2

Design and Implementation of the EEO-1 Component 2 Instrument

BACKGROUND OF COMPONENT 2 DESIGN

EEOC’s data collections capture a wide range of information to assess differences in employment by sex, race/ethnicity, and occupation. Although pay data have been collected from public-sector employees using the EEO-4 report since the 1970s (see Chapter 3), EEOC had not previously collected pay data from the private sector. The Component 2 data collection included three significant departures from earlier pay data collected from public-sector employees. The EEO-4 report limits pay-data information to full-time employees, uses fewer and wider pay bands, and does not collect information on hours worked.

Importantly, no federal data collection has captured the breadth of information from private-sector employers that is needed for EEOC’s enforcement purposes.¹ Although some federal statistical data collections of private employers include pay, sex, race/ethnicity, and occupation information, none collect all these variables, and the resulting data cannot be used for enforcement purposes by law (44 U.S.C. § 3501, 2010) (see Table 2-1).

To address this gap in private-sector pay data and better achieve its mission, EEOC explored multiple options. Several of these are described

___________________

¹ The Office of Federal Contract Compliance Programs requires private employers that contract with the federal government to make pay data available, but it does not require these data to be collected.

TABLE 2-1 Federal Data Collections with Pay and Demographic Measures

^a EEOC, 2016a.

^b EEOC, 2021d.

^c Bureau of Labor Statistics (BLS), 2021a.

^d Holt and Wagoner, 2021; Holt, personal communication, May 10, 2022.

^e BLS, 2022b.

^fBLS, 2022c.

^g Census Bureau, 2022.

^h Age is incorporated from the Social Security Administration’s numident files.

ⁱ Education is incorporated from the American Community Survey.

^j Census Bureau, 2021b.

^k The EEO tabulation is sponsored by four federal agencies: EEOC, the Employment Litigation Section of the Civil Rights Division at the Department of Justice, Office of Federal Contract Compliance Programs (OFCCP) at the Department of Labor, and the Office of Personnel Management (Census Bureau, 2021c).

in detail in the 2013 National Research Council report on this topic and therefore are not recounted here except to note final design recommendations (see Chapter 1). Subsequently, EEOC consulted with Sage Computing for additional design suggestions. EEOC ultimately chose a mixture of these measures and methods for its private-sector pay-data collection (hereafter referred to as Component 2) (see Table 2-2).

TABLE 2-2 Previous Measurement Recommendations and Decisions

^a “[A]nnual monetary compensation is defined as an employee’s base rate (wage or salary), plus other earnings such as cost-of-living allowance, hazard pay, or other increment paid to all employees regardless of tenure on the job. Annual monetary compensation should not include the value of benefits, overtime, or one-time payments such as relocation expenses. Annual

monetary compensation should be expressed in terms of an annual amount” (Office of Federal Contract Compliance Programs, 2000).

^b “Earnings in the OES survey are defined as straight-gross pay, exclusive of premium pay. The definition includes a base rate of pay, cost-of-living allowances, guaranteed pay, hazardous-duty pay, incentive pay (including commissions and production bonuses), and tips. The definition excludes overtime pay, severance pay, shift differentials, nonproduction bonuses, employer cost for supplementary benefits, and tuition reimbursements” (NRC, 2013, p. 56).

^c The Sage Computing report does not identify which box of the W-2 it recommends. However, it notes on p. 6 that the Internal Revenue Service (IRS) defines gross income “as including wages, salaries, fees, commissions, tips, taxable fringe benefits, and elective deferrals.” The inclusion of elective deferrals suggests Box 5 is in greater alignment with the spirit of its recommendations (Sage Computing, 2015, p. 6).

^d “Wages, tips, other compensation. “IRS directions for that entry are “show the total taxable wages, tips, and other compensation that you paid to your employee during the year. However, do not include elective deferrals (such as employee contributions to a section 401(k) or 403(b) plan) except section 501(c)(18) contributions.” Employers are expected to use payroll reports for the previous four quarters to generate the necessary data. Please refer to the IRS website for specific examples (IRS, 2022).

^e The 2000 Equal Opportunity survey instrument collected race as American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, and White. The instrument collected ethnicity as Hispanic or Latino (all races), Hispanic or Latino (White race only), and Hispanic or Latino (all other races). It also provided a category for “race unknown” (Office of Federal Contract Compliance Programs, 2000, pp. 4–5).

2013 National Research Council Report

EEOC asked the National Research Council, through the Committee on National Statistics, to convene a panel to review methods for measuring and collecting pay information from U.S. employers by sex and race/ethnicity. In its 2013 consensus report, the Panel on Measuring and Collecting Pay Information from U.S. Employers by Gender, Race, and National Origin considered suitable data-collection instruments, procedures for reducing reporting burden on employers, and issues of confidentiality protection and data access (NRC, 2013).

The 2012 National Research Council panel concluded that the collection of earnings data would be a significant undertaking for EEOC and cautioned the possibility of an increased reporting burden on some employers. The panel noted the absence of specific plans that it deemed necessary before the launch of data collection. To address this absence, the panel provided specific recommendations relevant to the study design and, subsequently, to quality of data for intended use.

Among the central findings of the 2013 report was that EEOC would benefit from articulating the planned use of earnings data. Pre-identification of a plan was necessary to design an information collection that would

adequately address the intended use of the data, particularly for decision making related to enforcement. Although EEOC described its intention to use these data for policy purposes, it did not describe the comparisons that would be made or the level of precision needed for decision making.

Once EEOC described its planned use for data analysis, an information collection could be designed. The 2012 panel recommended that this proposed information collection be tested in a pilot study to examine the feasibility, acceptability, and utility of the data collection for its intended purpose. A pilot study of a data collection is a field test that allows the selection of the most accurate and reliable measures and collection mechanisms in real-world conditions but on a smaller scale than the full study (e.g., in smaller geographic areas). Experiments with measures and methods are often embedded in a pilot study to inform design decisions and to estimate the actual burden and costs of the full study. Such a pilot study was not undertaken prior to Component 2 data collection.

The 2012 panel considered several measures of compensation for use in an EEOC pay-data collection. That panel recommended that EEOC use the same definition of earnings as the Bureau of Labor Statistics’ (BLS’s) Occupational Employment Survey (OES, later renamed to Occupational Employment and Wage Statistics [OEWS]), in a way that would allow the computation of valid measures of central tendency and dispersion. Prior to 2004, OES accepted printouts of payroll reports (which contained individual-level information) but did not request such information. In 2004, OES first began asking by email for the electronic submission of employee-level data. In 2011, OES began sending letters to the largest establishments asking for online or email submissions of employee-level data (rather than the completion of a paper form)² (Holt, personal communication, May 10, 2022).

The panel discouraged the use of pay bands because of the loss of information on within-group variance (NRC, 2013). The panel recommended that, if pay bands were used, the bands should be the same as those used by OES, because most employers were familiar with the requirements of that large-scale data collection. As described in the National Research Council’s 2013 report, “[e]arnings in the OES survey are defined as straight-time, gross pay, exclusive of premium pay. The definition includes a base rate of pay, cost-of-living allowances, guaranteed pay, hazardous-duty pay, incentive pay (including commissions and production bonuses), and tips. The

___________________

² In 2015, OES expanded its request for online or email submissions of employee-level data to include medium- and large-sized establishments; starting in November 2020, all firms were asked to report employee-level data. Thus, establishments initiated the process of submitting employee-level data, and OES (later renamed OEWS) responded by progressively making that format the default. However, then and now, OEWS has continued to allow reporting using pay bands.

definition excludes overtime pay, severance pay, shift differentials, nonproduction bonuses, employer cost for supplementary benefits, and tuition reimbursements” (NRC, 2013, p. 56). This definition of earnings was recommended because employers already were responding to the OES survey, which required this same information and earnings calculation.

EEOC adopted OES pay bands in the Component 2 data collection but used a more expansive definition of pay, based on Box 1 gross earnings as defined in federal tax law and reported via W-2 reports to the Internal Revenue Service. The W-2 Box 1 earnings definition includes all OES earnings components, plus bonuses, overtime, and shift-differential pay.

For both the field test and the full study, the 2012 National Research Council panel recommended that EEOC contract with a professional, large-scale data-collection organization to collect pay information. The 2012 panel was aware that EEOC’s past information collections were conducted by a smaller organization with less experience managing high volumes of sensitive, complex data. The accuracy, reliability, and timeliness of the new information collection would be best supported by a more experienced data-collection organization.³

The 2012 National Research Council panel understood that pay data would be used to assist with initial intake assessments as well as to prioritize among EEOC programs to address emerging enforcement needs. In addition to use in research on pay equity, pay data could be used for employer self-assessment. Analysts could examine pay differentials by type of establishment, location, occupation, sex, and race/ethnicity, which could not be done with existing data (NRC, 2013).

To enable this use, “…systems managing the data in a protected environment; statistical disclosure risk reduction methods; and statistical policy protections would have to be developed” (NRC, 2013, p. 87). Although EEOC has significant legal protections to maintain confidentiality of EEO data, technical and policy controls need to be implemented to improve appropriate access to these data, both for employer self-assessment and for research to inform public policy. Therefore, given the potentially disclosive and sensitive nature of pay data, and EEOC’s intention to make the resulting data available for employer self-assessment and research in anonymized form, the 2012 panel recommended EEOC improve its data-security systems and develop appropriate data-access policies.

___________________

³ EEOC did not follow this advice when the pay-data collection was launched in 2017; when it was relaunched in 2019, however, EEOC contracted with the National Opinion Research Center at the University of Chicago.

2015 Sage Computing Pay-Data Study

After the 2013 National Research Council report was released, EEOC commissioned Sage Computing to prepare a report that would provide EEOC’s Program Research and Surveys Division of the Office of Research, Information and Planning⁴ with the most efficient means of collecting compensation data from employers.

It is important to note that, while the Sage Computing report was titled a “pilot” study, it does not describe the design, implementation, or results of a field test for the collection of pay data. Therefore, the Sage Computing report does not provide a means for the objective assessment of collection cost, respondent burden, and data quality for enforcement purposes. Its discussion of estimated burden was hypothetical, and its recommendations regarding measures and data quality were not based on field experience. Thus, the 2017 and 2018 Component 2 data collections are the current panel’s primary source of information on field experience for EEOC’s pay-data collection.

The Sage Computing report addressed three aspects of the 2013 National Research Council report: the appropriate measurement of compensation data, potential statistical methods for EEO-1 pay-data analyses (see Chapters 6 and 7 of this report), and proposed modernization of EEOC information systems.

The Sage Computing report diverged from the recommendations in the 2013 National Research Council report. The Sage Computing report concluded that EEOC should use the W-2 definition of compensation, arguing that W-2 data are a more comprehensive measure of compensation than that used by OES and would be familiar and accessible to human resource information systems specialists completing EEO surveys. EEOC followed this advice in the Component 2 data collection.

Furthermore, the Sage Computing report suggested that, rather than individual-level estimates or pay rates, pay bands should be used—under the dual assumptions that they would reduce reporting burden and address confidentiality concerns (Sage Computing, 2015) with minimal loss of accuracy (Micklewright and Schnepf, 2007).

The Sage Computing report recommended that EEOC follow the methodology used by OES to establish appropriate pay bands. The number of pay bands could be reconsidered to balance data utility and response burden. EEOC followed these suggestions in the Component 2 data collection and used the 2017–2018 OES pay bands.

The Sage Computing report did not specifically address whether the 10 EEO-1 job categories should be used but instead focused on how to use them (i.e., collecting aggregate W-2 information using pay bands).

___________________

⁴ Now the Office of Enterprise Data and Analytics (OEDA).

In addition to collecting pay data, Sage Computing suggested that total hours worked for each sex, race/ethnicity, and occupation group be collected to account for pay differences due to differences in number of hours worked. EEOC followed these suggestions in the Component 2 data collection.

The Sage Computing report also provided a number of recommendations to improve information systems and resulting data quality. Sage Computing recommended updating the EEO-1 system data-upload functionality, noting that respondents would benefit from the real-time ability to correct and fix errors. Additionally, the report recommended that the responding employer’s address be validated, and a geocoded address stored in the database. EEOC implemented data-validation rules differently for separate reports used in the Component 2 data collection (EEOC, 2020e).

2016 EEO-1 Pay-Data Collection

After considering recommendations from the 2013 National Research Council report and the 2015 Sage Computing report, EEOC filed a request⁵ to OMB to collect data using the EEO-1 form. Specifically, EEO-1 would be fielded as one collection with two related but distinct components, each with slightly different eligibility criteria. Component 1 would comprise the historic form only and Component 2 would add pay data and hours worked to the historic form.⁶

Intended Use

In its OMB request, EEOC stated that collecting summary pay data with the EEO-1 form would “improve the EEOC’s ability to effectively assess allegations of pay discrimination and focus investigations, as well as strengthen the Office of Federal Contract Compliance Programs’ (OFCCP’s) ability to select appropriate federal contractors and subcontractors for review of their compliance with equal employment opportunity mandates”

___________________

⁵ Typically, an information-collection request submitted for clearance under the Paperwork Reduction Act should have a Supporting Statement B, which describes the statistical procedures to be used. EEOC’s 2016 information-collection request did not include a Supporting Statement B. However, statistical methods were described briefly in the Supporting Statement [A].

⁶ In 2018 and 2019, EEO-1 data collections occurred in two components. The customary EEO-1 instrument containing composition data (Component 1) was collected in 2018 for reporting year 2017, and in 2019 for reporting year 2018. The addition of pay data (Component 2) was collected for reporting years 2017 and 2018 during 2019 and 2020. In this report, we refer to historical EEO-1 data collections as EEO-1, and the 2017 and 2018 EEO-1 Component 1 and Component 2 instruments as “Component 1” and “Component 2,” respectively. All information collected by the Component 1 instrument is also collected by the Component 2 instrument.

(EEOC, 2016b, p. 1). In addition, EEOC would use these data in aggregated form to develop studies of the private-sector workforce. Summary pay data would encourage employers to voluntarily address unjustified pay disparities. EEOC might also provide these data to researchers for academic studies, subject to Title VII of the Civil Rights Act of 1964 (hereafter, Title VII) confidentiality and data-security requirements.

In terms of using these data to inform initial investigations of charges, EEOC indicated that Component 2 data would be examined by staff investigators using an expanded version of its existing desktop tool. This tool would display the distribution of various demographics (e.g., sex, race, and ethnicity) across job categories and across pay bands within job categories. EEOC noted that, where appropriate, enforcement staff could conduct a Kruskall-Wallis test to determine whether statistically significant disparities exist in an occupation by sex or race/ethnicity. In addition, using analytical software, staff could conduct an analysis of hours-worked data using an interval regression to determine whether pay disparities remain after data on hours worked are included in the analysis. These methods were also suggested in the Sage Computing report.

EEOC proposed that, when evaluating a charge, investigating staff could compare EEO-1 pay data to other available data (e.g., Census Bureau statistics regarding comparable workers).⁷ In addition, EEOC stated its intent to publish reports on pay disparities by race, sex, industry, occupational groupings, and metropolitan statistical area to assist employers wishing to engage in voluntary self-assessment of pay practices (EEOC, 2016b).

Reporting Requirements

As described in Chapter 1, the EEO-1 is a mandatory annual collection.⁸

• Component 1 (the historical collection) was required from all federal contractors with 50 or more employees.⁹

___________________

⁷ The particular Census Bureau data source for comparison was not identified.

⁸ In the OMB information request, when describing the impact of collecting data less frequently, EEOC noted that it considered collecting EEO-1 data every two years. This approach was rejected because EEOC concluded that the utility of the data would be diminished before new data became available. In the private sector, workforce changes are frequent. A delay of two years in collecting data reflecting these changes could undermine EEOC’s enforcement efforts.

⁹ Specific eligibility criteria were “federal contractors who (1) are not exempt as provided for by 41 CFR 60-1.5; (2) have 50 to 99 employees; (3) are prime contractors or first-tier subcontractors; and (4) have a contract, subcontract, or purchase order amounting to $50,000 or more or serve as depositories of Government funds in any amount, or are financial institutions which are issuing and paying agents for U.S. Savings Bonds and savings notes” (EEOC, 2016b, pp. 1–2).

• Components 1 and 2 (the historical collection plus the pay-data collection) were required from the following:
  • Private firms with 100 or more employees;
  • Private establishments with fewer than 100 employees if the establishment was owned by or affiliated with another establishment and the entire firm employed a total of 100 or more employees; and
  • Federal contractors with 100 or more employees (EEOC, 2020d).¹⁰

Single and multi-establishment firms had different reporting requirements (Figure 2-3).

• “Single-establishment firms (i.e., employers doing business at only one establishment in one location) had to complete a single EEO-1 online Component 2 Report for each reporting year” (Type 1; EEOC, 2020e, p. 271).
• “Multi-establishment firms (i.e., employers doing business at more than one establishment) had to complete as many as five online reports:
  • A consolidated report (Type 2) (mostly auto-populated);
  • A report covering the principal or headquarters office (Type 3);
  • A separate report for each establishment employing 50 or more persons (Type 4);
  • A separate report for each establishment with more than 50 employees which were not reported in the previous year (Type 9), if applicable; and
  • Either a separate report for each establishment employing fewer than 50 employees (Type 8); or

___________________

¹⁰ Specific eligibility criteria were “(i) all private employers who are (1) subject to Title VII, as amended, with 100 or more employees EXCLUDING State and local governments, public primary and secondary school systems, institutions of higher education, American Indian or Alaska Native tribes and tax-exempt private membership clubs other than labor organizations; OR (2) subject to Title VII who have fewer than 100 employees if the company is owned or affiliated with another company, or there is centralized ownership, control or management (such as central control of personnel policies and labor relations) so that the group legally constitutes a single enterprise, and the entire enterprise employs a total of 100 or more employees; and (ii) all federal contractors who (1) are not exempt as provided for by 41 CFR 60-1.5; (2) have 100 or more employees; (3) are prime contractors or first-tier subcontractors; and (4) have a contract, subcontract, or purchase order amounting to $50,000 or more; or (b) serve as depositories of Government funds in any amount; or are financial institutions which are issuing or paying agents for U.S. Savings Bonds and or savings notes. Establishments located in the District of Columbia and the 50 states are required to submit Standard Form 100 [the Component 1 Form]. No reports should be filed for establishments in Puerto Rico, the Virgin Islands, or other American Protectorates” (EEOC, 2020f, p. 2).

• • An establishment list showing the name, address, and total employment for each establishment employing fewer than 50 persons (Type 6)” (EEOC, 2020e, p. 271).

EEO-1 Frame (“Master List”)

The EEO-1 frame is a master list of employer companies that are required to file the EEO-1 Component 1 report(s) (EEOC, 2021g).

The EEO-1 frame is maintained by the Office of Enterprise Data and Analytics.¹¹ The firms on the frame are notified annually of their obligation to file the EEO-1 Component 1 report(s) (EEOC, 2021g). Firms could be added to the master list through firm self-identification, and, although infrequently, firms could also be added as a result of EEOC charge investigations (EEOC, 2021g,h).

Firms could also be removed from the frame. “If filers go out of business or undergo organizational changes such as mergers, acquisitions, or spin-offs, they must formally report these changes directly to the EEOC for the master list to remain accurate and up to date” (EEOC, 2021g, pp. 1–2). “Once the status of … ineligible filers is determined, updates are made in the EEO-1 Component 1 online filing system and to the master list… when a filer is determined to be ineligible, such a designation is indicated in the online filing system for the EEO-1 Component 1. The final cleaned and validated data file indicates which filers are now ‘ineligible’ (EEOC, 2021g, p. 3).

Historically, management of the frame relied largely on individual firms indicating that they were newly eligible, had closed or opened locations, or had become ineligible for the EEO data collection. “Periodically, the frame has been compared against commercial data, but this has not occurred on a regular or frequent basis” (EEOC, 2021h, p. 8). “Historically, the EEOC did not conduct additional canvassing efforts to inquire with filers about changes to their filing status” (EEOC, 2021g, p. 3).

Component 2 Instrument

The Component 2 instrument extends the Component 1 instrument by adding pay bands and hours-worked data components (see Figures 2-2, 2-3, and 2-4). Verbatim instructions to filers of the Component 2 instrument are provided for each data component, followed by a note regarding measurement considerations, where applicable.

___________________

¹¹ Formerly the Office of Research, Information and Planning.

Consistent with the Sage Computing report recommendations, EEOC elected to use the W-2 definition of income, to use pay bands that track the wage intervals used by BLS’s OES, and to collect summary hours-worked information for employees. Table 2-2 provides a comparison of measurement recommendations and decisions. Comparison of specific measures used in the Component 2 instrument to other federal data collections measuring employment are provided in Table 2-1.

Employees:

“Reports must include all full-time and part-time employees who were employed during the payroll period selected by the employer between October 1 and December 31 ([i.e., the] workforce snapshot)” (EEOC, 2020d, p. 6). Employment figures from any pay period in October through December could be used in this snapshot.¹²

The panel noted that there does not appear to be a data-verification process in place to ensure filers select a pay period during the specified “snapshot” time. The panel also noted that if the selected period doesn’t match the period for tax reporting, the increased reporting burden for filers could contribute to misreporting. If filers followed the instructions by selecting a snapshot period from October 1 to December 31 of the reporting year, the collected data were easier to interpret. If filers did not select that snapshot period, it was more difficult to compare pay and hours worked across establishments. Additionally, using the same snapshot pay period when reporting for Components 1 and 2 might lessen filer burden, improve prospects for error checking between the two components, and thereby improve data quality.

If filers did not use the same snapshot pay period when reporting for Components 1 and 2, when data are merged by establishment to evaluate coverage by industry, size, and geography (as described in Chapter 4), discrepancies in these files are more difficult to interpret. Both of these aspects—using a reporting period different from the tax period reporting period, and using different reporting periods for filing Component 1 and Component 2 data—can therefore affect data quality for intended use.

The panel examined the month of pay reporting periods selected for filing Component 1 and 2 instruments for 2017 and 2018¹³ (see Table 2-3).

___________________

¹² At the time of the proposed collection, Components 1 and 2 were to be fielded concurrently. It may have been assumed that the workforce snapshot referenced by an establishment would be the same for both components. However, the pause in data collection (discussed later in this section) may have increased the likelihood that the workforce snapshots used by an establishment differed when reporting Component 1 and 2.

¹³ The panel considered the unit of analysis (firm versus establishment) to be used when analyzing pay-period snapshots. Since establishments are the primary focus of this report, as they are understood to be the unit of greatest interest to EEOC for review of initial charges, the panel chose establishments for the analysis.

The table suggests that most filers picked a snapshot period consistent with the instruction, even without a formal check to ensure compliance: only 89 filers in 2017 and 65 in 2018 chose a snapshot period outside of the requested window. Snapshot periods outside of the requested window tended to be in January or September. For both years, filers who chose a snapshot period within the requested window were most likely to choose a snapshot period in December (46% of total filers).

However, most filers used different snapshot pay periods for Components 1 and 2 and, when the pay periods differed, the average number of days between their start dates ranged from 22–30, with small variations from one month to another. Among those records for which Components 1 and 2 could be matched, the proportion of filers that chose the same pay period for both Components 1 and 2 was 27.3 percent for 2017 and 37.1 percent for 2018.

Additionally, note there is no specific provision for identifying full-time, part-time, and newly hired employees (as is provided in EEO-4). There is also no distinction between Fair Labor Standards Act (FLSA) exempt and non-exempt employees (see below for discussion of hours worked).

Sex and race/ethnicity:

employees must be counted by sex and race/ethnicity for each of the ten occupational categories. Every employee must be accounted in only one of the seven EEOC race/ethnicity categories: Hispanic or Latino, White, Black or African American, Native Hawaiian or Other Pacific Islander, Asian, American Indian or Alaska Native, and two or more races. Categories provided for sex were male and female (EEOC, 2020e). Non-binary employees could be accounted for in the remarks section (see below; EEOC, 2020d).

Occupation:

employment data must be reported by EEOC job category. The 10 EEOC job categories are: executive/senior-level officials and managers; first- or mid-level officials and managers; professionals; technicians; sales workers; administrative support workers (formerly office and clerical workers); craft workers (formerly craft workers, skilled); operatives (formerly operatives, semi-skilled); laborers and helpers (formerly laborers, unskilled); and service workers. Each employee must be reported in only one job category (EEOC, 2020e).

The EEO-1 Job Classification Guide and the EEO-1 census codes crosswalk website were provided as job aides to assist filers completing the instrument. Filers were also referred to the Alphabetical and Classified Indices of Industries and Occupations (Census Bureau, 2010).

Pay:

filers were instructed to report W-2 Box 1 earnings for the year for all employees identified in the workforce snapshot. If an employee is not

TABLE 2-3 Reporting Periods Used for Establishments When Completing EEO-1 Data Forms

SOURCE: Component 2 firm and establishment files for 2017 and 2018.

NOTE: Excludes firms that reported more than 1.4 million employees (i.e., larger than the largest employer), and establishments that reported only employee count data and not pay data (i.e., Type 6 reports).

included in the workforce snapshot, that employee’s W-2 Box 1 earnings are not tallied. If an employee is included in the workforce snapshot, his or her W-2 Box 1 earnings are tallied, even if that employee is no longer employed at the end of the year. Note that income earned after the conclusion of the last full pay period in December is often paid to employees in the next calendar year. For EEO-1 purposes, that income would be counted in the next year, when it is reported on the W-2 (EEOC, 2020d).

Pay bands:

filers were instructed to report total employees in the workforce snapshot for each job category and pay band. Every employee identified in the workforce snapshot must be accounted for in one of the twelve pay bands for an EEO-1 job category. The pay bands are: (1) $19,239 and less; (2) $19,240–$24,439; (3) $24,440–$30,679; (4) $30,680–$38,999; (5) $39,000–$49,919; (6) $49,920–$62,919; (7) $62,920–$80,079; (8) $80,080–$101,919; (9) $101,920–$128,959; (10) $128,960–$163,799; (11) $163,800–$207,999; and (12) $208,000+ (EEOC, 2020d).

Hours worked:

filers were instructed to report hours worked for all employees in the snapshot in their job category and pay band. Hours worked are reported as an aggregate value for each job category and pay band, representing the total of all hours worked that year by all employees reported in that job category and pay band. If wages are paid in the next calendar year for work done in the last days of the past calendar year, these hours are counted in the next year, because that is when the pay is reported on the W-2.

For employees who are not exempt from the minimum wage and/or overtime provisions of the FLSA, filers are instructed to count the “hours worked” as recorded for FLSA purposes during the reporting year.

For employees who are exempt from the minimum wage and/or overtime provisions of the FLSA, filers are instructed to either (1) report a proxy of 40 hours per week for full-time exempt employees and 20 hours per week for part-time exempt employees, multiplied by the number of weeks the individuals were employed during the EEO-1 reporting year; or (2) provide actual hours worked by exempt employees if the employer already maintains accurate records of this information (EEOC, 2020d).¹⁴

Major activity:

filers were instructed that the major activity reported for the establishment should be sufficiently descriptive to identify the industry and product produced or service provided. If an establishment is engaged

___________________

¹⁴ To the extent that the use of the proxy numbers causes deviation from an exempt employee’s actual hours worked, the certification of the report as accurate would be considered appropriate.

in more than one activity, filers were instructed to describe the activity at which the greatest number of employees work. The description of the major activity indicated on the Headquarters Report (Type 3) must reflect the dominant economic activity of the company in which the greatest numbers of employees are engaged (EEOC, 2020f).

Remarks:

in this section, filers were instructed to include any remarks, explanations, or other pertinent information regarding this report (EEOC, 2020e). Employers were directed to note employees self-identifying as non-binary gender in the remarks section (EEOC, 2020d).

Certification:

if all reports were completed at headquarters, the authorized official was instructed to check Item 1 and sign the consolidated report (i.e., the Type 2 report) only. If the reports were completed at the individual establishments, the authorized official was instructed to check Item 2 and sign the establishment report (EEOC, 2020e).

Note that the panel interprets the certification instructions as directing the authorized official to check Item 1 and sign the consolidated report (i.e., the Type 2 report) if all reports were at headquarters. If reports have been completed at individual establishments of a multi-establishment firm, the authorized official was instructed to check Item 2 and sign an individual Type 8 report.

Submission Methods and Information Security

Filers of the EEO-1 Component 2 instruments were required to use the EEO-1 online-entry mode or to upload a data file using the data-upload mode.

Data-validation rules were implemented differently for separate reports (EEOC, 2020e). As noted earlier in this chapter, establishments with fewer than 50 employees could choose to complete either a Type 6 report or a Type 8 report.

• Type 6 reports do not require reporting individual counts by sex, race/ethnicity, job category, and pay band. When using the online-entry mode, companies using establishment-list reports (Type 6) must still enter all employment data from these smaller establishments into the consolidated firm-level report (Type 2).
• Type 8 reports require reporting individual counts by sex, race/ethnicity, job category, and pay band. When using the online-entry mode, all keyed employment data, including data from the Type 8 reports, automatically transfer to populate the consolidated report (EEOC, 2020d).

Filers were instructed that the total number of employees indicated on the headquarters report (Type 3), plus the establishment reports (Types 4 and 8), plus the list of establishments employing fewer than 50 employees (Type 6) must equal the total number of employees shown on the consolidated report (Type 2; EEOC, 2020e). If the filer included no Type 6 reports, the consolidated report was produced automatically, and thus no errors should occur. Filers using the data-upload mode did not need to submit a consolidated report if they had no Type 6 reports. Filers who included a Type 6 report were required to enter the Type 6 data into the consolidated report (Type 2); in those cases, users would see an error warning if there were discrepancies in the totals, and they were required to correct those errors before they could certify their data. “Employment data for multi-establishment firms, including parent corporations and their subsidiary holdings, must report all employees working at each establishment or subsidiary establishment”¹⁵ (EEOC, 2020e, p. 13).

In the 2016 OMB information-collection request, the total estimated annual burden-hour costs for employers and contractors completing the EEO-1 pay-data collection were estimated to be $53,546,359.08 (EEOC, 2016).

In 2019, EEOC re-estimated the burden calculation, taking into account federal guidance from the General Accounting Office and OMB. The improved methodology considered that the reporting burden differed based on the number of forms required for a particular filer. The recalculated burden for filing both Component 1 and 2 instruments of the EEO-1 data collection was estimated as $614,391,388 and $622,015,798 for reporting years 2017 and 2018.¹⁶

Regarding information security, EEOC stated in its 2016 OMB information-collection request that it had completed a privacy impact assessment and would be fully compliant with Circular A-130 before launch of the EEO-1 pay-data collection. However, it did not address secure access recommendations made by the 2013 National Research Council report.

___________________

¹⁵ For purposes of this report, “the term ‘parent corporation’ refers to any corporation that owns all or the majority stock of another corporation so that the latter relates to it as a subsidiary” (EEOC, 2020f, pp. 2–3).

¹⁶ In May 2018, EEOC created OEDA. In March 2019, EEOC began preparing to seek continued approval of the EEO-1 data collection under the Paperwork Reduction Act. OEDA staff revisited the previous methodology for calculating burden estimates utilized by EEOC for EEO-1, taking into consideration Government Accountability Office and OMB guidance on the appropriate methodology for calculating burden estimates in federal information collections. In light of these considerations, OEDA staff believed that EEOC’s previous burden estimate for EEO-1 was insufficient (EEOC, 2019b).

SPECIAL CIRCUMSTANCES OF THE EEO-1 COLLECTION

OMB approved EEOC’s request to collect EEO-1 pay data (Components 1 and 2) in September 2016. Concerns about the Component 2 collection led to OMB staying its prior approval of Component 2. The two components were separated after data collection began, and they continued as separate collections when data collection resumed under court order in 2019. Component 1 continued with the existing data-collection organization (Sage Computing). When the Component 2 data collection resumed by court order, it was not immediately clear to filers when, how, and for which year they were required to report. Subsequently, a data-collection contract for Component 2 was awarded to the National Opinion Research Center (NORC) at the University of Chicago, a professional survey research organization with considerable experience in large-scale collection of survey data.

Ultimately, two data-collection contractors collected EEO-1 reports (i.e., Sage Computing for Component 1 and NORC for Component 2), which may have created confusion for filers. Additionally, most employers were required to participate in both Component 1 and 2 data collections but at different time periods, which also may have created confusion. Last, significant time had passed between the reporting period and the continuation of data collection, creating possible data-retrieval problems. It is important to note that all information collected by the Component 1 instrument is also requested by the Component 2 instrument.

2017 Pause in Component 2 Data Collection

In January 2017, a private-sector firm that advises employers on legal compliance with EEO law, the OFCCP Institute,¹⁷ sent a letter to the Department of Labor and EEOC regarding the status of the EEO-1 pay-data collection (Cohen, Fortney, and Silberman, 2017). In February 2017, the Chamber of Commerce sent a letter to OMB requesting OMB to review and rescind EEOC’s previously approved collection of Component 2 data on the grounds of substantially underestimating public burden and therefore not meeting the justification of burden requirements under the Paperwork Reduction Act (PRA; Johnson and Plunkett, 2017). In March 2017, the Equal Employment Advisory Council (EEAC),¹⁸ which represents employers, also requested OMB to exercise its authority under

___________________

¹⁷ This text was changed after release of the pre-publication version of the report to correct an error regarding the name of the firm referenced. Now The Institute for Workplace Equality, this organization, previously doing business as the OFCCP Institute, is not affiliated with the Department of Labor’s Office of Federal Contract Compliance Programs.

¹⁸ Now the Center for Workplace Compliance: https://www.cwc.org/CWC/About_/History.aspx

the PRA to review EEOC’s previously approved collection of Component 2 data. EEAC urged OMB to immediately rescind its approval of the Component 2 instrument because EEOC’s underlying PRA justification was in material error, and because new information disclosed after the form was approved generated significant compliance questions (Eastman, 2017). These concerns were summarized into three categories. First, the resulting data would be of low utility, meaning they were highly unlikely to be of real help in identifying unlawful or improper pay practices.¹⁹ Second, EEOC’s respondent-burden estimates were inaccurate, such that the proposed pay-data collection would be tremendously burdensome to implement, so was not justifiable by the anticipated data utility. Third, confidentiality concerns regarding the disclosure of potentially sensitive compensation data had not been properly addressed. EEAC noted that EEOC had not addressed the recommendations made in the 2013 National Research Council report.

In April 2017, Lamar Alexander (Senator, R-TN), then Chairman of the Committee on Health, Education, Labor, and Pensions; and Pat Roberts (Senator, R-KS) requested OMB to rescind approval for the EEO-1 pay-data collection (Alexander and Roberts, 2017). In their letter to the OMB director, Senators Alexander and Roberts cited undue respondent burden concerns, concerns regarding the calculation of burden estimates as submitted in EEOC’s 2016 information-collection request to OMB, and concerns regarding respondent privacy and data confidentiality.

The same day, the National Women’s Law Center (NWLC) and 90 additional signatories wrote to OMB in opposition to requests to revoke the previously approved data collection (National Women’s Law Center, 2017). The Leadership Conference on Civil and Human Rights also wrote to OMB opposing revocation of approval (Henderson and Zirkin, 2017). In May 2017, Senators Cory Booker (D-NJ), Patty Murray (D-WA), and Kirsten Gillibrand (D-NY), joined by more than two dozen other senators, wrote to OMB also opposing the revocation of approval (Booker, Murray, and Gillibrand, 2017). These arguments can be summarized in two main themes. First, since OMB had already approved the proposed data collection under the PRA, OMB could only review its previous approval if relevant circumstances changed or if the burden estimates as submitted by EEOC at the time of initial approval were in material error, and neither of these conditions had been met. Second, the pay-data collection would assist EEOC in investigations of pay disparities, assist employees in advocating for fair pay, and assist employers by providing data that could be used to benchmark and compare performance.

___________________

¹⁹ As relevant, concerns about utility from a data-quality standpoint are examined in Chapters 3–7.

In May 2017, a hearing by the House Committee on Education and Labor, Subcommittee on Workforce Protections was held, titled The Need for More Responsible Regulatory and Enforcement Policies at the EEOC (Education and Labor Committee, 2017). Testimony was given by advocates for each viewpoint.

In August 2017, OMB stayed its approval of the EEO-1 pay-data collection. In September 2017, EEOC issued a stay for the collection of Component 2 data but instructed filers to continue to submit the Component 1 instrument.

It is important to note that, during the pause in collection of Component 2 data, data collection continued for Component 1, managed by a small computing research organization on behalf of EEOC. EEOC instructed filers to continue to use the currently opened EEO-1 portal to submit 2018 Component 1 data by May 31, 2019.

2019 Court-Ordered Reinstatement of Component 2 Data Collection

In November 2017, NWLC and the Labor Council for Latin American Advancement (LCLAA) sued OMB; John Michael Mulvaney, Director of OMB; Neomi Rao, Administrator of Office of Information and Regulatory Affairs; EEOC; and Victoria A. Lipnic, Acting Chair of EEOC.

NWLC claimed, among other injuries, that if OMB had not stayed the pay-data collection, NWLC “would have been able to make its reports and advocacy more robust with additional data and analysis” (NWLC et al. v. OMB et al., 2019a, p. 10). Possession of aggregate EEO-1 pay data would allow NWLC to “focus its resources, analysis, and advocacy on the jobs, industries, and regions where intervention is most urgent” (NWLC et al. v. OMB et al., 2019a, p. 10). LCLAA similarly claimed injury from OMB’s decision to stay the pay-data collection (NWLC et al. v. OMB et al., 2019a). With information that otherwise would have been collected, LCLAA would “have presented statistics on pay equity within industries and across the nation, based on these data. This information would have materially improved LCLAA’s and its members’ ability to negotiate with and educate employers and to fulfill LCLAA’s mission of improving the condition of Latinos and Latinas in the workforce” (NWLC et al. v. OMB et al., 2019a, p. 10).

On March 4, 2019, the U.S. District Court for the District of Columbia found in the plaintiffs’ favor. It found that the stay in Component 2 data collection materially injured NWLC and LCLAA by increasing costs to achieve their organizations’ missions. The court found there was public expectation, based on OMB Federal Register notices, that Component 2 data in aggregate form would be made available to the public such that NWLC, LCLAA, and others could use those data to address pay disparities on behalf of their stakeholders (NWLC et al. v. OMB et al., 2019a). OMB

did not provide an analysis of burden to justify its claims of material error regarding the calculation of reporting burden. Noting the absence of OMB’s analysis, the court found that OMB’s argument that data-file formatting requirements for Component 2 data understated the actual respondent burden did not represent a change in circumstance or material error that justified stay of the data collection for good cause (NWLC et al. v. OMB et al., 2019a>).

Accordingly, in Civil Action No. 17-cv-2458 (D.D.C.) document 46 (NWLC et al. v. OMB et al., 2019b) the Court ordered the Component 2 data collection to resume immediately. In D.D.C. document 71 (NWLC et al. v. OMB et al., 2019c, p. 2) it further ordered the following:

• “By April 29, 2019, EEOC must issue a statement on its website and submit the same for publication in the Federal Register notifying EEO-1 filers that they should prepare to submit Component 2 data no later than September 30, 2019;”
• “In lieu of collection of Component 2 data for calendar year 2017, EEOC could satisfy the court’s order requiring two years of data by collecting Component 2 data for 2019 during the 2020 EEO-1 reporting period. EEOC must alert filers if it has decided by April 29, 2019 whether to collect 2017 or 2019 data;”
• “If EEOC has not decided by April 29, 2019 whether to collect 2017 or 2019 data, it must issue a statement of its decision on its website and submit the same for publication in the Federal Register by May 3, 2019;”
• “Beginning on May 3, 2019, and continuing every 21 days thereafter, EEOC must provide reports to plaintiffs and the court of all steps taken to implement the Component 2 data collections since the prior report, of all steps to be taken during the ensuing three-week period, and indicating whether EEOC is on track to complete the collection(s) by September 30, 2019;” and
• The court ordered that the Component 2 data collection(s) would not be deemed complete until the “percentage of EEO-1 reporters that submitted Component 2 data equaled or exceeded the mean percentage of EEO-1 reporters that submitted EEO-1 reports in each of the past four reporting years.”

2019 Compressed Component 2 Data Collection

In April 2019, EEOC notified EEO-1 filers to prepare to submit Component 2 data for calendar year 2018 by September 30, 2019. EEOC stated that it expected to begin collecting Component 2 data for calendar year

2018 in mid-July 2019 and would notify filers of the precise date the survey would open as soon as it was available. EEOC also stated that it would collect Component 2 data for either calendar year 2017 or 2019, noting that it would publish in the Federal Register by May 3, 2019 an additional notice announcing which year would be required (EEOC, 2019c).

In May 2019, the Component 2 data-collection contract was awarded to NORC. Shortly after, EEOC announced its decision to collect 2017 Component 2 data in addition to 2018 Component 2 data (EEOC, 2019d). Data collection for Component 2 began in July. The court determined that a reasonable response rate had been achieved, and the Component 2 data collection was permitted to conclude in February 2020 (NWLC et al. v. OMB et al., 2020). This represented a much longer response window than anticipated compared to past collections of Component 1 data.

POSSIBLE EFFECTS ON DATA QUALITY

Specific design decisions and field experiences may have affected the quality of Component 2 data. Several of these experiences relate to response rates and item reliability. Response rates affect data quality, even if population generalizations are not the primary use of the data. If eligible filers do not respond, nonresponse affects EEOC’s ability to locate and investigate some establishments for which individuals bring charges. Nonresponse also affects EEOC’s ability to compare peer establishments to the establishment under investigation, and to explore systemic charges. Reliability of data items matters for data quality. Even if subsequent investigations access more complete data for a given establishment, poor reliability of initial data impairs EEOC’s ability to describe the establishment size (by employee count) and industry, and to account for hours worked in pay comparisons. In addition, allowing establishments and firms to choose different reporting periods may hamper within- and between-firm data comparability.

Under the PRA, agencies must balance the utility of the data against the burden the data collection imposes on the employers that must submit it. EEOC needs to consider the expected utility of the Component 2 data collection before deciding whether to submit a pay-data collection to OMB.

SUMMARY

This chapter provided a brief history of the Component 2 data collection. It began with a description of key design considerations and subsequent decisions. It then described the special circumstances surrounding the implementation of the Component 2 data collection, which was launched in 2016, paused in 2017, and resumed under court order in 2019. Design decisions and implementation experiences that may have affected the

quality of Component 2 data are restated below. Some of these potential concerns were identified by EEOC in early discussions with the Committee on National Statistics to develop the panel’s charge. These are marked with an asterisk (*). Some potential concerns, although discussed at a conceptual level, could not be investigated empirically by the panel due to data, budget, and time limitations. These are marked with an exclamation point (!).

Investigation of these potential concerns and their relevance to future Component 2 data collections (where relevant) are presented in subsequent chapters, as noted in adjacent parentheses.

Component 2 Instrument Design Decisions and Implementation Experiences to Be Examined

Sample Frame

An incomplete firm master list or sample frame can affect EEOC’s ability to identify establishments that should file EEO-1 reports. This can lead to missing firm and establishment data important when individual charges are filed, as well as the inability to correctly identify peer establishments for comparison when examining individual or systemic charges. (Chapters 4, 6, 7)*

Measures

• Measuring compensation as Box 1 of form W-2 may not provide a comprehensive measure of pay as required by EEOC statute. (Chapter 3)
• Collecting pay data in pay bands is not suitable for analysis of central tendency or dispersion. This limits comparison of pay by masking differences within wide pay bands. (Chapters 3, 6, 7)*
• The absence of indicators for part-time/full-time/new hires and exempt/non-exempt employees affects interpretation and therefore utility of hours for pay analysis. (Chapters 3, 6, 7)
• When job categories and race/ethnicity measures are overly broad, comparison to similarly situated employees can be difficult. Additionally, when job categories and race/ethnicity measures are not consistent with federal data standards, comparisons to other federal data for investigation benchmarking purposes and further analysis are limited. (Chapter 3, 6, 7)
• Gender identity is treated as a protected group through case law under Title VII. Collecting non-binary gender identity under a “remarks” section may be stigmatizing to some employees. This

• can reduce the likelihood of response and the utility of the data to describe similarly situated employees, thus diminishing data quality for enforcement purposes. (Chapter 3)!

Instrument and Submission

• The Component 2 instrument featured six report types or forms. Each was required under different circumstances, and each featured specific data validation checks. The complexity of the reporting requirements could affect filer response rates, individual question (i.e., item) response, and reliability, thereby decreasing the amount and quality of data collected. (Chapters 4, 5)
• The smallest eligible establishments for the Component 2 data collection could be reported on using Type 6 reports. Type 6 reports do not require employee data on sex, race/ethnicity, or job category, nor pay or hours-worked data. To the extent that the smallest eligible establishments chose to submit Type 6 reports, these data are not available for the examination of pay disparities. These smallest establishments may be the most vulnerable to pay disparities. Type 6 reports may increase respondent burden and introduce confusion in producing consolidated reports for multi-establishment firms. (Chapters 4, 6, 7)
• Data collected using the Component 2 Type 2 report required manual checks to confirm that employee counts across establishments by sex, race/ethnicity, and job category were correct. Errors in counting employees across sex, race/ethnicity, and job categories could lead to misclassification of establishment and firm size, and therefore affect comparisons to similarly situated employees. (Chapters 5, 6, 7)
• Hours-worked data were reported on a separate form from that used to report employee counts. Separate forms, as opposed to a side-by-side format, may have caused confusion for filers, leading to error in reporting hours worked and therefore possibly resulting in inaccurate employee counts and/or calculations of hourly pay. (Chapters 3, 5, 6)
• Employers could not download reports prior to certification. This limited their ability to check responses and thereby reduce error. (Chapter 5)
• The submission process allowed professional employer organizations (PEOs) to submit Component 2 data on behalf of their filer clients. However, filers are responsible for certifying their submissions. Use of PEOs may have reduced the opportunity for filers to correct errors in their filings, ranging from employee counts and

• associated hours worked to major activity of the establishment or firm. (Chapters 3, 4, 5, 6, 7)

Information Security and Modernization

• When the Component 2 data collection was approved by OMB in 2016, it was conducted by a research organization without large-scale data-collection experience. Component 2 data collection resumed in 2019 under the aegis of a more experienced information-collection research organization. Collection and management of Component 2 data by two different research organizations may have affected data file management and therefore data quality—including matching firms and establishments to examine the reliability of responses. (Chapters 4, 5)*
• From the Component 2 information-collection request materials sent to OMB in 2016, it is not clear which system protocols were enhanced to protect the sensitive nature of pay data. To the extent that these enhancements were not communicated well to filers, this concern could affect filers’ willingness to respond, even for a mandatory collection. (Chapters 4, 5)

Court-Ordered Data Collection

• When OMB stayed approval of the EEO-1 pay-data collection, EEOC continued data collection for Component 1 (the historical collection). All measures included in the Component 1 instrument are also included in the Component 2 instrument. When the Component 2 data collection resumed, filers may have been confused, thinking they had responded to the filing requirement by filing Component 1 data, when they also needed to complete the Component 2 instrument. This confusion may have affected response rates. (Chapters 4, 5, 6)*
• EEOC contracted with two separate research organizations to collect Component 1 and Component 2 data. This difference in contractors may have confused some filers and affected response rates to the Component 2 instrument. (Chapters 4, 5, 6)*
• The prolonged data-collection window (and uncertainties around it) may have affected response rates to the Component 2 instrument. (Chapter 4)*

RECOMMENDATIONS

Improvements could be made to the current Component 2 data-collection instrument to substantially reduce and possibility eliminate many of the errors and weaknesses identified by the panel.

RECOMMENDATION 2-1: EEOC should combine the Component 1 and 2 instruments into a single data-collection instrument, thus lessening respondent burden and reducing the chances for inconsistencies or reporting errors.

RECOMMENDATION 2-2: EEOC should eliminate Type 6 reports and mandate Type 8 reports for all establishments in multi-establishment firms of 100 or more employees. Consolidated reports (Type 2) could then be eliminated, and firm-level data created by summing across establishment reports. These actions would increase coverage, simplify reporting, and reduce respondent burden.

This page intentionally left blank.