 2019
When Google or the U.S. Census Bureau publish interested in technical problems associated with detailed statistics on browsing habits or neigh- protecting privacy. Economists should join the borhood characteristics, some privacy is lost for discussion, first, to determine where to balance everybody while supplying public information. To privacy protection against data quality; a social date, economists have not focused on the privacy choice problem. Furthermore, economists must loss inherent in data publication. In their stead, ensure new privacy models preserve the validity of these issues have been advanced almost exclu- public data for economic research. sively by computer scientists who are primarily
40 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-10 IMMIGRANTS’ EARNINGS GROWTH AND RETURN MIGRATION FROM THE UNITED STATES: EXAMINING THEIR DETERMINANTS USING LINKED SURVEY AND ADMINISTRATIVE DATA Randall Akee Maggie R. Jones
March 2019
Using a novel panel data set of recent immi- related to emigration decisions; time-variant grants to the United States (2005–2007) from unobserved characteristics may be more impor- individual-level linked U.S. Census Bureau survey tant in determining out-migration than previously data and Internal Revenue Service administrative known. We also show that wage assimilation with records, we identify the determinants of return native-born populations occurs fairly quickly; migration and earnings growth for this immigrant after 10 years there is strong convergence in arrival cohort. We show that by 10 years after earnings by several characteristics. Finally, we arrival, almost 40 percent have return migrated. confirm that the use of stock-based panel data Our analysis examines these flows by educa- lead to estimates of slower earnings growth tional attainment, country of birth, and English than is found using repeated cross-section data. language ability separately for each gender. However, we also show, using selection-correc- We show, for the first time, that return migrants tion methods in our panel data, that stock-based experience downward earnings mobility over 2 to panel data may understate the rate of earnings 3 years prior to their return migration. This find- growth for the initial immigrant arrival cohort ing suggests that economic shocks are closely when emigration is not accounted for.
Center for Economic Studies Working Paper 19-11 STATISTICS ON THE SMALL BUSINESS ADMINISTRATION’S SCALE-UP AMERICA PROGRAM C.J. Krizan
April 2019
This paper attempts to quantify the difference in growth. However, publicly available survey results performance, of “treated” (program participant) have shown that entrepreneurs have a variety of and “nontreated” (nonparticipant) firms in the goals in-mind when they start their businesses. Small Business Administration’s (SBA) Scale-Up Two prominent, and potentially contradictory, ones initiative. I combine data from the SBA with admin- are work-life balance and greater income. That istrative data housed at the U.S. Census Bureau means that not all firms may want to grow and I using a combination of numeric and name and am unable to completely control for owner motiva- address matching techniques. My results show that tions. Finally, I do not find a statistically significant after controlling for available observable char- relationship between participation in Scale-Up and acteristics, a positive correlation exists between firm survival once other business characteristics participation in the Scale-Up initiative and firm are accounted for.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 41 Center for Economic Studies Working Paper 19-13 RELEASING EARNINGS DISTRIBUTIONS USING DIFFERENTIAL PRIVACY: DISCLOSURE AVOIDANCE SYSTEM FOR POST-SECONDARY EMPLOYMENT OUTCOMES (PSEO) Andrew Foote Ashwin Machanavajjhala Kevin McKinney
April 2019
The U.S. Census Bureau recently released data on and the application of the Laplace mechanism to earnings percentiles of graduates from post- recover a differentially private cumulative density secondary institutions. This paper describes and function of earnings. We demonstrate that our evaluates the disclosure avoidance system devel- algorithm can release earnings distributions with oped for these statistics. We propose a differen- low error, and our algorithm out-performs prior tially private algorithm for releasing these data work based on the concept of smooth sensitivity based on standard differentially private building from Nissim, Raskhodnikova, and Smith (2007). blocks, by constructing a histogram of earnings
Center for Economic Studies Working Paper 19-14 THE ANTIPOVERTY IMPACT OF THE EITC: NEW ESTIMATES FROM SURVEY AND ADMINISTRATIVE TAX RECORDS Maggie R. Jones James P. Ziliak
April 2019
Evaluations of the Earned Income Tax Credit tax records. We find that significantly more actual (EITC), including its antipoverty effectiveness, are EITC payments flow to childless tax units than pre- based on simulated EITC benefits using either the dicted by the tax simulators, and to those whose U.S. Census Bureau’s tax module or from exter- family income places them well above official pov- nal tax simulators such as the National Bureau erty thresholds. However, actual EITC payments of Economic Research’s TAXSIM or Jon Bakija’s appear to be target efficient at the individual tax model. Each simulator utilizes model-based unit level, whether correctly paid or not. We then assumptions on who is and who is not eligible for compare the antipoverty impact of the EITC across the EITC, and conditional on eligibility, assumes the survey and administrative tax measures of that participation is 100 percent. However, recent EITC benefits. We find that in the full CPS ASEC, evidence suggests that take-up of the EITC is con- the tax simulators overestimate the antipoverty siderably less than 100 percent, and thus claims effects of the EITC by about 1.8 million persons in regarding the impact of the program on measures a typical year. Restricting to a harmonized sample of poverty may be overstated. We use data from of filers, we find that the antipoverty estimates the Current Population Survey Annual Social and derived from the TAXSIM and Bakija models align Economic Supplement (CPS ASEC) linked to more closely to actual EITC payments compared Internal Revenue Service tax data on the EITC to to the CPS, suggesting a discrepancy in assign- compare the distribution of EITC benefits from ment of tax filers between the tax simulators. three tax simulation modules to administrative
42 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-17 FOREIGN VS. U.S. GRADUATE DEGREES: THE IMPACT ON EARNINGS ASSIMILATION AND RETURN MIGRATION FOR THE FOREIGN BORN Randall Akee Maggie R. Jones
June 2019
Using a novel panel data set of recent immigrants same country of birth over the period 2005–2015. to the United States, we identify return migration In Census-Internal Revenue Service administrative rates and earnings trajectories of two immigrant data, we found that downward earnings trajecto- groups: those with foreign graduate degrees and ries are predictive of return migration for immi- those with a U.S. graduate degree. We focus on grants with degrees acquired abroad. Meanwhile, immigrants (of both genders) to the United States immigrants with U.S.-acquired graduate degrees who arrive in the same entry cohort and from the experience mainly upward earnings mobility.
Center for Economic Studies Working Paper 19-18 PREDICTING THE EFFECT OF ADDING A CITIZENSHIP QUESTION TO THE 2020 CENSUS J. David Brown Misty L. Heggeness Suzanne M. Dorinski Lawrence Warren Moises Yi
June 2019
The addition of a citizenship question to the 2020 ACS-census difference in response rates for house- Census could affect the self-response rate, a key holds that may contain noncitizens (more sensitive driver of the cost and quality of a census. We find to the question) with the difference for households that citizenship question response patterns in the containing only U.S. citizens. We estimate that the American Community Survey (ACS) suggest that it addition of a citizenship question will have an 8.0 is a sensitive question when asked about administra- percentage point larger effect on self-response rates tive record noncitizens but not when asked about in households that may have noncitizens relative to administrative record citizens. ACS respondents those with only U.S. citizens. Assuming that the citi- who were administrative record noncitizens in 2017 zenship question does not affect unit self-response frequently choose to skip the question or answer in all-citizen households and applying the 8.0 that the person is a citizen. We predict the effect on percentage-point drop to the 28.1 percent of hous- self-response to the entire survey by comparing mail ing units potentially having at least one noncitizen response rates in the 2010 ACS, which included a would predict an overall 2.2 percentage-point drop citizenship question, with those of the 2010 Census, in self-response in the 2020 Census, increasing costs which did not have a citizenship question, among and reducing the quality of the population count. households in both surveys. We compare the actual
U.S. Census Bureau Center for Economic Studies Research Report: 2019 43 Center for Economic Studies Working Paper 19-20 AUTOMATING RESPONSE EVALUATION FOR FRANCHISING QUESTIONS ON THE 2017 ECONOMIC CENSUS Joseph Staudt Yifang Wei Lisa Singh Shawn Klimek J. Bradford Jensen Andrew L. Baer
July 2019
Between the 2007 and 2012 Economic Censuses in 2012. In this paper, we examine the potential (EC), the count of franchise-affiliated establish- of using external data harvested from the Web in ments declined by 9.8 percent. One reason for this combination with machine learning methods to decline was a reduction in resources that the automate the process of evaluating responses to U.S. Census Bureau was able to dedicate to the the franchise section of the 2017 EC. Our method manual evaluation of survey responses in the allows us to quickly and accurately identify and franchise section of the EC. Extensive manual recode establishments have been mistakenly classi- evaluation in 2007 resulted in many establish- fied as not being franchise-affiliated, increasing the ments, whose survey forms indicated they were unweighted number of franchise-affiliated estab- not franchise-affiliated, being recoded as franchise- lishments in the 2017 EC by 22 percent–42 percent. affiliated. No such evaluation could be undertaken
Center for Economic Studies Working Paper 19-22 REENGINEERING KEY NATIONAL ECONOMIC INDICATORS Gabriel Ehrlich John Haltiwanger Ron Jarmin David Johnson Matthew D. Shapiro
July 2019
Traditional methods of collecting data from busi- be collected or aggregated simultaneously at the nesses and households face increasing challenges. source. This new architecture for economic sta- These include declining response rates to surveys, tistics creates challenges arising from the rapid increasing costs to traditional modes of data col- change in items sold. The paper explores some lection, and the difficulty of keeping pace with recently proposed techniques for estimating price rapid changes in the economy. The digitization of and quantity indices in large-scale, item-level data. virtually all market transactions offers the potential Although those methods display tremendous for reengineering key national economic indica- promise, substantially more research is necessary tors. The challenge for the statistical system is before they will be ready to serve as the basis for how to operate in this data-rich environment. This the official economic statistics. Finally, the paper paper focuses on the opportunities for collecting addresses implications for building national statis- item-level data at the source and constructing key tics from transactions for data collection and for indicators using measurement methods consistent the capabilities and organization of the statistical with such a data infrastructure. Ubiquitous digi- agencies in the twenty-first century. tization of transactions allows price and quantity
44 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-24 GENDER DIFFERENCES IN SELF-EMPLOYMENT DURATION: THE CASE OF OPPORTUNITY AND NECESSITY ENTREPRENEURS Adela Luque Maggie R. Jones
September 2019
A strand of the self-employment literature sug- the wage sector for five consecutive entry cohorts, gests that those “pushed” into self-employment out including two cohorts who entered self-employment of necessity may perform differently from those during the Great Recession. Severely limited labor- “pulled” into self-employment to pursue a business market opportunities may have driven many in opportunity. While findings on self-employment the recession cohorts to enter self-employment, outcomes by self-employed type are not unani- while those entering self-employment during the mous, there is mounting evidence that performance boom may have been pursuing opportunities under outcomes differ between these two self-employed favorable market conditions. To more explicitly test types. Another strand of the literature has found the concept of “necessity” versus “opportunity” important gender differences in self-employment self-employment, we also examine the wage labor entry rates, motivations for entry, and outcomes. attachment (or weeks worked in the wage sector) in Using a unique set of data that links the American the year prior to becoming self-employed. We find Community Survey to administrative data from that, within the cohorts we examine, there are gen- Form 1040 and W-2 records, we bring together der differences in the rate at which men and women these two strands of the literature. We explore depart self-employment for either wage work or whether there are gender differences in self-employ- nonparticipation, but that the patterns are depen- ment duration of self-employed types. In particular, dent on pre-self-employment wage-sector attach- we examine the likelihood of self-employment exit ment and cohort effects. towards unemployment versus
Center for Economic Studies Working Paper 19-25 DID TIMING MATTER? LIFE CYCLE DIFFERENCES IN EFFECTS OF EXPOSURE TO THE GREAT RECESSION Kevin Rinz
September 2019
Exposure to a recession can have persistent, nega- experience the largest earnings losses in percent tive consequences, but does the severity of those terms (up to 13 percent), in part because recession consequences depend on when in the life cycle exposure makes them persistently less likely to a person is exposed? I estimate the effects of work for high-paying employers even as their over- exposure to the Great Recession on employment all employment recovers more quickly than older and earnings outcomes for groups defined by year workers’. Younger workers also experience reduc- of birth over the 10 years following the beginning tions in earnings and employment due to changes of the recession. With the exception of the oldest in local labor market structure associated with the workers, all groups experience reductions in earn- recession. These effects are substantially smaller in ings and employment due to local unemployment magnitude but more persistent than the effects of rate shocks during the recession. Younger workers unemployment rate increases.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 45 Center for Economic Studies Working Paper 19-26 HIGH LABOR FORCE ATTACHMENT, BUT FEW SOCIAL TIES? LIFE-COURSE PREDICTORS OF WOMEN’S RECEIPT OF CHILDCARE SUBSIDIES Rachel Shattuck
September 2019 The U.S. federal Child Care and Development Fund into subsidy receipt, and how subsidy recipiency (CCDF) childcare subsidy represents the largest is situated in women’s broader work and family source of means-tested assistance for U.S. families trajectories. My study links administrative records with low incomes. The CCDF subsidy aims to help from the CCDF to the American Community mothers with low incomes gain employment and Survey (ACS) to construct a longitudinal data set education, with implications for women’s labor from 38 states that observes CCDF recipients in force participation, and the well-being of their the 1–2 years before they first received the subsidy. children. Because recipients of the CCDF subsidy I compare women who subsequently received the are either already employed, or seek the subsidy CCDF subsidy to other women with low incomes in with the goal of gaining employment or school- the ACS who did not go on to receive the sub- ing, this group may represent the public assistance sidy, with a total of roughly 641,000 individuals. I recipients who are best able to succeed in the find that CCDF recipients are generally positively low-wage labor market. However, existing research selected on employment history and educational on the CCDF observes recipients only after they attainment, but appear to have lower levels of begin receiving the subsidy, thus giving an incom- social support than nonrecipients. plete picture of whether recipients may select
Center for Economic Studies Working Paper 19-27 A TASK-BASED APPROACH TO CONSTRUCTING OCCUPATIONAL CATEGORIES WITH IMPLICATIONS FOR EMPIRICAL RESEARCH IN LABOR ECONOMICS Julia Manzella Evan Totty Gary Benedetto
September 2019
Most applied research in labor economics that nuanced interpretation for grouping occupations examines returns to worker skills or differences and permits quantitative assessments of similari- in earnings across subgroups of workers typically ties in task compositions across occupations. We accounts for the role of occupations by control- also replicate a recent analysis and find that our ling for occupational categories. Researchers often task-based occupational categories explain more aggregate detailed occupations into categories of the gender wage gap than the SOC-based based on the Standard Occupation Classification approaches explain. Our study enhances the (SOC) coding scheme, which is based largely on Federal Statistical System’s understanding of the narratives or qualitative measures of workers’ SOC codes, investigates ways to use third-party tasks. Alternatively, we propose two quantitative data to construct useful research variables that task-based approaches to constructing occupa- can potentially be added to U.S. Census Bureau tional categories by using factor analysis with data products to improve their quality and versatil- O*NET job descriptors that provide a rich set of ity, and sheds light on how the use of alternative continuous measures of job tasks across all occu- occupational categories in economics research pations. We find that our task-based approach may lead to different empirical results and deeper outperforms the SOC-based approach in terms understanding in the analysis of labor market of lower occupation distance measures. We show outcomes. that our task-based approach provides an intuitive,
46 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-28 ADDRESSING DATA GAPS: FOUR NEW LINES OF INQUIRY IN THE 2017 ECONOMIC CENSUS Emek Basker Randy A. Becker Lucia Foster T. Kirk White Alice Zawacki
September 2019
We describe four new lines of inquiry added to current Census Bureau products concerning the the 2017 Economic Census regarding (1) retail U.S. economy. The new content addresses, such health clinics, (2) management practices in health issues as the rise in importance of health care and care services, (3) self-service in retail and service its complexity, the adoption of automation tech- industries, and (4) water use in manufacturing and nologies, and the importance of measuring water, mining industries. These were proposed by econo- a critical input to many manufacturing and mining mists from the U.S. Census Bureau’s Center for industries. Economic Studies in order to fill data gaps in
Center for Economic Studies Working Paper 19-29 WHO GAINS FROM CREATIVE DESTRUCTION? EVIDENCE FROM HIGH-QUALITY ENTREPRENEURSHIP IN THE UNITED STATES Astrid Marinoni John Voorheis
October 2019
The question of who gains from high-quality returns. In both fixed effects and IV models using a entrepreneurship is crucial to understanding Bartik-style instrument, we find that entrepreneur- whether investments in incubating potentially ship increases income inequality. Further, we find innovative start-up firms will produce socially that this increase in income inequality arises due beneficial outcomes. We attempt to bring new to the fact that almost all of the individual gains evidence to this question by combining new associated with increased entrepreneurship accrue aggregate measures of local area income inequal- to the top 10 percent of the income distribution. ity and income mobility with measures of entre- While we find mixed evidence for small positive preneurship from Guzman and Stern (2017). Our effects of entrepreneurship lower on the income new aggregate measures are generated by link- distribution, we find little, if any, evidence that ing American Community Survey data with the entrepreneurship increases income mobility. universe of Internal Revenue Service 1040 tax
U.S. Census Bureau Center for Economic Studies Research Report: 2019 47 Center for Economic Studies Working Paper 19-32 FOUNDING TEAMS AND STARTUP PERFORMANCE Joonkyu Choi Nathan Goldschlag John Haltiwanger J. Daniel Kim
November 2019
We explore the role of founding teams in account- that the exogenous separation of a founding team ing for the postentry dynamics of startups. While member due to premature death has a persis- the entrepreneurship literature has largely focused tently large, negative, and statistically significant on business founders, we broaden this view by impact on postentry size, survival, and productiv- considering founding teams, which include both ity of startups. While we find that the loss of a the founders and the initial employees in the first key founding team member (e.g. founders) has an year of operations. We investigate the idea that especially large adverse effect, the loss of a non- the success of a startup may derive from the orga- key founding team member still has a significant nizational capital that is created at firm formation adverse effect, lending support to our inclusive and is inalienable from the founding team itself. To definition of founding teams. Furthermore, we find test this hypothesis, we exploit premature deaths that the effects are particularly strong for small to identify the causal impact of losing a founding founding teams but are not driven by activity in team member on startup performance. We find small business-intensive or high-tech industries.
Center for Economic Studies Working Paper 19-33 MATERNAL LABOR DYNAMICS: PARTICIPATION, EARNINGS, AND EMPLOYER CHANGES Danielle Sandler Nichole Szembrot
December 2019
This paper describes the labor dynamics of and labor force participation after having their first U.S. women after they have had their first and child. The penalty grows over time, driven by the subsequent children. We build on the child pen- birth of subsequent children. Non-White mothers, alty literature by showing the heterogeneity of the unmarried mothers, and mothers with more educa- size and pattern of labor force participation and tion are more likely to return to work following the earnings losses by demographic characteristics of birth of their first child. Conditional on returning mothers and the characteristics of their employ- to the labor force, women who change employ- ers. The analysis uses longitudinal administrative ers earn more after the birth of their first child earnings data from the Longitudinal Employer- than women who return to their prebirth employ- Household Dynamics database combined with ers. The probability of returning to the prebirth the Survey of Income and Program Participation employer and industry is heterogeneous over both survey data to identify women, their fertility tim- the demographics of mothers and the characteris- ing, and employment. We find that women experi- tics of their employers. ence a large and persistent decrease in earnings
48 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Center for Economic Studies Working Paper 19-34 NONEMPLOYER STATISTICS BY DEMOGRAPHICS (NES-D): EXPLORING LONGITUDINAL CONSISTENCY AND SUBNATIONAL ESTIMATES Adela Luque Michaela Dillon Julia Manzella James Noon Kevin Rinz Victoria Udalova
December 2019
Until recently, the quinquennial Survey of Business at the subnational level, (4) exploring estimates Owners (SBO) was the only source of information by industrial sector, (5) examining demographics for U.S. employer and nonemployer businesses by estimates of business receipts as well as of counts, owner demographic characteristics such as race, and (6) implementing imputation of missing ethnicity, sex, and veteran status. Now, however, demographic values. the Nonemployer Statistics by Demographics series (NES-D) will replace the SBO’s nonem- Our current results are consistent with the main ployer component with reliable and more frequent findings in Luque et al. (2019), and show that high (annual) business demographic estimates with coverage and demographic assignment rates are no additional respondent burden, and at lower not the exception, but the norm. Specifically, we imputation rates and costs. NES-D is not a survey; find that AR coverage rates are high and stable rather, it exploits existing administrative and cen- over time for each of the 3 years we examine, sus records to assign demographic characteristics 2014–2016. We are able to identify owners for to the universe of approximately 25 million (as of approximately 99 percent of nonemployer busi- 2016) nonemployer businesses. nesses (excluding C-corporations), 92 to 93 percent of identified nonemployer owners have no Although only in the second year of its research missing demographics, and only about 1 percent phase, NES-D is rapidly moving towards produc- are missing three or more demographic character- tion, with a planned prototype or experimental istics in each of the 3 years. We also find that our version release of 2017 nonemployer data in 2020, demographics estimates are stable over time, with followed by annual releases of the series. After the expected small annual changes that are consistent first year of research, we released a working paper with underlying population trends in the United (Luque et al., 2019) that assessed the viability of States. Due to data limitations, these results do estimating nonemployer demographics exclusively not include C-corporations, which represent only 2 with administrative records (AR) and census data. percent of nonemployer businesses and 4 percent That paper used 1 year of data (2015) to produce of receipts. preliminary tabulations of business counts at the national level. This year we expand that research Without added respondent burden and at lower in multiple ways by: (1) examining the longitudinal imputation rates and costs, NES-D will provide consistency of administrative and census records high-quality business demographics estimates at coverage, and of our AR-based demographics a higher frequency (annual vs. every 5 years) than estimates, (2) evaluating further coverage from the SBO. additional data sources, (3) exploring estimates
U.S. Census Bureau Center for Economic Studies Research Report: 2019 49 This page is intentionally blank. Appendix 4. CENTER FOR ECONOMIC STUDIES WORKING PAPERS: 2019 CES Working Papers are available at
19-01 “Nonemployer Statistics by 19-08 “Optimal Probabilistic Record Linkage: Demographics (NES-D): Using Best Practice for Linking Employers in Administrative and Census Records Data Survey and Administrative Data,” by in Business Statistics,” by Adela Luque, John M. Abowd, Joelle Abramowitz, Renuka Bhaskar, James Noon, Kevin Margaret C. Levenstein, Kristin McCue, Rinz, and Victoria Udalova, January Dhiren Patki, Trivellore Raghunathan, 2019. Ann M. Rodgers, Matthew D. Shapiro, and Nada Wasi, March 2019. 19-02 “Predictive Analytics and Organizational Architecture: Plant-Level Evidence from 19-09 “Why the Economics Profession Must Census Data,” by Eva Labro, Mark Lang, Actively Participate in the Privacy and Jim Omartian, January 2019. Protection Debate,” by John M. Abowd, Ian M. Schmutte, William N. Sexton, and 19-03 “Why Are Employer-Sponsored Health Lars Vilhuber, March 2019. Insurance Premiums Higher in the Public Sector than in the Private Sector?” by 19-10 “Immigrants’ Earnings Growth and Alice M. Zawacki, Jessica P. Vistnes, and Return Migration from the US: Examining Thomas C. Buchmueller, February 2019. their Determinants using Linked Survey and Administrative Data,” by Randall 19-04 “The Effect of Child Support on Akee and Maggie R. Jones, March 2019. Selection into Marriage and Fertility,” by Daniel I. Tannenbaum, February 2019. 19-11 “Statistics on the Small Business Administration’s Scale-Up America 19-05 “Do Institutions Determine Economic Program,” by C.J. Krizan, March 2019. Geography? Evidence from the Concentration of Foreign Suppliers,” 19-12 “Fraudulent Financial Reporting and the by Fariha Kamal and Asha Sundaram, Consequences for Employees,” by Jung February 2019. Ho Choi and Brandon Gipper, March 2019. 19-06 “Immigrant Entrepreneurs and Innovation in the U.S. High-Tech Sector,” 19-13 “Releasing Earnings Distributions by J. David Brown, John S. Earle, Mee using Differential Privacy: Disclosure Jung Kim, and Kyung Min Lee, February Avoidance System for Post Secondary 2019. Employment Outcomes (PSEO),” by Andrew Foote, Ashwin Machanavajjhala, 19-07 “Downward Nominal Wage Rigidity in and Kevin McKinney, April 2019. the United States: New Evidence from Worker-Firm Linked Data,” by André 19-14 “The Antipoverty Impact of the EITC: Kurmann and Erika McEntarfer, February New Estimates from Survey and 2019. Administrative Tax Records,” by Maggie R. Jones and James P. Ziliak, April 2019.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 51 19-15 “Managing Trade: Evidence From China 19-24 “Gender Differences in Self-employment and the U.S.,” by Nicholas Bloom, Kalina Duration: the Case of Opportunity and Manova, John Van Reenen, Stephen Necessity Entrepreneurs,” by Adela Teng Sun, and Zhihong Yu, May 2019. Luque and Maggie R. Jones, September 2019. 19-16 “Property Rights, Place-Based Policies, and Economic Development,” by Laurel 19-25 “Did Timing Matter? Life Cycle Wheeler, June 2019. Differences in Effects of Exposure to the Great Recession,” by Kevin Rinz, 19-17 “Foreign vs. U.S. Graduate Degrees: The September 2019. Impact on Earnings Assimilation and Return Migration for the Foreign Born,” 19-26 “High Labor Force Attachment, but by Randall Akee and Maggie R. Jones, Few Social Ties? Life-Course Predictors June 2019. of Women’s Receipt of Childcare Subsidies,” by Rachel Shattuck, 19-18 “Predicting the Effect of Adding a September 2019. Citizenship Question to the 2020 Census,” by J. David Brown, Misty L. Heggeness, 19-27 “A Task-based Approach to Constructing Suzanne M. Dorinski, Lawrence Warren, Occupational Categories with and Moises Yi, June 2019. Implications for Empirical Research in Labor Economics,” by Julia Manzella, 19-19 “The Two-Income Trap: Are Two- Evan Totty, and Gary Benedetto, Earner Households More Financially September 2019. Vulnerable?” by Jonathan Fisher and Nathaniel Johnson, June 2019. 19-28 “Addressing Data Gaps: Four New Lines of Inquiry in the 2017 Economic Census,” 19-20 “Automating Response Evaluation for by Emek Basker, Randy A. Becker, Lucia Franchising Questions on the 2017 Foster, T. Kirk White, and Alice Zawacki, Economic Census,” by Joseph Staudt, September 2019. Yifang Wei, Lisa Singh, Shawn Klimek, J. Bradford Jensen, and Andrew L. Baer, 19-29 “Who Gains from Creative Destruction? July 2019. Evidence from High-Quality Entrepreneurship in the United States,” 19-21 “Demographic Origins of the Startup by Astrid Marinoni and John Voorheis, Deficit,” by Fatih Karahan, Benjamin October 2019. Pugsley, and Aysegül Sahin, July 2019. 19-30 “Human Capital, Parent Size and the 19-22 “Re-engineering Key National Economic Destination Industry of Spinouts,” Indicators,” by Gabriel Ehrlich, John by Mariko Sakakibara and Natarajan Haltiwanger, Ron Jarmin, David Johnson, Balasubramanian, September 2019. and Matthew D. Shapiro, July 2019. 19-31 “What Do Establishments Do When 19-23 “Pay, Employment, and Dynamics of Wages Increase? Evidence from Young Firms,” by Tania Babina, Wenting Minimum Wages in the United States,” Ma, Christian Moser, Paige Ouimet, and by Yuci Chen, November 2019. Rebecca Zarutskie, July 2019.
52 Center for Economic Studies Research Report: 2019 U.S. Census Bureau 19-32 “Founding Teams and Startup 19-34 “Nonemployer Statistics by Performance,” by Joonkyu Choi, Nathan Demographics (NES-D): Exploring Goldschlag, John Haltiwanger, and Longitudinal Consistency and Sub- J. Daniel Kim, November 2019. national Estimates,” by Adela Luque, Michaela Dillon, Julia Manzella, James 19-33 “Maternal Labor Dynamics: Participation, Noon, Kevin Rinz, and Victoria Udalova, Earnings, and Employer Changes,” by December 2019. Danielle Sandler and Nichole Szembrot, December 2019.
U.S. Census Bureau Center for Economic Studies Research Report: 2019 53 This page is intentionally blank. Appendix 5. LONGITUDINAL EMPLOYER–HOUSEHOLD DYNAMICS (LEHD) PARTNERS Under the Local Employment Dynamics (LED) Partnership, the Longitudinal Employer-Household Dynamics (LEHD) research team at the Center for Economic Studies produces new, cost effective, public- use information combining federal, state, and U.S. Census Bureau data on employers and employees. The LED Partnership works to fill critical data gaps and provide indicators increasingly needed by state and local authorities to make informed decisions affecting their economies and workforces.
LOCAL EMPLOYMENT DYNAMICS (LED) Mountain-Plains (Colorado, Kansas, Missouri, STEERING COMMITTEE Utah, Wyoming) As of January 2020. Jeffrey Drake, Labor Market Information Manager New England (Connecticut, Maine, Massachusetts, Missouri Economic Research and Information New Hampshire, Rhode Island, Vermont) Center Patrick Flaherty, Assistant Director Missouri Department of Higher Education and Office of Research and Information Workforce Connecticut Department of Labor Southwest (Arkansas, Louisiana, New Mexico, New York/New Jersey (New York, New Jersey, Oklahoma, Texas) Puerto Rico, U.S. Virgin Islands) Rachel Moskowitz, Chief Leonard Preston, Chief Economic Research and Analysis Bureau Labor Market Information New Mexico Department of Workforce New Jersey Department of Labor and Solutions Workforce Development Western (Alaska, Arizona, California, Guam, Mid-Atlantic (Delaware, District of Columbia, Hawaii, Idaho, Nevada, Oregon, Washington) Maryland, Pennsylvania, Virginia, West Virginia) Robert Uhlenkott, Division Director Tim Kestner, Director Workforce and Economic Research Economic Information and Analytics Division Oregon Employment Department Virginia Employment Commission
FEDERAL PARTNERS Southeast (Alabama, Florida, Georgia, Kentucky, Mississippi, North Carolina, South Carolina, U.S. Department of Agriculture Tennessee) U.S. Department of Commerce, National Oceanic Adrienne Johnston, Chief and Atmospheric Administration, National Bureau of Labor Market Statistics Hurrican Center Florida Department of Economic Opportunity U.S. Department of Homeland Security, Federal Emergency Management Agency Midwest (Illinois, Indiana, Iowa, Michigan, U.S. Department of the Interior Minnesota, Nebraska, North Dakota, Ohio, U.S. Office of Personnel Management South Dakota, Wisconsin) U.S. Bureau of Labor Statistics Coretta Pettway, Chief U.S. Geological Survey, Geospatial Multi-Agency Labor Market Information Bureau Coordination Ohio Department of Job and Family Services Internal Revenue Service U.S. Army
U.S. Census Bureau Center for Economic Studies Research Report: 2019 55 STATE EDUCATION PARTNERS District of Columbia Saikou Diallo, Chief Economist University of Texas System Office of Labor Market Policy and Information Colorado Department of Higher Education District of Columbia Department of Employment Institute for Research on Innovation and Science, Services in partnership with: University of Michigan Florida University of Wisconsin—Madison Adrienne Johnston, Chief Bureau of Workforce Statistics and Economic STATE PARTNERS Research Florida Department of Economic Opportunity As of January 2020. Georgia Alabama Mark Watson, Director Jim Henry, Director Workforce Statistics and Economic Research Labor Market Information Division Georgia Department of Labor Alabama Department of Labor Guam Alaska Gary Hiles, Chief Economist Dan Robinson, Director Government of Guam Research and Analysis Section Department of Labor Alaska Department of Labor and Workforce Development Hawaii Phyllis Dayao, Chief Arizona Research and Statistics Office Doug Walls, Labor Market Information Director Hawaii Department of Labor and Industrial Arizona Office of Economic Opportunity Relations Arkansas Idaho Robert S. Marek, Chief Salvador Vazquez, Labor Market Information Labor Market Information Division Director Arkansas Department of Workforce Services Research and Analysis Bureau California Idaho Department of Labor Amy Faulkner, Chief Illinois Labor Market Information Division George Putnam, Labor Market Information Manager California Employment Development Department Economic Information and Analysis Division Colorado Illinois Department of Employment Security Michelle Morelli, Director Indiana Office of Labor Market Information Charlie Baer, Acting Director Colorado Department of Labor and Employment Research and Analysis Connecticut Indiana Department of Workforce Development Andrew Condon, Director Iowa Office of Research Jeremy Hamp, Director Connecticut Department of Labor Labor Market Information Division Delaware Iowa Department of Workforce Development Thomas Dougherty,Acting Chief Office of Occupational and Labor Market Information Delaware Department of Labor
56 Center for Economic Studies Research Report: 2019 U.S. Census Bureau Kansas Missouri Angela Berland, Director Jeffrey Drake, Labor Market Information Manager Labor Market Information Services Missouri Economic Research and Information Kansas Department of Labor Center Missouri Department of Higher Education and Kentucky Workforce Jessica Cunningham, Interim Executive Director Kentucky Center for Statistics Montana Mike Peery, Director Louisiana Labor Market Information Alison Ocmand, Assistant Secretary Montana Department of Labor and Industry Office of Occupational Information Services Louisiana Workforce Commission Nebraska Scott Hunzeker, Research Administrator Maine Nebraska Department of Labor Ruth Pease, Acting Director Center for Workforce Research and Information Nevada Maine Department of Labor David Schmidt, Chief Economist Research and Analysis Bureau Maryland Nevada Department of Employment, Training, Carolyn Mitchell, Director and Rehabilitation Office of Workforce Information and Performance Maryland Department of Labor New Hampshire Brian Gottlob, Director Massachusetts Economic and Labor Market Information Bureau Rena Kottcamp, Director of Research New Hampshire Employment Security Massachusetts Division of Unemployment Assistance New Jersey Chester Chinsky, Director Michigan Economic and Demographic Research Jason Palmer, Director New Jersey Department of Labor and Workforce Bureau of Labor Market Information and Development Strategic Initiatives Michigan Department of Technology, Management, New Mexico and Budget Rachel Moskowitz, Chief Economic Research and Analysis Bureau Minnesota New Mexico Department of Workforce Solutions Oriane Casale, Acting Director BLS Cooperative Programs New York Minnesota Department of Employment and Bohdan Wynnyk, Deputy Director Economic Development Division of Research and Statistics New York State Department of Labor Mississippi Mary Willoughby, Bureau Director North Carolina Labor Market Information Meihui Bodane, Interim Assistant Secretary of Mississippi Department of Employment Security Policy, Research, and Strategy Labor and Economic Analysis Division North Carolina Department of Commerce
U.S. Census Bureau Center for Economic Studies Research Report: 2019 57 North Dakota Texas Marcia Havens, Manager Mariana Vega, Director Labor Market Information Center Labor Market Information Job Service North Dakota Texas Workforce Commission
Ohio Utah Coretta Pettway, Chief Collin Petersen, Director Labor Market Information Bureau Research and Analysis Ohio Department of Job and Family Services Utah Department of Workforce Services
Oklahoma Vermont Lynn Gray, Director Mathew Barewicz, Director Economic Research and Analysis Economic and Labor Market Information Section Oklahoma Employment Security Commission Vermont Department of Labor
Oregon Virgin Islands Robert Uhlenkott, Division Director Gary Halyard, Director Workforce and Economic Research Bureau of Labor Statistics Oregon Employment Department U.S. Virgin Islands Department of Labor
Pennsylvania Virginia Ed Legge, Director Tim Kestner, Director Center for Workforce Information and Analysis Economic Information and Analytics Division Pennsylvania Department of Labor and Industry Virginia Employment Commission
Puerto Rico Washington Elda Pares, Director Steven Ross, Director Bureau of Labor Statistics Labor Market and Economic Analysis Department of Labor Washington Employment Security Department
Rhode Island West Virginia Donna Murray, Assistant Director Joseph Jarvis, Director Labor Market Information Research, Information and Analysis Division Rhode Island Department of Labor and Training Workforce West Virginia
South Carolina Wisconsin Brian Nottingham, Director Dennis Winters, Director Business Intelligence Department Bureau of Workforce Information and Technical South Carolina Department of Employment Support and Workforce Wisconsin Department of Workforce Development
South Dakota Wyoming Melodee Lane, Administrator Tony Glover, Manager Labor Market Information Center Research and Planning South Dakota Department of Labor and Regulation Wyoming Department of Workforce Services
Tennessee Lisa Howard, Assistant Commissioner Workforce Insights and Reporting Engine Division Tennessee Department of Labor and Workforce Development
58 Center for Economic Studies Research Report: 2019 U.S. Census Bureau sk i
oo n ha skar ph ic and ks trom, nn ial Lead ec e og ra J.David Brown D ea rch LeahClark Brad Foster James N en uka B es earch Group Erik Vic R R SuzanneDorin Dem
Marta Murray-Close GenevieveDenoeux
ph icRes og ra ll a ill on NikolasPharris-Ciurej Dem oo rheis, ey and Assistant Center Chief for AssistantCenter ae la D Lead ne Walkerli ne a li a Manze Economic Kevin Rinz Surv oh n V es earch Group Ju J Mach Caro Victoria Udalova R GarretChristensen ne s nd ou p pb ell Pitts t on Gr au dia Perez Lead ng Zhengli ng JimSpletzer op menta es earch Cl ob er Kimberly Jo pp lications Joyce Hahn ea thHayward R ha o A od y Hoon-Starr Carl Anderson ev el BenCam H C J nno vati Matthew Graham, D I il ey Employer- l
Keith Ba ud ina an dyBecker ck i R hold R Dynamics ge r Jani Lon git an dusky Assistant Center Chief for AssistantCenter t Stai
Lead t ou se en ryHyatt H Moises Yi r bo r Markets ub er Claire Hou Lee Tucker H ep henTibbets Mat Seth Murray Larry Warren Andrew Foote es earch Group H La Kevin McKinney St Kristin S Erika McEntarfer, R ono mist Lucia S.Foster Chief Ec Chief li ty
eb eccaTurner, Admin Assistant on na Myers Lead R an Johnson, SierraNoland Morta ah n Hakes D ea rch ss istant J es earch Group an dace Cosgrove Matthew Neiman R Norm C he rylGrim iplin aryRes C ga n- ne s nd ler, an dez di sc qu e li nary tt y
eb orah Wright,Admin A Assistant Center Chief for AssistantCenter Lead In ter Doyle e ll e Sa D gg ieJo dd Gardner Raj Che Adela Lu oh n Friedman Ma To es earch Group en iseFlana J Nathan Hendren Interdiscip on ya Porter,Lead R D LeticiaFern Danie David Dreisigmeyer S il lo ho w ll o-Tr ng siri
a ss a C ha Stinson an iel Kim ss istant D Chris Goetz Wei Ouyang Sean Streiff T.Kirk White Jeff Po FarihaKamal Meli Jerome Davis Mart li sabeth Perlman Ti mothy Beggs MarinaKrylova Statistics Group E Cristina Te Nathan Goldschlag Business Dynamics Kristin McCue,Lead EminDinlersoz, Lead es earch li mek es sR ShawnK oo d og rams en sen tt le Busin li vares
Assistant Center Chief for AssistantCenter Dawn Anderson,Admin A od y Tu li ce Zawacki C Zoltan Wolf Teresa Fort Luke Pardue Emek Basker Zachary Kroff NikolasZolas A es earch Group KaivonTaylor Joseph Staudt Mee JungKim Scott Ohlmacher Edward O R GlennBlackw J.Bradford J EconomicPr Appendix 6. CHART ORGANIZATIONAL STUDIES ECONOMIC CENTER FOR 2020) (February
U.S. Census Bureau Center for Economic Studies Research Report: 2019 59