Written Testimony of Camille Olson Seyfarth Shaw LLP on behalf of U.S. Chamber of Commerce

Good morning Madam Chair, Commissioners, and General Counsel. Thank you for the opportunity to participate in today's hearing. On behalf of the U.S. Chamber of Commerce, I am pleased to testify on the Equal Employment Opportunity Commission's (EEOC's or Commission's) Proposed Revisions to the Employer Information Report (EEO-1) (Proposed Revisions).[1] I am Chair of the Chamber's equal employment opportunity policy subcommittee.[2] The Chamber is the world's largest business federation, representing more than three million businesses and organizations of every size, sector, and region, with substantial membership in all 50 states. The Chamber's mission is to advance human progress through an economic, political, and social system based on individual freedom, incentive, initiative, opportunity, and responsibility.

In today's testimony, I discuss the impact of the EEOC's Proposed Revisions to the EEO-1 Report, specifically with respect to the proposed "Component 2" requirements. According to the EEOC's proposed changes, beginning in 2017, employers with 100 or more employees would be required, for the first time, to compute, record, aggregate and submit data regarding employees' W-2 earnings and hours worked for each of its establishments. For each of the 10 EEO-1 job categories, by location, employers would be required to report the number of employees by ethnicity, race, and sex that fall within twelve pay bands. Employers would be required to provide W-2 earnings data from any pay period between July 1st and September 30^th.

According to the EEOC, both the Commission and the Office of Federal Contract Compliance Programs (OFCCP) will use this data "to more effectively focus agency investigations, assess complaints of discrimination, and identify existing pay disparities that may warrant further examination." The EEOC has also stated that in conjunction with the OFCCP, it will develop statistical software to analyze the EEO-1 data, as well as guidance to be used by their staff when reviewing such data.

The impact of the new EEO-1 report would be substantial, both in the millions of hours that private employers would be required to spend completing the new report and in the false results that may be generated. This report, allegedly borne out of a laudable desire to ensure that employees are paid fairly, will do no such thing. Instead, in practice, it would impose enormous burdens and risks on employers who base complex compensation decisions on factors other than membership in a particular EEO-1 category. However, at the outset I must highlight that the U.S. Chamber of Commerce and its member employers have always been supportive of the non-discrimination laws including the Equal Pay Act (EPA), Title VII and the Americans with Disabilities Act (ADA). We have held many seminars and meetings in which the requirements of the laws have been explained and which afford our members the latest information regarding compliance. And indeed it was in large part through the efforts of the U.S. Chamber and our other employer organizations that the 2008 Americans with Disabilities Amendments Act was enacted with bi-partisan support and with the active involvement of all stakeholders. That is why we are so concerned with the current attempt to revise the EEO-1 form without any input from the employer community and in a process so violative of the Paperwork Reduction Act.

Summary of Submission

On February 1, 2016, without any prior notice to the regulated community, the EEOC published a proposed revision to the EEO-1, Employer Information Report. This data request, to which every employer with 100 or more employees and every government contractor with 50 or more employees must respond annually, has been in existence for 50 years. However, for the first time, the EEOC is proposing that employers submit data showing the W-2 wages and hours worked of all of their employees divided into 12 arbitrary pay bands, in addition to the demographic makeup of their employment rosters. To get a sense of the magnitude of the proposed revision, the existing EEO-1 report requires 128 data points. Pursuant to the changes proposed by the EEOC, covered employers will have to submit forms for each establishment, and each establishment report would consist of 3,360 data points.

Ironically, EEOC advances this brand new and burdensome data request pursuant to, of all laws, the Paperwork Reduction Act (PRA or Act).[3] The PRA requires that any request for data by a government agency meet three basic criteria: (1) the request minimizes the burden on responders to reply; (2) the request results in data which is meaningful to the government for policy and enforcement purposes; and (3) the data request is designed to ensure that the data is securely and confidentially obtained and retained. The OMB is charged with reviewing the data request to ensure that the requesting agency complies with the PRA.

Although the EEOC admits that it began working on this project four years ago (and perhaps even longer ago), it permitted employers only five weeks to prepare these comments for submission at a public "hearing" in which selected employer representatives are permitted five minutes each to "testify" about the proposal. The Chamber's request for an extension of the comment period due date was summarily denied. The Chamber nevertheless retained noted labor economists and statisticians to review the EEOC's proposal and two law firms to further analyze the legal compliance of the EEOC with the PRA and the new requirements it is proposing to impose on employers. The conclusions of these reviewing experts and law firms is stunning:

The EEOC has cavalierly refused to comply with its responsibilities under the Paperwork Reduction Act in the following ways:

The EEOC has produced an "analysis" of the burden its proposal will impose which is completely lacking in any substance and has no basis in fact.

The EEOC suggests that a "revised form" with almost 26 times the number of data points to complete will impose no additional burden and cost 50% less than the previous form which was approved in 2015.
The EEOC and its consultant admit that there was no testing of the form or the time that would take to complete it, but rather that it used "synthetic data" compiled from fictitious companies to produce an estimate of the time required to complete the new forms.
The EEOC refers to proposals by other agencies which have never been completed and which have never been published to sustain its estimate of burden.
The EEOC, through sleight of hand, arbitrarily eliminated from its analysis of the burden the time and effort required to submit data relating to more than 250,000 employer establishments. Under the EEOC's proposal, employers will still be required to submit data for the 250,000 establishments that have been omitted from the Agency's burden analysis. The EEOC simply ignores this fact.

The EEOC offers no argument that the mass of data to be submitted will be useful for any law enforcement or policy enhancement.

The laws that the EEOC enforces do not permit the consideration of broad aggregates of data from dissimilar jobs combined into artificial groupings. There can be no legal or enforcement use of this data. Indeed, the EEOC's own compliance manual and its consultant recognize that these broad aggregations of data are essentially useless. Myriad federal courts have reached the same conclusion.
The EEOC is requiring the combining of completely dissimilar jobs to determine if there is pay discrimination. For instance, the proposed revised forms will require a reporting hospital to combine lawyers, doctors, nurses and dieticians - all grouped as "professionals" - to somehow determine whether there are pay disparities based on gender, race or ethnicity. No law permits comparisons of such diverse workers to prove discrimination.
In order to meet its own bureaucratic timetable, the EEOC will require employers to combine two distinct years of W-2 data to create a fictitious W-2 amount for employees. This combination of W-2 data over a two year period will yield completely useless information. It does not take into account job changes, promotions, annual pay adjustments, different working conditions or locations or the many other factors that go into compensation.
The EEOC will require employers to collect and report the hours worked for all employees. While the EEOC suggests that it will not require collection of new information from employers, they have not addressed the critical fact that employers do not currently collect hours information for exempt employees. The EEOC suggests that employers may use a "default" number of 40 hours for each exempt employee. In the private sector, exempt employees regularly work more than 40 hours, thus the hours information would be inaccurate, and therefore, of limited use. A legitimate study before this proposal was published would have revealed that fact.

The EEOC offers absolutely no discussion of the threats to confidentiality or privacy of the information it is requiring employers to submit.

The PRA requires that the requesting agency and the OMB ensure that data collected will be treated with complete confidentiality. The EEOC did not even attempt to take this responsibility seriously. When the OPM cannot even protect the personnel data of 21 million federal employees or applicants, the EEOC should at least be cognizant that confidentiality of the pay data of more than 70,000 employers--encompassing millions of employees--deserves some consideration. It offers none.
This data, which is shared with the Department of Labor, is subject to demand for production under the Freedom of Information Act. The EEOC states that the data will be protected to the extent permitted by law. Of course, employers will have to expend significant resources to assert their right under law to protect this data. And for some of the smaller employers, the identity and the compensation of employees will be easily ascertained; the same is true for larger employers with small establishments. The EEOC devotes only two paragraphs to discuss these privacy and confidentiality issues.

In short, the EEOC has sprung upon employers a proposal that would (1) impose significant new, costly administrative burdens; (2) yield data of no utility; and (3) fail to protect confidential information. For these reasons, the Chamber submits that EEOC should withdraw this proposed data collection and, if it refuses to do so, the OMB should exercise its authority and refuse to approve the revised form.

I. Paperwork Reduction Act

As previously noted, the EEO-1 Revision process is being conducted pursuant to the PRA. The PRA, which was reauthorized in 1995, was promulgated in order to bring a degree of coherence and prudence to the Government's rather voracious quest to collect data from responding employers (and other responders). See Dole v. United Steelworkers of America, 494 U.S. 26 (1990) (recognizing that the PRA was enacted in response to the federal government's "insatiable appetite for data."). The purposes of the PRA set forth in direct terms what the Act was designed to accomplish:

The purposes of this chapter are to--

(1) minimize the paperwork burdens for individuals, small businesses…Federal contractors…and other persons resulting from the collection of information by or for the Federal Government;

(2) ensure the greatest possible public benefit from and maximize the utility of information created, collected, maintained, used, shared and disseminated by or for the federal Government.

(4) improve the quality and use of Federal information to strengthen decisionmaking, accountability and openness in Government and society…

The PRA established within the Office of Management and Budget ("OMB") the Office of Information and Regulatory Affairs ("OIRA") whose Director is charged with the administration of the PRA. Livestock Marketing Ass'n v. U.S. Dept. of Agr., 132 F. Supp. 2d 817, 830 (D.S.D. 2001) ("Among other things, the Act establishes the Office of Information and Regulatory Affairs within the Office of Management and Budget, with authority to" facilitate and manage the PRA). OIRA has a substantial role in the federal regulatory process. No data collection instrument which is directed to more than nine responders can be issued without first receiving the approval of OMB and OIRA. CTIA-The Wireless Ass'n v. F.C.C., 530 F.3d 984, 987 (2008) ("The need for OMB approval of information collections derives from the Paperwork Reduction Act"); Gossner Foods, Inc. v. E.P.A., 918 F. Supp 359, 361-62 (D. Utah 1996) ("The Act institutes a second layer of review by the OMB for new paperwork requirements.") (quoting Dole, 494 U.S. at 32-33).

The Director, in turn, is mandated to review data collection requests in accordance with the direction of the PRA to (1) minimize the burden on those individuals and entities most adversely affected and (2) maximize the practical utility of and public benefit from information collected by or for the Federal Government and establish standards for the agencies to estimate the burden of data collection. See Dole, 494 U.S. at 32 (explaining that the PRA charges the OMB with responsibility for minimizing the burden on individuals and establishing standards to reduce federal collection of information). See also Tozzi v. U.S. E.P.A., No. Civ. 98 - 0169 (TFH), 1998 WL 1661504, *1 (D.D.C. April 21, 1998). The Director is also charged with developing and promulgating standards to insure the privacy, confidentiality and security of information collected or maintained by agencies. In re French, 401. B.R. 295 (E.D. Tenn. 2009) (noting that, amongst other things, the PRA's purpose is to ensure that information is collected consistent with privacy and security laws) (quoting 44 U.S.C. § 3501).

It is under this specific statutory framework that the proposal to revise the EEO-1 form and collection procedures must be reviewed. There is no exemption from the mandates of the PRA and neither the agency head nor the Director of OIRA may exempt data collection efforts from the constraints and mandates of the PRA. Simply put, data collection efforts of the size and scope of the proposed EEO-1 revision must be reviewed independently to insure that the burdens placed upon responders is minimized, the information collected has the maximum utility, and that the security and confidentiality of the information is assured. See Dole, 494 U.S. at 32-33 (explaining that all federal agency regulations that require the collection of paperwork must be reviewed by the OMB in accordance with the goals and purpose of the PRA).

These are simple precepts and should be reviewed independently. In other words, the burden placed upon responders must be determined only with respect to the efforts necessary to collect and report the data. The obligation to minimize the burden should not be reviewed in relevance to the purposes of the data. The PRA does not create a burden/benefit analysis but rather requires each standard to be reviewed on its own terms. Similarly, the benefit to be derived from the data collection requirement is not dependent upon the degree of burden the effort creates, but rather the utility of the data collected. The government is not permitted to require the collection of data with no utility regardless of the extent of the burden imposed to collect the data. In short, there is no license for the Government to simply collect data without a clearly articulated purpose and legitimate use. Finally, in conjunction with the type and sensitivity of the data being collected, OIRA and the requesting agency must insure that the requested data be collected and retained in a manner to insure its confidentiality and privacy.

These requirements are neither difficult nor complex. Indeed, they establish a commonsense framework by which the requesting agency in the first instance and Director of OIRA must review requests for authorization to collect data. As will be shown in this submission and in our formal comments filed by April 1, 2016, the EEOC has not met any of the statutory requirements. Rather, it has proposed a data collection protocol which will place a significant burden upon responding employers and which the EEOC has remarkably understated or ignored. The EEOC is proposing to collect extensive data heretofore never collected by the federal government without any developed framework to review the data, or use the data for any legally authorized or recognized purpose. In addition, the EEOC has totally ignored the obvious security and confidentiality issues which it is directed to consider.

II. The EEOC's PRA Burden Estimate is Insufficient and Unsupported

The EEOC should rescind its Proposed Revisions because its PRA burden estimate is wholly deficient in three critical ways: (1) it underestimates the employer burden associated with generating the EEO-1 Report in its current form and is inconsistent with prior EEOC estimates of the burden associated with generating the EEO-1 Report; (2) it underestimates the employer burden associated with compiling, analyzing, and reporting the W-2 information the Proposed Revisions would require; and (3) it vastly underestimates the burden associated with compiling, analyzing and reporting the hours information that would be required, particularly as to exempt employees.

A. The EEOC's Estimate of the Burden Associated with Completing the Current EEO-1 Report is Inconsistent with its Prior Burden Analyses and Otherwise Lacks Factual Support

The EEOC makes the following claims and assumptions in connection with its estimate of the burden associated with generating the EEO-1 Report in its current form:

There are 67,146 employers who file EEO-1 Reports ("EEO-1 Filers"), based on 2014 filing figures. [4]
Each EEO-1 Filer spends just 3.4 hours generating its EEO-1 Reports each year, across all establishments, with 30 minutes dedicated to reading the instructions and 2.9 hours dedicated to generating the data required to be reported and populating the cells in the Form itself. Thus total hourly burden incurred by employers filing the EEO-1 Report is 228,296.4 hours.
All work done to complete the EEO-1 Report is performed by "Administrative Support" personnel who are paid, according to the Bureau of Labor Statistics ("BLS") publication "Employer Costs for Employee Compensation," on average, $24.23 per hour.

Based on these assumptions, the EEOC estimates that the total annual cost burden incurred by employers filing the current EEO-1 Report is just over $5.5 million. This figure represents the number of EEO-1 Filers (67,146) multiplied by 3.4, then multiplied by $24.23.

The EEOC's approach for assessing the actual current burden imposed on EEO-1 Filers is deficient for several reasons.

First, and perhaps most importantly, the "EEO-1 Filer"-based approach adopted by the EEOC in its Comment Request is at odds with the approach the EEOC has used to quantify the burden associated with EEO-1 filings since at least 2009. From at least 2009 through 2015, the EEOC, when estimating the burden associated with filing EEO-1 Reports, based its assessment on the number of responses filed by the employer community, not on the number of EEO-1 Filers. See Exhibits 4 -7 (EEOC OIRA filings). The EEOC's approach over at least the last six years appropriately accounted for the fact that many EEO-1 Filers are required to generate multiple "responses" or reports: one for each physical establishment with 50 or more employees and one consolidated report. As a result, the EEOC burden estimates for the years 2009 through 2015 are significantly higher than the burden estimate identified in the Comment Request. See Exhibits 4 - 7 (EEOC OIRA filings). For instance, in 2015, the EEOC estimated the annual employer burden associated with the EEO-1 filings as follows:

307,103 EEO-1 reports filed, based on actual 2013 filings.
4 hours per report, which yields 1,044,150 total burden hours.
Average cost per hour estimated to be $19.00 per hour, based on the hourly rate paid to "Human Resources Assistants" according to the BLS publication "Occupational Employment Statistics, Occupational Employment and Wages, May 2010" - $18.22 - "rounded to $19.00 to account for instances where higher paid staff perform this work."

Based on those assumptions, the EEOC estimated in 2015 - less than one year ago - that EEO-1 Filers would incur costs of $19.8 million annually to generate the EEO-1 filings. That cost figure is 350% higher than the EEOC's new estimate of what the current burden is without any of the proposed changes.[5]

Attempting to explain this $14.3 million reduction in the estimated burden, and perhaps anticipating criticism of this change, the EEOC includes only the following statement in its Proposed Revisions:

The reporting hour burden calculations in this notice reflect a departure from the manner in which EEOC traditionally estimated reporting burden. In the past, the EEOC estimated the reporting hour burden based on the number of total cells in the report(s) that a firm had to complete. This approach viewed each report filed by a firm as a separate reporting requirement, analogous to a paper report. In reporting year 2014, however, the number of paper reports declined to just three. In addition, employers now rely extensively on automated HRIS to generate the information they submit on the EEO-1 report. As a result, each additional report filed has just a marginal additional cost. To accurately reflect the manner in which employers now collect and submit the data for filing, the estimated reporting burden set forth in this notice is calculated per firm, rather than per report. This burden calculation is based on the time spent on the tasks involved in filing the survey, rather than on 'key strokes' or data entry."

81 Fed. Reg. 5120 (February 1, 2016) (emphasis added).

The EEOC's explanation rings hollow, as it is contradicted by its 2015 OIRA filing and is wholly inconsistent with the experience of EEO-1 Filers. The burden estimate reached by the EEOC in 2015 expressly noted that "98%" of the 70,070 respondents in 2013 "file [their EEO-1 Reports] on-line." Ex. 7 at Paragraph 3. Thus, the claim that the shift from a response-based approach that takes into account the number of establishments on which reports are filed to an EEO-1 Filer-based approach was prompted by a sudden shift from paper filings to online filings is not supported by the EEOC's own submissions. The PRA requires that the imposed burden be accurately computed. It does not countenance fictitious assertions of burden bereft of factual support or historical experience.

Second, the assumption that "each additional report" for those employers with multiple establishments "has just a marginal cost" is based on no analysis whatsoever. The EEOC seems to assume that the electronically fillable PDF format relieves employers of the necessity of manually entering data when filing their EEO-1 Reports. That assumption is contrary to fact.[6] Even when using the EEOC's on-line system, EEO-1 Filers must manually enter the data for each cell and for each establishment.[7] Notably, the EEOC's Proposed Revisions acknowledge that only 2% of EEO-1 Filers (1,449 of 67,146) submitted their data by uploading a data file in 2014 rather than manually completing the online submission. Thus, the process remains a manual one for most employers, even if the EEO-1 Report is submitted electronically, and the actual burden remains unchanged from recent years.[8]

Third, the EEOC nowhere explains its estimate that EEO-1 Filers spend a total of just 2.9 hours collecting, verifying, validating, and reporting" their current EEO-1 data. As part of its burden under the PRA, the EEOC should identify its current basis for this estimate. [9]

Fourth, the EEOC bases its cost estimate on an improper assumption that EEO-1 data and the ultimate reports are developed by "Administrative Support" personnel alone, at a cost of $24.23 per hour, without any internal or external assistance. The EEOC's assumption is flatly contradicted by employer experience. In fact, personnel ranging from HRIS and information technology personnel to senior human resources officials and legal professionals often assist in the development of the data, the compilation of the EEO-1 Report, review of the Report, and the submission of the Report. Because the EEO-1 Report requires certification by a company official, subject to penalties for false reporting, the EEOC's assumption that the entire EEO-1 filing process is left to clerical personnel is simply false. Because the EEOC has not included the wage rates of senior human resources, HRIS and information technology personnel, or legal professionals in the development, compilation, review, certification, and submission of the report, the assumption that the wage rate of an Administrative Support personnel alone, is flawed.

In short, the EEOC's estimate that EEO-1 Filers will incur costs of just $5.5 million in 2016 to complete the EEO-1 Report as currently configured - representing a 350% reduction in the EEOC's own estimated total costs of $19.8 million incurred in 2015 - is fatally flawed. Because that erroneous estimate impacts the Agency's estimate of the anticipated costs of completing the proposed revised EEO-1 Report in 2017 and 2018, the EEOC's entire burden estimate associated with the Proposed Revisions is completely inaccurate and misleading.

B. The EEOC's Analysis of the Burden Associated with Completing the Proposed Revised EEO-1 Report Lacks Factual Support

The EEOC's Comment Request includes the PRA-required estimate of both the one-time burden of complying with the proposed revisions and the annual burden of complying with those revisions. As for the "One-Time Implementation Burden," the EEOC's analysis consists of three sentences, one of which is contained in a footnote. The EEOC estimates a one-time burden of $23,000,295, which is based on the following assumptions:

Eight (8) hours of time "per filer" to "develop[ ] queries . . . in an existing human resources information system."
An hourly wage rate of $47.22, which is the average compensation for a Professional as identified in the BLS publication "Employer Costs for Employee Compensation," issued in 2013.

The EEOC's total estimate for the one-time burden is calculated by multiplying the number of EEO-1 Filers that would be required to submit W-2 and hours information (60,886) by eight hours, and multiplying that total by $47.22.

As for the recurring, annual burden employers will incur completing the proposed revised EEO-1 Report, the EEOC restates its current burden analysis for the estimated 6,260 EEO-1 Filers that have 50-99 employees. For that group of employers, the EEOC maintains its estimate that each filer will spend just 3.4 hours completing and submitting its EEO-1 Report, that those 3.4 hours will be spent by Administrative Support personnel at a cost of $24.23 per hour, and that the total cost burden to the 6,260 EEO-1 Filers who need not submit W-2 wage and hours data will be $515,711. That estimate is flawed, for the reasons identified above.

For employers with 100 or more employees, the EEOC estimates that each EEO-1 Filer will spend 30 additional minutes reviewing the instructions and just 2.7 additional hours per year "collecting, verifying, validating, and reporting" the W-2 wage data and the hours data for its employees. The EEOC continues to assume that all of these tasks will be performed by Administrative Support personnel, again at a cost of $24.23 per hour. In total, the EEOC's annual burden estimate for employers with 100 or more employees is as follows:

6 hours per EEO-1 Filer to collect, verify, validate, and report both the representational data currently reported and the W-2 wage data and hours data that would be required beginning in 2017.
An hourly rate of $24.23.

Based on these assumptions, the EEOC estimates that the total annual cost burden on EEO-1 Filers with 100 or more employees will be $9,736,767, which is the number of such filers (60,886) multiplied by 6.6, then multiplied by $24.23. When coupled with the estimated cost burden for EEO-1 Filers with 50-99 employees, the total annual estimated cost identified by the EEOC for all EEO-1 Filers in 2017 is $10,252,478.

The EEOC's estimates are wholly unrealistic. As detailed below, absent from the EEOC's analysis of the burden associated with the Proposed Revisions is any explanation of the basis for its estimate of the hours that would be incurred by EEO-1 Filers. [10] As an initial matter, while the EEOC claims it reviewed "the public comments relating to the burden calculation for OFCCP's proposal to collect pay data and consulted with OFCCP about burden estimates," the EEOC fails to state how it reconciled the information supplied to the OFCCP with the EEOC's current estimates.[11] More importantly, the EEOC failed to identify any other basis for its hours estimates, noting that its Pilot Study sought data from private employers "about the possible cost of collecting pay information but few employers responded, and the employers that did respond did not provide quantitative feedback." 81 Fed. Reg. 5120 (February 1, 2016). Without more, one can only conclude that the EEOC's estimate of the hours employers will spend "collecting, verifying, validating, and reporting" W-2 wage data was "pulled out of thin air."

To make matters worse, the EEOC provides no explanation for its estimate of the hours burden associated with supplying hours-worked data. The EEOC does not even address this issue in its Comment Request, stating simply that it is seeking input from employers as to "the anticipated estimated burden to also submit . . . hours-worked data." As detailed below, this omission should sound the death knell for the EEOC's Proposed Revisions, given that the collection of hours-worked data for exempt employees who do not currently capture their hours worked would be staggering.

1. The EEOC's Estimate of the One-Time Burden Bears No Relationship to Reality

The EEOC's estimate of the one-time burden associated with the Proposed Revisions cannot withstand scrutiny. First, its assumption that employers will be able to generate W-2 and hours data after an HRIS professional spends just eight hours "developing queries . . . in an existing human resources information system" underscores how distant the EEOC's experience with employer HRIS systems is from reality. The EEOC's underlying assumption - that a single HRIS houses all the data necessary to generate the W-2 and hours data - is plainly false. Most employers maintain gender, race, and ethnicity data in an HRIS that is different from the system that houses payroll information, including W-2 wage information. Hours data for non-exempt employees are likewise captured outside of the HRIS and hours data for exempt employees simply do not exist for most employers. Determining how to marry gender, race, and ethnicity data housed in an HRIS with W-2 wage data housed in a payroll system, often by a third party, in and of itself, will take more than eight hours.

Second, the EEOC's estimate is off-base because it ignores the fact that the W-2 data to be submitted is different from the W-2 data generated annually by employers for income tax reporting purposes. Because the 12-month reporting period proposed by the EEOC crosses tax years, employers cannot simply rely on the data they generate for the IRS each year. Instead, employers will be required to generate data spanning two different tax reporting years and will be required to make one-time changes to their systems in 2016 to permit such reporting.

Third, the EEOC's estimate of the one-time burden is inaccurate because the hourly rate on which it is based - $47.22 for a "Professional" - fails to account for the fact that senior information technology personnel, legal personnel, and others will be involved in any effort to develop the processes necessary to generate the required data. The identified hourly rate is not commensurate with the rates of pay provided to such individuals. [12]

Fourth, as detailed in Section II.B.3, below, the EEOC estimate fails to account for the one-time burden associated with requiring employers to develop, for the first time, a system that captures the hours worked by exempt employees and to train all exempt employees on any new system.[13] Those costs would be massive, and should be quantified by the EEOC before the Proposed Revisions are considered.

The Chamber is in the process of surveying its members to obtain information regarding the one-time hour and cost burden associated with the Proposed Revisions. The Chamber's written comments to the Comment Request will include information from that survey.

2. The EEOC's Estimate of the Annual Burden of Compliance with the Proposed Revisions is Likewise Unrealistic

The Agency's prediction that the Proposed Revisions will require only 3.2 hours of additional work per EEO-1 Filer - 30 additional minutes to read the instructions and just 2.7 hours to "collect, verify, validate, and report" all of the required W-2 and hours data for every establishment - is ludicrous.

First, the EEOC dramatically underestimates the time it will take each EEO-1 Filer to collect, verify, validate and report on data that must be pulled from various systems and sources. As noted above, often the required data is not housed in a single HRIS and cannot be generated with the push of a button. Once generated, a combination of HRIS professionals and HR professionals will have to expend time verifying and validating the data. Legal professionals will likewise be involved in the verification process, given that the EEOC's stated purpose for the collection is to target its, and the OFCCP's, enforcement efforts, and given the requirement that a company official certify the filing, subject to penalties.

Second, the EEOC bases its burden estimate for the generation and reporting of W-2 data and hours data on the wage rate of $24.23, which is - again - the BLS wage for Administrative Support personnel. As detailed above, employees other than Administrative Support personnel would be engaged in the effort to collect, verify, validate, and report the W-2 and hours data - legal professionals and others are, and will continue to be, involved in the EEO-1 filing process.[14]

Third, the EEOC's annual burden estimate for filing the Revised Reports is improperly based on the number of EEO-1 Filers, rather than the number of EEO-1 Reports filed. The "per EEO-1 Filer" approach is inappropriate, for the reasons identified above, and the approach the EEOC used between 2009 and 2015 - in which the burden estimate takes into account the fact that many EEO-1 Filers must file reports on all of their establishments individually - remains the appropriate approach. If the per response approach were applied here, the EEOC's total cost estimate would jump from $10,252,478 to $49,111,297, and that calculation assumes that the EEOC's hours estimates of 6.6 hours and cost estimate of $24.23 per hour are correct, which they clearly are not.

Fourth, the EEOC's assumption that filing on-line, through fillable PDFs, alleviates the burden of manual data entry is a false assumption. As detailed above, fillable PDFs still require the employer to manually enter the relevant data. With the addition of W-2 data and hours data, reported in twelve different pay bands within each EEO-1 category, each EEO-1 Filer will now be required to populate as many as 3,360 separate cells of data. The EEOC's burden estimate fails to account for this reality in any way.

Fifth, the EEOC does not account, in any way, for the costs employers will incur responding to investigations and enforcement actions that are prompted by "false positives" that flow from comparing employees within broad EEO-1 categories.[15] Because the EEOC and OFCCP intend to use the results of analyses of W-2 wages by EEO-1 categories to target their enforcement efforts and because those analyses will be fundamentally flawed, the substantial cost to employers who must respond to such investigations is not theoretical. Employers will be forced to expend resources producing additional data to the EEOC and/or OFCCP, retaining labor economists to run their own analyses of pay, and engaging legal counsel to correct the Agencys' presumption of discrimination.

3. The EEOC Ignores the Substantial Burden Associated with Collecting Hours Data for Exempt Employees

The EEOC proposes to collect "the total number of hours worked by employees" included in each of 12 pay bands within each EEO-1 category, to "allow analysis of pay differences while considering aggregate variations in hours." With respect to exempt personnel, the Commission "seeks employer input with respect to how to report hours worked for salaried employees" and states that "[o]ne approach would be for employers to use an estimate of 40 hours per week for full-time salaried workers." The Commission further states that it is "not proposing to require an employer to begin collecting additional data on actual hours worked for salaried workers," and the EEOC's "initial conclusion is that requiring employers to provide the total number of hours worked would impose a minimal burden."

As detailed in Section III.C.2, below, assuming that all full-time exempt employees work 40 hours per week is neither an accurate nor a fair assumption. Thus, to the extent the EEOC intends to consider "aggregate variations in hours," the only way for the EEOC to accurately capture hours for exempt employees would be to require employers to track hours of exempt employees. That option would be incredibly burdensome.

First, employers would be required to implement a time system to track all hours worked by exempt employees. Because there is no continuous workflow for exempt employees, and because they often work outside of normal business hours, any such system would be much more complex than the standard time-clock, punch-clock, or timesheet system currently used by employers for their non-exempt employees. The cost of developing such a system would be immense.

Second, the hours burden associated with exempt employees recording time would be substantial. Unlike non-exempt employees who may use a punch clock system or work regular hours, each exempt employee would be required to capture the "starts" and "stops" of his or her work. Even if that effort required just 15 minutes of time each day for each exempt employee, the result would be significant new burden for employers. The Chamber is seeking additional data from its members to quantify this potential burden.

Finally, the EEOC has failed to abide by its obligation under the PRA, which is to propose a methodology for, and also assess the burden of, collecting hours for exempt employees. Because the EEOC has failed to do so, commenters - including the Chamber - are unable to examine the utility and burden of producing the hours report for exempt employees.

III. The Proposal will Not provide a public benefit or maximize the utility of the requested data as Required by The Paperwork Reduction Act.

Despite the significant burden that employers with more than 100 employees will face as a result of the requirements in the Proposed Revisions, the EEOC has failed to articulate any rationale or evidence whatsoever to justify the need for the proposed EEO-1 Revisions. Nor has the agency identified the specific benefit that the collection of aggregated wage and hours data would provide to the agency. The EEOC's wholesale failure to articulate the proposed benefit is significant given that one of the purposes of the PRA is to "ensure the greatest possible public benefit from and maximize the utility of information created, collected, maintained, used, shared and disseminated by or for the federal Government."

The Sage Report, which the EEOC used to "formulat[e] the proposal and guide the development of analytical techniques to make full use of the data to be collected," recognized the inherent limited use of the requested data. Specifically, the Sage Report recognized that "[s]ummary data at the organization level will likely be of very limited use in EEOC practice."[16] The reasons for this observation are numerous, as articulated below.

To evaluate the utility of the proposed EEO-1 Revision, one must consider the legal framework that governs compensation discrimination. Since 1963, the EPA has required employers to pay men and women at the same establishment equal pay for equal work. In addition to the protections against wage discrimination based on sex afforded by the EPA, Title VII prohibits pay discrimination on the basis of race, color, national origin as well as other protected factors.[17] Employees who work for Federal contractors and subcontractors share similar protections under Executive Order 11246. Since 1978, the EEOC is the administrative agency with enforcement authority of the EPA and Title VII, while the OFCCP enforces EO 11246.

Notwithstanding these protections, both the EPA and Title VII recognize that there are legitimate reasons for differences in compensation. In this regard, the EEOC has recognized that differences in education, experience, training, shift differentials, job classification systems, temporary assignments, "red circling," revenue production, and market factors, to name a few, can legitimately explain compensation differences.[18] Thus, mere differences in pay even as between comparable employees are insufficient to infer unlawful discrimination.

A. The EEOC's Proposed Tool Will Not Advance Investigations Under the EPA

1. The EEOC's Proposed Broad Comparison Groups is Inconsistent with Equal Pay For Equal or "Substantially Similar Work"

As noted above, the EPA prohibits sex-based compensation discrimination for employees working at the same establishment if they perform "equal work on jobs the performance of which requires equal skill, effort, and responsibility, and which are performed under similar working conditions."[19] In other words, the relevant comparators are only those employees who perform "equal work" in the "same establishment." The EEOC itself describes the prima facie elements of an EPA claim as follows:

Prima Facie Case: (1) the complainant receives a lower wage than paid to an employee of the opposite sex in the same establishment; and (2) the employees perform substantially equal work (in terms of skill, effort, and responsibilities) under similar working conditions.[20]

Despite the requirement that pay comparisons can only be made between employees who perform "equal work," the proposed EEO-1 Revision would require that employers provide W-2 wage data by EEO-1 job category. Because there are only ten EEO-1 job categories or groupings of jobs,[21] employers would be forced to categorize employees who perform wildly different work into these groupings. The job groupings are exceedingly broad, and will necessarily capture a wide range of positions that are not capable of meaningful compensation comparisons.

For instance, as Economist Dr. J. Michael DuMond observed, one of the most serious deficiencies with the proposed EEO-1 Revisions are the "very broad occupational group[ings]" that would "result in comparisons of employees who work in very different jobs and who may perform different work." Dr. DuMond uses the following example:

[O]ne of the 10 EEO-1 job categories is "Professionals", which encompasses a wide range of occupations such as lawyers and registered nurses. Hospitals that employ both nurses and lawyers would nevertheless be required to include both of these occupations together in the proposed EEO-1 survey. This grouping of nurses with lawyers ignores the fact that a nursing degree does not require a post-graduate college degree whereas a lawyer will almost surely have post-graduate education. Moreover, the knowledge, skills and abilities required of a nurse differ greatly from those factors that required of a lawyer. In fact, there is actually very little in common between nurses and lawyers beyond the sharing of a common EEO-1 category. While it is undeniable that nurses and lawyers are tied to very different labor markets, the EEOC's proposal ignores this reality and assumes pay should be similar for these types of occupations simply because they are both "Professionals."[22]

The EEOC's instruction booklet outlines other "Professional" positions that similarly cannot form a proper comparison for compensation purposes under the EPA. For instance, using the Professionals job category for positions typically found within a hospital, we would also find accountants, computer programmers, dieticians, physicians and surgeons.[23] But these jobs are simply not comparable under the EPA.

The EEOC itself must recognize the inappropriateness of such groupings under the EPA. Indeed, in investigating and analyzing the "equal work" requirement, the EEOC advises as follows:

[A]n inquiry should first be made as to whether the jobs have the same common core of tasks, i.e., whether a significant portion of the tasks performed is the same. If the common core of tasks is not substantially the same, no further examination is needed and no cause can be found on the EPA violation.[24]

Common sense tells us that accountants, computer programmers, dieticians, registered nurses, lawyers, physicians and surgeons do not share the same common core of tasks. But to be more specific, the Department of Labor, Bureau of Labor Statistics makes those distinctions clear in the descriptions it assigns to these workers: [25]

Accountants & Auditors: Examine, analyze, and interpret accounting records to prepare financial statements, give advice, or audit and evaluate statements prepared by others. Install or advise on systems of recording costs or other financial and budgetary data.

Computer Programmers: Create, modify, and test the code, forms, and script that allow computer applications to run. Work from specifications drawn up by software developers or other individuals. May assist software developers by analyzing user needs and designing software solutions. May develop and write computer programs to store, locate, and retrieve specific documents, data, and information.

Dieticians: Plan and conduct food service or nutritional programs to assist in the promotion of health and control of disease. May supervise activities of a department providing quantity food services, counsel individuals, or conduct nutritional research.

Registered Nurses: Assess patient health problems and needs, develop and implement nursing care plans, and maintain medical records. Administer nursing care to ill, injured, convalescent, or disabled patients. May advise patients on health maintenance and disease prevention or provide case management. Licensing or registration required. Includes Clinical Nurse Specialists. Excludes "Nurse Anesthetists" (29-1151), "Nurse Midwives" (29-1161), and "Nurse Practitioners" (29-1171).

Lawyers: Represent clients in criminal and civil litigation and other legal proceedings, draw up legal documents, or manage or advise clients on legal transactions. May specialize in a single area or may practice broadly in many areas of law.

Physicians:[26] Physicians who diagnose and provide non-surgical treatment of diseases and injuries of internal organ systems. Provide care mainly for adults who have a wide range of problems associated with the internal organs. Subspecialists, such as cardiologists and gastroenterologists, are included in "Physicians and Surgeons, All Other" (29-1069).

Surgeons: Physicians who treat diseases, injuries, and deformities by invasive, minimally-invasive, or non-invasive surgical methods, such as using instruments, appliances, or by manual manipulation. Excludes "Oral and Maxillofacial Surgeons" (29-1022).

And of course, the BLS job descriptions do not take into account the many specific job descriptions and duties employers are likely to have in their own workforces.

Aside from the "same common core of tasks" analysis, assessing whether required "skills" of comparator jobs are substantially equal further demonstrates that comparisons of pay by EEO-1 job groupings can serve little utility. As the EEOC's Compliance Manual provides:

Two jobs require equal skill for purposes of the EPA if the experience, ability, education and training required are substantially the same for each job.

Again, one need only look at the instructions the EEOC has provided to employers to see that the EEO-1 job groupings take into account a wide range of experience and educational requirements within the same job group. For instance, employers are advised that most jobs in the "Professionals" job group "require bachelor and graduate degrees, and/or professional certification." The EEOC further provides that "[i]n some instances, comparable experience may establish a person's qualifications." Thus, the explicit directions make clear that jobs that do not require the same education, experience and training will be grouped together. On its face, the EEO-1 job groupings are not capable of any meaningful analysis under the EPA.

B. The EEOC's Proposed Tool Will Not Advance Investigations Under Title VII

The proposed EEO-1 Revisions are also inconsistent with Title VII and, therefore, cannot support any meaningful analysis probative of further investigation. Specifically, providing W-2 data by EEO-1 job categories is not consistent with the requirement that compensation differences only matter if the comparators are "similarly situated." As articulated by the EEOC in its compliance manual, "similarly situated employees are those who would be expected to receive the same compensation because of the similarity of their jobs and other objective factors."[27] And when reviewing similarity of jobs, EEOC investigators are instructed that the "actual content of the jobs must be similar enough that one would expect those who hold the jobs to be paid at the same rate or level."[28]

For the reasons set forth in Section III.A.1 above, the EEO-1 job groups are exceedingly broad and will necessarily include a wide range of positions that are not capable of meaningful compensation comparisons under Title VII. Using the same example described above, it is simply implausible to expect that accountants, computer programmers, dieticians, registered nurses, lawyers, physicians and surgeons are "paid at the same level or rate."

Moreover, courts across this country have held that although similarly situated employees need not be "identical", they must be "directly comparable to the plaintiff in all material respects...."[29] Under these legal standards, any compensation analysis based upon EEO-1 job groupings would be meaningless.

We see a similar result in the EEO-1 "Sales Workers" job grouping. In that group we find advertising sales agents, insurance sales agents, real estate brokers and sales agents, securities, commodities, financial services sales agents, and telemarketers. Many of these positions could reasonably be found within a financial services organization. Yet one would not expect that these jobs would be similar enough to suggest that those who hold the positions would be paid at the same rate or level. Indeed, sales workers are often compensated based on varying levels of base salary and commission plans. Thus, unless they were truly in similar jobs, it would be impossible to conduct pay comparisons based on an overall broad-based job grouping that would include, for example, telemarketers and securities and commodities brokers.

And aside from the overly broad EEO-1 job groupings themselves, the EEOC has acknowledged that "differences in job titles, departments, or other organizational units may reflect meaningful differences in job content or other factors that preclude direct pay comparisons between employees."[30] But neither these factors nor the legitimate factors that explain compensation discrimination as described in Section III.C.1 below, are captured under the proposed EEO-1 Revisions.

C. There are Fatal Limitations with the Information Sought and the Statistical Tool Proposed to be Utilized by the EEOC to Analyze this Information

For reasons addressed in more detail below, there is simply no circumstance under which broad-brush, aggregate compensation and hours data can be used effectively on a grand scale to target employers for review.[31] Every employer's compensation system is unique and numerous factors impact compensation decisions and results. The authors of the Sage Report, commissioned by the EEOC, detailed a theoretical framework for identifying pay disparities that is consistent with standard labor economic theory.[32] More specifically, the authors demonstrated that an employee's pay is estimated as a function that includes "control variables that can have a justifiable impact on difference in pay (such as education, certification, and work experience)." The inclusion of the control variables is critical in explaining why some employees are paid more than others. As the authors correctly note: "market forces determine one's level of pay, and variation in pay is both necessary and inevitable."[33]

Nonetheless, the two tests that the EEOC specifically references in the Proposed Revisions are simply distributional tests that do not control for any legitimate factors that explain pay such as those identified in the pilot study. Besides pay band, the only other factor that a covered employer will be providing to the EEOC is the aggregate number of work hours associated with employees in each pay band and demographic group. However, the pilot study's authors noted that simply adjusting for the number of work hours, as recommended by the EEOC, could generate misleading results:

[W]e have to recognize the varying patterns of compensation for employees who work different hours. As a result, pooling together into a single cell the employees who may have received the same compensation from working different hours and analyzing them with a single offset as the format of the proposed EEOC form suggests, may lead to biases that are difficult to quantify…"[34]

As the EEOC has found with similar efforts to collect compensation data on a broad scale, compensation cannot be subjected to a normalized, one-size-fits-all method of interpretation. As a result, the proposed data collection will have no "meaningful value" and certainly no benefit that rises to the level required by the PRA. Simply put, the EEOC cannot achieve the objectives identified in the proposal through any mechanism, and certainly not through the mechanism it has identified.

1. The Pay Proposal Ignores Legitimate Reasons for Differences in Compensation

Employers have an inherent right to value jobs differently based on non-gender or race/ethnicity based standards.[35] The EEOC's proposal does not account for the explicitly-permitted differentiation in compensation based on factors like experience, performance, work productivity, skills, scope of responsibility, market, and education - to name just a few. This is particularly troublesome because differences in sex and race/ethnicity correlate with some of these factors. For example, the Bureau of Labor Statistics indicates that the average amounts of company-specific tenure differ between men and women and among different race/ethnic groups.[36] Notwithstanding, the statistical tests proposed by the EEOC do not account for these permitted differences.

Moreover, the EEOC's proposal ignores the role of working conditions and how that affects an employee's compensation. For example, employees who work night shifts, swing shifts and/or weekends are often paid a differential to account for less desirable work schedules.[37] For the same reason, jobs that require employees to work outside or exert atypical physical effort may also command a wage premium.[38] Nevertheless, the statistical tests proposed by the EEOC using the collected data will not and cannot account for differences in working conditions.

The concerns with the EEOC's failure to account for differences in working conditions is compounded because the Commission is proposing that employers use W-2 data, which includes not only base salary but also commissions, tips, overtime pay, shift differentials and bonus payments in the aggregate. In doing so, the EEOC is ignoring that some types of pay, specifically commissions and tips, are determined more by an employee's skill and effort than an employer's pay policies.[39] For example, Economist Dr. J. Michael DuMond notes that differences in W-2 earnings among servers at a restaurant will be based on their ability to provide quality service and earning the associated tips/gratuities, and is largely undetermined by the employer.[40]

2. The EEOC's Approach to Reporting of Hours for Salaried, Part-Time and Terminated Employees Further Denigrates the Efficacy of the Data

The proposal would require employers to report hours worked as well as data about W-2 earnings. The Commission stated that this will allow the EEOC and OFCCP to "meaningfully analyze"[41] pay differences because "collection of hours-worked data will account for the fact that some individuals are employed for less than the entire reporting year, and therefore, may work fewer hours."[42] The proposed approach to reporting data for employees exempt from the overtime provisions of the FLSA, part-time employees, and employees who are hired or terminated mid-year, however, is fundamentally flawed.

First, as the EEOC admits in the proposal, it is unsure how to count hours worked for full-time, salaried exempt employees, noting that the Agency "seeks employer input" on how to report hours worked for these exempt employees.[43] The EEOC indicated that it is "not proposing" that employers begin collecting additional data on actual hours worked for salaried workers "to the extent that the employer does not currently maintain such data," but rather is considering a standard such as estimating 40 hours per week for all full-time salaried workers in all industries. In our experience and in the experience of Dr. DuMond, very few employers track the number of actual hours worked for their salaried workers and even HRIS systems that maintain a standardized or default value for "work hours" for salaried exempt employees (such as 40 hours per week) does not reflect an employee's actual hours worked.[44] As such, the Agency will be forced to apply an across-the-board rule that does further harm to the efficacy of the data and that would be uninformative and unreliable for purposes of identifying pay disparities.[45] The impact of this data limitation is serious: according to data from the Bureau of Labor Statistics, 59% of the US workforce is paid by the hour, meaning that 41% of the US workforce is paid on a basis for which no accurate work hours may be available and for which the EEOC does not have a recommended method for measuring.[46]

Further, combining employees who work part-time or partial year with full-time and full-year employees will result in very misleading results. The variation is explained by Dr. DuMond in his example of an employee who has an annual salary of $120,000 per year but was recently hired and only worked 1 of 12 months, he or she will be placed in the lowest pay band since he/she only earned $10,000. Under the EEOC's proposal, this employee could be grouped with lower wage workers at the same company, even though the hourly rate for this employee would be much higher than the other employees in the same pay band. [47] Similarly, an employee who works a part-time schedule of 20 hours per week will be grouped and counted with FT employees who earn half of that hourly rate but work 40 hours per week.[48]

Simply collecting the aggregate number of hours worked does not fix the bigger underlying problem with the EEOC's proposal: employees are being counted in the "wrong" pay band. However, the EEOC might again infer discrimination from differences in the imputed hourly rates for employees in the same pay band, not recognizing that those differences can just as easily arise from part time and partial year employees, and the agency will not be able to make that distinction with the aggregated data. Additionally, the proposal does not suggest a methodology by which the EEOC will distinguish between intermixed part-time, partial year or full-time employees in the aggregate W-2 data. Moreover, since the total number of hours worked by employees in a pay band does not incorporate any of the factors known to affect pay, the data collected under this proposal will not provide any useful insight into the actual nature of employers' pay practices.

3. The EEOC's Proposed Statistical Approach Will Not be Useful in Identifying Pay Differences

The EEOC Proposes using two statistical tests to analyze the collected data: the Mann-Whitney and the Kruskal-Wallis tests.[49] The EEOC indicates that it can "compute the statistical tests within job categories and then proceeding [sic?] to more closely investigate companies and establishments with low p-values" and may also "compare a particular firm's regression coefficients for the hours worked, race and gender variables to those derived from an analysis of the relevant labor market as a whole."[50] The use of these measures could easily lead to both "false positives" (i.e., flagging a company for further review even when all employees working in the same job are paid exactly the same) and also "false negatives" (i.e., concluding that pay disparities do not exist at a company even in the presence of unambiguous pay discrimination).

The reason that the two statistical tests proposed by the EEOC are likely to lead to both false positives and false negatives is that the aggregated data collected by the EEOC will not include one of the most critical factors that determine pay, specifically the job level/job grade of an employee.[51] According to Dr. DuMond, it is commonly understood that all employees, regardless of their race, ethnicity or gender, will experience increases in pay as they move "up" or "higher" in an organization.[52] He provides the example that Pharmaceutical Sales Representatives are paid less than their District Sales Managers who are paid less than their Regional Sales Directors. [53] While it would not be reasonable to assume that an inexperienced Pharmaceutical Sales Representative would be paid similarly to an experienced Regional Sales Director simply because they are in the same EEO-1 job category (Sales), the statistical measure the EEOC proposes utilizing could lead to this conclusion because the EEOC's proposal will not include any information or breakdown relating to an employee's job level/job grade. Despite this lack of consideration for an employee's job level/grade, the two statistical tests will nevertheless ascertain whether gender and racial groups are equally distributed across all the pay levels of a company without any consideration of the employee's job level/grade.

The Pilot Study commissioned by the EEOC is very clear as to what these two statistical tests assume: "Although the (Mann-Whitney) test is sensitive to pure shifts, it has the more general interpretation of a test of the differences between two or more distributions."[54] Thus, because the Mann-Whitney and Kruskal-Wallis tests are determining whether men and women, for example, are distributed similarly across all pay bands (and hence all levels of an organization), these tests do not determine whether there are improper disparities within a pay band.[55] Instead, it is a test of whether racial and gender groups of employees are similarly distributed across all pay bands, even when the employees in these groups may have very different levels of experience, may hold very different jobs (e.g., nurses and lawyers) and may be at very differently job and pay levels within an organization.[56] If the EEOC proceeds with its stated proposal to use these tests on the collected data to "detect discrimination," then the likelihood that they would lead to false positives and false negatives is greatly increased.

Economist J. Michael DuMond conducted the Mann-Whitney test on two sets of simulated employment data.

In the first simulation, an employee's pay is completely determined by two factors: his or her job grade/level and the number of years he or she has worked at that level.[57] For example, an employee entering the first job grade receives a starting annual salary of $30,000 and is awarded a raise of $1,000 each year he or she remains in that position. Similarly, an employee hired into the next grade level receives a starting salary of $40,000 and receives a raise of $2,000 each year he/she or she remains in their job. A similar method of pay determination exists for the next three job grades at this fictitious company, though the starting pay and annual pay adjustments are naturally greater in the higher-level grades. In this simulation, it is clearly impossible for a pay disparity to exist between women and men in the same job grade, after taking into account any differences in job tenure. A standard multiple regression analysis of this simulated data confirmed this obvious conclusion. Using the Mann-Whitney test on these simulated data, however, yielded a statistically significant result, which the EEOC may incorrectly interpret as an indication of pay disparities that are adverse to women.[58] That is, even in a simulation in which an employee's pay is 100% determined by a formula that allows no room for discrimination, the test that the EEOC is proposing to use still yielded a statistically significant difference. Put simply, the EEOC's proposal won't be able to test whether or not similarly situated men and women are paid equally.

This point is further illustrated in the second simulation. At this second fictitious company, there is an explicit, racially discriminatory policy that all African-American employees in grade level 1 receive a starting salary of $30,000 and all white employees in grade level 1 receive a higher starting salary of $30,500. Both employees receive a pay adjustment of $1,000 for each year worked. This blatantly discriminatory policy is found at all other job grades as well: white employees in grade level 2 receive a starting pay of $40,500 compared to $40,000 for African-Americans. At the highest job grade in this hypothetical company, the starting pay gap is even more pronounced: African-Americans have a starting salary of $100,000 compared to $105,000 for white employees.[59] As would be expected, the existence of such an overt policy of pay discrimination was readily detectible through a standard multivariate regression analysis, which indicated that African-American employees were paid statistically significantly less than their white counterparts. On the other hand, applying the EEOC's proposed Mann-Whitney test to these same data did not reveal any statistically significant differences between white and African-American employees.

These two simulations call into question the usefulness of the data that would be collected under the EEOC's proposal. These examples illustrate that the proposed statistical tests may not identify actual pay differences that are consistent with discrimination when it truly exists, and may incorrectly conclude that there is evidence consistent with discrimination when employees are actually paid equivalently. But perhaps more importantly, the first simulation shows that a company with a disproportionately greater number of female or minorities at the lower grade levels is likely to "fail" the Mann-Whitney test. In fact, failing the Mann-Whitney test could occur while companies are making good-faith efforts to increase their minority and female representation. That is, many firms have worked diligently to increase the participation of workers from historically disadvantaged groups by hiring these workers as they became available. The result of this effort is that higher percentages of female, African American, and Hispanic workers are found in some of the lower job levels as they gain additional experience to be eligible for promotion.

Accordingly, the EEOC's proposed statistical methodology provides no benefit and does not meet the standards of the PRA.

D. The EEOC's Proposal to Compare the Pay of a Firm or Establishment to Industry or Metropolitan-Area Data Serves No Purpose Under Title VII or the EPA.

The EEO-1 Revisions state that it will develop a software tool that will allow its investigators to analyze "W-2 pay distribution within a single firm or establishment, and by comparing the firm's or establishment's data to aggregate industry or metropolitan-area data" which would "highlight statistics of interest."

However, neither Title VII, the EPA nor EO 11246 requires employers to pay its employees according to industry or geographical standards. Certainly, employers cannot use as a defense to claims of inequitable pay practices that others in its industry are also paying women or minorities in a particular EEO-1 category less than white men. Conversely, it is not discriminatory for an employer to decide to pay lower wages for certain positions than its competitors may choose to pay for the same positions. As such, the mere fact that a particular employer's aggregate compensation data is below the pay of the industry or metropolitan area is irrelevant to an investigation of whether an employer's pay practices are discriminatory.

Employers already have many different, much more finely-tuned, methods by which they can benchmark their compensation against others for particular jobs. Many companies, large and small alike, look to market data that is specific to the jobs in their workforce to assess their position vis-a-vis their competitors. The data that could be assembled through the proposed EEO-1 revisions will lack the reliability of such targeted market data - data that employers can and do already access.

Moreover, the EEO-1 Revisions suggestion that reporting pay data may "encourage self-monitoring" and "voluntary" compliance by employees if they uncover pay disparities misses the mark. As an initial matter, federal contractors and subcontractors are already required to monitor their pay practices as part of their compliance obligations.[60] And any contractor subject to an OFCCP compliance evaluation is required to submit employee-level detailed compensation data for the establishment under review. Thus, there would be no added utility to federal contractors and subcontractors who already self-monitor pay within their organizations.

And, generally speaking, the EEO-1 Revisions would not otherwise encourage self-monitoring efforts because, for the reasons explained above: (1) EEOC cannot effectively target its enforcement efforts based on the data; and (2) the data is too broad and generalized to be of any practical use to those employers who endeavor to establish equitable pay practices.

IV. Confidentiality & Privacy

A. EEOC Should Take All Possible Actions to Protect the Confidentiality and Security of the Highly Confidential, Proprietary Data Being Requested

EEOC gives rather short shrift to the privacy and confidentiality issues raised by the Proposed Revisions. The notice published in the Federal Register merely reiterated that Section 709(e) of Title VII, 42 USC § 2000e-8(e) prohibits disclosure of any information contained in the EEO-1 report "prior to the institution of any [Title VII] proceeding." The proposal notes that while the OFCCP is not subject to the restrictions of section 709(e), it nevertheless will hold the information contained in the EEO-1 confidential "to the extent permitted by law, in accordance with Exemption 4 of the Freedom of Information Act and the Trade Secrets Act." While these assertions state the law as codified and practiced, they understate the significant privacy and confidentiality concerns raised by the proposal.

We only have to refer to the draft report of the EEOC Survey System Modernization Work Group (Working Group) Meeting conducted on March 8-9, 2012, which was published with the Proposed Revisions to confirm these concerns. While this Working Group performed its work nearly four years prior to the publication of the Proposed Revisions and while the draft report was deficient in many respects, the sparse published report highlighted issues raised which were not addressed in the Proposed Revisions. The Working Group noted that "the Privacy Act is a concern."[61] The report stated that: "Statistical Controls would be required to ensure the confidentiality of data to companies with smaller job categories."[62] The proposal is devoid of any discussion of data confidentiality. The draft report notes as well that if the "cell" size was lower than 5, we do not publish data for that cell."[63] There is no discussion of the implication of small cells, particularly in the upper pay bands.

In fact, there is no discussion at all of confidentiality issues in the proposal. In addition, the general reference to the Exemption 4 exclusion from the OFCCP releasing this information does not reference a crucial point. Unlike the section 709(e) restriction on EEOC disclosure, the OFCCP procedure requires contractors to follow the regulatory constraints found in the OFCCP and Department of Labor FOIA regulations. The Department of Labor FOIA Regulations, 29 CFR Part 70, create a complex process for submitters to request that business confidential information not be disclosed. Rather than a simple prohibition of disclosure, the regulations state:

(b) Designation of business information. A submitter of business information will use good-faith efforts to designate, by appropriate markings, either at the time of submission or at a reasonable time thereafter, any portions of its submission that it considers to be protected from disclosure under Exemption 4. These designations will expire ten years after the date of the submission unless the submitter requests, and provides justification for, a longer designation period.

(c) Notice to submitters. A component will provide a submitter with prompt written notice of a FOIA request that seeks its business information whenever required under paragraph (d) of this section, except as provided in paragraph (g) of this section, in order to give the submitter an opportunity to object in writing to disclosure of any specified portion of that information under paragraph (e) of this section. The notice will either describe the business information requested or include copies of the requested records or record portions containing the information. When notification to a voluminous number of submitters is required, notification may be made by posting or publishing notice reasonably likely to accomplish such notification.

(e) Opportunity to object to disclosure. A component will allow a submitter a reasonable time to respond to the notice described in paragraph (c) of this section. If a submitter has any objection to disclosure, it is required to submit a detailed written statement. The statement must show why the information is a trade secret or commercial or financial information that is privileged or confidential. In the event that a submitter fails to respond to the notice within the time specified, the submitter will be considered to have no objection to disclosure of the information. Information provided by a submitter under this paragraph may itself be subject to disclosure under the FOIA.

Thus, in every instance where the Department of Labor receives a FOIA request for the extended EEO-1 information, it must timely notify the employer submitter which in turn must submit a detailed written statement explaining why some or all of the EEO-1 information is a trade secret or commercial or financial information. The regulations then provide a process whereby the requester or the submitter have to undertake a lengthy appeal process where each presents its arguments. Final determinations are subject to judicial review. Further, even if the Department of Labor applies Exemption 4, that exemption from disclosure expires 10 years after it is granted. As noted in the draft Working Group report, information contained in small cells or smaller job categories may well disclose individual pay or other personal information. Thus, the submission to the OFCCP is not resolved by the glib assertion in the proposal that the OFCCP holds this information confidential "to the extent permitted by law." In fact, there is a significant burden placed on contractors to protect the confidentiality of the new EEO-1 information.

To the extent that the OFCCP shares the EEO-1 information, disclosure concerns would be alleviated to some extent if the EEOC requests that the Director of OIRA treats all information submitted to the EEOC as covered by Section 3510 of the Paperwork Reduction Act[64] and directs the EEOC that the statutory restrictions on disclosure of EEO-1 data will apply to the OFCCP.[65] In this regard, the direct prohibition on disclosure will not be subject to regulatory considerations and determination of trade secrets or commercial or financial information by OFCCP personnel not trained in these issues or a laborious and expensive regulatory and judicial review process.

B. Potential Data Breaches

As noted above, the proposal to collect employee pay data through the EEO-1 reports presents significant confidentiality issues related to the potential disclosure of this data. The EEOC publishes aggregate data collected from the EEO-1 reports and also shares original EEO-1 data with other federal and state agencies and individual researchers. With regard to the aggregate data, there are concerns that those who receive it will be able to reverse-engineer the aggregate data. With regard to original data, there are concerns that those who receive it might not take appropriate steps or have appropriate procedures in place to maintain its confidentiality. Furthermore, the EEOC must be transparent about the steps that it will take as an agency to insure that its information security program is now, and will be in the future, able to protect this sensitive data from access or acquisition by unauthorized individuals.

Prior to issuing the proposed rule, the EEOC engaged the National Academy of Sciences (NAS) to conduct a study, which, inter alia, looked at confidentiality concerns raised by the EEOC's collection of employee pay data in EEO-1 reports and its subsequent disclosure of this data in aggregate and original form. As a threshold matter, the report issued by the NAS (NAS Report) recognized -- and we wish to underscore -- that "[e]mployee compensation data are generally considered to be highly sensitive; they are even considered proprietary information by many private-sector employees."[66]

The NAS Report underscored that "there will be a great demand on the part of other federal agencies, researchers, analysts, compensation-setting bodies and others for access to these powerful new data . . . and the EEOC would be well advised to start taking steps now to develop policies to provide access in a protected environment."[67] The NAS Report went on to note that, despite the sensitive nature of employee pay data, the "EEOC provides [this] data to agencies that do not have the same level of confidentiality protections."[68] With regard to both its routine disclosure of data from the EEO-1 reports to other federal and state agencies, as well as researchers, the NAS Report recommended that the EEOC:

(1) consider implementing appropriate data protection techniques, such as data perturbation and the generation of synthetic data to protect the confidentiality of the data, and it should also consider supporting research for the development of these applications; and

(2) seek legislation that would increase the ability of the agency to protect confidential data. The legislation should specifically authorize data-sharing agreements with other agencies with legislative authority to enforce antidiscrimination laws and should extend Title VII penalties to nonagency employees.[69]

EEOC's proposal does not address these recommendations in any way.

Moreover, these is no indication in the proposal that the EEOC requires those to whom it provides the EEO-1 reports to (1) to maintain the same level of confidentiality that the EEOC does with respect to this information (other than the Department of Justice) (2) demonstrate that their information security programs are sufficient to protect this data from malicious attacks targeted at such data or (3) provide notification to the EEOC in the event their data security is compromised or the entity or individual experiences as a data breach. Moreover, the proposal is silent as to how the data will be transferred from the EEOC to the various federal or state agencies or individuals.

Although the NAS Report did not make any specific recommendations about a review of, or improvements to, the EEOC's information security protocols in connection with requiring employers to provide pay data in the EEO-1 forms, the NAS Report did emphasize that "the consequences of a breach in the protection of data provided in confidence are, as other federal agencies have discovered, painful and of lasting consequence."[70] This has most recently been seen with the data breach experienced last year by the Office of Personnel Management. This massive data breach was one of the largest in recent memory and was specifically targeted at employee, applicant and former employee information. OPM has now had to make major changes to its personnel and its policies and procedures in an attempt to earn the trust of the American public when it comes to the information entrusted to this agency. A major source of criticism of OPM has been its decision to transfer the personally identifiable information entrusted to it to a third party vendor. The class action complaints filed against OPM and the vendor allege that OPM did not do enough to vet the security measures employed by the vendor or require the vendor to provide ongoing updates to OPM regarding its security program and any breaches.

The OPM breach holds lessons for all federal agencies and should inform how the EEOC handles the disclosure and transfer of employee pay data to third parties. The EEOC needs to assure employers that it has reviewed its information security protocols and that they can and will safeguard the pay data employers are providing. In the hands of the wrong people, the original pay data from the EEO-1 report regarding pay could cause significant harm to a company and as previously noted subject employees to potential violation of their privacy. Indeed, while the Working Group noted potential Privacy Act concerns, the EEOC apparently gave absolutely no consideration to this potential problem. And even though no information in the EEO-1 report is tied to the name of any one individual, it is not hard to imagine a scenario in which individual employees within a small group listed on the EEO-1 report could be identified. Indeed, this is why the EEOC has suppression protocols in place for aggregate data where a group has three or less employers' information in it or any one employer accounts for 80% of the group data being aggregated.

V. CONCLUSION

In conclusion, the Chamber has serious concerns with the proposed EEO-1 Revisions. In view of the multitude of deficiencies which we have illuminated in this testimony and which will be further highlighted in our formal comments, we request that the EEOC withdraw the proposed revisions and commence a cooperative effort with all stakeholders to deal with the issues. Madam Chair, Commissioners, and General Counsel, we thank you for the opportunity to share some of those concerns with you today. Please do not hesitate to contact me or the Chamber of Commerce, if we can be of further assistance in this matter.

Exhibit 1

DECLARATION OF J. MICHAEL DUMOND, PH.D.

I, J. Michael DuMond, do hereby declare as follows:

I am over the age of 18. I have personal knowledge of the facts contained in this declaration and if called as a witness I would testify truthfully to the matters stated herein.
My name is Jon Michael DuMond. I hold a Ph.D., M.S. and B.S. in economics, all from Florida State University. I am currently a Vice President at Economists Incorporated, a position I have held since 2014. Economists Incorporated ("EI") is an economic consulting firm that offers consulting services to a wide variety of clients, including law firms, businesses, trade associations, government agencies and multilateral organizations. From 2009 to 2014, I worked as a Principal within the labor and employment practice of Charles River Associates ("CRA"). Prior to CRA, I worked as an economist with ERS Group from 2001 to 2008, an economic consulting firm which specializes in the economic analysis of labor and employment issues.
Since 2002, I have also been an adjunct professor with the Department of Economics at Florida State University. Beginning in 2006, my teaching has consisted solely of graduate-level courses, including computer programming, applied research methods and the econometric analysis of data. My academic research has focused on employee compensation and selection procedures and has appeared in peer-reviewed professional economic journals such as Economic Inquiry, Industrial Labor Relations Review, Managerial and Decision Economics, and the Journal of Sports Economics.
During my tenure with EI, CRA and ERS Group, I have regularly been engaged to analyze employee compensation for potential disparities in pay related to gender or race/ethnicity. These engagements have encompassed a wide variety of industries and occupations, and have included both private-sector and public-sector employees. I have been retained on behalf of both plaintiffs and defendants. A copy of my curriculum vitae is attached to this declaration as Exhibit A.
I was retained by the United States Chamber of Commerce in connection with the Equal Employment Opportunity Commission's proposed changes to the EEO-1 survey, which would affect federal contractors and private employers with at least 100 employees. More specifically, I was asked to evaluate whether the data that would be collected by the Equal Employment Opportunity Commission (EEOC) would be useful and informative with respect to identifying pay discrimination.[71]
If adopted, the new reporting requirements would require covered employers to "collect aggregate W-2 data in 12 pay bands for the 10 EEO-1 job categories. Employers will simply count and report the number of employees in each pay band.[72]" Separate counts would be made in order to show the number of male and female employees in each of the pay bands as well as the number of employees in seven race/ethnic groups.[73] In addition to the number of employees in each pay band, covered employers would also report the total number of hours worked by those same employees in the prior 12 months.
The EEOC and Office of Federal Contract Compliance Programs (OFCCP) have stated they intend to rely on these data to guide their investigations, "identify employers with existing pay disparities" and "detect discrimination.[74]" The proposal does not define "discrimination".
In my opinion, the collection of such limited and aggregated data is highly unlikely to provide meaningful information for detecting discriminatory pay disparities and even less useful in determining whether a contractor or covered employer is providing "equal pay for equal work." The reasons that these data would be uninformative in identifying potential pay discrimination are numerous, and I have detailed six of the most serious deficiencies in the following paragraphs.
First, the aggregated reporting structure proposed by the EEOC only distinguishes employees based on very broad occupational groups. That is, these groupings result in comparisons of employees who work in very different jobs and who may perform different work. As a result, these groupings will necessarily result in comparisons of employees with large differences in skills, training, education and other qualifications. For example, one of the 10 EEO-1 job categories is "Professionals", which encompasses a wide range of occupations such as lawyers and registered nurses. Hospitals that employ both nurses and lawyers would nevertheless be required to include both of these occupations together in the proposed EEO-1 survey. This grouping of nurses with lawyers ignores the fact that a nursing degree does not require a post-graduate college degree whereas a lawyer will almost surely have post-graduate education. Moreover, the knowledge, skills and abilities required of a nurse differ greatly from those factors that required of a lawyer. In fact, there is actually very little in common between nurses and lawyers beyond the sharing of a common EEO-1 category. While it is undeniable that nurses and lawyers are tied to very different labor markets, the EEOC's proposal ignores this reality and assumes pay should be similar for these types of occupations simply because they are both "Professionals."
Second, the EEOC's proposal does not recognize that pay is highly correlated with job experience. For example, some employers have an explicit seniority-based pay system. Also, pay generally increases the longer an employee has been with a company. Recent data from the Bureau of Labor Statistics indicates that the average amounts of company-specific tenure differ between men and women and among different race/ethnic groups.[75] Notwithstanding, the statistical tests proposed by the EEOC do not account for differences between employees based on length of service.
Third, The EEOC's proposal ignores the role of working conditions and how that affects an employee's compensation. For example, employees who work night shifts, swing shifts and/or weekends are often paid a differential to account for less desirable work schedules. For the same reason, jobs that require employees to work outside or exert atypical physical effort may also command a wage premium. Nevertheless, the statistical tests proposed by the EEOC using the collected data will not and cannot account for differences in working conditions.
Fourth, the EEOC is proposing that employers use W-2 data rather than an employee's "base" pay as the former includes commissions, tips, overtime pay, shift differentials and bonus payments. However, the EEOC is not proposing that employers report each of these compensation types separately, but instead is requiring that employers use the total of all these earnings when assigning an employee to one of the 12 proposed pay bands. In doing so, the EEOC is ignoring that some types of pay, specifically commissions and tips, are determined more by an employee's skill and efforts rather than an employer's pay policies. For example, differences in W-2 earnings among servers at a restaurant will be based on their ability to provide quality service and earning the associated tips/gratuities, and is largely undetermined by the employer. Unfortunately, the data that the EEOC is proposing to gather will not account for such differences.
The fifth reason that the data collected through the EEOC's proposal would be uninformative and unreliable for purposes of identifying pay disparities arises from the inherent problem in determining the number of work hours for salaried employees. As previously noted, covered employers would be required to also determine the total number of annual work hours for employees in each of the 12 pay bands. While such data is obviously available for employees paid by the hour, in my experience very few employers track the number of actual work hours for their salaried workers. In reviewing data from numerous HRIS data systems over the course of my professional career, I have learned that even though these systems often maintain a standardized or default value for "work hours" for salaried exempt employees (such as 40 hours per week) this default value frequently does not reflect an employee's actual work hours. For most employers, there is no need to track actual hours worked for salaried exempt employees, as they are not eligible for overtime pay. Put another way, accurate data on work hours for salaried or commissioned employees are typically not maintained in the normal course of business.
This limitation on accurate data for non-hourly employees is openly acknowledged within the EEOC's own proposal, as they invite "specific, detailed input on this aspect of its proposed data collection.[76]" The impact of this data limitation is serious: according to data from the Bureau of Labor Statistics, 59% of the US workforce is paid by the hour, meaning that 41% of the US workforce is paid on a basis for which no accurate work hours may be available and for which the EEOC does not have a recommended method for measuring.[77]
The proposal is also unclear as to how the EEOC will use or analyze the data relating to work hours. I assume that this information will be used to undertake a rough conversion of the pay range counts into hourly rates for workers in each of the pay bands, most likely using the midpoint of each of the proposed pay bands divided by the number of hours per employee. This approach, however, could be very misleading. For example, an employee that has an annual salary of $120,000 per year but was recently hired and only worked 1 of 12 months will be placed in the lowest pay band since he/she only earned $10,000. Under the EEOC's proposal, this employee would be grouped with minimum wage workers at the same company, even though the hourly rate for this employee would be much higher than the other employees in the same pay band. Similarly, an employee who works a part time schedule of 20 hours per week will be grouped and counted with FT employees who earn half of that hourly rate but work 40 hours per week. Simply collecting the aggregate number of hours worked doesn't fix the bigger underlying problem with the EEOC's proposal: employees are being counted in the "wrong" pay band. However, the EEOC and OFCCP might again infer discrimination from differences in the imputed hourly rates for employees in the same pay band, not recognizing that those differences can just as easily arise from part time and partial year employees, and the agencies will not be able to make that distinction with the aggregated data. Additionally, the proposal does not suggest a methodology by which the EEOC and the OFCCP will distinguish between intermixed part time, partial year or fulltime employees in the aggregate W-2 data. Moreover, since the total number of hours worked by employees in a pay band does not incorporate any of the factors known to affect pay, the data collected under this proposal will not provide any useful insight into the actual nature of employers' pay practices.
The sixth reason that the data that would be collected by the EEOC would not be useful in identifying pay disparities is that the two statistical tests that the EEOC suggests they would use in analyzing the collected data (i.e., the Mann-Whitney tests and the Kruskal-Wallis test) could easily lead to both "false positives" and "false negatives." In this context, an example of a "false positive" would be an inference that a company has gender pay disparities although women and men working in the same job are paid exactly the same (i.e., "equal pay for equal work.") On the other hand, an example of a "false negative" is a conclusion that pay disparities do not exist at a company even in the presence of unambiguous pay discrimination.
The reason that the two statistical tests proposed by the EEOC are likely to lead to both false negatives and false positives is that the aggregated data collected by the EEOC will not include one of the most critical factors that determine pay, specifically the job level/job grade of an employee. It is commonly understood that all employees, regardless of their race, ethnicity or gender, will experience increases in pay as they move "up" or "higher" in an organization. In my experience, most every company defines salary ranges for positions based on the level of the job within the organization. For example, Pharmaceutical Sales Representatives are paid less than their District Sales Managers who are paid less than their Regional Sales Directors. It would not be reasonable to assume that an inexperienced Pharmaceutical Sales Representative would be paid similarly to an experienced Regional Sales Director simply because they are in the same EEO-1 job category (Sales). But since the data that the EEOC is proposing to collect will not include any information or breakdown relating to an employee's job level/job grade, these relevant data will be ignored.
Despite a lack of consideration for an employee's job level/grade, the two statistical tests will nevertheless ascertain whether gender and racial groups are equally distributed across all the pay levels of a company without any consideration of the employee's job level/grade. The Pilot Study commissioned by the EEOC is very clear as to what these two statistical tests assume:

Although the (Mann-Whitney) test is sensitive to pure shifts, it has the more general interpretation of a test of the differences between two or more distributions.[78]
Because the Mann-Whitney and Kruskal-Wallis tests are determining whether men and women, for example, are distributed similarly across all pay bands (and hence all levels of an organization), these tests do not determine whether there are improper disparities within a pay band. Instead, it is a test of whether racial and gender groups of employees are similarly distributed across all pay bands, even when the employees in these groups may have very different levels of experience, may hold very different jobs (e.g., nurses and lawyers) and may be at very differently job levels of an organization. If the EEOC proceeds with their stated proposal to use these tests on the collected data to "detect discrimination," then the likelihood that they would lead to false positives and false negatives is greatly increased.
I conducted the Mann-Whitney test on two sets of simulated employment data[79] that the EEOC is proposing to collect in order to illustrate the inherent flaws in the analytical framework in the EEOC's proposal. In the first simulation, an employee's pay is completely determined by two factors: their job grade/level and the number of years they have worked at that level. For example, an employee entering the first job grade receives a starting annual salary of $30,000 and is awarded a raise of $1,000 each year he or she remains in that position. Similarly, an employee hired into the next grade level would receive a starting salary of $40,000 and receives a raise of $2,000 each year he/she or she remains in their job. A similar method of pay determination exists for the next three job grades at this fictitious company, though the starting pay and annual pay adjustments are naturally greater in the higher level grades. In this simulation, it is clearly impossible for a pay disparity to exist between women and men in the same job grade, after taking into account any differences in job tenure. A standard multiple regression analysis of this simulated data confirmed this obvious conclusion.
Using the Mann-Whitney test on these simulated data, however, yielded a statistically significant result, which the EEOC may incorrectly interpret as an indication of pay disparities that are adverse to women. That is, even in a simulation in which an employee's pay is 100% determined by a formula that allows no room for discrimination, the test that the EEOC is proposing to use still found a statistically significant difference.
This seemingly contradictory result occurs because in this simulation there are (by design) proportionately more female employees than male employees in the lower job grades, and therefore also in the lower pay bands. This demonstrates an important issue about the EEO-1 Pay Reporting Proposal: Since employees' compensation (and therefore their pay band) is highly correlated with their job grade, the statistical significance of both the Mann-Whitney and Kruskal-Wallis tests are going to depend on whether men and women are similarly distributed across all levels of a company's job grades/levels. This of course, is a very different type of question than whether pay disparities exist among employees in jobs requiring substantially equal skill, effort, and responsibility. Put simply, the EEOC's proposal won't be able to test whether or not similarly situated men and women are paid equally.
This point is further illustrated in the next simulation. At this second fictitious company, an African-American employee in grade level 1 receives a starting salary of $30,000. In contrast, a white employee in grade level 1 receives a higher starting salary of $30,500. Both employees receive a pay adjustment of $1,000 for each year worked. This blatant policy of pay discrimination is found at all other job grades as well: white employees in grade level 2 receive a starting pay of $40,500 compared to $40,000 for African-Americans. At the highest job grade in this hypothetical company, the starting pay gap is even more pronounced: African-Americans have a starting salary of $100,000 compared to $105,000 for white employees.
As would be expected, the existence of such an overt policy of pay discrimination was readily detectible through a standard multivariate regression analysis, which indicated that African-American employees were paid statistically significantly less than their white counterparts. On the other hand, applying the EEOC's proposed Mann-Whitney test to these same data did not show any statistically significant differences between white and African-American employees.
As before, this contradiction occurs because the Mann-Whitney test does not measure actual pay differences between groups of employees, but instead it tests if groups of employees are found in relatively similar proportions across all pay levels of that company. To illustrate this point, the second simulation included a similar percentage of African-Americans and white employees in each of the EEO-1 pay bands, even though the actual pay of African-Americans was always less than white employees in the same job grade.
These two simulations call into question the usefulness of the data that would be collected under the EEOC's proposal. These examples illustrate that the proposed statistical tests may not identify actual pay differences that are consistent with discrimination when they truly exist, and may incorrectly conclude that there is evidence consistent with discrimination when employees are actually paid equivalently.
But perhaps more importantly, the first simulation shows that a company with a disproportionate greater number of female or minorities at the lower grade levels is likely to "fail" the Mann-Whitney test. In fact, failing the Mann-Whitney test could occur while companies are making good-faith efforts to increase their minority and female representation. That is, many firms have worked diligently to increase the participation of workers from historically disadvantaged groups by hiring these workers as they became available. The result of this effort is that higher percentages of female, African American, and Hispanic workers are found in some of the lower job levels as they gain additional experience to be eligible for promotion. Accordingly, such companies would "fail" the Mann-Whitney or Kruskal-Wallis tests. The EEOC's proposal is likely to have the unintended and ironic effect of harming employers seeking to increase the participation of historically disadvantaged workers at all levels of their organization.
In summary, the aggregated data that the EEOC is proposing to collect and the analyses using the Mann-Whitney and Kruskal-Wallis tests are extremely unlikely to be useful or informative in addressing the laudable and important goal of eliminating gender and race-related pay disparities. The EEOC's proposal, as currently designed, will result in misleading comparisons, will not take into account known and accepted factors that influence pay, and will not lead to any useful determination as to whether employees are truly being paid the same for equal work or are otherwise being subjected to pay practices which violate Title VII.

I declare under penalty of perjury under the laws of the United States that the foregoing is true and correct. Executed at Tallahassee, Florida on March 4, 2016.

______________/s/_______________
J. Michael DuMond

Exhibit A to the Declaration of J. Michael DuMond, Ph.D.

J. Michael DuMond
1276 Metropolitan Blvd. Suite 303 · Tallahassee, Florida 32312 · (850) 558-6036
dumond.m@ei.com

Professional Experience

Economists Incorporated
Vice President (2014 - Present)

Apply economic, econometric and statistical analysis to pay equity, employment litigation, EEOC investigations, OFCCP audits and pro-active self-monitoring studies. Consult with corporations, government contractors and law firms on employment discrimination matters to monitor and assess the risk of litigation or government investigation for various occupations and industries. Work closely with clients to design statistical models consistent with the employer's hiring, promotion, performance evaluation, compensation, termination and reduction-in-force decisions. Assist clients in identifying the data required to conduct thorough analyses of their workforce and excels at preparing and analyzing extremely large and complex human resources data.

Conduct quantitative data analysis designed to help attorneys assess the value and merits of Fair Labor Standards Act (FLSA) and state wage and hour claims including misclassification, missed meal/rest periods, donning/doffing, off-the-clock work, unpaid overtime and regular rate calculations. Assist companies with extracting, compiling and summarizing archived data from payroll and timekeeping systems, as well as unconventional systems such as computer logs, to evaluate wage and hour claims. Compute waiting time penalties and other PAGA penalties in California wage claims.

Prepare written reports and declarations relating to economic analyses, data production and/or the calculation of economic exposure; provide support of these analyses in the form of sworn testimony.

Charles River Associates
Principal (2009 - 2014)

Manage and design economic analyses relating to labor and employment issues, including matters in preparation for litigation, audits, mediation, monitoring, and settlement. Supervise the activities and analyses of a team of Ph.D. economists and other junior staff. Direct the construction of computerized databases for use in analyses. Extensive experience in matters relating to Fair Labor Standards Act compliance (as well as state-specific wage and hour laws), including calculation of economic exposure relating to allegations of unpaid overtime and the miscalculation of the regular rate of pay.

Economic Research Services (ERS) Group, Inc.
Client Relationship Manager (2008 - 2009)
Research Economist (2001 - 2008)

Worked as a client relationship manager at an economic consulting firm that specializes in labor and employment issues. Casework included matters in preparation for litigation, arbitration, monitoring and settlement. Conducted economic and statistical analyses involving allegations of gender, race, and age discrimination in a variety of employment practices, including selection, termination, and compensation as well as Fair Labor Standards Act compliance.

Case management experience included the supervision of economists and analysts, data receipt and preparation, database construction and analysis, summarization of relevant economic literature and direct contact with clients. Past clients represent a wide variety of employers, including state and local governments, multi-facility retailers, production facilities, and educational institutions.

Florida State University
Adjunct Professor (2002 - Present)

Courses taught include Labor Economics, Economic Analysis of Data, SAS Programming, and MS Project. Since 2006, all courses have been taught at the post-graduate level. MS Project is the capstone course that simulates a consulting project, including the preparation of written and oral presentations.

Blockbuster, Inc.
Director of Franchise Finance (1999 - 2001)
Manager/Director of Modeling and Research (1997 - 1999)
Senior Demographic Research Analyst (1996 - 1997)

Responsible for all budgeting, forecasting and financial analysis of the 800+ store franchise division. Designed and implemented projects to improve profitability of both franchisees and franchisor, such as optimal labor allocation methods and alternative methods of acquiring rental product. These programs and models helped franchisees lower their labor and product costs without sacrificing top-line revenue.

Developed and supervised the implementation of predictive models for use by the real-estate, product and special format divisions. Oversaw the worldwide rollout of site-selection and cannibalization modeling applications and assisted in the valuation and pricing of company stores for sale to franchise units. The site-selection models were used extensively during the years in which Blockbuster opened 500 stores annually, resulting in first-year store revenues that exceeded the projected rate of return. Expanded the scope of the department by introducing econometric models into business segments that had not previously relied on statistical tools. Subsequently, took on the additional responsibility of designing a product allocation system that simultaneously determines the optimal aggregate purchase amount for the entire company. Managed a staff of analysts, including Ph.D economists. Prepared written reports and presented results to executive management.

Education

Ph.D., Labor Economics, Florida State University (1997)

M.S., Economics, Florida State University (1994)

B.S., Florida State University (1991) - Summa Cum Laude

Publications and Research Papers

"Stockwell v. City & County of San Francisco: What it Doesn't Say about Statistics in Age Discrimination Cases," (with Kenneth W. Gage), Daily Labor Report, Bloomberg BNA, July 2, 2014.

"Evidence of Bias in NCAA Tournament Selection and Seeding," (with Jay B. Coleman and Allen K. Lynch), Managerial and Decision Economics, Vol. 31, March 2010.

"An Examination of NBA MVP Voting Behavior: Does Race Matter?" (with Jay B. Coleman and Allen K. Lynch), Journal of Sports Economics, Vol. 9, No. 6, December 2008.

"An Economic Model of the College Football Recruiting Process," (with Allen K. Lynch and Jennifer Platania), Journal of Sports Economics, Vol. 9, No. 1, February 2008.

"Estimating Wage Differentials: When Does Cost-of-Living Matter?" (with Barry Hirsch and David Macpherson), Economic Inquiry, Vol. 37, No. 4, October 1999.

"Two Essays on Wage Differentials," Ph.D. Dissertation, Department of Economics, Florida State University, April 1997."

"Workers Compensation Recipiency in Union and Nonunion Workplaces," (with Barry Hirsch and David Macpherson), Industrial and Labor Relations Review, Vol. 50, No. 2, January 1997.

Presentations/Professional Meetings

"Navigating the New Frontier of Steering Claims," National Industry Liaison Group 2015 Annual Conference, July 2015, New York, NY. (Panel Discussant).

"Early Mediation of Wage and Hour Claims," American Conference Institute's 20th National Forum on Wage and Hour Claims and Class Actions, January 2014, Miami, FL. (Panel Discussant).

"Prob(it)ing the NCAA: Three Empirical Models on Collegiate Athletics," (with Allen K. Lynch and Jay B. Coleman), 2005, presented at the SAS M2005 Data Mining Conference. Invited Presentation, Mid-day Keynote Address.

"An Economic Model of the College Football Recruiting Process," paper presented at the Eastern Economic Association meetings, 2005.

"Customer Discrimination in Major League Baseball," (with Allen K. Lynch) paper presented at the Southern Economic Association meetings, 1999.

Expert Reports & Testimony

Johnny Reynolds v. State of Alabama. This case involved the termination of a consent decree between the State of Alabama and the U.S. Department of Justice relating to the hiring and promotions of African-American employees. Submitted report and provided testimony at judicial hearing.

Simmons, et al. v. Comerica Bank. This matter involved allegations of additional overdraft fees resulting from changes in debit posting procedures during nightly settlement. Submitted a written report that examined plaintiff's expert's statistical approach and its applicability for purposes of class certification and provided deposition testimony.

Coordinated Proceedings Special Title, Sutter Health Wage and Hours Cases and Coordinated Actions. These matters involved allegations of missed meal and rest periods for nurses and surgical care technicians at approximately 20 affiliates within the Sutter Health system. Submitted a declaration in relation to the motion for class certification.

Tammy Garcia v. MAKO Surgical Corporation. This case involved allegations of an unlawful termination due to plaintiff's gender. Submitted a written report that provides estimates of economic loss, arising from differences in base salary, incentive compensation, and stock options.

Angel Corona v. Time Warner Cable. This matter involved calculation of unpaid overtime due to an alleged improper calculation of the regular rate of pay. Written report submitted.

Selene Prado v. Warehouse Demo Services, et al. This report detailed the necessary calculations relating to the determination of rest period violations on a class-wide basis and damages relating to the miscalculation of the regular rate of pay for overtime purposes. Submitted a written report in relation to the motion for class certification.

Victor Guerrero v. California Department of Corrections and Rehabilitation. Plaintiff alleges that he was disqualified from consideration of a Correctional Officer position due to his past usage of a false Social Security number, a pre-employment criteria that allegedly had a disparate impact on Latino applicants. Submitted written reports and provided testimony at deposition and trial.

Daisy Vazquez and Bryan Joseph, et al.v. TWC Administration, LLC. Plaintiff alleges that putative class members were not properly paid overtime due to the inclusion of non-working hours in the calculation of the regular rate of pay and an improper allocation of commissions to the time periods in which they were earned. Submitted a written report in relation to the motion for class certification.

Fernando Vega v. Hydraulics International, Inc. Submitted a written report in regards to allegations that non-exempt employees were not paid for all hours they were logged into the timekeeping and project management system.

Representative Engagements as a Consulting Expert - Wage and Hour

Regular Rate Miscalculation

Inside sales employees at a multi-site retailer alleged that their overtime true-ups were based only on weekly overtime hours without consideration of additional daily overtime hours. Prepared estimates of economic exposure, including the valuation of potential offsets from factors that were included in the regular rate calculation beyond what was legally required.

Customer Service Representatives at multiple facilities filed collective actions involving failure to properly calculate the regular rate relating to overtime pay. Prepared economic loss estimates for use in mediation and settlement. Prepared measures of overpayments that were used as financial offsets to the estimated liability.

First line supervisors at a national landscaping company had overtime pay calculated using the fluctuating work week (FWW) methodology, in violation of wage and hour laws in some states, and the paid overtime did not properly include shift differential payments. Calculated economic exposure relating to these claims using the proper regular rate of pay.

Misclassification of Exempt Status

Inside sales representatives for an online retailer alleged they were misclassified as salaried exempt. Analyzed millions of records of time-stamped activity in order to ascertain estimates of alleged overtime hours. Calculated potential economic losses under alternate scenarios for use at mediation.

Store managers at a nationwide multi-site retailer claimed misclassification as salaried exempt. Prepared economic loss estimates of alleged unpaid overtime for use in mediation. (Multiple similar engagements)

Outside engineers at a national telecommunications company alleged that they were misclassified as exempt. Prepared an analysis of alleged overtime work and the related economic exposure for mediation and eventual settlement.

First-line construction and maintenance supervisors challenged their exempt status; prepared estimates of potential liability exhibits for trial.

Missed Meal/Rest Periods

Delivery drivers of a nationwide medical supply distributor alleged that their schedule necessitated that they work through their unpaid meal periods. Combined multiple data sources, including GPS data to assess the validity of these claims.

Major financial institution with multiple collective actions involving pay stub violations, deductions, wait-time penalties, overtime, and meal and rest periods. Prepared economic loss estimates for use in mediation and settlement.

Assistant store managers at a nationwide multi-site retailer alleged they were working without compensation during meal periods. Prepared an analysis to assess liability by integrating time-clock data with time-stamped cash register activity records.

Employees of on-campus restaurants at a state university alleged they were denied meal and rest breaks in violation of state law. Analyzed time-punch records to assess violation incidence and prepared estimates of economic exposure that were relied upon during mediation to reach settlement.

Off-the-clock/Time-Rounding allegations

Call center operators at several facilities filed collective actions involving failure to pay overtime as required under the Fair Labor Standards Act (FLSA). Prepared estimates of the amount of alleged "off-the-clock" work and the associated economic loss estimates; identified and corrected numerous errors made by plaintiff's expert. (Multiple similar engagements)

Customer Service Representatives at multiple facilities filed collective actions alleging a failure to pay for pre-shift activities. Integrated employees' computer time-stamp records with their time-clock data in order to determine the potential range of pre-shift work. Prepared estimates of potential exposure for use at mediation.

Non-exempt employees alleged that their Employees of a restaurant chain alleged that paid hours were systematically lower than their actual work hours due to the employer's time-rounding policies. Analyzed and compared actual work activity with payroll data to assess validity of these claims. (Multiple similar engagements)

Representative Engagements as a Consulting Expert - Equal Employment Opportunity

A major computer manufacturer was faced with allegations of Title VII violations, specifically involving discriminatory compensation and promotions. Assisted client with data production and responses to discovery requests. Conducted statistical analyses of compensation, promotions and estimates of potential economic exposure. Developed methodologies and analytical tools used to assess differences in performance ratings, annual bonuses, and merit pay adjustments on an on-going basis. (Multiple similar engagements)

Prepared analyses relating to a comprehensive audit of the major employment decisions at a global healthcare company, including promotions, terminations, performance evaluations, compensation, and disciplinary actions. Provided remediation recommendations for sub-groups of employees based on results of the audit.

Servers at a nationwide chain of restaurants alleged disparate treatment with respect to work assignments. Prepared a statistical analysis of allegations that resulted in a favorable settlement for client.

Store managers at a nationwide multi-site retailer alleged compensation differences under the Equal Pay Act. Prepared analytical databases and prepared estimates of potential economic exposure.

Multiple engagements involving analyses of potential adverse impact by race and gender relating to the conversion of salary grades to salary bands and the associated reclassification of employees.

Reviewed and identified errors made by opposing counsel's expert relating to the valuation of stock-options in a case involving a termination allegedly based on the plaintiff's age.

Multiple retentions for a variety of clients relating to potential adverse impact by race and gender of employees' annual performance rating assignments and the associated merit increases/bonus payments.

Prepared analyses of compensation for a large technology company. Investigated whether the observed pay differentials were the result of initial (starting) pay or whether they developed during the course of their employment tenure with the company. Calculated remediation amounts and devised a procedure in which the pay adjustments were incorporated into the annual merit pay adjustment cycle.

Data Sampling and Production

Processed millions of data records and prepared numerous databases for opposing counsel for a class of over 600 opt-ins in a wage and hour matter after limiting the data to relevant work dates and employment spells. (Multiple similar engagements)

Prepared a stratified sampling plan of stores for a nationwide, multi-site retailer who was alleged to have improperly classified assistant managers as salaried exempt. The sample was used for purposes of challenging conditional certification.

Exhibit B to the Declaration of J. Michael DuMond, Ph.D.

The EEOC's pay reporting proposal ("Proposal") recommends the use of the Mann-Whitney or Kruskal-Wallis test to evaluate the pay band data that will be submitted by companies. The Proposal does concede that the accuracy of these tests "needed to be addressed" as there is a concern that using them may yield "false positives."[80] Put another way, a statistically significant result from these tests may give rise to an inference of discrimination, even when no such discrimination exists. This is not a trivial concern. As will be shown in the following two examples, the Mann-Whitney test can be shown to find "discrimination" against women even when a company has a gender-neutral and formulaic pay policy. The second example shows just the opposite: the Mann-Whitney test fails to identify evidence consistent with discrimination against African-Americans even when a company chooses to pay them less than white employees in every job.

Example 1 - "False Positive"

Company XYZ sets their employee's salary based on two factors. The first factor is the employee's job grade. Like most companies, jobs at the higher grade levels pay more than jobs at the lower grade levels in accordance with the greater responsibilities, supervisory authority and requisite skills and experience that the higher grade jobs require. Second, company XYZ increases an employee's salary for each additional year of experience in their job grade. The salary structure of Company XYZ is shown in the following table.

Job Grade Level	Starting Salary	Additional Salary per (Full) Year in Job
1	$30,000	$1,000
2	$40,000	$2,000
3	$70,000	$3,000
4	$90,000	$4,000
5	$100,000	$5,000

As would be expected, a standard regression analysis of employee compensation that controls for job grade and the number of years an employee has worked in their current grade would reveal that there is not a gender pay difference. That is, the average female employee at Company XYZ is paid almost identically to the average male employee, after considering the effect of job grade and the number of years worked, and the measured gender pay difference is not statistically significant.

Total Number of Employees	Number of Female Employees	Estimated Pay Difference for Female Employees	Number of Standard Deviations
777	399	-$44.54	-0.19

Despite a pay policy that is, by construction, completely gender-neutral, the Mann-Whitney test of the EEO-1 pay bands would show a statistically significant result in favor of male employees, as depicted in the following chart.

This seemingly contradictory result stems only from the fact that there are proportionately more female employees than male employees in the lower grades, and therefore also in the lower pay bands. For example, the percent of women in Pay Band 4 ($30,680 - $38,999) is noticeably higher than the percent of men, while the opposite is true in Pay Bands 7 and above. This raises an important issue about the EEO-1 Pay Reporting Proposal: Since employees' compensation (and therefore their pay band) is heavily correlated with their job grade, the statistical significance of both the Mann-Whitney and Kruskal-Wallis tests are going to depend on whether men and women are similarly distributed across all levels of a company's pay grades.

Put simply, the Proposal doesn't test whether or not similarly situated men and women are paid equally, but rather, whether women are hired, promoted or otherwise placed into levels within the company at rates similar to men. This point is further illustrated in the second example.

Example 2 - "False Negative"

Like the first example, Company ABC uses only three factors to determine an employee's pay: job grade, the number of years of worked in the job and the race of an employee.[81] The salary structure for Company ABC is as follows:

Grade Level	Starting Pay Whites	Starting Pay African-Americans	Additional Salary per Year in Job
1	$30,500	$30,000	$1,000
2	$40,500	$40,000	$2,000
3	$71,000	$70,000	$3,000
4	$95,000	$90,000	$4,000
5	$105,000	$100,000	$5,000

Naturally, a regression analysis that controls for job grade and the number of years an employee has worked in their current job would identify the existence of a pay disparity. More specifically, African-Americans are paid, on average, $923 less than otherwise similar white employees, a difference that is statistically significant.

Total Number of Employees	Number of Female Employees	Estimated Pay Difference for Female Employees	Number of Standard Deviations
777	79	-$923	-2.34

In spite of this pay policy, the Mann-Whitney test does not reveal a statistically significant difference in the pay bands between whites and African-American employees, even in a situation in which the company is blatantly discriminating against African-Americans.

These examples call into question the usefulness of the data that would be collected under the EEOC's Pay Data Reporting proposal. First and foremost, these examples illustrate that the proposed statistical tests may not identify differences that are consistent with discrimination when those differences exist, and may incorrectly conclude that there is evidence consistent with discrimination when the evidence definitively is not. Secondly, while it is often the case that female and minority employees are paid less than male and white employees on average, the aggregated data that would be collected ignores any of the possible and often legitimate reasons for pay differences between gender or race/ethnicity groups. It only depicts the pay difference in a different manner.

Exhibit 2

Declaration of Ronald Bird, Ph.D.
Regarding the Equal Employment Opportunity Commission
Proposal to Revise the Annual Employer Report EEO-1 Survey
To Add Earnings and Hours Data

March 8, 2016

Statement of Qualification:

I, Ronald Edward Bird, was awarded the degree of Doctor of Philosophy (Ph.D.) in economics by The University of North Carolina at Chapel Hill on May 9, 1974. Subsequently I served as a faculty member teaching economic theory, economic benefit/cost analysis, economic statistics, labor economics and financial economics at North Carolina State University, The University of Alabama, Meredith College, and Wesleyan College.

Subsequently I served as Chief Economist for DynCorp Government Services conducting research regarding the costs and benefits of existing and proposed regulations and Paperwork Reduction Act information collection requests as a contractor to various Federal agencies, including the U.S. Departments of Labor, Energy, Treasury, and Defense and the U.S. Environmental Protection Agency.

From 1999 to 2005, I served as Chief Economist for the Employment Policy Foundation, a non-profit, non-partisan educational foundation that conducted research, inter alia, regarding the compliance costs to employers of Federal regulations and information collection mandates.

From 2005 to 2009, I served as Chief Economist of the U.S. Department of Labor, in which capacity I directly advised the Secretary of Labor regarding various economic policy and statistical issues, including

the compliance costs and benefits of existing and proposed regulations;
methods of estimating the compliance costs of Federal information collection requirements;
the economic theories of employment and wage determination; and
the statistical/econometric analysis of earnings differentials in relation to occupations, industries, education, experience, hours of work, race, ethnicity, gender and other factors.

As Chief Economist of the U.S. Department of Labor, I also served as liaison for the Department to the Executive Office of the President , the President's Council of Economic Advisers, and the Office of Management and Budget's Office of Information and Regulatory Affairs regarding various economic policy and analysis topics, including the conformity of the Department's regulatory economic analyses with the requirements of Executive Order 12866 regarding analysis of the economic costs and benefits of regulations, the requirements of The Paperwork Reduction Act, regarding the time and monetary cost burdens of Federal information collection activities, and the Regulatory Flexibility Act, regarding the economic impact of regulation compliance costs and of Federal information collection costs on small businesses and organizations.

Currently, I serve as Senior Regulatory Economist for the United States Chamber of Commerce. I conduct economic research and analysis regarding the costs and benefits of Federal regulatory and information collection activities as they affect employers. This work includes detailed review, evaluation and analysis of the benefits and costs of existing and proposed regulations and information collection activities of various Federal agencies and programs, including the Department of Labor's Office of Federal Contract Compliance Programs and the Equal Employment Opportunity Commission.

Summary of EEOC Information Collection Burden Estimate:

The Equal Employment Opportunity Commission has proposed (81 FR 20, p. 5113) to revise the existing annual "Employer Information Report" (Form EEO-1), by adding to the existing form requirements that employers of 100 or more total employees (1) report the number of employees in each of the 140 standard EEOC occupation/gender/race/ethnicity categories with a further cross-tabulation of each of these categorical employee counts by 12 annual earnings categories based on employer tax form W-2 data for the 12 months preceding the reference date of the report, and (2) report the total hours worked of employees counted in each of the resulting 1,680 categories for the 12 months prior to the reference date of the report. These requirements (which EEOC identifies as "Component 2") are in addition to the continuing requirement to file the currently specified tabulation of total employees in each of 10 EEOC occupations and 14 gender/race/ethnicity categories (140 categorical cell counts) , which EEOC identifies as "Component 1" of the proposed report.

The proposed revised EEO-1 form, including both components 1 and 2, would be required of all private industry employers of 100 or more total employees across all of the employer's establishments and wholly owned subsidiaries. The proposed revision would become effective for the annual employer information reports due September 30, 2017.

EEOC claims that the proposed combination of components 1 and 2 (see Table 4, 81 FR 20, p. 5120) of the proposed information collection will impose an average annual burden of 6.6 hours on each covered employer, including 5.6 hours per employer for collecting, verifying, validating and reporting data and 1 hour for "reading instructions." This amounts to 3.2 hours of additional labor effort per employer, 0.5 hours for additional instruction reading and 2.7 additional hours for collecting, verifying, validating and reporting data. The EEOC asserts that the total average per hour labor cost will be $24.23, which implies a total cost per employer of $159.92 per employer for components 1 and 2 combined. The additional cost, according to EEOC calculations, for the increment of the earnings and hours data required in proposed component 2 is $77.54. EEOC also asserts that the number of employers (filers) affected by the proposed new requirements (component 2) is 60,886, resulting in a national aggregate on-going annual burden subject to the Paperwork Reduction Act of 401,847.6 hours or $9,736,767.35.

EEOC recognizes in its proposal notice at 81 FR 20, p. 5120, a one-time implementation burden for developing new or revised standard database queries to existing human resources management information systems. EEOC asserts that an employer (filer) affected by the proposed requirements will "take 8 hours per filer at a wage rate of $47.22 per hour." This cost element amounts to $377.76 per employer who would be required to file the proposed EEO-1 reports. Based on EEOC's assertion that 60,886 employer filers would be affected, the calculated national one-time additional cost burden is $23,000,295.

EEOC also estimates the additional cost on itself to process the expanded data that would be submitted as $290,478, as the cost of needed internal staffing needs and costs.

In addition to reviewing the referenced Federal Register notice, I have reviewed the following documents that EEOC references in support of its calculation of time and cost burdens of its proposed information collection request under the requirements of the Paperwork Reduction Act.

National Research Council, "Collecting Compensation Data from Employers," report of the Committee on National Statistics, Panel on Measuring and Collecting Pay Information from U.S. Employers by Gender, Race and National Origin. Washington: National Academies Press, 2012.
Equal Employment Opportunity Commission, "EEOC Survey System Modernization Work Group Meeting, March 8-9, 2012, Draft Report," Washington, D.C., prepared by Sage Computing, Inc., March 19, 2012.
Sage Computing, "Final Report [of task order] To Conduct a Pilot Study for How Compensation Earning Data Could Be Collected From Employers on EEOC's survey Collection Systems and Development of Burden Cost Estimates for Bothe EEOC and Respondents for Each of EEOCC Surveys (EEO-1, EEO-4, and EEO-5).

I have also reviewed the burden estimation supporting statements and other documents submitted by EEOC to OMB/OIRA in connection with the information collection clearance requests for the currently approved version of the EEO-1 form and for previous requests going back to 2009.

Findings and Opinion:

Based on my review of the burden calculations and supporting materials and justifications presented by EEOC in its Federal Register notice and my review of the supporting documents cited by EEOC, based on my knowledge and experience regarding the requirements of the Paperwork Reduction Act, OMB guidance documents and requirements for submission of information collection requests and regulatory impact analyses, and based on my knowledge and experience of the generally accepted standards of research in the economics profession, it is my finding and opinion, to a reasonable degree of economic certainty, that the cost burden estimate presented by EEOC in support of its proposal to add earnings and hours worked data requirements to the current EEO-1 report grossly and significantly under-estimates the likely cost burden that the proposal will impose on affected employers. Both elements of the cost burden, (1) the one-time cost of modifying information systems and administrative procedures and (2) the on-going annual costs of collecting, compiling, tabulating, verifying, validating, and submitting the data have been inaccurately calculated and grossly and significantly underestimated by EEOC.

The errors and under-estimation of the reporting burden arise from at least six fundamental deficiencies in EEOC's economic analysis:

EEOC has not conducted any credible empirical study to validate the accuracy of its estimate of the compliance time burden of the current EEO-1 reporting requirement. Because the cost burden of the proposed new reporting elements (Component 2), is presented by EEOC as an addition of about 96% (near doubling) to the 3.4 hour reporting burden ascribed to the existing report (Component 1), the accuracy of the underlying current report burden is of critical importance to the final cost result. To establish, to a reasonable degree of economic certainty, the time and cost of the current reporting format, the Commission could have either conducted a survey of current filers or an experiment using itself and other government agencies as proxy filers. No evidence of either sort has been presented. EEOC, in its Federal Register notice states that its "pilot study" contractor, Sage, "approached some private employers to seek data" about the additional cost of the proposed new elements, but received too few responses. My experience in research to obtain similar data for information collection burden and regulatory impact analysis purposes has been that surveys, field audits and experiments to collect such compliance time and cost information are feasible and relatively economical to conduct. EEOC's claim of non-response to its request may be mostly a reflection of the ineffectiveness of its initial effort, and it is no excuse for putting forward a baseline burden estimate that is without empirical or reasonable foundation.
EEOC has also not conducted any credible empirical study to validate its estimate of the additional time burden of the proposed new earnings and hours report elements. The claim that the additional burden would be 3.2 hours on top of the current format report (Component 1) burden is not based on any evidence cited explicitly in the Commission's notice. One possible conclusion is that EEOC invented the number arbitrarily and capriciously from its imagination. The Commission's consultant, Sage, clearly states on page 109 of its report that its failed survey attempt yielded no results and that "it is not possible to provide estimates of the burden." It seems that EEOC has chosen to attempt to do the impossible by presenting a rather precise seeming burden estimate (3.2 hours for Component 2) that its own consultant declared to be impossible to do.[82]
EEOC has not provided any credible empirical evidence as a basis for its estimate that the one-time cost burden of the proposal is represented by 8 hours of total labor time. Again, EEOC could have conducted surveys or experiments to obtain empirical estimates of the time and resources needed to adjust information systems and administrative procedures to facilitate the proposed earnings and hours reporting. Again, EEOC has presented a number arbitrarily and capriciously produced from nothing. EEOC's analysis on this and other points fails to meet the most fundamental standard of economic research - that the methods by which data is obtained be transparent to and reproducible by independent observers. EEOC's analysis is not transparent.
EEOC uses two wage data parameters in its burden cost calculations: $24.23 per hour for labor employed to collect, compile, tabulate, validate, verify, and submit required data, and $47.22 per hour for the labor to make one-time modifications to information systems to facilitate the new reporting elements. In the first case the amount is the average hourly compensation reported by the Bureau of Labor Statistics from its Employer Cost of Employee Compensation Survey of December 2013 for an "administrative support" employee. In the second case the amount is the BLS compensation survey amount for a professional employee. EEOC has failed to account for the fact that the labor involved in both the annual report preparation and submission activity and in the one-time adjustment of systems to facilitate new requirements is not restricted to a single occupational category. A correct analysis of the information cost burden would recognize that the activity involves the blended effort of workers across an array of occupations, levels of responsibility, training, experience and compensation. An average hourly compensation rate used in any information collection burden calculation is meaningless unless it is constructed as a weighted average of the time and cost associated with each member of the overall task team. In particular, EEOC should have considered that an official report to the government by an employer from which potential legal charges and liabilities could emanate will require some review and approval at the highest levels of corporate responsibility. It is plainly absurd for EEOC to assume, without any empirical basis reflecting the practice of actual filers, that the annual reporting responsibility is fully represented by the hourly pay rate of the lowest rank of administrative support worker. A further error in the analysis is EEOC's assumption that employee compensation is the full measure of the compliance burden on the employer of allocating labor time to the task. In addition to direct labor compensation, EEOC should add allowance for physical and indirect labor overhead as parts of an estimate of the full economic opportunity cost of a regulatory or paperwork compliance mandate.
EEOC's calculation of the aggregate national time and cost burden of the proposed information collection is based on an assumption that 60,886 filers will be subject to the revised reporting requirement. The proposed requirement applies to all private sector employers of 100 or more workers. According to authoritative data published by the Office of Advocacy of the U.S. Small Business Administration, the number of private employers in 2012 (the latest year available) with 100 or more employees totaled 101,642 and these firms operated a total of 1,559,581 separate establishments (39 establishments per employer firm). To the extent that the EEOC's burden calculation identifies the number of employers subject to the current EEO-1 filing requirement, that number is, 101,642. The Paperwork Reduction Act does not contain any provision allowing an agency to "discount" its estimate of information collection burden imposed on the public because of public non-response or because of the agency's own ineffectiveness in collecting the subject data. The EEOC is asking for data from all firms with 100 or more employees. That number of firms being asked for data is the number that EEOC is required to use to calculate its information collection burden estimate. That number is 101,642 according to the most recent available authoritative data. This means that regardless of other issues discussed herein, EEOC's burden estimate is in error by about 40 percent. Furthermore, to the extent that individual filers must submit multiple reports, the number of responses and attendant burden is greater.
EEOC's analysis assumes, contrary to its own reporting instructions, that every covered employer will file only a single report. EEOC's current and proposed instructions clearly require that each employer of 100 or more total workers, company-wide, submit a separate report form for each separate establishment employing 50 or more employees, and that each multi-establishment employer submit an additional composite report for aggregating the data across all establishments. Furthermore, covered employers are required to prepare and submit for establishments with less than 50 employees either a small establishments consolidated report or individual establishment reports. EEOC's own data indicates that the average filer in 2013 submitted 4.5 distinct reports, including all separate establishment reports. My experience in analysis of hundreds of information collection and regulatory recordkeeping burdens is that the number of separate establishments within a company is a significant parameter affecting the total compliance cost burden, and it has remained so despite reporting efficiency improvements associated with newer information technology and systems. Total employment is a similarly important parameter. EEOC has presented no meaningful empirical evidence to support its contention that the time and cost burdens of the existing and proposed EEO-1 reporting requirements will be invariant with respect to a filer's number of establishments or number of employees.

In all prior calculations of the reporting burden for the EEO-1 form, the Commission recognized the principle that number of establishments for which data must be compiled, tabulated and reported is a significant factor affecting the total time and cost burden for a filing entity. In fact, the commission previously used the exact number of 3.4 hours per report filed in all previous burden calculations, beginning with the earliest on record in 2009. Now EEOC proposes to shift that number and apply it to the burden per filer (i.e., per employer firm) instead of per report. EEOC claims that its shift from the per report basis to the per filer basis is a reflection of its consideration of the effect of increasing automation of human resource information systems, but EEOC does not present any quantitative empirical data regarding the supposed change in circumstances.

It is notable that the exact same 3.4 hour number, which previously was used as a per report burden and multiplied by the 307,103 reports filed in 2013 in EEOC's supporting statement for its information collection clearance filed in 2015, is now being used by EEOC as a per filer hour burden and multiplied against the 67,146 filers of Component 1 of the proposed information collection. Not only does the Commission lack real data to support its putative estimate of a per filer burden, it appears that the Commission lacks the imagination to invent a number that is different from the one it used previously for the per report calculation. The numbers and calculations presented by the Commission give the appearance of an effort to manipulate the results to make it appear that the burden is low. The error is compounded by the fact that EEOC has no credible evidence to support the contention that 3.4 hours is the correct burden even on a per report basis. There is reason to suspect that for many filers the burden per report is significantly greater than 3.4 hours per report, and for filers of multiple reports, the total cost per filer would likely be even greater.

I also find that EEOC has not presented credible evidence that the proposed revision will yield any meaningful or measurable welfare benefit. Indeed, the study by Sage Computing, cited above, which EEOC commissioned to inform and advise its decision to put forth the proposed EEO-1 revision, points toward the conclusion that the proposal will likely reduce benefit in comparison to the current reporting format. On page 61 of their report to EEOC, Sage states

…we have to recognize the varying patterns of compensation for employees who work different hours. As a result, pooling together into a single cell the employees who may have received the same compensation from working different hours, and analyzing them with a single offset as the format of the proposed EEOC form suggests, may lead to biases that are difficult to quantify because of the model misspecification with respect to the time commitments required by the different positions.

This flaw in the design of the proposed new reporting requirement, which EEOC's own contract consultant identified, could significantly impact the rate of false-positive results obtained. Increased false positives from using the proposed form to inform investigations will divert EEOC's scarce investigative resources in fruitless directions, with the possible result that the Commission's effectiveness to identify and eliminate discrimination may be reduced by adoption of the proposed revision. Reduction in the Commission's effectiveness because it allows its scarce resources to be wasted by investigations based on a report format that generates excessive numbers of false positives would mean that the welfare benefit of the proposed revision is negative compared to the status quo reporting requirement.

EEOC is required to seek public input regarding its information collection and regulatory proposals, and the requirement is that EEOC provide the public with at least 60 days to comment before it sends its proposal to OMB/OIRA. EEOC is not prevented from providing a longer public comment period, and the size and complexity of the proposal to add earnings and hours worked data to the EEO-1 reporting form is such that meaningful public comment is curtailed by the 60 day limit. The 60 day limit has been further truncated by EEOC's decision to hold hearings on the proposal at an earlier date and to require submission of testimony and supporting documents (such as this declaration) a further week in advance of the hearing. EEOC rejected requests for an extension of the comment deadline, and EEOC has presented no reasoned basis for its rush to proceed with the proposed revision.

Because EEOC has arbitrarily limited the time available for public response and comment, the analysis presented here is limited. In particular, it would have been useful to provide empirical data from surveys of employers, which are currently underway and being tabulated. Preliminary results confirm the findings presented here based on limited time and information: EEOC's estimates of the information collection burden are inaccurate and grossly under-estimate the likely compliance cost burden. This analysis will be revised and resubmitted as additional information becomes available from on-going private surveys and as further research into existing sources reveals additional empirical evidence.

I declare under penalty of perjury under the laws of the United States that the foregoing is true and correct. Executed this 9th day of March, 2016, at Washington, D.C..

_____________/s/________________
Ronald Bird

Exhibit 3

DECLARATION OF ANNETTE TYMAN

I, Annette Tyman, do hereby declare as follows:

Qualifications

I am over the age of 18. I declare that the statements in this declaration are correct of my own personal knowledge and I am competent to testify concerning them.
My name is Annette Tyman. I am an attorney and Partner at Seyfarth Shaw LLP. I also serve as Co-Chair of the Firm's OFCCP Compliance, Affirmative Action & Diversity Consulting Practice Group. In that role, I oversee a team of Advisors and Analysts who prepare Affirmative Action Plans and EEO-1 Reports on behalf of companies across the United States.

Overview of the EEO-1 Filing Process

A general overview of the EEO-1 filing submission process is summarized as follows.
Typically, the process to prepare and submit EEO-1 Reports begins with employers who provide employee information, including employee name or employee identification number, location, race, gender, EEO-1 job category (if tracked by the employer), and job title.
At that point, one must analyze and prepare the data. This step includes validating, coding and creating summary tables in order to input or submit the EEO-1 required data to the EEO-1 Survey site. The quality and consistency of the data set necessarily impacts the time it takes to prepare the data for submission to the EEO-1 Survey site.
Pursuant to the EEO-1 instructions, multi-establishment employers are required to submit (1) a report covering the principal or headquarters office ("Headquarters Report"); (2) a separate report for each establishment employing 50 or more persons ("Type 4" report); (3) a separate report ("Type 8" report - which provides similar information to the Type 4 report) for each establishment employing fewer than 50 employees, or an Establishment List (Type 6 report), showing the name, address, and total employment for each establishment employing fewer than 50 persons; and (4) a Consolidated Report which summarizes the employers workforce.

Submission of EEO-1 Data - Two Formats

The EEOC allows employers to submit data using two electronic formats methods.
1. The first format, which is most commonly used by employers, is most easily described as a "fillable pdf." The data requires manual entry of employee summary counts by race/ethnicity and gender, using the EEO-1 job groups. Pursuant to Footnote 62 of the EEOC's Proposed Revision of the Employer Information Report (EEO-1) and Comment Request ("proposed EEO-1 Revision"), 98% of all employers use this format ((60,886 - 1449)/60,886) for completing the EEO-1 Survey.
2. The second format involves the submission of an electronic data file. The EEOC's parameters are specific and technical and include a few data file formats. While the proposed EEO-1 Revision refers to this methodology as a data file "upload," the data file is actually emailed to upload@EEOC.gov.

Submission by "PDF Fillable Format"

If employers use the EEO-1 electronic "fillable pdf" format for submitting to the EEO-1 Survey, which is described in 7.a. above, the steps identified below are generally followed.
1. After analyzing the data and preparing summary totals for input into the EEO-1 Survey, employers log into the EEO-1 Survey site using employer specific login information. Once employers log in, they see a listing of the EEO-1 reports that were filed for each establishment in the location that was filed in the previous year. The company will be required to "select" each location and either complete a new form or delete the form. EEO-1 forms for establishments may be deleted for various reasons, including, for example, if the location closed during the last reporting period.
2. The employers then answer or verify the response to the three EEO-1 Survey questions found in Section C of the EEO-1 Survey regarding "Employers Who Are Required to File" and verify NAICS codes.
3. In our experience, most employers do not track EEO-1 survey "unit numbers" which identify the establishment for previously submitted reports. Instead, this information is obtained by downloading prior year EEO-1 reports before submission.
4. For previously existing establishments, employers review and revise, if necessary, the address information of the location and complete the ethnicity/race and gender totals by location and EEO-1 Category. If the location total has changed by more than 20%, a prompt will appear to verify the entries are correct.
5. If there are new locations, employers are required to complete a new EEO-1 establishment form for those locations.
  1. For locations with more than 50 employees, the company is required to complete Type 4 establishment reports. For locations with less than 50 employees, the company has the option to complete a Type 8 report, which is the similar to the Type 4 establishment report (but is coded as a Type 8 report to signify it is being submitted for a location with less than 50 employees), or a Type 6 establishment list. A Type 6 establishment list includes a listing of the address and total employee count for each location with less than 50 employees.
  2. A key difference between Type 6 establishment list and Type 8 establishment report (both of which relate only to establishments with less than 50 employees) is found when preparing the employer's Consolidated Report. The Consolidated Report automatically populates if employers submit Type 8 establishment reports. In other words, if employers choose to provide the more detailed information that includes race/ethnicity and gender, even for its locations with less than 50 employees, the Consolidated report is automatically populates the detailed counts provided for each establishment and the headquarters location. If the employer instead completes Type 6 Establishment lists for locations with less than 50 employees (which includes only address location and total employee count at that location), then it must manually complete the Consolidated Report, to ensure the total counts, including race/ethnicity and gender by location and EEO-1 Category for both locations with more than 50 employees and locations with less than 50 locations are included in the Consolidated Summary. This manual process must be carefully reviewed because the employer must ensure that the totals for each location and establishment match the data entered into the Consolidated Report submission.
Once the data has been entered and the employer receives confirmation from the EEO-1 Survey site that the forms are complete (as noted by a change in status from "Red" to "Green"), the data is ready to be "certified." At that point, the authorized official for the employer certifies the submission of the report.

Submission by Data File "Upload" and One-Time Implementation Cost

As previously noted in 7.b. above, once the data has been collected, validated and coded, employers also have the option to "upload" a data file to the EEO-1 Survey tool. All data files must comply with the precise data specifications prescribed by the EEOC.
Because Seyfarth Shaw provides specialized consulting services to a multitude of employers, it has developed an analytical tool to facilitate the uploading of data files needed to complete the current EEO-1 Survey submission. The development of the tool for the current EEO-1 data file "upload" required a one-time expenditure of over 110 hours data analyst hours to implement the data requirements of the current EEO-1 report. To facilitate the data file upload for employers, Seyfarth utilizes this data tool to submit multi-establishment EEO-1 reports. Pursuant to Footnote 62 of the proposed EEO-1 Revision, the EEOC recognizes that only 2% of all employers have availed themselves of the tools necessary to use this format (1,449/60,886) for completing the EEO-1 Survey. In our experience, information does not go "nearly directly from an electronic file generated by the HRIS to the survey data base" as set forth in Footnote 62 of the proposed EEO-1 Revision.
Before data submission using an "upload" file, employers must first test the data file to ensure it complies with the EEO-1 Survey data upload specifications. During the testing phase, employers receive a summary report detailing the specific issues, if any, with the upload file. After correcting any identified issues, the file is e-mailed to upload@EEOC.gov for upload to the Survey site. Once confirmation is received from the EO Survey team that the data file has been "uploaded" to the EEO-1 Survey site, the process continues on as if the Company had manually keyed in the data using the "electronic fillable pdf." Specifically, the company will receive a notification that the forms are ready for certification. The company representative must then review all the reports that are flagged on the EO Survey site (identified as red). The flagged reports are typically those for new establishments, closed establishments, or locations establishments with workforce counts that have changed by more than 20%. Once all the reports that were flagged have been reviewed and finalized, the Company will then need to certify their reports to successfully file them.

I declare under penalty of perjury under the laws of the United States that the foregoing is true and correct. Executed this 8th day of March, 2016, at Chicago, Illinois.

__/s/ Annette Tyman________________________
Annette Tyman

Exhibit 4

Chamber of Commerce Exhibit 4

Exhibit 5

Chamber of Commerce Exhibit 5

Exhibit 6

Chamber of Commerce Exhibit 6

Exhibit 7

Supporting Statement
Recordkeeping and Reporting Requirements for
Employer Information Report (EEO-1)

A. Justification

The legal basis for the Employer Information Report (EEO-1) form and recordkeeping requirements is Section 709(c) of Title VII of the Civil Rights Act of 1964, as amended, 42 U.S.C. 2000e-8(c), which imposes the requirement that "[e]very employer, employment agency, and labor organization subject to this subchapter shall (1) make and keep such records relevant to the determinations of whether unlawful employment practices have been or are being committed, (2) preserve such records for such periods, and (3) make such reports there from as the Commission shall prescribe by regulation or order. . ." Accordingly, the EEOC issued a regulation, 29 C.F.R. §1602.7, which sets forth the reporting and related recordkeeping requirements for private industry employers with 100 or more employees. The U.S. Department of Labor's Office of Federal Contract Compliance Programs has imposed the same reporting requirement on certain Federal Government contractors and first-tier subcontractors with 50 or more employees. The individual reports are confidential and may not be made public by the Commission prior to the institution of a lawsuit under Title VII in which the individual reports are involved.
EEO-1 data are used by EEOC to investigate charges of employment discrimination against employers in private industry and to provide information about the employment status of minorities and women. The data are used to evaluate and prioritize charges under the Commission's charge processing system and to determine the appropriate investigative approaches. The data can be analyzed to develop statistical evidence as the investigation proceeds. EEOC uses the data to develop studies of private sector work forces (see www.eeoc.gov/statistics for some examples), and to assist researchers requesting data for academic studies.

The data are shared with the Office of Federal Contract Compliance Programs (OFCCP), U.S. Department of Labor, and several other Federal agencies. Pursuant to §709(d) of Title VII of the Civil Rights Act of 1964, as amended, EEO-1 data are also shared with ninety- four State and local Fair Employment Practices Agencies (FEPAs) for their enforcement efforts.
The EEO-1 report is collected through a web based on-line filing system. There are 70,070 respondents reporting annually[83] and 98% of these respondents file on-line. The on-line filing system has reduced the burden hours.
We are not aware of any duplicative or related data collection efforts.
The EEO-1 report is collected from all private employers with 100 or more employees and certain government contractors with 50 or more employees, so there is no burden on small business.
Because the data are an integral part of the Title VII enforcement process, failure to collect the data would reduce our ability to enforce Title VII. The data has been integrated into the enforcement process and computer applications for retrieving EEO-1 reports and conducting statistical comparisons of such reports to the external labor market are available to the enforcement staff at their desktops. Collecting the data less often would impair enforcement decisions by reducing the reliability of the data, as there will be a lag between the employment statistics provided by employers when reporting and the time when the data is used. This problem is likely to be most pronounced among industries and employers with fluctuations in employment. It is important to make certain that employment decisions are consistent with law when increases or decreases in employment occur. A gap of more than a year between data collections would also impose some processing costs on EEOC because more work would be needed to update mailing lists. The data is only collected annually.
None of the above special circumstances will be used to collect the EEO-1 Report.
See attached Federal Register Notice dated June 30, 2014 (79 FR 36802). Only one comment was received from the public, and it supports the continued use of the EEO-1 without change to the form. However, the commenter suggests making a change to the reporting procedures that currently prevent parent companies from electronically submitting EEO-1 reports for different subsidiary companies operating at the same physical location within the same industry classification. EEOC has made contact with this organization to begin implementation of this recommendation for the EEO-1 report. Finally, EEOC is planning to make this change by 2015 reporting cycle.
EEOC's employees are prohibited by law from providing any payment or gifts to respondents, other than remuneration of contractors or grantees.
All reports and information from individual reports are subject to the confidentiality provisions of Section 709(e) of Title VII, and may not be made public by EEOC prior to the institution of any proceeding under Title VII. However, aggregate data may be made public in a manner so as not to reveal any particular employer's statistics. All state and local FEPAs with whom we share the data must agree to maintain the confidentiality of the data.
The EEO-1 Report does not solicit any information of a sensitive nature from respondents.

ANNUAL RESPONDENT BURDEN HOURS	1,044,150
ANNUAL EMPLOYER COSTS	$19.83 million

REPORTS FILED 2013	ESTIMATED BURDEN HOURS PER REPORT	ESTIMATED TOTAL BURDEN HOURS	COSTS PER HOUR	ESTIMATED TOTAL ANNUAL EMPLOYER COSTS
307,103	3.40	1,044,150	19.00	19,838,850

Burden hours are assumed to be 3.4 hours per report at a cost of $19.00 per hour. [84]

There are no cost changes. Private employers have been completing this form for a number of years.
Estimated cost to the federal government will be: $650,000.00 dollars contract cost (based on a competitive bid process from prior years.) The government cost has decreased because the government has used base year contract plus two option years.
There are no program changes. However it should be noted that the burden hours estimated in question 12 above have been revised since the approval of the report for the 2011 reporting period. The total burden hour estimate represents an update from the 2011 estimates when total burden hours were estimated at 987,394. This increase in total burden is due to the creation of new firms filing EEO-1 reports that are subject to the reporting requirement, leading to an increase in reports filed with the agency.
The time schedule for information collection and publication is as follows:

Filing deadline September 30

First Follow-up October 15

Second Follow-up November 15

Preliminary Data Periodic data audits

Final Data June 30

In each survey year a publication, Job Patterns for Minorities and Women in Private Industry is posted on our web site. This consists primarily of non-confidential aggregations of the data based on various geographic and industrial criteria. So for example, a table combining all EEO-1 reports for all food and beverage stores in the New York metropolitan area is provided (http://www.eeoc.gov/eeoc/statistics/employment/jobpat-eeo1/index.cfm). Similar data sets are available on data.gov. Some of the primary users of this data are employers (for self assessment and affirmative action purposes) and researchers.

EEOC is not seeking approval of this nature.
No exceptions are requested.

Camille A. Olson

SEYFARTH SHAW LLP
2029 Century Park East
Los Angeles, CA 90067
310-277-7200

131 S. Dearborn Street
Chicago IL 60603
312-460-5831
colson@seyfarth.com

Camille A. Olson is a partner of Seyfarth Shaw LLP, Co-Chair of its National Complex Litigation Practice Group, and National Chairperson of its Complex Discrimination Litigation Practice Group. She serves on the firm's National Labor and Employment Law Steering Committee and is the past National Chairperson of the Labor and Employment Practice Department. In November 2013, Ms. Olson was appointed Chairperson of the United States Chamber of Commerce's Equal Employment Opportunity ("EEO") Subcommittee. For nearly 30 years, she has represented companies nationwide in all areas of litigation, with emphasis on employment discrimination and harassment, wage and hour matters, and independent contractor status. Her trial experience includes lead defense trial counsel in "bet the company" harassment, discrimination, independent contractor, and wage and hour cases. In addition, Ms. Olson represents employers in designing, reviewing, and evaluating their pay practices to ensure compliance with federal and local equal employment opportunity laws. Ms. Olson also works closely with Seyfarth Shaw's OFCCP & Affirmative Action Compliance team, which has defended clients in every OFCCP region and virtually all district offices with clients ranging from Fortune 100 companies to small employers in all industries. She has served as outside counsel and independent counsel to Boards of Directors and Executive Team Members in connection with internal investigations and highly sensitive litigation matters. Ms. Olson is on the Board of Directors of Inland Press Association, and Co-Chair of Inland's Human Resource Committee.

Ms. Olson is the recipient of numerous accolades. Ms. Olson was named one of 2016's Top Attorneys in Illinois as well as one of the Top 50 Women Attorneys in Illinois by Chicago Magazine. She was named one of the top 10 employment litigators in the country by Legal 500, which praised her 'superb representation' and named her 'one of the best employment litigators in the country.' Law360 has named Ms. Olson one of seven Employment Litigator MVPs for her exemplary work on critical litigation. She has been named as one of the Nation's Most Powerful Employment Attorneys by Human Resource Executive and LawDragon, and is a Fellow of the American Bar Foundation. Chambers USA consistently rates her at the highest level, recently noting: "Specializing in complex discrimination litigation, Camille Olson is 'just terrific.' Everything you hear about her is excellent and she handles the really cutting-edge stuff." Also in 2013, Inland Press Association named Ms. Olson the recipient of its Ray Carlsen Distinguished Award; one of the Association's highest honors, the award recognizes "members who have distinguished themselves in service to the association and its affiliated foundation, who have been exemplary in service to their communities and their companies, and who deserve the recognition of their peers and colleagues." In 2014 the Columbus Dispatch inaugurated the Camille A. Olson Award of Excellence, to be awarded annually to one of their managers in recognition of Ms. Olson's commitment to excellence in managing independent contractor relationships.

ENDNOTES

[1] The Testimony was prepared on behalf of the U.S. Chamber of Commerce by Seyfarth Shaw LLP attorneys Lawrence Lorber, Annette Tyman, Richard Lapp, Christine Hendrickson, Karla Grossenbacher, Sam Sverdlov, and Arielle Eisenberg with assistance from case assistant Kali Froh as well as attorneys Kris D. Meade and Rebecca Springer of Crowell & Moring. In addition, the Testimony is based on the Declarations of Economist Dr. J. Michael DuMond (see Attachment Exhibit 1), Economist Dr. Ronald Edward Bird (see Attachment Exhibit 2) and Annette Tyman, J.D. (see Attachment Exhibit 3).

[2] I am also a partner with the law firm of Seyfarth Shaw LLP, where I chair the Labor and Employment Department's Complex Discrimination Litigation Practice Group. Seyfarth Shaw LLP is a global law firm of over 800 attorneys specializing in providing strategic, practical legal counsel to companies of all sizes. In addition to my litigation practice, which has specialized in representing local and national companies in federal court litigation involving claims of employment discrimination, I also represent employers in designing, reviewing, and evaluating their pay practices to ensure compliance with federal and local equal employment opportunity laws. Nationwide, over 350 Seyfarth attorneys provide advice, counsel, and representation to employers in pay equity matters.

[3] For underlying authority to require employers to submit the EEO-1 form, EEOC relies on the recordkeeping provisions of 42 U.S.C. § 2000-e8(c) and 29 CFR 1602.7. See 81 Fed. Reg. 5113. Nothing in the authorizing statute exempts the EEOC from complying with the PRA.

[4] This figure underestimates by 40% the number of private employers with 100 or more employees. Ex. 2, Ronald Bird Decl. p. 6.

[5] In 2009, the EEOC used this approach and estimated 599,000 burden hours. Ex. 4. That figure rose to 987,394 burden hours in both 2011 and 2014. Exs. 5 and 6. Using the $19.00 per hour rate used by the EEOC in 2015, those estimates would have identified annual burdens of $11.3 million in 2009 and $18.8 million in both 2011 and 2014.

[6] Ex. 3, Annette Tyman, Decl., generally describing the EEO-1 submission process.

[7] Ex. 3, Annette Tyman Decl. ¶7.a.

[8] The EEOC's suggestion that uploading a data file means that "the information goes nearly directly from an electronic file generated by the HRIS to the survey data base" is a gross misunderstanding of the process required for gathering and validating data for submission to the EEO-1 survey. Ex. 3, Annette Tyman Decl. ¶11

[9] Ex. 2, Ronald Bird Decl. p. 4.

[10] Ex. 2, Ronald Bird Decl. pp. 4-5.

[11] Insofar as the OFCCP has apparently abandoned its review of the submitted comments, there is no agency representation as to the input it received regarding the time and burden attendant to its proposal. Thus, the EEOC is citing to non-existent or non-published data to support its burden estimates.

[12] Ex. 2, Ronald Bird Decl. pp.5-6.

[13] As noted in Section III.C.2, any approach that assumes each exempt employee works 40 hours per week is not realistic and would only further undermine the efficacy of the data collected.

[14] Ex. 2, Ronald Bird Decl. pp. 5-6.

[15] See Section III (describing lack of efficacy of any analysis that compares employee pay within broad EEO-1 categories); Ex. 1, Declaration of Dr. Michael DuMond (same).

[16]" Sage Computing Final Report: To Conduct a Pilot Study for How Compensation Earning Data Could Be Collected from Employers on EEO's Survey Collection Systems (EEO-1, EEO-4, EEO-5 Survey Reports) and Develop Burden Cost Estimates for Both EEOC and Respondents for Each of EEOC Surveys (EEO-1, EEO-4, and EEO-5), p. 57 (Sept. 2015).

[17] Against this backdrop, it is important to note that Congress and courts have explicitly rejected the notion of pay equity based on "comparable worth" theories under which employers would be required to set wages to reflect differences in the "worth" of different jobs. (In the 1960s, the Kennedy Administration proposed a ban on sex discrimination in wages "for work of comparable character on jobs the performance of which requires comparable skills," with the assumption that job evaluation systems were available to evaluate the comparative worth of different jobs. Equal Pay Act of 1963: Hearings Before the Subcomm. on Labor of the Senate Comm. on Labor and Public Welfare, 88th Cong., 1st Sess. 2 (1963) (quoting S. 882, 88th Cong., 1st Sess., 109 CONG. REC. 2770 (1963) and S. 910. 88th Cong., 1st Sess., 109 CONG. REC. 2886 (1963)). Congress resisted the proposal for the simple reason that it did not want the government or judges to invade the workplace and tell employers what to pay their employees.) Under federal law, the value of the relative worth of one job as compared to another job is a matter to be decided within the province of the employer offering the employment opportunities to workers. For instance, the Sixth Circuit rejected the comparable worth theory stating, "Title VII is not a substitute for the free market, which historically determines labor rates." Int'l Union v. Michigan, 886 F.2d 766, 769 (6th Cir. 1989). Likewise, the Ninth Circuit rejected the theory on the grounds that it found "nothing in the language of Title VII or its legislative history to indicate Congress intended to abrogate fundamental economic principles such the laws of supply and demand or to prevent employers from competing in the labor market." AFSCME v. Washington, 770 F.2d 1401, 1407 (9th Cir. 1985).

[18] EEOC Compl. Man. Ch. 10.

[19] 29 U.S.C. § 206(d).

[20] EEOC Comp. Man. Ch. 10, at p. 20, available at www.eeoc.gov/policy/docs/compensation.html

[21] The ten EEO-1 Job groups include: Executive/Senior Level Officials and Managers; First/Mid-Level Officials and Managers; Professionals; Technicians; Sales Workers; Administrative Support Workers; Craft Workers; Operatives; Laborers and Helpers; and Service Workers.

[22] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 9.

[23] See EEO-1 instruction booklet, available at http://www.eeoc.gov/employers/eeo1survey/2007instructions.cfm. Notably, the examples provided omit the myriad other jobs that would also fall within the Professionals EEO-1 job group.

[24] EEOC Comp. Man. Ch. 10, at p. 22, available at www.eeoc.gov/policy/docs/compensation.html, (emphasis added) citing Stanley v. University of S. Cal., 178 F.3d 1069, 1074 (9th Cir.) (EPA requires two-step analysis: first, the jobs must have a common core of tasks; second, court must determine whether any additional tasks incumbent on one of the jobs make the two jobs substantially different), cert. denied, 120 S. Ct. 533 (1999); Stopka v. Alliance of Am. Insurers, 141 F.3d 681, 685 (7th Cir. 1998) (critical issue in determining whether two jobs are equal under the EPA is whether the two jobs involve a "common core of tasks" or whether "a significant portion of the two jobs is identical"); Brewster v. Barnes, 788 F.2d 985, 991 (4th Cir. 1986) (same).

[25] Standard Occupational Classification system overview, March 5, 2016, http://www.bls.gov/soc/home.htm, "The 2010 Standard Occupational Classification (SOC) system is used by Federal statistical agencies to classify workers into occupational categories for the purpose of collecting, calculating, or disseminating data. All workers are classified into one of 840 detailed occupations according to their occupational definition. To facilitate classification, detailed occupations are combined to form 461 broad occupations, 97 minor groups, and 23 major groups. Detailed occupations in the SOC with similar job duties, and in some cases skills, education, and/or training, are grouped together."

[26] See Internists, SOC 29-1063 given the multiple listings categories of "Physicians and Surgeons."

[27] EEOC Comp. Man. Ch. 10, at p. 6, available at www.eeoc.gov/policy/docs/compensation.html.

[28] Id. at 7.

[29] Eskridge v. Chicago Bd. of Educ., 47 F. Supp. 3d 781, 790-91 (N.D. Ill. 2014). Although a similarly situated employee need not be "identical," Caskey v. Colgate-Palmolive Co., 535 F.3d 585, 592 (7th Cir.2008), he must be "directly comparable to the plaintiff in all material respects...." citing Naik v. Boehringer Ingelheim Pharm., Inc., 627 F.3d 596, 600 (7th Cir.2010); Lopez v. Kempthorne, 684 F.Supp.2d 827, 856-57 (S.D.Tex. 2010) ("'Similarly situated' employees are employees who are treated more favorably in 'nearly identical' circumstances; the Fifth Circuit defines 'similarly situated' narrowly. Similarly situated individuals must be 'nearly identical' and must fall outside the plaintiff's protective class. Where different decision makers or supervisors are involved, their decisions are rarely 'similarly situated' in relevant ways for establishing a prima facie case."); Alexander v. Ohio State University College of Social Work, 697 F.Supp.2d 831, 846-47 (S.D. Ohio 2012) (To be similarly situated, a plaintiff's purported comparators must have the same responsibilities and occupy the same level position.)

[30] EEOC Comp. Man. Ch. 10, at p. 8, available at www.eeoc.gov/policy/docs/compensation.html.

[31] Bazemore v. Friday, 478 U.S. 385, 400 n.10 (1986) (noting that "some [are] regressions so incomplete as to be inadmissible as irrelevant . . . .''); Sheehan v. Purolator, 839 F.2d 99, 103 (2d Cir. 1988) (affirming the district court's denial of class certification because the regression analysis relied upon was deemed to be ''flawed'' for failing to take into account non-discriminatory factors, such as "education and prior work experience."); Griffin v. Board of Regents, 795 F.2d 1281, 1292 (7th Cir. 1986) (holding that ''the explanatory power of a model is a factor that may legitimately be considered by the district court in deciding whether the model may be relied upon.''); EEOC Compliance Manual, Section 10: Compensation Discrimination ("[Employers often] assert[] that pay disparities are caused by nondiscriminatory factors. Such factors could include the employees' education, work experience with previous employers, seniority in the job, time in a particular salary grade, performance ratings, and others. The Commission will need accurate information about all the variables on which the employer relies, for each employee similarly situated to the charging party. The employer should be asked to provide and explain all of its reasons for a compensation differential to reduce the need for burdensome repetitive requests.")

[32] Sage Computing "Final Report: To Conduct a Pilot Study for How Compensation Earning Data Could Be Collected from Employers on EEO's Survey Collection Systems (EEO-1, EEO-4, EEO-5 Survey Reports) and Develop Burden Cost Estimates for Both EEOC and Respondents for Each of EEOC Surveys (EEO-1, EEO-4, and EEO-5), p. 16 (Sept. 2015).

[33] Id., at 16, emphasis added.

[34] Id., at 61, emphasis added.

[35] 29 U.S.C. 206(d) and 42 CFR 2000e-2(h).

[36] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 10.

[37] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 11.

[38] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 11.

[39] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 12.

[40] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 12.

[41] EEOC's Notice of Proposed Changes to the EEO-1 to Collect Pay Data from Certain Employers, available at http://www.eeoc.gov/employers/eeo1survey/2016_eeo-1_proposed_changes_qa.cfm.

[42] 81 Fed. Reg. 5117, fn. 45 (February 1, 2016).

[43] 81 Fed. Reg. 5117-5118 (February 1, 2016) (emphasis added).

[44] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 13.

[45] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraphs 13-15.

[46] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 14.

[47] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 15.

[48] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 15.

[49] 81 Fed. Reg. 5118, fn. 47 (February 1, 2016).

[50] 81 Fed. Reg. 5118, fn. 47 (February 1, 2016).

[51] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 17.

[52] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 17.

[53] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 17.

[54] EEOC pilot study on collecting pay data, available at:

pay-pilot-study.pdf

, at p. 27.

[55] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 19.

[56] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 19.

[57] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 20.

[58] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 20.

[59] Ex. 1, Dr. J. Michael DuMond Declaration at Paragraph 23.

[60] 41 CFR 60-2.17 (b) "Identification of problem areas. The contractor must perform in-depth analyses of its total employment process to determine whether and where impediments to equal employment opportunity exist. At a minimum the contractor must evaluate … (3) Compensation system(s) to determine whether there are gender-, race-, or ethnicity-based disparities."

[61] Draft Report pg 10, Question 6. Group 1

[62] Draft Report page 11 group 1

[63] Draft Report page 11, group 2.

[64] 44 U.S.C. 3510

[65] 44 U.S.C. 3510 (b)(1)(2)

[66] National Research Council, 2012, Collecting Compensation Data from Employers, Washington D.C., National Academies Press, p. 84, available at http://www.nap.edu/catalog/13496.

[67] Id. at 90.

[68] Id.

[69] Id. at 91.

[70] Id.

[71] The details of the EEOC's proposal are presented in the Federal Register, Vol. 81, No. 20, pp. 5113 - 5121.

[72] Federal Register, Vol. 81, No. 20, page 5117.

[73] These seven groups are: 1) Hispanic/Latino, 2) White (not Hispanic/Latino), 3) African-American (not Hispanic/Latino), 4) Native Hawaiian or Other Pacific Islander (not Hispanic/Latino), 5) Asian (not Hispanic/Latino), 6) American Indian or Alaskan Native (not Hispanic/Latino) and 7) Two or More Races (not Hispanic/Latino).

[74] Federal Register, Vol. 81, No. 20, pages 5115, and 5118.

[75] See "Employee Tenure in 2014"; News Release; Bureau of Labor Statistics, US Department of Labor; http://www.bls.gov/news.release/pdf/tenure.pdf

[76][76] Federal Register, Vol. 81, No. 20, page 5118, footnote 46.

[77] http://www.bls.gov/opub/reports/cps/characteristics-of-minimum-wage-workers-2014.pdf

[78] Sage Computing Pilot Study, September 2015, page 27. Emphasis added.

[79] A summary of these two simulations is provided within this declaration. The details of these simulations are included as Exhibit B.

[80] Federal Register, Vol. 81, No. 20, page 5118. The Proposal also suggests that the statistical tests would be used to "detect discrimination," implying that a statistically significant result from either a Mann-Whitney or Kruskal-Wallis would be sufficient evidence thereof. However, statistical evidence cannot be used to find discrimination-it can only identify outcomes that are consistent or inconsistent with discrimination.

[81] For this simulation, an employee's race was randomly assigned such that 12.5% of the employees were designated as African-American. The random nature of this assignment results in a similar job grade distribution of "white" and "African-American" employees.

[82] Although EEOC does not cite any source for its 3.2 hour burden increment, careful examination of the Sage (2015) report provides a clue. Sage reports on page 104 of its report that a 2012 EEOC working group forum on surveys modernization included some representatives of EEO-1 filers and that these filers on average opined that the addition of compensation data to the report would increase their reporting burden by "approximately 96 per cent. It is an interesting coincidence that 3.2 hours, EEOC's estimate of the incremental burden for component 1, is 96 percent of the 3.4 hours listed for the Component 1 (existing form) burden in EEOC's notice Table 4 on p.5120. A problem for EEOC, however, is that this 96% burden increase estimate is relative to the respondents' own current burden, not the EEOC estimate of 3.4 hours as contained in the ICR supporting documents that EEOC has filed with OIRA since 2009. Another important fact that is not presented is how many forum attendees were responsible for the 96 percent burden estimate and how statistically representative they were of the total filer universe. These are questions that should be important to an agency whose mission is closely linked with the importance of statistical data and analysis.

[83] These respondents often file multiple reports.

[84] Estimated burden hours were calculated by multiplying the number of reports expected to be filed annually (307,103 in 2013) by the estimated average time to complete and submit each report (3.4 hours), for a total of 1,044,150 hours. Relying on an estimate of $19 per hour results in a total cost of $19.83 million (1,044,150 burden hours multiplied by 19.00 per hour). The rate of $19 per hour is based on the hourly pay rate of human resources assistants of $18.22 (Occupational Employment Statistics, Occupational Employment and Wages, May 2010, 43-4161 Human Resources Assistants, Except Payroll and Timekeeping, https://www.bls.gov/oes/current/oes434161.htm, Last Modified Date: May 17, 2011, U.S. Bureau of Labor Statistics, Division of Occupational Employment Statistics). $18.22 was rounded to $19 to account for instances where higher paid staff perform this work.

Breadcrumb