Nonemployer Statistics By Demographics (NES -D): By Adela .

filing tax forms for sole proprietorships (Form 1040, Schedule C), partnerships (Form 1065), orcorporations (the Form 1120 series). The vast majority of nonemployers are sole proprietorsfollowed by partnerships, S-corporations and C-corporations (approximately 87, 7, 4 and 2percent respectively as of 2016). 6 Please note that our current work does not include Ccorporations because owners of these businesses cannot unequivocally be identified withadministrative records data. Assigning demographics to C- corporations will be addressed infuture work. A more detailed discussion on this topic can be found in section VI of this paperand Luque et al. (2019).NES-D will include key demographic characteristics (i.e., sex, race, Hispanic origin, andveteran status) 7 that were collected by the SBO and imputed if missing, as well asdemographics that the SBO collected but did not impute if missing (i.e., age, place of birth, andcitizenship). The demographic characteristics as well as the universe of nonemployer businessesitself come from a spectrum of administrative records and census data sources including theBusiness Register (BR), tax data from the IRS, the Decennial Census and American CommunitySurvey (ACS), Census Numident files,8 and administrative records on veteran status from theDepartment of Veteran Affairs (VA) . Our objective is for future versions of NES-D to expand incontent to include additional characteristics that will help us improve our understanding ofnonemployers behavior and dynamics. Examples may include: gig-economy relatedcharacteristics (e.g., does the nonemployer also work for wages?), household attributesobtainable through tax data (e.g., marital status, number of dependents, home ownership),Sole proprietorships are businesses owned and managed by one individual. The owner or sole proprietor doesnot pay separate income tax on the company, but instead reports all losses/profits from the business on his/herindividual IRS 1040 tax return. A partnership is a business with two or more owners, each receiving a share of theprofits/losses of the business. A partnership must file an annual information return (Schedule K-1) to report theincome/losses from its operations, but it is not subject to income tax itself. Instead, it "passes through" anyprofits/losses to its partners (hence their “pass-through entities” alias). S-corporations are corporations owned byone or more individuals (up to 75), and they are also pass-through entities. C-corporations are corporations inwhich the owners, or shareholders, are taxed separately from the entity. The taxing of profits from the business isat both corporate and personal levels, creating a double taxation situation. C-corporations are discussed furtherbelow.7 These characteristics were referred to as “core demographics” in the SBO.8 These files are derived from the Social Security Administration Numident files.66

transitions from nonemployer to employer status, etc. See Figure II below for a snapshot ofNES-D content.Tabulations of:Figure II: NES-D Content Number of nonemployer businesses Number of business owners Business ReceiptsBy: Demographics of businessowner Sex, race, Hispanic origin, veteran status, age, placeof birth, citizenship status Potential future characteristics: Gig-economyrelated (e.g., does nonemployer also work for asalary? Does he/she hire contractors?), transitionsfrom nonemployer to employer status, householdcharacteristics (e.g., marital status, number ofdependents, home ownership), exports, etc. Legal Form of Organization (LFO) Receipt-size class Geography detail NAICS1 industry detail Cross-tabulations of dimensions above (e.g., state-sector)1NAICS stands for North American Industry Classification System.NES-D Timeline & Upcoming Prototype Release in 2020NES-D is currently in the second of year (2019) of its research phase. As shown in Figure IIIbelow, a prototype or experimental version is planned to be released in 2020 with data fromthe 2017 reference year of nonemployers, and annual releases will follow after that. From thispoint forward the terms “prototype” and “experimental version” will be used interchangeably.Background work began in April 2018 and resulted in a working paper (Luque et al., 2019),which assessed the viability of estimating nonemployer demographics exclusively with7

administrative and census data. The paper included discussions of the data and methodologythat could be used to create NES-D, the challenges and limitations we faced, and also providedpreliminary tabulations of owner and business counts by legal form of organization (LFO) at thenational level with one year of data (2015). 9 In the current paper, we revisit relevantbackground information and highlights in order to acquaint any new audience with NES-D, butour primary purpose is to expand last year’s work by: i) examining the longitudinal consistencyof administrative and census records coverage, and of our AR-based demographics estimates,ii) evaluating further coverage from additional data sources, iii) exploring estimates by state, iv)exploring estimates by industry sector, v) tabulating demographics estimates of businessreceipts as well as of counts, and vi) implementing imputation of missing demographics values.The 2020 prototype will include estimates of nonemployer demographics by LFO, receiptsize class, geography, and industry (see Figure III below). Geographic detail in the 2020prototype will consist of top 50 MSAs (Metropolitan Statistical Areas), state and nationalestimates while industry detail will include 2-digit NAICS. The plan is to increase the level ofboth geography and industry detail in future releases, 10 and to enrich and augment the set ofcharacteristics that describe nonemployer businesses.9No imputation of missing values was conducted.The level of detail possible will depend on statistical quality standards and disclosure avoidance rules.108

Figure III: NES-D TimelineYear 1 (2018): Research phase AR coverageassessment Preliminarydemographicsestimates Counts National 1 year of data(2015) Working paper IYear 2 (2019): Research phase Longitudinalconsistencyassessment Demographicsestimates Counts & receipts LFO National & State 2-digit NAICS 3 years of data(2014-2016) Imputation Working paper IIYear 3 (2020):Beyond Year 3: Release of NESD prototypewith 2017reference year LFO National, state& top 50 MSAs 2-digit NAICS Receipt-sizeclass Crosstabulationspossible Annual NES-Dreleases Additionalgeography &industry detail Longer-term:Additional nondemographiccharactersitics(e.g., ions toemployer status)Findings SummaryOur previous findings indicated that AR coverage of nonemployer business owners is veryhigh (see Luque et al., 2019). We were able to identify owners for approximately 99 percent of2015 nonemployer businesses (not including C-corporations), and obtain demographicinformation for approximately 90 to 99 percent of them -- depending on the demographic.Specifically, matching to the Census Numident provided sex, age, place of birth and citizenshipstatus for approximately 99 percent of identified owners while Decennial and ACS data suppliedrace and Hispanic origin for approximately 90 percent of identified owners. Furthermore, about90 percent of identified owners had no missing demographics and only about 1 percent wasmissing three or more demographic characteristics.The findings in this paper are consistent with our previous results. They indicate that ARcoverage rates remain high and stable over time for each of the three years in the 2014 through2016 period. We are again able to identify owners for approximately 99 percent of9

nonemployer businesses (not including C-corporations), and find that 92 to 93 percent ofidentified nonemployer owners have no missing demographics in each of the three years. Weare also able to increase AR coverage for some demographics thanks to the use of secondarydata sources. Specifically, using the Census Numident as a secondary source in addition toDecennial and ACS data increases the rate of Hispanic origin assignment by 5 percentage points(from 90 to 95 percent), and the rate of race assignment by 2 to 3 percentage points –depending on the year. The Data and AR Coverage Sections of this paper discuss these issues indetail, including how Numident data are used, and why the additional coverage for race islower than for Hispanic origin. Finally, as mentioned earlier, our work in this paper also includesthe imputation of missing values when we cannot obtain a given demographic characteristicfrom AR or census data. Following the imputation methodology used by business demographicssurveys (i.e., ABS and SBO), imputation is performed using a hot-deck procedure. 11For all demographic characteristics, our demographics estimates show stability and nosharp fluctuations over the time period under examination at the national, state and sectorlevels, and display some variation over time that are consistent with underlying population andindustry trends. We also observe (expected) heterogeneity in demographics distributions acrossLFOs, sectors and states for most demographic characteristics. While an in-depth study of thesepatterns or a formal comparison to the underlying national and state populations are beyondthe scope of this paper, we examine estimates by state and sector to check for unexpectedpatterns that do not align with demographic trends, and/or the AR demographics researchliterature. A general discussion of our results (for each demographic) as well as of comparabilityissues with the 2012 SBO follows below. A detailed discussion is provided in the FindingsSection of this paper.Regarding race, our estimates show stability and no sharp fluctuations over the three-yearperiod at the national, state and sector levels, with heterogeneity observed across LFOs, statesand sectors. Approximately 19 to 20 percent of nonemployer businesses in the 2014 throughThe imputation algorithm was provided by Robin Kurec at the Economic Statistical Methods Division, CensusBureau.1110

2016 period are not white-owned, and 30 to 32 percent are minority-owned. 12 We see slightincreases in minority-owned nonemployer businesses over time in line with demographicchanges in the underlying U.S. population. Variation across states follows general patterns ofracial distribution in underlying state populations. For instance, in 2016 white-owned firmsaccount for over 97 percent of nonemployer firms in Montana, Vermont, Maine, South Dakota,and Idaho while accounting for a smaller share of firms in Hawaii (52 percent).Similarly to race, our firm ownership results by Hispanic origin show no sharp fluctuations orirregularities over time at the national, state or sector levels, and display heterogeneity acrossLFOs, sectors and states. Approximately 13 to 14 percent of nonemployer firms are Hispanicowned, 86 to 87 percent are non-Hispanic-owned and less than 1 percent are equally Hispanic/non-Hispanic-owned. The observed slight increase in Hispanic-owned firms is consistent withU.S. population trends. Across states, the pattern also generally follows the underlying statepopulation. New Mexico (33-34 percent), Florida (30-32 percent), Texas (30-31 percent) are thestates with the highest share of Hispanic-owned nonemployer firms while Maine and Vermonthave the lowest with approximately 1 percent.Firm ownership by sex changes little over time between 2014 and 2016. Each year about 42percent of firms are female-owned, 56 percent of firms are male-owned and the rest areequally male-female owned. We observe variation across LFOs, sectors and states, withvariation across states being relatively smaller –as expected given underlying state populations.Maine, New Hampshire, Pennsylvania and Alaska have the highest shares of male-ownedbusinesses at about 59 to 60 percent while the District of Columbia has the lowest atapproximately 51 percent.Firm-level results concerning veteran status are, again, stable over time between 2014 and2015 at the national, state and sector levels, with heterogeneity observed across LFOs, statesand sectors. 13 Nationally, about 6 percent of all nonemployer firms are veteran-owned, about93.5 percent are non-veteran owned and the rest are equally-owned. Cross-state variation inThe minority category is comprised of individuals who are Hispanic (of any race), Black or African American,American Indian or Alaska Native, Asian, and Native Hawaiian or Other Pacific Islander.13 At the time this paper was written 2016 VA AR data were not yet available.1211

veteran-owned nonemployer firms spans approximately 8 percentage points (from 11 percentin Alaska and 10 in South Dakota to about 3 percent in New York and New Jersey).Regarding firm ownership by place of birth at the national, state and sector levels, 14 weagain see no sharp fluctuations over the 2014-16 period, but observe heterogeneity acrossLFOs, states and sectors. The percentage of nonemployer businesses owned by people bornoutside the U.S. is 21 percent in 2014 and 22 percent in 2016. Across-state variation spansapproximately 30 percentage points and reflects underlying state populations. The highestrates of U.S.-born ownership at about 96 percent are in Mississippi, West Virginia, Montanaand South Dakota compared to the lowest rates at approximately 60 percent in Florida.Turning now to firm ownership by U.S. citizenship, again we observe stable estimates overthe period under examination at the national, state and sector levels. U.S. citizens ownedapproximately 86 percent and non-citizens about 14 percent of all nonemployer businesses. Aswith other demographic characteristics, we observe variation in firm ownership by U.S.citizenship across LFOs, sectors and states. The highest rates of citizen ownership reach justover 98 percent in South Dakota, Montana and West Virginia compared to the lowest rate ofcitizen ownership just below 74 percent in Florida and New York.Age classifications of the ownership of nonemployer firms is also stable over the three yearswe consider at the national, state and sector levels, with observed age heterogeneity acrossLFOs, states and sectors. States with older owners of nonemployer firms include Maine andNew Hampshire as well as Vermont, where between 42 to 45 percent of firms are owned bypeople 55 and older. By contrast, nonemployer firms in the District of Columbia tend to haveyounger owners –approximately 27 percent of firms are owned by people 55 and older.Finally, and as discussed in Luque et al. (2019), it is worth noting that our AR-based and SBOfirm ownership estimates are not comparable for race, veteran status, and firm ownership bysex. 15 For veteran status, the concept of veteran captured by the SBO is broader than VA’sdefinition of a veteran. As a result, AR-based estimates are expected to be, and are, lower thanPlace of birth refers to whether a person has been born in the U.S. or outside the U.S.Also note that the SBO did not provide firm-level ownership estimates by owner’s age, place of birth or U.S.citizenship, so no comparison is possible for firm ownership of these demographic characteristics.141512

SBO estimates. Regarding firm ownership by sex, the survey response allows for soleproprietorships to be equally owned by a man and a woman (usually married couples) while ARcan only consider the sex of the person that appears as the owner of the sole proprietorship intax data. Consequently, the AR-based equally-owned category is expected to be, and is, lowerthan the SBO estimate. Regarding race, i) the SBO included a “Some-Other-Race” category(which is no longer allowed in business statistics or surveys), 16 and ii) AR research finds thatagreement rates for race between AR and survey responses are high, but tend to be lower forsmall population groups (e.g., American Indian and Alaska Native, Native Hawaiian and OtherPacific Islander) relative to other race groups. 17 In general, our AR-based and previous SBO raceestimates are within 5 percentage points at the national level, and our state level and sectorlevel estimates behave according to our expectations based on the comparability issuesmentioned above and prior AR research. The Comparison to SBO sub-section in the FindingsSection of this paper thoroughly discusses comparability issues between the 2012 SBO and ourAR-based estimates, and presents a detailed comparison of the two.Conclusions, Limitations & Next StepsOur primary purpose in this paper is to examine the longitudinal consistency of AR andcensus data coverage, and identify any red flags in our AR-based nonemployer demographicsestimates regarding patterns that are not consistent with national or state demographic trendsand/or AR demographics research. Our findings indicate that AR and census data can providenonemployer demographics statistics. We were able to identify owners for approximately 99percent of 2015 nonemployer businesses (not including C-corporations), and obtaindemographic information for approximately 90 to 99 percent of them -- depending on theThe revisions to Statistical Policy Directive No. 15, Race and Ethnic Standards for Federal Statistics andAdministrative Reporting issued by the Office of Management and Budget do not allow business demographicsstatistics to contain a race category of “Some Other Race. The Revisions to the Standards for the Classification ofFederal Data on Race and Ethnicity were issued in 1997 and can be accessed at 1997.pdf.17 Studies show that this can be largely attributed to the fact that racial fluidity is more prevalent among thesepopulations. Racial fluidity refers to the idea that an individual can be observed as having different races over timeor across data sources. See, for instance, Ennis et al. (2015), Liebler et al. (2014) ts/socio-econ-demo/race-fluidity.html.1613

demographic. These AR coverage rates remain stable in each of the three years underexamination (2014-2016). Demographics estimates show no sharp fluctuations over time, anddisplay expected small annual changes that are consistent with underlying U.S. population andindustry trends. These estimates are also stable within sector and state in each year underexamination, and as expected, are heterogeneous across legal form of organization, sectors andstates.There are still some issues to resolve and some limitations regarding what type ofinformation can be provided with AR and census data. One of those issues is how and/orwhether demographics for C-corporations can be obtained through AR. We plan to researchthis issue in the coming year, reach a decision, and if needed, provide some alternatives. NES-Dexperimental version release in 2020 will not include C- corporations, but fortunately, this typeof firm only makes-up 2 percent of all nonemployer businesses and accounts for 4 percent oftotal receipts. In addition, a substantial share of C- corporations may not eligible fordemographic classification because, for instance, they may be owned by other companies(instead of people). 18 Other issues pertain to differences between AR-based and survey-basedestimates, and these deserve further research. For instance, we plan to address the currentmisalignment between the survey-based (SBO and ABS) and AR-based veteran concept, and willexplore the use of Department of Defense AR data as a supplementary source with the goal ofbetter aligning the two. Another issue relates to differences in race self-responses in surveys vs.AR, and how these differences are more pronounced in small population groups (e.g., AmericanIndian and Alaska Native). Throughout this work, we continue to be mindful of concerns relatedto potential non-sampling errors in AR and census data sources (e.g., coverage and bias issues),and also of issues regarding data agreements and delivery schedules. These issues werediscussed in Luque et al. (2019) and are also discussed in detail in the Findings Section and theLimitations & Challenges Section of this paper.18The Methodology Section of this paper discusses the topic of firms not eligible for demographic assignment.14

This work will start transitioning into the production phase next year with the release ofa NES-D experimental version in 2020 with 2017 nonemployers. 19 This experimental version willinclude demographics estimates for nonemployer owners and their firms by LFO and receiptsize class at the national, state, top 50 MSA levels, and industry detail (likely 2-digit NAICS).Futu

filing tax forms for sole proprietorships (Form 1040, Schedule C), partnerships (Form 1065), or corporations (the Form 1120 series). The vast majority of nonemployers are sole proprietors followed by partnerships, S- corporations and C -corporations (approxim