U.S. Department of Labor

Office of Administrative Law Judges
90 Seventh Street, Suite 4-800
San Francisco, CA 94103-1516
(415) 625-2200
(415) 625-2201 (FAX)

Issue Date: 14 July 2017
CASE NO. 2017-OFC-00004
In the Matter of
OFFICE OF FEDERAL CONTRACT
COMPLIANCE PROGRAMS,
U.S. DEPARTMENT OF LABOR,
Plaintiff,
v.
GOOGLE INC.,
Defendant.

RECOMMENDED DECISION AND ORDER
I begin with an explanation of what this case is and what it is not. The Office of Federal
Contract Compliance Programs is the agency of the Department of Labor charged with auditing
government contractors to determine whether they are complying with certain contractuallyimposed anti-discrimination and affirmative action obligations.1 OFCCP’s auditing activities
generally are not “complaint-driven”; rather, OFCCP opens audits of federal contractors based
on neutral criteria.2 That is how OFCCP selected Google for this audit,3 not because any of the
more than 25,000 potentially affected employees (or anyone else) filed a complaint with
OFCCP.4
When OFCCP determines after an audit that a government contractor is discriminating or failing
to meet affirmative action obligations, it must try to resolve the violation voluntarily and without
litigation.5 According to OFCCP’s Regional Director (Pacific Region), these efforts lead to
voluntary resolutions of about 99 percent of OFCCP’s cases.6 But when there is no voluntary
1

See Tr. 32. “Tr.” refers to the hearing transcript. “J.Ex.” refers to the parties’ joint exhibits. “P.Ex.” refers to
Plaintiff OFCCP’s exhibits. “D.Ex.” refers to Defendant Google’s exhibits. “ALJ Ex.” refers to the ALJ’s exhibits.
2

Tr. 33:5. OFCCP also may initiate an audit after receiving a complaint, Tr. 33:23-24, but this is not what happened
in this case.
3

ALJ Ex. 1, ¶ 8.

4

Tr. 126-27.

5

Tr. 34.

6

Id.

 resolution, OFCCP brings an enforcement action before an administrative law judge at the U.S.
Department of Labor.7
This case is not an enforcement action after a finding of discrimination. OFCCP insists –
persuasively – that it has made no determination whether Google is complying with its antidiscrimination and affirmative action obligations.8 Rather, OFCCP is auditing Google’s
headquarters operations, an endeavor that OFCCP began in September 2015 and has not
concluded.9 No one – including OFCCP – knows whether the audit will reveal a violation of
Google’s non-discrimination or affirmative action obligations.
For nearly a year, the parties were able to agree on extensive information and documentation that
Google would give OFCCP for this audit. Google also made available a number of managers for
OFCCP investigators to interview.10 The amount of information involved now makes this the
largest ongoing audit in OFCCP’s Pacific Region and one of the ten largest that OFCCP has
conducted.11
But the process halted when, in June 2016, OFCCP requested in two letters a large amount of
additional information and materials. Google agreed to produce some of what OFCCP requested
and protested the remainder, raising various legal challenges.12 The parties compromised on
several of the disputed requests but reached an impasse as to three of them.13 I will discuss the
three disputed requests in the Findings of Fact below.
With the parties at an impasse, OFCCP initiated this action to require Google to provide its
information and materials responsive to the three requests. OFCCP seeks only an order requiring
Google to provide this information. OFCCP seeks no finding that Google has engaged in
employment discrimination or has failed to meet its affirmative action obligations. It seeks no
penalties and agrees that Google’s good faith is not at issue.14

7

Id.

8

See Order denying Google’s motion to dismiss (May 2, 2017).

9

See J.Ex. 5 (scheduling letter opening audit).

10

Tr. 67. According to OFCCP, on April 27 and 28, 2016, it conducted onsite interviews of Google’s “HR
personnel, compensation director, recruiter, [and] hiring managers.” Id.
11

Tr. 40. OFCCP’s Pacific Region is one of six regions nationally. It covers California, Washington, Oregon,
Alaska, Hawaii, Idaho, Nevada, and Arizona.
12

J.Ex. 6, 7.

13

J.Ex. 8-10. For purposes of this hearing, Google does not assert a defense that OCFFP failed to conciliate about
the scope of its requests. ALJ Ex. 1, ¶ 28.
14

Tr. 7. In its complaint, OFCCP drafted a “Prayer for Relief” more expansive that what it now seeks. Compare
Complaint at 5-6 with Tr. 7. Penalties could become appropriate – after another complaint and hearing – if Google
were to fail to comply with any final order requiring it to produce further materials or information.

2

 I will order Google to provide only some, but not all, of what OFCCP seeks.15 I set out the
procedural history in the margin.16
Findings of Fact
Google’s obligation to comply with OFCCP’s audit requests. Google’s obligation to submit to
the OFCCP audit arose on June 2, 2014, when the General Services Administration accepted
Google’s bid on a contract for “Advertising and Marketing Solutions” (“the AIMS contract”).17
Google admits that, as a result of this contract award, it must comply with Executive Order
11246.18
15

The parties agree that the administrative law judge is not limited to an order either granting OFCCP all of the
relief it seeks or denying all of that relief. On the contrary, the ALJ may order Google to provide some, but not all,
of the materials OFCCP seeks (i.e., “blue pencil” the requests). Tr. 6-7; briefs filed by each party on Mar. 8, 2017.
16

OFCCP filed its complaint on January 4, 2017. It was assigned to the undersigned ALJ on January 13, 2017.
With an agreed extension of time, Google filed its answer on January 24, 2017. On February 21, 2017, I granted
OFCCP’s motion to expedite under 41 C.F.R. §§ 60-30.31, et seq. Expedited case handling narrowly restricts
discovery to requests for admission plus, if the ALJ permits, depositions. I allowed Google one specific, narrowly
circumscribed deposition. Each party propounded one set of requests for admission, which the other party
answered. There was no other discovery permitted or undertaken.
Also on February 21, 2017, I noticed the hearing for March 10, 2017, in San Francisco. On February 27, 2017, the
parties jointly requested a continuance, which I allowed, eventually resetting the hearing for April 7, 2017. The
parties stipulated to this date knowing that, if the hearing did not conclude in one day, there would be an
adjournment for some time until it could be resumed.
On March 15, 2017, I issued a written order denying OFCCP’s motion for summary decision. Thereafter, the parties
submitted required pre-hearing materials, including 32 written stipulations submitted on March 28, 2017 (ALJ Exhs.
1 and 2). I conducted a telephonic pre-hearing conference on the record on April 5, 2017.
The hearing began as scheduled on April 7, 2017. Owing to extraordinary events, I adjourned the hearing midafternoon to allow Google time to file a motion to dismiss. I denied that motion on May 2, 2017, and resumed the
hearing by stipulation on May 26, 2017. The hearing concluded on that day. Both parties offered closing oral
argument on the record.
Over the two hearing days, OFCCP called as witnesses its Regional Director Janette Wipper, its Deputy Regional
Director Jane Suhr, and contractor Michael Brunetti, Ph.D., who is an economist and statistician. Google called its
Vice President of Compensation Frank Wagner and Senior Legal Operations Manager Kristin Zrmhal. The parties
offered Joint Exhibits 1-16, all of which I admitted. Tr. 12. OFCCP initially offered exhibits numbered 201-223.
After Google moved to seal certain exhibits, OFCCP withdrew its exhibits 221 and 222. It withdrew exhibit 218
except for pages 15-17, 145-51, 158-60, 170-72, 196-222, and 236-47. Without objection, I admitted OFCCP’s
remaining exhibits. Google withdrew before the hearing certain exhibits rendered irrelevant after the parties reached
stipulations at the pre-hearing conference. Google offered its remaining exhibits: 101-02, 106-11, 113, 118, and
121-22. Without objection, I admitted each of these exhibits.
After the hearing, the parties requested that I allow them time to file simultaneous briefs on the merits after the
transcript became available. The regulations for expedited hearings are to the contrary, see 41 C.F.R. § 60-30.35,
but by this point, in my view, the case was no longer on an expedited schedule. I therefore allowed the parties the
time they requested, and both submitted closing briefs.
17

J.Ex. 1-4.

18

P.Ex. 223. Although OFCCP has authority to audit government contractors for compliance with section 503 of
the Rehabilitation Act of 1973, 29 U.S.C. § 793, and the Vietnam Era Veterans’ Readjustment Assistance Act of
1974, 38 U.S.C. §§ 4211 and 4212, OFCCP has not offered any evidence or argument that relies on these statutes.
Nothing about OFCCP’s evidence or argument concerns disabled workers (whom section 503 protects) or Vietnam

3

 In particular, under the AIMS contract, Google agreed to “comply with Executive Order 11246,
as amended, and the rules, regulations, and orders of the Secretary of Labor” and to
permit the Government to inspect and copy any books, accounts, records
(including computerized records), and other material that may be relevant to the
matter under investigation and pertinent to compliance with Executive Order
11246, as amended, and rules and regulations that implement the Executive
Order.
ALJ Ex. 1, ¶ 5.
Consistent with that obligation, Google produced information and documentation responsive to
demands that OFCCP made at the time it notified Google of the audit (in a “scheduling letter”)
on September 20, 2015; Google also produced considerable further information in response to
OFCCP’s follow-up demand on June 1, 2016, but it did not produce everything OFCCP
requested.19
To collect and produce the information that it did make available to OFCCP, Google initially
relied on its own employees in a litigation support unit. When these employees were too few to
handle the task, Google brought in an outside contractor to collect and review information and
data responsive to OFCCP’s requests. Google also asked outside attorneys for advice and
counsel.
In all, Google expended about 2,300 person hours on these tasks. Nearly half of the time was
spent on attorneys who reviewed the materials to be certain that private employee information
was properly protected. One of the complicating factors was that, to protect employee privacy,
Google stores much information about individual applicants and employees in different locations
and in ways that are intentionally difficult to access. Google engineers had to develop tools to
access the information OFCCP wanted.
Google produced nearly 1.3 million data points about its applicant flow; 400,000 to 500,000 data
points on compensation; and 329,000 documents, totaling about 740,000 pages.20 The project
cost Google about $500,000, a significant amount when compared to the $600,000 gross total
that GSA paid Google under the AIMS contract from June 2014 through December 29, 2016.21
An OFCCP compliance review must evaluate, not only the contractor’s hiring and employment
practices, but also “the written affirmative action program, and the results of the affirmative
action efforts undertaken by the contractor.” 41 C.F.R. § 60-1.20(a)(1). Thus, to complete the
or other veterans. I therefore focus the analysis on Executive Order 11246 and its implementing regulations at 41
C.F.R. Part 60.
19

See demands in J.Ex. 5, 6.

20

A data point is the same as a cell on an Excel spreadsheet. The description of Google’s production efforts and
results as set out in the text above is based on hearing testimony. See Tr. 264-75.
21

See Tr. 6, 275.

4

 affirmative action aspects of a compliance review, OFCCP’s audit generally extends to the
employees at the “establishment” defined in the contractor’s affirmative action plan.22
A central reason that the information OFCCP sought was so extensive (and required so much
effort to produce) is that the Google employees who worked on the AIMS contract were located
at Google’s headquarters at 1600 Amphitheatre Parkway in Mountain View, California. Thus,
for purposes of its affirmative action plan, Google took its headquarters as the relevant
“establishment,” meaning that the affirmative action plan covered headquarters employees.23 In
September 2015, when OFCCP issued its “scheduling letter,” Google’s headquarters employees
within the affirmative action plan numbered more than 21,000. This explains how the Google
compliance review became the largest in the OFCCP’s Pacific Region.24
Background – the relevant portions of Google’s production to date. OFCCP has narrowed its
requests for purposes of this litigation. Although one focus of the audit has been Google’s hiring
practices (“applicant flow”), OFCCP does not seek applicant flow data. Similarly, OFCCP has
not pursued information focused on any requirements in Google’s affirmative action plan.25
22

See Tr. 35.

23

See 41 C.F.R. § 60-1.40.

24

Had OFCCP approved, Google could have adopted a “functional affirmative action plan,” rather than a plan
defined by geographical workplace “establishment.” Id. But that might well have accomplished nothing.
Considering that Google has a very large number of software engineers across many locations and that software
engineers were involved in the AIMS contract, there is no evidence that adopting a functional AAP would have
resulted in a narrower OFCCP audit. Tr. 121:12-122:12.
OFCCP’s Regional Director also asserted that Google could have created several smaller affirmative action plans by
obtaining a variety of different addresses at its Mountain View facility and using each address as a separate
establishment with a separate affirmative action plan. This neglects that Google has many times represented – in
documents that must be accurate – that its headquarters is the facility at 1600 Amphitheatre Parkway, not a variety
of addresses. See, e.g., Google’s Form 10-Ks filed with the Securities and Exchange Commission, on which Google
assures accuracy to the SEC and investors. P.Ex. 201, 212. In any event, Google’s affirmative action plan is what it
is; the reason is not relevant to the issues pending here.
25

In their prehearing exhibit lists, neither party identified the affirmative action plan as an exhibit they would offer.
At the prehearing conference, the parties confirmed that they did not intend to offer the plan as an exhibit. I
explained that, without the plan, there would be no evidence of what steps the plan requires of Google and thus no
way to know whether requests OFCCP might make would be relevant to whether Google was complying with the
plan. (It might even be possible that the affirmative action plan placed no requirements on Google other than
monitoring its own compliance.) I asked OFCCP if it asserted that some of the information and documents it was
requesting concerned Google’s compliance with the affirmative action plan, or whether OFCCP’s focus was entirely
on the Executive Order’s non-discrimination provision.
OFCCP’s response was that the “case primarily goes to a non-discrimination analysis that OFCCP is undertaking.
And basically the analysis is whether or not Google pays its employees in a non-discriminatory fashion. So it is
going to that broader issue of is Google complying with its non-discrimination obligation” (emphasis added).
This answer was evasive because OFCCP’s counsel would say only that this was “primarily” OFCCP’s focus. I
asked for clarification. OFCCP did not clarify. It said only that the affirmative action plan and non-discrimination
“go hand-in-hand,” a meaningless response. OFCCP added, however: “That said, as we know, the AAP isn’t part
of the exhibit list and, you know, our intent is to make our case without it.” (See Tr. 43:24-48:20.)
I then issued the following ruling: “If you don’t put on any evidence of what the affirmative action plan requires, I
will not find that the requested information relates to the affirmative action plan. Prehearing Tr. 48:16-20.

5

 From its pre-hearing statements, it appears that OFCCP’s focus is on compensation
discrimination, and from OFCCP’s evidence it appears that OFCCP is especially concerned
about indications that Google compensates women less than similarly situated men. As OFCCP
has explained, and I discuss below, an indication of a disparity in compensation is not a finding
of unlawful discrimination; rather, it is part of OFCCP’s explanation for requesting more
information to determine whether unlawful discrimination is occurring.
Given OFCCP’s focus, I need address as background only one major collection of data that
Google has already produced: a “snapshot” of employees as of September 1, 2015.26 The term
“snapshot,” while apropos, could be misleading. It is a snapshot in the sense that it looks at data
about Google’s employees as of a moment frozen in time (September 1, 2015). But it is far from
a simple, 3” x 3” photograph.
Initially, the snapshot required Google to collect, organize, and produce data in about 14
categories for each of the approximately 21,114 people Google employed at its headquarters on
September 1, 2015. Following guidelines the Office of Management and Budget approved after
considering the burden it would place on contractors,27 OFCCP initially requested for each
employee28 on the snapshot the following: gender, “race/ethnicity,” hire date, job title, EEO-1
category (such as clerical or executive), job group, base salary or wage rate, hours worked in a
typical workweek, and other compensation or adjustments to salary (bonuses, incentives,
commissions, merit increases, locality pay, and overtime).29 Google was to produce
documentation and policies related to compensation practices, “particularly those that explain the
factors and reasoning used to determine compensation.”30 Google was invited to submit
additional data related to compensation, “such as education, past experience, duty location,
performance ratings, department or function, and salary level/band/ range/grade.”31 Google
produced a responsive “snapshot” in November 2015.32
On June 1, 2016, OFCCP went beyond the OMB-approved list to request Google to supplement
the snapshot with data in the following categories: name, date of birth, bonus earned, bonus
At the hearing, OFCCP had a copy of the affirmative action plan with it. It has marked the plan with an exhibit
number. It stated, however, that – at least initially – it would not offer the plan as an exhibit because “it may not be
necessary.” Tr. 12:15-22. OFCCP never offered the affirmative action plan as an exhibit, and it is not on the record.
I therefore find that OFCCP abandoned any argument that it is entitled to any of the requested information or
materials because they are relevant to Google’s compliance with its affirmative action plan or the results of action
Google has taken under its plan. That leaves only the Executive Order’s non-discrimination provision within the
scope of OFCCP’s investigation for purposes of this case.
26

These are the employees within the scope of Google’s affirmative action plan for its headquarters facility in
Mountain View.
27

Tr. 39.

28

This includes full-time, part-time, contract, per diem, day labor, and temporary employees.

29

J.Ex. 5 at 5-006 (Item 19).

30

Id.

31

Id.; Tr. 39:2-5.

32

J.Ex. 12 at 12-017.

6

 period covered, campus hire or industry hire (i.e., hired directly out of school or from another
employer), whether the employee had a competing offer, current “CompaRatio,” 33 current job
code, current job family, current level, current manager, current organization, department hired
into, education, equity adjustment, hiring manager, job history, locality, long-term incentive
eligibility and grants, market reference point, market target, performance rating for past 3 years,
prior experience, prior salary, referral bonus, salary history, short-term incentive eligibility and
grants, starting CompaRatio, starting job code, starting job family, starting level, starting
organization, starting position/title, starting salary, stock monetary value at award date, target
bonus, and total cash compensation, and “any other factors related to Compensation.”34 Google
produced much of this additional data between August 2016 and February 2017.35
On September 19, 2016, OFCCP asked for “snapshot” data in yet additional categories: the
employees’ ID,36 country of citizenship, secondary country of citizenship, visa (yes/no), visa
type, and place of birth. J.Ex. 12 at 12-013. Google produced this data around February 1, 2017.
ALJ Ex. 1, ¶ 30.
In all, Google states that it has produced 844,560 compensation data points for the 21,114
employees on the snapshot for September 1, 2015.37
OFCCP’s currently requested additional documents and information. In this litigation, OFCCP
seeks an order requiring Google to provide data falling into the following three categories:


A “snapshot” as of September 1, 2014 – a year earlier than the first snapshot. The 2014
snapshot must address each of the 19,539 people Google employed at its headquarters on that
date. It must contain the same categories of data as did the snapshot as of September 1,
2015, including all those added on June 1, 2016.38



A salary history (a list of starting salary and each salary change) and job history (a list of
starting job and each change in job) for each person whom Google employed at its
headquarters on either of the two snapshot dates.39 The histories must cover the entire time
Google employed each person, going back for its longest-term employees to the founding of
the corporation in 1998.

33

Compa-ratio is the ratio between the employee’s salary and the market reference point for the job the employee
holds.
34

J.Ex. 6 at 6-002, 003.

35

ALJ Ex. 1, ¶¶ 18, 29, 30.

36

“Employees’ ID” is undefined on the record.

37

J.Ex. 12 at 12-017.

38

J.Ex. 6 at 6-003; ALJ Ex. 1, ¶ 14.a. OFCCP has withdrawn the category for “any other factors related to
Compensation.” Prehearing Tr. at 50.
39

ALJ Ex. 1, ¶ 14.b; Tr. 43:6-10, 46:5-16.

7

 

The name, address, telephone number, and personal email of every employee reflected on
either the 2014 snapshot or the 2015 snapshot.40

Evidence going to OFCCP’s authority. The AIMS contract is a contract of $100,000 or more.
At all times since the AIMS contract was formed, Google had 50 or more employees. ALJ Ex. 1,
¶¶ 2, 3.41 But as will become relevant in the discussion below, the record is silent as to when
Google first had 50 employees or even had the 15 employees necessary for the antidiscrimination provisions of Title VII of the Civil Rights Act of 1964 to apply.42
Google was first a government contractor in 2007. Tr. 64-65. It was a government contractor at
various unspecified times between 2007 and 2012.43 But the evidence of record does not
establish that Google was a government contractor or subcontractor in 2013 or any of 2014 prior
to being awarded the AIMS contract.44
Evidence going to relevance of the requests. When auditing compensation practices, OFCCP
follows the principles underlying Title VII of the Civil Rights Act of 1964. Tr. 160-61. It has
40

ALJ Ex. 1, ¶ 14.c; Prehearing Tr. at 50.

41

Google made similar (and broader) admissions when answering OFCCP’s requests for admission. P.Ex. 223.

42

See 42 U.S.C. § 2000e(b).

43

I notified the parties at the hearing on April 7, 2017, that I found it potentially relevant whether Google had other
government contracts immediately before the June 2, 2014 contract. See Tr. 110:18 to 111:22. If the parties were
unaware of the potential relevance before then, they had time to gather any needed evidence together before the
hearing resumed on May 26, 2017. Both Google and OFCCP should have had ready access to evidence of former
government contracts with Google.
At the hearing, OFCCP Deputy Regional Director (Pacific Region) Jane Suhr testified that OFCCP conducted audits
of Google in 2007, 2010, 2011, and 2012, and that OFCCP does not begin an audit without being certain that the
party being audited is a federal contractor. Tr. 65:14-66:4. Suhr thus concluded, and I accept, that Google was a
federal contractor at certain unspecified times during those four years. See also, P.Ex. 204-209.
44

Suhr testified on direct examination that Google was a government contractor immediately before GSA awarded it
the AIMS contract on June 2, 2014. Tr. 66:6-21. As Suhr initially characterized this testimony as a guess (Tr. 66:912), defense counsel followed up on cross. Suhr admitted that she didn’t really know if Google had some other
government contract immediately before GSA gave it the AIMS contract. Tr. 87:25-88:19. I therefore reject Suhr’s
direct testimony on this point because she recanted it on cross. The record contains no evidence of any other
contract between a government agency and Google in 2013 or before the AIMS contract in 2014.
The closest evidence falls short of establishing a government contract or subcontract. In particular, the parties
stipulated that “in 2014 and 2015, Google has received approximately $30 million from its licensing contracts with
distributors and resellers as a result of the resale of Google’s technology products by the distributor/reseller to the
federal government” (emphasis added). Nothing on the records shows (or even suggests) that these licensing
agreements were government contracts or that they required Google to comply with Executive Order 11246 or with
OFCCP’s requests for information. The stipulated facts are irrelevant.
Moreover, the stipulation does not state that any of the payments to Google occurred before June 2, 2014, when the
AIMS contract was formed. The stipulation states only that there were payments (at unidentified times) in 2014 and
2015. Those could be co-extensive with the current contract; they could even have begun after the current contract
was formed.
In all, there is no evidence of Google’s being a government contractor or subcontractor after sometime in 2012 until
it was awarded the AIMS contract on June 2, 2014.

8

 published procedures for these audits. See Directive 307 (Feb. 28, 2013).45 According to its
Regional Director, OFCCP defines the “review period” for audits as running back in time two
years from the date OFCCP issues the “scheduling letter” that initiates the audit. Tr. 35-36. For
this case, that defines the review period as September 21, 2013, through September 20, 2015.46
OFCCP contends that its current requests are relevant to Google’s pay practices at its
headquarters in Mountain View during this review period. Tr. 37-38, 41-44; J.Ex. 5, Item 19.
OFCCP states that it analyzed the “snapshot” Google submitted for September 1, 2015; this
included multiple regression analyses. Tr. 47-48. It found “systemic compensation disparities
against women pretty much across the entire workforce.” Tr. 48. If OFCCP is correct, this
means that its analysis showed a statistical correlation between wage disparity and sex.47
OFCCP contends that, given the disparity as of the September 2015 snapshot, it needs data for a
second point in time to determine whether the disparity was ongoing throughout the two-year
review period. That is why OFCCP requested the September 2014 snapshot. Tr. 48-49.
In addition, as Director Wipper explained:
Usually after you find a disparity in pay level, the second question you want to
answer is the cause of the disparity. So, what you’ll do is look back at every
decision that impacted pay, from starting salary to every change going forward.
This is something that not only we do, but Google itself says it does when they do
pay equity analysis.
Tr. 41. One of the reasons OFCCP wants to know the cause of the disparity is that it could
reveal a way to correct the disparity.48 Tr. 48. To evaluate every decision that impacted pay,
OFCCP argues, requires it to look back before the review period to determine the cause of the
pay disparity.49 Tr. 46-47. For example, as OFCCP contends, it might have to look back to the
time of hire if there are indications that a practice at that time is the root cause of a pay disparity
that has persisted into the review period. Tr. 46-47.
Here, OFCCP believes that it has found indications that Google’s practices at the time of hire –
at least as to some of its employees – could well be the cause of the sex-based pay disparity
shown in the September 2015 snapshot. These indications especially – though not exclusively –
relate to occasions when Google hires a person who has been out of school and working
45

Tr. 36-37; https://www.dol.gov/ofccp/regs/compliance/directives/dir307.htm.

46

As Director Wipper explained, OFCCP might also look forward, past the date of the scheduling order, to
determine if violations found within the review period have been corrected. Tr. 36. If not, OFCCP might seek
remedies for those continuing violations. Id.
47

To meet OFCCP’s internal standards, the disparity must be statistically significant, i.e., at least two standard
deviations, consistent with principles applied in Title VII adverse impact cases. See OFCCP Directive 307 n.8.
OFCCP offered no evidence going to whether the disparity it found was statistically significant.
48

As I discuss below, the cause of any statistical disparity likely also is relevant to liability.

49

OFCCP concedes that it would not seek back wages or remedies for conduct before the review period. It gathers
pre-review period data only as needed to understand the cause of disparities occurring within the review period.

9

 elsewhere in the industry. Tr. 42-44. Google refers to these as “industry hires,” as opposed to
“campus hires,” whom Google employs directly out of schools. I will use the same terms. I
therefore turn to the evidence of how Google sets starting compensation at the time of hire as
well as how Google adjusts compensation for personnel already in its employ.
Google’s compensation policies and practices. Google states that it has developed compensation
practices aimed at: (1) attracting and retaining “the world’s best talent” by targeting pay “well
above the market”; (2) fostering innovation by linking salary increases, bonuses, and stock grants
to performance; and (3) sharing profits with employees through bonuses and stock grants. Tr.
167-68. As Google’s Vice President of Compensation describes the company’s practices, it
would seem that compensation does relate to these factors.
Google has followed a complex and carefully conceived compensation plan since at least April
2007, when it hired its current Vice President of Compensation. Tr. 165-66, 255-56. For most
employees, compensation includes on an annual basis: salary, equity grants, and bonus.50 New
hires may also receive a sign-on bonus and a one-time equity (stock) grant.51 Tr. 218-21.
For each job in its workforce,52 Google researches compensation rates in the local market. It
determines for each job a “market reference point” that corresponds to a particular percentile on
the local market survey.53 Tr. 169-73, 174-75. Google attempts to bring in all new hires at a
starting salary at a particular percentage of the market reference point for that job regardless of
other factors.54 Id. Google does not offer less than this percentage of the market reference point,
but under some circumstances, it offers more. Tr. 172-73.
A “compensation team” sets the salary for each hire. Tr. 165-69. The team has no direct contact
with the applicant and does not have the applicant’s name, gender, race, or ethnicity. Id. For
industry hires, the committee might be given the applicant’s current compensation but no earlier
compensation data. Tr. 175-77.
For the employees included in the September 2015 “snapshot,” about 20 percent were campus
hires. Tr. 197-98. The only circumstance under which Google might increase the starting
compensation for a campus hire is when the applicant has a competing offer greater Google’s.
Google will not offer a larger salary or annual bonus plan, but it might offer a larger sign-on
bonus or one-time stock grant. Tr. 207, 210-16, 223. There are no starting pay negotiations, and
Google considers no other factors. Tr. 197-98.

50

Certain employees receive a sales bonus plan rather than the company bonus plan most employees get. The sales
bonuses are paid quarterly. Tr. 212-13.
51

Equity grants vest over time.

52

The record does not address Google’s top 200 executives. Tr. 169. In addition, new hires in certain support roles
(e.g., IT help desk) all get the same starting pay. Tr. 198.
53

The same percentile level is used for new hires on all jobs across-the-board.

54

The same percent of the market reference point is used as a goal for new hires on all jobs across-the-board.

10

 Google refers to people hired with relevant industry experience (i.e., not directly out of school)
as “industry hires.” Many industry hires ask to negotiate a starting package, but Google will not
negotiate in any traditional sense. It might increase its offer, but only if the applicant presents
either: (1) a higher competing offer, or (2) a greater current compensation at the job the
applicant would be leaving.55 That describes about 10 to 15 percent of industry hires. Tr. 199200.
To attract applicants in this limited group, Google might offer more salary but tries instead to
emphasize stock grants. Tr. 200-05, 226-31. The magnitude of the applicant’s current salary as
well as the applicant’s education and prior job experience could affect the decision to offer a
larger starting salary, but a starting offer at full market reference point probably occurs in no
more than 0.5 percent of cases. Tr. 226-31. Again, it is the compensation team that modifies the
offer without meeting with the applicant, having discussions with the applicant, or knowing any
identifying information about the applicant (including name, sex, race, ethnicity, national origin,
etc.)
Promotions. Google has a similar – but not identical – practice for setting compensation after a
promotion. No negotiation is involved. The salary component for the promoted job begins as a
specified percentage of the market reference point for that job.56 Tr. 200-04. That will be the
starting salary in the new job except for a single consideration that applies in a limited number of
cases. Specifically, there is an adjustment if the new salary would give the promoted employee
less than a 5 percent raise or more than a 20 percent raise. Id. If so, the raise will be increased to
meet the 5 percent minimum or decreased to limit it to the 20 percent maximum. Id. The only
remaining factor is that the manager involved may make further adjustments, but only for the
purpose of bringing the promoted employee into pay “alignment” with their peers. Id.
Google states that about 80 percent of promoted employees begin with the standard salary equal
to the set percentage of the market reference point. Id. In over 90 percent of the occasions when
managers adjust the starting salary for the promoted job, the adjustment is no greater than one
percent. Id.57
Annual merit increases. Employees are eligible for annual merit salary increases. Tr. 233-34.
Increases depend mostly on two factors: the employee’s performance rating and the percentage
of the market reference point at which the employee is currently being paid (before the increase).
Id.; Tr. 55, 77; P.Ex. 216 at 83.

55

Tr. 172-73, 199-200, 202; 223 (for industry hires only competing offer or higher current compensation affect the
initial equity grant), 226-31.
56

Across-the-board, the percentage of market reference point for promoted jobs is higher than the rate Google would
give a new hire in the same job. Id. The point is to slot in the promoted employee in proper “alignment” with her
peers; that is, more than a newly starting Google employee. Id.
57

OFCCP argues in its closing brief that Google makes promotions available to employees annually. There is
nothing on the record to support this or any other particular frequency of promotion opportunities for individual
employees.

11

 The policy is intricately designed to bring people on the same job with the same job performance
rating to the same salary over time. Tr. 245-46, 251-52. To achieve this, the merit increases are
less for the same performance if the employee is already being paid a higher percentage of the
market reference point. Id. For example, if two employees are in the same job and one of them
is being paid at 85 percent of the market reference point while the other is being paid at 80
percent, and if both employees receive a “meets expectation” rating, the higher paid employee
will get a smaller merit increase than the lower paid employee. Id. The result is that salaries for
two similarly performing employees converge more closely with each year’s merit increase. Id.;
P.Ex. 216 at 83.
Managers can reduce the time for convergence by making discretionary adjustments. Tr. 239-40.
Again, managers are directed to do this for the purpose of bringing workers into better
“alignment” with their peers who have similar performance. Id.; Tr. 42-44. But Google
acknowledges, for example, that a discretionary adjustment might also help Google retain highperforming employees even if that is not the stated purpose. Tr. 240.
Stock grants. The starting stock grant is a one-time event at hire. Tr. 220-21. After that, stock
grants depend on job performance and the employee’s grant history. Id.
OFCCP’s rational for its requests: Salary and job histories. OFCCP Regional Director Wipper
explained that OFCCP needs data going back to 1998 because research shows that (1) women
don’t negotiate as well as men, and (2) if starting salaries are negotiated, female employees
remain behind the better negotiators for their entire career.58 Tr. 157-58. Director Wipper
contends that, because Google pegs merit increases to market percentages, this carries forward
the adverse impact from the negotiation of staring salaries year-after-year so long as the
employee works at Google. Id. According to Wipper, this suggests that a consideration of prior
pay has led to “anchoring,” or an ongoing weight keeping women’s pay less, and that OFCCP
routinely seeks a complete salary history in such instances. Tr. 130-35.
More generally, Wipper asserts that OFCCP needs salary and job histories going back as much
as nineteen years because OFCCP must look at every decision that impacted pay. Tr. 40-42. For
example, managers are allowed discretion in merit increases and promotion pay, and this could
be the cause of sex-based pay disparity. Id.59

58

In its pre-trial disclosures of exhibits, in its exhibit list, and at the hearing, OFCCP offered no academic literature
or other research to support Wipper’s contention about how women’s negotiating skills adversely affects their
starting pay. In its closing brief, OFCCP cites an article at slate.com and a Washington Post article. P. Closing Br.
at 20 n. 12. This comes too late for consideration; it should have been disclosed to Google pre-hearing and should
have been offered at the hearing. This would have allowed Google to submit opposing material and to object to
OFCCP’s offers. Footnote 12 is stricken from OFCCP’s closing brief. I similarly strike articles that OFCCP cites at
footnotes 9, 13, and 14. OFCCP must try its case at the hearing, not by footnoting unsubmitted evidence in its
closing brief. OFCCP’s citation in footnote 15 to a National Labor Relations Act charge is stricken as irrelevant. A
charge is merely an allegation, and whether Google has engaged in anti-union activity is not a reason to order it to
produce what OFCCP seeks in this litigation.
59

Wipper states as well that Google also considers all this historical data when it evaluates its own pay equity. Id.

12

 The requested “snapshot” as of September 1, 2014. Director Wipper explained that, while
OFCCP generally begins with a single “snapshot,” it inquires further when that data shows a
disparity in compensation. Tr. 40-42. OFCCP then seeks an earlier “snapshot” to determine
whether the disparity also existed at that time, which would suggest a continuing violation. Id.
The point is to cover the full review period. Tr. 47-48. Wipper also explained that OFCCP
would have narrowed its requests but for the fact that the indications of sex-based pay disparity
were so widespread across Google’s headquarters. Tr. 130-35.
Employee contact information. Director Wipper explained that OFCCP asks for contact
information for all employees to get the employees’ perspective on Google’s employment
practices without Google’s knowing who OFCCP interviewed. Tr. 57-58. She views this as
consistent with the informant’s privilege and asserts that it protects employees against retaliation.
Id. She testified that the regulations require OFCCP to keep confidential all information
provided in a compliance review. Tr. 59-60. She said she was unaware of any data breaches at
OFCCP. Id.
As to managerial employees, Wipper stated that it would only ask managers “about their specific
experiences with potential discrimination. We wouldn’t be asking anything about confidential
information. And we probably would notify Google before we contacted the managers.” Tr. 4142 (emphasis added).
OFCCP concedes that, if Google does not have requested contact information, it need not try to
obtain it (such as from the employee). Tr. 89-90.
Google’s economic capacity to comply with OFCCP’s requests. OFCCP offered evidence to
show that Google is economically easily capable of complying with OFCCP’s requests; the cost
will be insignificant compared to Google’s resources. Google’s 10-K filings with the Securities
and Exchange Commission show operating income in 2015 of $19.36 billion and cash of
$16.549 billion. In 2016, Google was a subsidiary of Alphabet, Inc., but Google’s operating
income was $27.89 billion. In the opinion of OFCCP’s expert, a cost of $1 million for Google to
comply with OFCCP’s demands would have no meaningful impact on Google’s business. See
Tr. 99-104; P.Ex. 201, 212. The expert admitted on cross-examination that he did not know how
burdensome it would be for Google to comply. Tr. 104-05. More to the point, he admitted that
money alone does not correlate to burden. Id.
Director Wipper testified that OFCCP considers the burden that it is placing on contractors. Tr.
130-35. Its initial requests extended to applicant flow data, which OFCCP recognizes can be
more burdensome because of how the materials are stored. Id. But OFCCP expects that it is less
burdensome to gather compensation data because it is typically stored centrally and
electronically. Id. Also, there typically are far more applicants in a routine hiring case than there
are employees in a routine compensation case. Id.
Google offered testimony to show the expense and intrusion into its business operations
associated with the materials and interviews it has supplied. Tr. 264-77. I discuss much of this
in the introduction to this Decision. To estimate the cost of compliance with OFCCP’s current
request, Google’s Senior Legal Operations Manager collected the views of the different teams

13

 that would be involved. Tr. 277-78. Given their experience responding to OFCCP’s earlier
requests, they estimated that the additional compliance would take 400 to 500 person hours just
to collect the information. This would cost as much as $100,000.60 Then the information would
have to be reviewed to redact private employee information and be reviewed by outside counsel
before production. The cost of compliance with the previous requests (including attorney’s fees)
was just under $500,000. Tr. 273-75.
Credibility, weighing the evidence, and findings re: OFCCP’s misconceptions about Google’s
practices. I find Wagner’s testimony credible. It was detailed, consistent, and not contradicted
on cross-examination. It reflected Wagner’s extensive, personal knowledge of the matters under
discussion that he has derived from his work in Google’s executive leadership on compensation
issues over the past ten years. Wagner understands Google’s compensation policies and
practices in a way that would be difficult for OFCCP to achieve at this point in its investigation.
Deputy Regional Director Suhr’s testimony lacked the foundation and understanding I observed
from Wagner. For example:


Suhr said she relied on Wagner for her conclusion that salary history and job history were
relevant to compensation. Tr. 77. From interviewing Wagner, Suhr concluded that complete
salary histories were relevant “because every time there’s a salary change or promotion or
merit increase, that person’s market target is looked at. So every time there’s a change,
you’re looking at what the person is making compared to the market at that time.” Tr. 78.
That might explain the relevance of market data, but it does not explain the relevance to
OFCCP’s audit of employees’ salary histories at Google.



On job history, Suhr recalled Wagner’s telling her that this was relevant to compensation, but
she couldn’t recall whether he limited that to the job the person held immediately before or
instead extended to that person’s entire job history. Tr. 74-75. It thus cannot be a basis to
find the entire job history relevant.

60

The Senior Legal Operations Manager Zrmhal has worked in Google’s legal department since 2011 and has
relevant experience dating back to 2004. Tr. 261-64. She has an MBA from the University of California
(Berkeley). She currently manages a team of 21 project managers and technologists who collect the data and
documents, analyze them, make them available to reviewing attorneys, and then produce them. She sets an annual
budget for litigation support, including for individual cases. In her six years at Google, she has managed or
supervised over 500 matters.
On cross-examination, OFCCP showed that Zrmhal’s estimate was based on the opinions of her team members who
would be doing the work. For OFCCP’s particular requests, she personally does not know where the needed
information is stored or how many hours it will take to extract it. Instead, Zrmhal summarized her conversations
with the people who gave her estimates they pulled from multiple systems. This included time that would be
required to write scripts to access personnel information, including historical information that had “migrated” over
the years. The scripts will have to be tested to verify that they collect all of the required information.
I accept Zrmhal’s estimate. Given the actual amount expended on OFCCP’s prior requests, she and her staff had a
basis for their estimates, they reached an estimate less than for the earlier productions, and Google offers this as no
more than an estimate. OFCCP offered no alternative evidence of what burden its requests will impose. I find
Zrmhal’s process sufficient and accept the estimate as fair and with adequate foundation.

14

 

Suhr testified about what Wagner and one of Wagner’s team members (Mr. Nambiar) told
her about starting compensation ratio: that “for new hires, you try to bring the person in at the
85 percent of the market target.” Tr. 78-79. Wagner testified to a different number, and he
would know. I give more weight to his testimony than to Suhr’s apparently mistaken notetaking and recollection.



Suhr testified that Wagner told her it used salary surveys to determine the market reference
point for starting pay based on locality and on what Google terms “job family” and “job
code.” Tr. 79-84. Thus, as she testified, the job family determined the range for starting
compensation. Tr. 81. Wagner’s testimony differed. He said that “job family” is a
professional category such as “software engineer.” Tr. 174. It contains various levels; for
example, a software engineer hired at graduation from college is a “software engineer level
3.” Job code is a number that refers to the job family at the specific level; e.g., software
engineer level 3 corresponds to job code 3403. Id. Google’s market reference point is tied to
location and job code, not to job family or job level. Tr. 174-75. It appears that Suhr has a
vague sense – not a full and accurate understanding – of how Google sets starting salaries.
She does not seem to appreciate that Google conducts a market survey for each job code in
the given locality, sets a market reference point at a particular percentile on the market
survey results, and determines a starting salary as a percentage of the market reference point
for the each separate job code in that locality.



Suhr did not know whether the goal of setting starting pay at a specific percentage of the
market reference point applies both to campus hires and industry hires; from Wagner, we
know that it does. Id.



Suhr understood that the type of project or team a person starts on “may impact” pay, but she
couldn’t recall how. Tr. 84. This cannot be used to show the relevance of any OFCCP
request.



Although Suhr understood that it was the compensation team that sets starting compensation
for a new employee, she never confirmed with any member of the compensation team that
factors such as starting level, starting job family, or starting job code affected starting
compensation or whether any of them was tied to the market reference point. Tr. 85-86. Yet,
she concluded that each of these factors was relevant.



Suhr had to recant her testimony that Google was a federal contractor on a different contract
immediately before GSA awarded Google the AIMS contract on June 2, 2014. After initially
admitting that she was guessing, she changed her testimony to assert that she knew – without
guessing – that Google had another contract immediately before June 2, 2014, and then on
cross-examination, she admitted that – truth be told – she didn’t know one way or the other.
Tr. 64:21-65:22; 66:6-21; 87:25-88:19.

I accept Wagner’s testimony that, as he answered the questions Suhr posed when she interviewed
him onsite at Google, he didn’t say anything different from his testimony on those subjects. Tr.
183-85. To the extent that Suhr’s testimony is inconsistent with Wagner’s, I give greater weight
to Wagner’s testimony.

15

 I found Regional Director Wipper generally – though not uniformly – credible in the sense that
she was truthful. But Wipper had limited personal knowledge and was relying heavily on her
staff, such as Deputy Regional Director Suhr. The result was that Wipper had an overview of
Google’s compensation practices and policies that was accurate in generalities but lacking in
detail.
In particular, Wipper conceded that her review of the hundreds of thousands of documents
Google produced was limited. Tr. 145-47. As to OFCCP’s interviews of Google’s managers,
Wipper admitted that she was not present for any of them. Tr. 136. Instead, she read the notes
her staff took. Id. One of those staff (apparently the highest ranking interviewer) was Suhr. Tr.
66-69.
This led to errors in the details Wipper offers. For example, Wipper seemed to think that, when
promoting employees, Google managers had discretion to give a percentage pay increase within
a range of 5 to 20 percent. Tr. 42-43. As discussed above, and as Wagner explained, the pay
increase turns on a mechanical calculation of a specific percentage of the market reference point
for the new job. This is adjusted to bring the raise at least up to 5 percent but not more than 20
percent. That is not a discretionary raise of 5 to 20 percent. The only discretion involved is an
adjustment that individual managers can make, according to Google, to better align the promoted
employee with her new peers; in over 90 percent of cases, the discretionary adjustment is no
more than one percent. Tr. 200-04.
As another example, Wipper thought that Google managers had discretion to offer as a starting
salary anywhere “between 80 and 120 percent of their mid-point.” Tr. 43. As Wagner’s more
persuasive and detailed testimony showed, there is no “mid-point.” The compensation
committee (not hiring managers) sets the compensation for all new hires without knowing the
gender (or other protected characteristics) of the new hires, and there is no basis for the assertion
Google offers new hires as much annual pay as 120 percent of the market reference point.
Finally, there were occasions on cross-examination when Wipper was evasive, as though she was
advocating. She resisted admitting that OFCCP would not interview all of the 21,000 Google
employees whose records it has obtained. Tr. 151-52. Google opposes the disclosure of
personal contact information for these employees for privacy reasons; it is concerned about
possible data security breaches at OFCCP. But when asked about the major breach at the Office
of Personnel Management in 2015, Wipper said only that OFCCP gives high priority to data
security – as if OPM does not. Tr. 155-56.
Wipper would not agree that OFCCP’s June 1, 2016 requests asked for almost the same amount
of information as OFCCP requested in Item 19 on the scheduling letter. Tr. 139-42. She would
not agree that the June 1, 2016 requests asked “at the very least for a large amount of
information.” Id. She would not agree that the June 1, 2016 requests asked for thousands of data
points. Id. In fact, the June 1, 2016 requests do all of these things. Having sat through Suhr’s
testimony about what Google had produced, Wipper stated that she could not recall the
testimony adequately to say whether it was correct. Id.

16

 On the other hand, I accept Wipper’s testimony that OFCCP conducted multiple regression
analyses on the September 1, 2015 snapshot of compensation-related data and concluded that
there were systemic disparities in compensation that statistically correlated with sex (adversely to
women) “pretty much across the entire workforce.” For purposes of this proceeding, the
question is not whether this preliminary finding is correct and would stand up in litigation on the
merits. The question is whether OFCCP in fact reached this conclusion and is pursuing relevant
information to develop it further, to determine the period of time that the disparity occurred, or to
determine the cause of the disparity.61 Director Wipper is an attorney with an extensive
background litigating cases of this kind and other employment law cases. She understands the
process for determining whether the data reveal disparities. I have no reason to doubt her
testimony that OFCCP made this preliminary finding, and I accept that it did. But I make no
finding that OFCCP’s analysis is valid or persuasive.62
In all, I find to a degree the testimony of Wagner and Wipper (and to some extent Suhr) are
consistent, and I accept the consistent evidence. I accept Wipper’s testimony that OFCCP’s has
found – on a preliminary basis – that the September 1, 2015 snapshot shows sex-based pay
disparities. Otherwise, to the extent there is any inconsistency between the testimony of Wagner
on one hand and the testimony of either Wipper or Suhr on the other, I give greater weight to
Wagner’s testimony for the reasons stated above.
Discussion
In general. As amended, Executive Order 11246 places responsibility with the Secretary of
Labor for the administration and enforcement of its provisions concerning federal contractors.
E.O. 11246 § 201.63 Unless a contract is exempt,64 every federal contract must include a
provision requiring that:
The contractor will not discriminate against any employee or applicant for
employment because of race, color, religion, sex, sexual orientation, gender
identity, or national origin. The contractor will take affirmative action to ensure

61

For that reason, Google’s questioning at the hearing aimed at disputing OFCCP’s conclusions was irrelevant for
the purposes of this proceeding, while they would be centrally relevant in any case on the merits of an adverse
impact discrimination case.
62

As reflected in OFCCP’s Directive 307, the results of the statistical analysis depend on what jobs the analytical
model defines as similarly situated for comparison purposes. As discussed in the text above, Deputy Regional
Director Suhr was the highest ranking OFCCP investigator involved in the onsite interviews. She also had access to
the documents Google produced. Yet her testimony reflected little understanding of Google’s organization as
indicated with terms such as “job family,” “job title,” “job level,” “job code,” and similar terms. If OFCCP
misunderstood Google’s organization – as it appears that OFCCP might have – it is likely that OFCCP’s model
lacks precision when defining which employees are similarly situated for comparison and statistical analysis. That
raises questions about the validity of OFCCP’s analysis. But, as I said, I do not reach that question. Instead, I will
conclude that greater collaboration and conciliation would be helpful.
63

The Secretary also administers and enforces the provisions on federally assisted construction contracts. Id.
Executive Order 12086 (Oct. 5, 1978) placed OFCCP within the Department of Labor.
64

See E.O. 11246 § 204; 41 C.F.R. § 60-1.5. No party asserts that any exemption applies here.

17

 that applicants are employed, and that employees are treated during employment,
without regard to [these factors].
* * *
The contractor will comply with all provisions of Executive Order No. 11246 . . .
and of the rules, regulations, and relevant orders of the Secretary of Labor.
The contractor will furnish all information and reports required by Executive
Order No. 11246 . . . and by the rules, regulations, and orders of the Secretary of
Labor, or pursuant thereto, and will permit access to his books, records, and
accounts by the contracting agency and the Secretary of Labor for purposes of
investigation to ascertain compliance with such rules, regulations, and orders.
E.O. 11246 §§ 202(1), (5), (6). Google agreed to these provisions when it entered into the AIMS
contract on June 2, 2014. ALJ Ex. 1, ¶ 5.
I.

The Parties Must Engage In Further Conciliation.

Threshold issues: OFCCP’s criteria when selecting Google for audit. The parties agree that,
although OFCCP has no subpoena authority, its authority to access contractors’ records is akin to
an administrative subpoena.65 Some courts have held that one of the requirements for
administrative subpoenas is that the “the search is ‘pursuant to an administrative plan containing
specific neutral criteria.’” See, e.g., U.S. v. Mississippi Power & Light Co., 638 F.2d 899, 907
(5th Cir. 1981), quoting Camara v. Municipal Court, 387 U.S. 523, 538 (1967). For that reason,
I allowed Google a deposition of OFCCP’s person most knowledgeable about how OFCCP
selected Google for this compliance review. After taking the deposition, Google stipulated that it
would not, for purposes of this litigation, dispute that OFCCP based its selection of Google on
specific neutral criteria.
Conciliation. The Executive Order provides sanctions and penalties for violations, but only after
the Secretary of Labor makes reasonable efforts to secure compliance “by methods of
conference, conciliation, mediation, and persuasion.” E.O. 11246 § 209(a), (b). The regulations
are to the same effect. See 41 C.F.R. 60-1.20(b). Here, the parties exchanged views a number of
times before reaching an impasse. See J.Ex. 7-16. Google has stipulated that, for purposes of
this litigation, it would not assert that OFCCP failed to conciliate. ALJ Ex. 1, ¶ 28.
I accept the parties’ stipulation. But in my view, circumstances have changed such that further
conciliation is appropriate if the regulatory scheme is to be properly respected.
The primary reason for the parties’ impasse concerned Google’s demand to know what issues
OFCCP was continuing to investigate and in what part of Google’s operations these issues arose.
J.Ex. 7. Google argued that, during eight months of investigation, Google had provided
65

The contractual obligations that the Executive Order requires combine with OFCCP’s authority to request
information to produce the same effect as would an administrative subpoena.

18

 extensive compensation data for the 21,114 employees in the September 1, 2015 snapshot. Id.
Yet OFCCP wanted a snapshot for another date also wanted job and salary histories for both
groups of employees, going back for each employee to the date of hire – even if that was 18
years earlier, when Google was formed. Id. Google argued that OFCCP had offered no
information about the issues it was finding, which prevented Google from evaluating whether
OFCCP’s additional requests were relevant to the investigation. Google asserted that this made
collaboration more difficult. See J.Ex. 7.
OFCCP responded that it was “unable to share” preliminary findings or analyses because it had
realized only after the onsite interviews that it needed additional information to assess the data
Google had provided. J.Ex. 8. OFCCP stated that it would not agree to disclose its preliminary
findings in return for Google’s agreement to comply with the additional data requests. J.Ex. 8.
After further exchanges, OFCCP notified Google that it must show cause why OFCCP should
not bring an enforcement action.66 J.Ex. 11. Google responded with an offer to continue
producing certain information and discussing the parties’ disagreement. J.Ex. 12. It also
asserted that OFCCP failed to comply with its own regulations in its “repeated refusal to provide
any explanation whatsoever regarding the relevance” of the disputed requests. J.Ex. 12-002
(emphasis in original).
OFCCP offered no explanation why the additionally requested data was relevant to its
investigation; instead, it initiated this litigation.
As late as OFCCP’s motion for summary judgment, OFCCP stated:
To be clear, if any hearing is ordered, OFCCP witnesses will not offer any
testimony regarding its internal deliberations concerning the ongoing compliance
evaluation, including its preliminary findings. This testimony is wholly
unnecessary to determining whether the [materials OFCCP seeks] were properly
requested. Moreover, as explained in OFCCP’s opening brief, such testimony
would invade the agency’s deliberative process and investigative privileges and
any work product protection.
OFCCP reply brief (Mar 6, 2017).
But at the hearing, OFCCP reversed course. Regional Director Wipper testified (as discussed
above) that OFCCP’s analysis of the September 1, 2015 snapshot showed: “systemic
compensation disparities against women pretty much across the entire workforce.” Tr. 48. It
would seem that this was what Google had been requesting: a general description of the issues
OFCCP was finding and in what portion of Google’s workforce the issues were appearing.
Obviously surprised at Wipper’s testimony, Google asserts that this was the first time OFCCP
disclosed this information. I do not question that the testimony came as a surprise to Google.

66

OFCCP cited 41 C.F.R. §§ 60-1.26, 60-1.28.

19

 Nor do I question that OFCCP’s reversal was in good faith. It likely reflected OFCCP’s
reevaluation after I denied its motion for summary judgment. I commented in the order on
OFCCP’s demand for job and salary histories for 25,000 employees going back as far as 1998. I
found the relevance of this demand far from obvious. Given the burden of unearthing this
information and the lack of apparent relevance, I stated that OFCCP failed to make the showing
required for summary judgment. It is entirely possible that this was a factor in OFCCP’s
decision to disclose – in a limited, very general way – its preliminary findings of sex-based pay
disparity.
I expect there were also surprises for OFCCP at the hearing. OFCCP’s understanding of
Google’s organizational structure and pay policies and practices was lacking. Wagner corrected
errors and filled in gaps. He delivered an organized, complete description of those policies and
practices. It is evident that the questions posed to him at the onsite failed to elicit that kind of
description and that the interviewers did not entirely understand the information that they did
elicit. I need not evaluate the history of OFCCP’s possible misconceptions; the point is that,
during the hearing, OFCCP likely learned much about Google’s relevant policies and practices –
at least if OFCCP found Frank Wagner even nearly as credible as I found him. And OFCCP
cannot evaluate Wagner’s credibility without considering what he said.
Had OFCCP made its disclosures and had Google presented Wagner’s information earlier, it
might have made the present litigation unnecessary. Google had cooperated extensively when
making disclosures earlier in the investigation. Once OFCCP gave Google the basic information
about its preliminary findings, Google might have been more forthcoming with information such
as Wagner’s. Perhaps Google would have offered its own statistical analysis for OFCCP to
consider. If OFCCP understood better what Google’s compensation policies are, it might have
reconsidered some of its current information requests. But with the information exchange
occurring mid-trial, neither party could be expected to interrupt the process to resume informal
discussions.
Given the mandate in the Executive Order and the regulations for conciliation, and given these
changed circumstances, I will include in the Order a requirement for further good faith
conciliation.
II.

OFCCP Must Meet Fourth Amendment Standards For An Administrative Subpoena.

The Fourth Amendment requirement of probable cause is not required for administrative
subpoenas. See United States v. Powell, 379 U.S. 48, 57 (1964); U.S. v. Golden Valley Elec.
Ass’n, 689 F.3d 1108, 1115 (9th Cir. 2012). But, as the Supreme Court observed:
“It is now settled that, when an administrative agency subpoenas corporate books
or records, the Fourth Amendment requires that the subpoena be sufficiently
limited in scope, relevant in purpose, and specific in directive so that compliance
will not be unreasonably burdensome.”

20

 Donovan v. Lone Steer Inc., 464 U.S. 408, 415 (1983) (quoting See, 387 U.S. at 544).67 “[I]t is
sufficient if the inquiry is within the authority of the agency, the demand is not too indefinite and
the information sought is reasonably relevant.” U.S. v. Morton Salt Co., 338 U.S. 632, 652-53
(1950) (quoting Oklahoma Press Publishing Co. v. Walling, 327 U.S. 186, 208
(1946)).
The scope of judicial review in an administrative subpoena enforcement proceeding is “quite
narrow.” 68 But the government must still answer some questions. “The critical questions are:
(1) whether Congress has granted the authority to investigate; (2) whether procedural
requirements have been followed; and (3) whether the evidence is relevant and material to the
investigation.”69 An administrative subpoena may not be “too indefinite or broad.” Peters v.
United States, 853 F.2d 692, 699 (9th Cir. 1988). Even if other criteria are satisfied, “a Fourth
Amendment ‘reasonableness’ inquiry must also be satisfied.” See Reich v. Mont. Sulphur &
Chem. Co., 32 F.3d 440, 444 n. 5 (9th Cir. 1994). In an OFCCP compliance evaluation,
reasonableness requires that the data OFCCP seeks is “‘sufficiently limited in scope, relevant in
purpose, and specific in directive so that compliance will not be unreasonably burdensome.’”
United Space Alliance, 824 F. Supp. 2d at 93, citing Lone Steer, 464 U.S. at 415.
The Government bears the initial burden to show that these criteria have been
met, although the burden to make a prima facie case is “minimal.” Once the
Government has made a prima facie case, the burden of going forward shifts to
the party opposing the subpoenas.
U.S. v. Transocean Deepwater Drilling, Inc., 767 F.3d 485, 489 (5th Cir. 2014).
As an example, Title VII grants the EEOC subpoena authority to obtain evidence “relevant to the
charge under consideration,” but relevance alone is not enough: Courts will not enforce an
EEOC subpoena if it is too indefinite, is issued for an improper purpose [e.g., beyond the
agency’s authority], or is unduly burdensome. See McLane Co., Inc. v. E.E.O.C., 137 S.Ct.
1159, 1165 (2017). Enforcing such a subpoena would violate the “gist” of Fourth Amendment
protection “that the disclosure sought not be unreasonable.” Id., citing U.S. v. Morton Salt,
supra, 338 U.S. 652-53. The regulation implementing the Executive Order and the concomitant
contract provision here70 provides the OFCCP with no greater authority than the Title VII
provision at issue in McLane.
67

The parties agree that OFCCP’s current request for information is akin to an administrative subpoena. OFCCP is
not demanding an intrusion into Google’s offices or access to its personnel, a demand that could be viewed as an
administrative warrant. (Indeed, OFCCP completed a two-day onsite review to which Google consented. The
consent is a waiver of any need for an administrative warrant.) For an administrative subpoena, the government
meets Fourth Amendment demands by showing only reasonableness. See Donovan v. Lone Steer, Inc., 464 U.S.
408, 414 (1984); United Space Alliance, 824 F. Supp. 2d at 92.
68

EEOC v. Children's Hosp. Med. Ctr. of N. Cal., 719 F.2d 1426, 1428 (9th Cir. 1983) (en banc), overruled on
other grounds as recognized in Prudential Ins. Co. of America v. Lai, 42 F.3d 1299 (9th Cir. 1994).
69

EEOC v. Children's Hosp., supra.

70

Contractor must provide access to “such books and accounts and records . . . and other material as may be
relevant to the matter under investigation and compliance with the [Executive] Order, and the rules and regulations
promulgated pursuant thereto . . . .” 41 C.F.R. § 60-1.43 (emphasis added).

21

 OFCCP contends that Google waived its Fourth Amendment rights in their entirety when it
agreed in the AIMS contract to give the government access to its records and other material
relevant to the investigation and its compliance with the Executive Order. The argument is
without merit. See First Alabama Bank of Montgomery, N.A. v. Donovan, 692 F.2d 714, 718-19
(11th Cir. 1982) (contractual agreement to provide information does not waive Fourth
Amendment right limiting government to demands that are reasonable).
OFCCP misplaces its reliance on a World-War II Supreme Court case that precedes McLane v.
EEOC by 71 years: Zap v. U.S., 328 U.S. 624, 628 (1946). Zap is a criminal case involving the
FBI’s warrantless seizure of a falsified check that the defendant allegedly used to defraud the
Navy on a cost-plus contract. The Court’s holding does not support OFCCP’s contentions for
several reasons. First, the Zap Court vacated its decision on rehearing and ordered the district
court to dismiss the indictment.71 Zap v. U.S., 330 U.S. 800 (1947). Second, as the First
Alabama court explained, Zap concerns administrative warrants, not administrative subpoenas,
and the Court still required that the search be reasonable (which is the “gist” of Fourth
Amendment protection). 692 F.2d at 719.
In addition: (1) the Zap Court did not find that the contractual agreement to make records
available was a waiver of all Fourth Amendment protection – it acknowledged that there were
limits to the waiver; (2) the portions of the decision on which OFCCP relies are dicta – or at least
not required for the result; (3) Zap is an early application of the exclusionary rule, and the law
has changed since 1946; and (4) the Zap Court – acting less than a year after the end of World
War II – manifestly strained to affirm the conviction of a war profiteer who relied on what the
Court viewed as a “technicality” to escape the consequences of defrauding the Navy during the
existential crisis of World War II or anything similar to it.72 Were the currently understood
71

OFCCP contends that the Court vacated its initial (1946) Zap decision “on other grounds.” P. Closing Br. at 11.
It offers no support for this contention. Id. The Court’s order does not explain why the Court is vacating its initial
decision. See 330 U.S. at 800-01. The order states no more than that the Court grants rehearing, vacates its initial
order, reverses the judgment of the Court of Appeals, and orders the district court to dismiss the indictment. As the
Court does not disclose its grounds and OFCCP offers no further information, I cannot disregard the order vacating
the initial Zap decision in 1946 as based “on other grounds.”
72

In Zap, a wartime government contractor was performing a defense contract with the Navy on a cost-plus basis.
The contract included a term that required the contractor to open his accounts and records to the government at all
times. The contractor wrote a fraudulent expense check to mislead the Navy into making excessive cost-plus
payments.
Asserting the inspection rights in the contract, the FBI reviewed the contractor’s accounts and records. The
contractor’s employees provided the FBI with the documents it requested, including the fraudulent check. None of
this was questioned in the criminal trial.
But the FBI agent took one more step: he seized the fraudulent check without the contractor’s permission. When
the government introduced the check at trial, the contractor moved to strike it as improperly obtained, given that the
contract allowed the government only access to review the accounts and records, not to seize them. The trial judge
denied the motion, and the contractor was convicted.
On certiorari, the only question before the Court was whether the trial court erred by denying the motion to strike.
The Court looked to its 1914 decision, in which it first announced the exclusionary rule, as well as to an explication
of the rule in 1920. (Citing Weeks v. U.S., 232 U.S. 383 (1914), and Silverthorne Lumber Co. v. U.S., 251 U.S. 385
(1920).) Zap, 328 U.S. at 630. The Court acknowledged that the contractor did not agree that the government could

22

 exclusionary rule applied to an administrative subpoena today, the result – as in McLane v.
EEOC – would give more weight and certainty to Fourth Amendment guarantees.73
III.

OFCCP Is Acting Within Its Authority.

The Executive Order requires the Secretary of Labor to investigate whether contractors are in
compliance with the Order’s requirements, applying procedures that the Secretary adopts. E.O.
seize documents; the contract term allowed the government only to access and review records. Neither did the FBI
obtain a warrant.
While the seizure of the check might thus have been unlawful, the Court held that there were many reasons not to
exclude the check regardless of the then-existing exclusionary rule. 328 U.S. at 629-30. As the Court stated, “‘A
criminal prosecution is more than a game in which the Government may be checkmated and the game lost merely
because its officers have not played according to rule.’ To require reversal here would be to exalt a technicality to
constitutional levels.” Id. The Court explained essentially why the unlawful seizure was harmless.
As the Court stated, rather than taking the check, the FBI could have “taken photostats” or made copies and
introduced those at trial. The only difference was that, if the FBI got a search warrant, it would have “the means of
insuring the production in court of the primary source of evidence otherwise admissible.” 328 U.S. at 630. Or the
trial judge could have returned the check to the contractor, the FBI could have obtained a warrant, and the FBI could
seize the check again – lawfully this time.
As a third example of why excluding the check would elevate a technicality to a constitutional level, the Court found
– as was undisputed – that the contractor had agreed to give the government access to his accounts and records,
including the check. This, the Court said was a waiver of Fourth and Fifth Amendment rights. The Court held the
waiver was limited: It did not encompass all rights under these Amendments. As the Court wrote:
Whatever may be the limits of that power of inspection, they were not transcended here . . . .
There is some suggestion that the search was unreasonable because made by agents of the Federal
Bureau of Investigation who were not persons authorized to conduct those examinations. But they
acted under the auspices and with the authority of representatives of the Navy Department who
were authorized to inspect . . . . Moreover, the right to inspect granted by the contracts was not
limited to inspections by the Navy but extended to inspections by any authorized representatives
of the government among whom the agents of the Federal Bureau of Investigation are included.
Zap, 328 U.S. at 628-29. Given these considerations, the Court held only that the incriminating facts were not
“bar[red] absolutely”; their admissibility was left to the trial judge’s discretion.
Read most favorably to OFCCP, Zap thus holds that the contract language is a waiver of some Fourth Amendment
rights, with limitations that the Court did not reach. At the least, the waiver is limited to the materials described in
the contract, and as First Alabama holds, to searches that are reasonable (such as by the FBI rather than the Navy).
Zap does not answer the question before me here – it merely frames it.
OFCCP also misplaces reliance on cases that do no more than follow Zap’s holding that, whatever the limitations on
the contractual waiver of Fourth Amendment rights, evidence will not be excluded at a criminal trial because the
FBI was conducting a criminal investigation rather than the Navy conducting an administrative audit. See U.S. v.
Jennings, 724 F.2d 436, 448 (5th Cir. 1984) (documents seized under the guise of a civil audit and without
disclosing that the seizure was part of a criminal investigation); U.S. v. Brown, 763 F.2d 984, 988 (8th Cir. 1985)
(state agency investigator conducting audit acted in conjunction with undisclosed FBI agent who used the
information gathered to support a criminal investigation). In neither of these cases was there any dispute that the
regulatory agency had a right to access the materials, if it had been done for regulatory purposes, as opposed to
being part of an undisclosed criminal investigation. The courts hold only – as did Zap – that the contractor’s waiver
– whatever its scope – extended, not only to administrative agency investigations, but also to criminal investigations
(even if they were undisclosed).
73

See NLRB v. American Medical Response, Inc., 438 F.3d 188, 192-93 (2d Cir. 2006) (applying the Morton Salt
requirements to enforcement NLRB subpoena).

23

 11246 § 206(a). The contractor’s compliance is required only “during the performance of [the]
contract.” 41 C.F.R. § 60-1.4(a).
The Executive Order allows the Secretary of Labor to delegate his or her authority under the
Order “to any officer, agency, or employee in the Executive branch of the Government.” Id.
§ 401. The Secretary has delegated the relevant authority to the Deputy Assistant Secretary,
Office of Federal Contract Compliance Programs. 41 C.F.R. § 60-1.1, 60-1.2. In turn, the
Deputy Assistant Secretary may re-delegate his or her authority. 41 C.F.R. § 60-1.46.
Google does not dispute that OFCCP’s audit comes within the agency’s authority, and I find that
OFCCP’s current requests are within its authority.74
IV.

Portions Of OFCCP’s Requests Are Unreasonable And Fail Fourth Amendment
Requirements.

I turn to the central question whether OFCCP has made out a prima facie case that each or any of
the three requests for additional information and data meets Fourth Amendment’s requirements.
The requests must be reasonable in that that are relevant to the investigation, sufficiently limited
in scope, and not unduly burdensome.
Relevance. Generally, in the context of an administrative subpoena, the term “relevant” is given
a “generous construction.” McLane, supra, 137 S.Ct. at 1165 (quoting EEOC v. Shell Oil Co.,
466 U.S. 54, 68-69 (1984). But there is some distinction between the general rule and the
present case: The present audit, as Regional Director Wipper acknowledged was not “complaintdriven” and is currently focused on an adverse impact theory of sex discrimination in
compensation.75 The lack of a complaint is important because in Shell Oil relevance in an EEOC
Title VII investigation encompasses “virtually any material that might cast light on the
allegations against the employer” (emphasis added). A similar standard applies in NLRB
subpoena enforcement cases because a specific charge against the employer is required, thus
tying relevance to the allegations in the charge. See NLRB v. American Medical Response,
74

In its initial request, OFCCP asked for “race/ethnicity” information. Then, in a later request, it requested in
addition: place of birth, citizenship, and visa status. See D.Ex. 116 at 1. OFCCP’s investigative authority is limited
to discrimination because of race, color, religion, sex, sexual orientation, gender identity, national origin, disability,
or certain veterans’ status. “National origin” can be defined somewhat variously. But identifying employees’
national origins does not require knowing their place of birth, citizenship, or visa status.
This is information not relevant to the matters that come within OFCCP’s authority. Actions against employers
under the Immigration Control and Reform Act are not within the Department of Labor’s authority. Nor are
deportation or similar actions. The Department of Labor has a regulatory function in a variety of programs
involving nonimmigrant alien workers and applicants for employer-sponsored permanent labor certification (an
initial step that can lead toward permanent residence). But OFCCP has no authority to conduct investigations
related to these functions.
I understand that OFCCP is not seeking in this matter an order requiring Google to provide any further employee
data on place of birth, countries of citizenship, visa status, or any similar data beyond national origin or ethnicity.
Any such request is denied as exceeding OFCCP’s authority and not relevant to the investigation OFCCP has
described.
75

In its closing brief, OFCCP requests official notice of internet reports and

24

 supra, 438 F.3d at 193-94 (NLRB may not initiate its own investigation, but when the Board
issues a subpoena on a pending charge, the “court defers to the agency’s appraisal of relevance,
which must be accepted so long as it is not obviously wrong”).
In the present case, there is no pending charge or complaint. OFCCP is auditing Google’s
headquarters facility with over 21,000 employees for compliance with the Executive Order’s
anti-discrimination provision. This extends to discrimination because of race, color, religion,
sex, sexual orientation, gender identity, national origin, disability, or certain veterans’ status, and
it includes claims of either disparate treatment or adverse impact. Unlike an administrative
investigation of an EEOC charge or an NLRB charge, OFCCP’s investigation finds its limits
only in the expanse of these several anti-discrimination provisions; there is no focus, for
example, on discrimination against a person based on a single characteristic – or even
discrimination against a group of people of the same race or national origin or sex. The result is
an investigation in which a vast amount of information could be relevant.
The expansive scope of the audit here is likely at the broadest OFCCP encounters because
OFCCP is pursuing indications of adverse impact discrimination. OFCCP has found what it
views as a statistical disparity in compensation at Google that is adverse to women.
Statistical disparity, however, does not establish an adverse impact violation. Rather, Title VII
expressly requires a plaintiff on such a claim to show that the employer “uses a particular
employment practice that causes” the adverse impact. 42 U.S.C. § 2000e-2(k)(1)(A)(i). Only
then does the burden pass to the employer to show “that the challenged practice is job related for
the position in question and consistent with business necessity.” Id. The requirements are the
same under a common law analysis applied to other civil rights statutes that do not expressly
address the burdens in an adverse impact case.76 Thus, OFCCP’s investigation is incomplete
unless it can identify and prove what practice is causing the disparity it claims to have found.77
In most adverse impact cases, including those that EEOC pursues, there is an allegation from the
outset about what is causing a disparity. For example, people of a given racial or national origin
group might assert that a blanket exclusion of applicants with a history of any criminal
conviction at any time is adversely impacting them in hiring. Before same sex marriage was
legal, gay applicants might allege that a requirement that only married applicants will be hired
was having an adverse impact on them.
But, as there is no charge of discrimination that OFCCP is considering, there is no allegation of
what practice or policy is causing an adverse impact on women’s pay. As a further complication,
76

See, e.g., Hardie v. National Collegiate Athletic Ass’n, ___ F.3d ___, 2017 WL 2766096 (9th Cir. June 27, 2017),
slip op. at 4 (applying in a Title II case Wards Cove Packing Co. v. Atonio, 490 U.S. 642 (1982)), superseded by
statute on other grounds, 42 U.S.C. § 2000e-2(k)). As the Court stated in Wards: a plaintiff must “specifically
show[] that each challenged practice has a significantly disparate impact on employment opportunities [based on a
protected characteristic]. To hold otherwise would result in employers being potentially liable for ‘the myriad of
innocent causes that may lead to statistical imbalances in the composition of their work forces.’” 490 U.S. at 657.
77

Regional Director Wipper testified that OFCCP must determine the cause of the disparity so that the disparity may
be corrected. That might be true, but it misses the point that OFCCP must prove the cause of the disparity if it is to
establish a violation.

25

 OFCCP is not an advocate representing Google’s employees against their employer. It is
conducting an objective audit of a government contractor against which there have been no
allegations raised with OFCCP. OFCCP might conclude its investigation with a finding that,
though there is a pay disparity correlated to sex, the cause is job-related, consistent with business
necessity, and thus lawful.
The result is that OFCCP must search, not only for causes of the disparity that are actionable, but
also those that are lawful. It is difficult to imagine a broader search in the employment law
context.
To bring sufficient focus to its investigations and narrow the scope of relevance, OFCCP
Directive 307 gives guidance on the conduct of compensation audits. It describes an iterative
process involving a wide range of experts, tools, sources, steps, and case-by-case adjustments as
the OFCCP learns more. A primary source of information, of course, must be the contractor,
which will be asked what factors are relevant to compensation. OFCCP considers “whether
these factors, in conjunction with other legitimate factors, if any, actually explain pay, are
implemented fairly and consistently applied, and whether they should be incorporated into a
statistical analysis, on a case by case basis.” The investigation is to be tailored to the
contractor’s compensation practices. The model that OFCCP uses for analysis must be refined to
fit the information gathered.
Applying this regime, OFCCP should be able to identify specific areas that are relevant to its
investigation rather than willy-nilly search anywhere and everywhere for practices that might be
causing a disparity in the compensation data. Relevance need not and cannot be established on
conjecture or speculation. Fortunately, OFCCP has identified two areas of focus in the current
investigation.
First, OFCCP offered testimony listing factors that Google said were relevant to its
compensation decisions. OFCCP wants data relevant to those factors.
Second, as discussed above, OFCCP seeks information going to a theory of causation for which
it contends there are some indicators. The theory is that some publications not on the record
argue, suggest, or find that women, on average, are less successful negotiators that men; Google
negotiates starting pay; this has caused women to enter Google’s workforce with lower pay
relative to men; raises during employment are based on existing pay rates; and thus women’s
lower pay at hire leads to ongoing pay disparity even for female employees who have worked at
Google since the company was formed in 1998.78
To determine whether OFCCP’s requests are reasonable under the Fourth Amendment, I will
consider together the relevance, scope, and burden imposed in an OFCCP audit that is not
complaint-driven and that OFCCP has aimed at a potential adverse impact violation.

78

OFCCP contends that the negotiated starting pay has an adverse impact on women as they are hired and that the
system of pay increases based on the starting pay “anchors” women into the lower pay.

26

 Burden. Courts often recite that an administrative subpoena is unduly burdensome only if it
would seriously hinder normal operation of the business. While this is generally helpful, in the
context of an OFCCP audit, I find more guidance the analysis the Eleventh Circuit has adopted.
That court finds that, while seriously hindering normal operation is one formulation, it is not a
rigid rule. See EEOC v. Royal Caribbean, 771 F.3d 757, 763 (11th Cir. 2014). As the court
held:
[A] district court may consider a number of factors in this analysis, rather than
requiring specific types of evidence on a single factor. See also [EEOC v.]
United Air Lines, 287 F.3d [643] at 653 [7th Cir. 2002] (noting that cases such as
Bay Shipbuilding have suggested a party must show that compliance would
threaten normal business operations but explaining “that scenario is more
illustrative than categorical” and “[w]hat is unduly burdensome depends on the
particular facts of each case and no hard and fast rule can be applied to resolve the
question” (internal quotation marks omitted)); EEOC v. Ford Motor Credit Co.,
26 F.3d 44, 47 (6th Cir. 1994) (“Essentially, this court's task is to weigh the likely
relevance of the requested material to the investigation against the burden to Ford
of producing the material.”).
Royal Caribbean, 771 F.3d at 763 (citation omitted). This formulation directs me to consider
burden in context; for this case that includes circumstances where relevance can become
speculative if it is not tethered to some substantiated indicators, and where OFCCP’s request
burden – not only Google – but also Google’s employees.
I now turn to the OFCCP’s three requests.
Snapshot for September 1, 2014. OFCCP argues that an additional snapshot is relevant because
it will show whether the same indications of a possible adverse impact violation existed over
time, not just on the single day reflected on the September 1, 2015 snapshot. OFCCP states that
this is consistent with its general practice: It begins by asking for a single snapshot and requires
a second only if there are indicators of possible violations in the first snapshot. When asked
whether OFCCP considered narrowing the scope of the snapshot to fewer employees than all
those covered by Google’s headquarters affirmative action plan, OFCCP’s response was that it
did consider narrowing the request. The difficulty was that OFCCP’s findings from the first
snapshot were systemic and across much of the entire workforce.
OFCCP states that it generally starts with a single snapshot and then asks for a second only when
the first snapshot indicates disparities. Wipper stated that’s what the 2015 snapshot showed. But
OFFCP offered nothing more than this conclusory statement.
By contrast, in United Space Alliance, supra, 824 F. Supp. 2d 68 (D. D.C. 2011), OFCCP
offered much more to show relevance. It was seeking follow-up data after receiving the
equivalent of the Item 19 data here. At the hearing, OFCCP supervisor Miguel Rivera testified
in detail about the data analyses that OFCCP performed. 824 F. Supp. 2d at 75-76 and at 77
(“Miguel Rivera testified about his analysis of the United Space data”). He explained why
OFCCP did not rely solely on the algorithm it generally uses (the “threshold test”) but instead

27

 applied a second analysis (the “pattern analysis”). Id. He gave detailed numerical results of the
pattern analysis to show why OFCCP found the additionally requested information relevant to its
investigation. 824 F. Supp. 2d at 76.79
Here, OFCCP refused any explanation of its analysis; it asserted deliberative process privilege.
It relies mostly on Wipper’s conclusory statement that OFCCP conducted an analysis (including
multiple regression analyses) that showed widespread pay disparities that correlated with sex.
Nonetheless, for purposes of a September 1, 2014 snapshot, I find OFCCP’s showing of
relevance sufficient to meet the deferential standard that applies in the narrow Fourth
Amendment review appropriate to administrative subpoenas. OFCCP showed that Google
managers exercise discretion on several compensation decisions. In addition, about 17,000 of
the employees working for Google on September 1, 2014, also worked for Google on the date of
the snapshot Google already provided for September 1, 2015. For those employees, OFCCP will
get more information about specific merit increases, promotions, and any other changes affecting
identifiable individuals and possibly showing disparate treatment. The information sought
pertains to a time when Google was performing the AIMS contract. The Director adequately
explained the relevance of this request.
As to burden on Google, OFCCP cannot explain extending its request for the data to all 19,000
employees based on the scope of the affirmative action plan; OFCCP has not pursued the
affirmative action plan in this case. But, instead, Director Wipper testified that the breadth was
minimally required because of the statistical findings of disparities that extended across the
entire workforce. That is sufficient to render the burden reasonable (not an undue burden) and to
find the focus of the request sufficient as well.
I also find no reason to question the relevance of most of the data categories that OFCCP
requests Google include on the snapshot, but I will modify or exclude some for lack of relevance
or because of the burden imposed.


The categories that the Office of Management and Budget approved and are listed in Item 19
of the scheduling letter (J.Ex. 5 at 6) are relevant, not burdensome, and must be included.



The categories noted above (n. 74) concerning place of birth, citizenship, and visa status do
not appear to be part of OFCCP’s current request. If OFCCP is including these, its request
for an order requiring them is denied. The information exceeds OFCCP’s authority and is
not relevant to the characteristics that Executive Order 11246 protects.



As for the 38 additional categories that OFCCP requested on June 1, 2016 (J.Ex. 6 at 1-2), I
make the following determinations:

79

I rely on the district court’s published opinion cited above. Google attached to its closing brief excerpts of the
administrative record in United Space. Google should have listed the excerpts on its exhibit list, provided a copy
before the hearing to OFCCP, and presented the exhibit at the hearing. This would have allowed OFCCP to add
more of the administrative record, object to the exhibit, or offer countervailing evidence. The excerpt (Exhibit A to
Google’s closing brief) is stricken.

28

 

OFCCP has withdrawn the request for “any other factors related to compensation.” Of
course, Google may volunteer information; indeed, it might be in Google’s interest to do
so if it will explain what OFCCP perceives as pay disparities. But it is not required to
respond to a request that is so unfocused.



I will address OFCCP’s demands for department hired into; job history; salary history;
starting “compa-ratio”; starting job code; starting job family; starting level; starting
organization; and starting salary when I consider OFCCP’s demand for job and salary
histories below. Google need not include this information in the September 1, 2014
snapshot.



I find no relevance in OFCCP’s request for each employee’s date of birth. Age
discrimination is not an area of enforcement within OFCCP’s authority. OFCCP offered
no explanation. It offered no suggestion that age might explain the pay disparities it has
identified. Google need not supply date of birth information.80



The request for locality information is unduly burdensome because the single defining
characteristic of all of the employees is that, on the date of the snapshot, the employee
worked at Google’s headquarters, 1600 Amphitheatre Parkway in Mountain View,
California. That’s what includes the employee in the affirmative action plan that defines
the scope of OFCCP’s audit. Google need not provide this information.



Google must include in the snapshot for September 1, 2014, responses to each of the
other categories on OFCCP’s June 1, 2016 list for each employee within the headquarters
affirmative action plan, to the extent that Google has that information within its
possession, custody, or control.

Employee contact information. OFCCP seeks the name, personal address, personal telephone
number, and personal email for each of the more than 25,000 employees whom this Order will
affect.81 Director Wipper explained this would lead to information relevant to the investigation
because OFCCP talks to employees to get their points of view on compensation issues and needs
to do so without revealing to Google the identities of informants.
I am persuaded that anecdotal information obtained from employees is relevant to OFCCP’s
systemic adverse impact investigation. Although I have found Frank Wagner’s testimony
credible, he cannot know with certainty that Google’s managers are faithfully implementing
Google’s policies and practices; there might be exceptions – few or many. Women feeling they
80

Even if age information were to be relevant, OFCCP would at most be entitled to each employee’s year of birth,
not date of birth. Date of birth, when combined with other information that OFCCP seeks, creates personally
identifiable information that could lead to identity theft or similar adverse results. I discuss in the text below the
burden of this risk. It cannot be justified when year of birth would be equally informative on any issue related to
employment discrimination.
81

The contact information is for every employee on either “snapshot.” The 2015 snapshot included 21,114
employees. Google states that adding persons it employed on September 1, 2014 will increase the total number of
persons covered to more than 25,000. This takes into account that many of the employees on the 2014 snapshot
were still working for Google a year later, and were already included in the 21,114 on the 2015 snapshot.

29

 are not being paid fairly because of their sex might complain orally to their managers, and some
managers might respond by using their discretion under the policy to increase these employees’
pay. There will be no record of a complaint, and the final (adjusted) pay will not reflect a
disparity. Yet, if there is a pattern of women having to complain to get non-discriminatory pay,
that could affect OFCCP’s investigative findings. There are many other possibilities.
I am not suggesting that OFCCP will find any of this; the anecdotal evidence might be favorable
to Google. The question is only whether the contact information is relevant to the investigation.
It is. In addition, if OFCCP concludes that it must initiate an enforcement proceeding on the
merits, anecdotal testimony from adversely affected employees would be a crucial part of its
proof in a systemic adverse impact case.
Although evasive on the witness stand, Director Wipper admitted that OFCCP will not interview
all 25,000. Rather, OFCCP wants contact information for a large number of employees so that it
may interview some limited number of them while hiding the selected informants in plain sight:
Google will have no way of knowing who the informants are.82
When OFCCP explains this “hidden in plain sight” tactic to the employees it contacts, that
should give comfort and confidence to the employees that any chance of retaliation has been
minimized (assuming the employee had such a concern). This should permit employees to be
more forthcoming and candid with OFCCP investigators. That is why its value is significant
even if Google never would retaliate: the concern is with employees’ perceptions and fears,
irrespective of Google’s intentions.83
Consistent with this, the Ninth Circuit recently held as a matter of law that an EEOC subpoena
for contact information was relevant to the EEOC’s investigation, where a question in the
investigation was whether an individual charge of discrimination should be expanded to cover
systemic disparate treatment. EEOC v. McLane Co., Inc., 857 F.3d 813, 815-16 (9th Cir. 2017)
(district court abused its discretion: test is not whether EEOC “needs” the information; rather, it
is whether the information is relevant to the investigation). This left for the district court on
remand the question whether the request for contact information was unduly burdensome. 857
F.3d at 816-17.

82

The system is not perfect. In some instances, the content of the disclosed information might imply or suggest a
source.
83

With perhaps hundreds of managers and thousands of co-workers involved, neither Google nor any similarly
situated contractor could guarantee that none of these people would retaliate.
But the assurance of anonymity via the “hidden in plain sight” technique is somewhat overstated. Government
informant privilege attaches to departmental investigations. But the privilege is qualified. When a hearing on the
merits approaches, if OFCCP plans to rely on an informant’s testimony or the information the informant provided, it
will be required to reveal the identity of the informant and relevant documents, statements, and other materials
related to the informant.
Still, the technique helps. As Director Wipper testified, very few OFCCP investigations result in merits litigation;
any adverse findings generally are resolved by an agreement of the parties. Thus, in most cases, the identities of
informants remain undisclosed.

30

 Turning to that question, I find that, as stated, OFCCP’s request for contact information is
unreasonable in that it is over-broad, intrusive on employee privacy, unduly burdensome, and
insufficiently focused on obtaining the relevant information.84
My concern centers on to extent to which the employee contact information, once at OFCCP,
will be secure from hacking, OFCCP employee misuse, and similar potential intrusions or
disclosures. OFCCP has already collected for 21,114 employees information such as name, date
of birth, place of birth, citizenship status, visa status, salary, and stock grants. That information,
if hacked or misused, could subject tens of thousands of employees to risk of identity theft, other
fraud, or the improper public disclosure of private facts. Adding contact data, such as personal
phone numbers and email addresses, increases the risk of harm to Google’s employees. The
contact information could ease the efforts of malicious hackers or misdirected government
employees.
Director Wipper’s response is that, whatever might have happened elsewhere, OFCCP has not
sustained any data breaches (of which she is aware), and OFCCP gives “high priority” to data
security. Tr. 158. Be that as it may, data breaches in 2015 at the Office of Personnel
Management resulted in the exposure of private, personally identifiable information (including,
for example, fingerprints) for millions of current and former federal employees and applicants
for federal employment. Anyone alive today likely is aware of data breaches surrounding this
country’s most recent Presidential election. The Department of Labor, of which OFCCP is a
part, was recently attacked with ransomware. The same has occurred at other government
agencies and private businesses. Ransomware being used internationally is reportedly derived
from tools hacked from our national security agencies. This Office (OALJ) has been hacked. I
have presided in a case where attorneys from the Solicitor of Labor’s office asserted that a
Departmental investigator released confidential information from the investigative file of an
alleged whistleblower – a release the Solicitor’s office argued was in violation of the Privacy
Act, government ethical rules, and other provisions aimed at protecting confidential, private
information. Despite the efforts it has made, the federal government generally and the
Department of Labor in particular are not immune to hacking or to the improper release
(“leaking”) of confidential, private materials about people involved in departmental
investigations.
An Order from the Department requiring these disclosures is all the more burdensome on
individual employees because the employees are given no due process. These are 25,000 people
whom no one has told their personal contact information might be given to the federal
government, when the government has already collected (for most of them) information on their
reportable (to IRS) income at work and their place of birth, citizenship, and visa status, among
scores of other data points. These people have been given no opportunity to “opt in” or “opt out”
of any disclosure; the decision will be made without notice or any opportunity to be heard.
OFCCP sees its role as “protecting workers’ rights,” yet it has done nothing to ask if any of
Google’s employees objects to the disclosure. Nor does OFCCP appear to have considered

84

OFCCP stated that, if Google does not have the requested contact information for an employee, it need not ask the
employee for the information.

31

 whether it can gather the same relevant information without contact information for all 25,000
employees, and do this without sacrificing the goal of hiding its informants in plain sight.
The second burden is the effect OFCCP’s intrusion into Google’s employees’ privacy will have
on Google’s relationship with these employees. OFCCP asked Google to disclose information
on all of these employees. Complying with its contractual obligations, Google provided
considerable information. OFCCP now seeks individual employees’ contact information for
over 25,000 people. If Google complies with that too, it could well suffer a loss of many
employees’ trust. Even if Google produces the information under compulsion, it might suffer a
loss of trust. Google says that it expends very considerable effort and resources to attract and
retain the world’s best employees. Nothing on this record refutes that. Government action that
detracts from those efforts is a significant burden on Google.
In its closing brief, OFCCP argues that, although there might be unavoidable security risks
affecting the contact information, the federal government needs the information and must be
allowed to have it; otherwise, the Department cannot complete the audit with which it is charged.
I accept this argument in part. But I find that the burden on Google and the burden on its
employees can be mitigated without depriving OFCCP of any information relevant to its
investigation – all while safeguarding the integrity of the process. To accomplish this, OFCCP
must (1) take reasonable steps to protect the information it obtains, and (2) limit the number of
employees whose contact information Google will supply. OFCCP’s Regional Director has
testified under oath as to the first of these requirements; I will address the second.
Although OFCCP did not provide an estimate of how many employees it might actually
interview, I find based on my extensive experience litigating civil rights and employment cases –
both as a government attorney and as a private sector attorney – as well as adjudicating these
cases, that OFCCP likely will interview no more than 100 to 300 employees. There is no
indication that OFCCP has the resources or would choose to interview more. OFCCP can hide
in plain sight the identities of that many employees in a list of 5,000, rather than 25,000. Indeed,
if this wasn’t the largest audit in OFCCP’s Region and one of the largest OFCCP has done
nationally, OFCCP routinely would be satisfied with contact information for far fewer than
25,000 people. But OFCCP must be allowed to choose the 5,000 employees whose contact
information it will get, and it might be entitled to follow-up information depending on the results
of its initial interviews.
In particular, some employees might be able to supply information relevant to the investigation
while others might not.85 OFCCP will have the names of the employees whose information is on
the 2015 snapshot or will be on the 2014 snapshot. OFCCP might therefore be able to identify
particular employees whom it wants to interview. If it is not getting the contact information for
all employees, it must have a way to be certain that it gets the information for those whom it

85

Some employees might choose to decline an interview with OFCCP investigators. Director Wipper testified that
OFCCP wouldn’t want to “force anyone to speak with us if they didn’t want to.” Tr. 153-54. OFCCP has no
subpoena authority. I am unaware of any authority OFCCP has to compel non-managerial employees to submit to
interviews during a compliance review.

32

 knows it wants to interview. Providing OFCCP with a random selection of employees’ contact
information thus might deprive OFCCP of information relevant to the investigation.
For that reason, I conclude that, after receiving the Google’s second snapshot, OFCCP may
submit to Google the names of up to 5,000 employees from among those listed on either
snapshot. Google must provide OFCCP with a list organized by name, showing the personal
address, personal telephone number, and personal email address for each of the employees whom
OFCCP selected. Google is excused from providing any information that it does not have in its
records; by OFCCP’s agreement, Google need not ask its employees or third parties for the
information.
After OFCCP has reviewed the information that it will receive based on this Order and has
interviewed those employees whom it chooses to interview, OFCCP might decide that it wishes
to conduct follow-up interviews with other employees. If that occurs, OFCCP may have contact
information for some of these follow-up interviews but not all. Thus, OFCCP may submit a
second list of up to 3,000 additional names; Google must provide the contact information for
those additional persons.
Together, this should give OFCCP ability to contact – confidentially and without Google’s
knowledge – all employees whom OFCCP believes are likely to have information relevant to the
investigation (plus others whom OFCCP randomly selects), keep those employees hidden in
plain sight, and at the same time protect the private contact information of as many Google
employees as possible. This is calculated to reduce the risk to Google’s employees and the harm
to Google’s employment relationships, while allowing OFCCP to access all the same employees
whom it would interview even if it had contact information for all of Google’s 25,000 employees
listed on either snapshot.86
Salary and job histories and related data. OFCCP seeks the salary history and job history for
each of the 25,000 employees. OFCCP also seeks the starting compa-ratio, starting job code,
starting job family, starting pay level, starting organization, starting position or title, starting
salary, and stock awards for all of these people. All of this – the job and salary histories and the
starting information – is to go back to each employee’s hire date. For a few employees, that was
in 1998, the year Google was incorporated.
Even were I to accept the most generous construction of Fourth Amendment required relevance,
I would not find this request enforceable without a greater showing of relevance. Moreover, the
request raises serious questions about the burden on Google and its employees.
OFCCP asserts that, given the unidentified research about starting pay negotiations and what
OFCCP understands to be Google’s practice of negotiating starting salaries, the cause of the pay
disparity might be negotiations at the time of hire. OFCCP’s theory is that Google perpetuates
these disparities by giving pay increases in proportion to current salaries. OFCCP argues that, to
determine whether this is correct, it must have the requested data. The theory is legally
86

It differs from holdings in other cases because those cases did not involve 25,000 employees. The large number
of employees is what allows OFCCP to contact anyone whom it might want to interview from a list of 5,000 while
still keeping them hidden in plain sight.

33

 questionable and factually unsupported at this point. Even were it to overcome the legal hurdles,
factually OFCCP’s theory is little more than speculation.
OFCCP’s anti-discrimination role can overlap that of the Equal Employment Opportunity
Commission, which is the federal government’s primary enforcement agency for the
employment discrimination laws.87 OFCCP and EEOC exchange information that might be
relevant to the other agency’s responsibilities. Tr. 125.88 But, unlike EEOC, OFCCP’s authority
is grounded on Executive Order 11246, section 503 of the Rehabilitation Act of 1973, 29 U.S.C.
§ 793, and the Vietnam Era Veterans’ Readjustment Assistance Act of 1974, 38 U.S.C. §§ 4211
and 4212. EEOC’s statutory authority is entirely different.89 Executive Order 11246 does not
give OFCCP general authority to enforce anti-discrimination laws in the private sector; its
authority is limited to federal government contractors during the performance of their contracts.
In practice, this has meant that – acting without Congressional approval – U.S. Presidents
starting with Franklin Roosevelt (who issued a predecessor to the current E.O. 11246) have
extended anti-discrimination protection beyond statutory provisions when applied to federal
contractors.90 But it also means that this authority is contractually based and limited to instances
involving federal contractors “during the performance of [the] contract.” 41 C.F.R. § 60-1.4(a).
Consistent with this limitation, OFCCP generally limits compliance review investigations to the
two-year period preceding issuance of the scheduling letter. Tr. 35, 48-49, 121. That generally
leaves to the EEOC enforcement of statutory-based anti-discrimination provisions in the private
sector when the relevant facts are more remote in time. But OFCCP’s compliance manual adds
that OFCCP “may need to review information relating to periods more than two years prior to
the contractor’s receipt of the Scheduling Letter where the potential for continuing violations
exists.” Federal Contract Compliance Manual § 2L at 97. Maintenance of a discriminatory
policy or system into the review period is an example of a continuing violation as the
Compliance Manual defines it. Id. § 7B at 228.91
87

EEOC is the primary federal agency charged with enforcement of the anti-discrimination laws in the private
employment sector. See Equal Employment Opportunity Act of 1972 (Pub. L. No. 92-261, Mar. 24, 1972) (giving
EEOC litigation authority on private sector cases). The EEOC has expanded its long-standing systemic enforcement
program and currently litigates systemic cases in every one of its districts across the nation. See
https://www.eeoc.gov/eeoc/systemic/review. Between 2011 and 2016, the EEOC increased its systemic
investigations by 250 percent. Id. Over the 10 years from 2006 to 2016, EEOC was successful in 94 percent of its
systemic litigation. Id.
88

See 41 C.F.R. § 60–742.2, which formalizes information exchanges between EEOC and OFCCP. OFCCP also
exchanges information with state employment practices agencies. The relevant agency here is the California
Department of Fair Employment and Housing.
89

EEOC enforces Title VII of the Civil Rights Act of 1964, the Pregnancy Discrimination Act, the Age
Discrimination in Employment Act, Title I of the Americans with Disabilities Act of 1990, sections 501 and 505 of
the Rehabilitation Act of 1973, the Genetic Information Nondiscrimination Act of 2008, and other statutes amending
these provisions.
90

For example, in 2014, President Obama added to the list of prohibited acts: discrimination based on sexual
orientation or gender identity. E.O. 13672 (July 21, 2014).
91

Contractors are required to keep personnel records only for two years from the date the record was made or the
personnel action occurred, whichever is later. 41 C.F.R. § 60-1.12(a). It is therefore possible in a compliance
review extending back more than two years because of a potential continuing violation, documents may be
unavailable.

34

 There is no evidence on this record that Google was a federal contractor at any time before 2007.
Nor is there evidence that Google was a federal contractor at any time in 2013 or in the months
of 2014 prior to the formation on June 2, 2014, of the contract that OFCCP is enforcing here.
The record establishes no more than that the Executive Order covered Google at various noncontinuous times starting in 2007 and ending in 2012 and then starting again with the current
contract on June 2, 2014, onward. OFCCP is seeking huge amounts of sensitive data, not only
for many years before the generally applicable limit of two years on review periods; it is seeking
data for many years during which it has no explicit investigative role – unless that role may be
extended on a continuing violation theory. It is even possible that Title VII did not apply to
Google in the earliest years involved, making the relevance of OFCCP’s requests all the more
dubious.92
Legal infirmities. There are also several legal concerns that I will discuss because the parties
should consider them in their conciliation efforts. I will not resolve these concerns in this Order,
as the facts are insufficiently developed and the merits of any findings OFCCP might make
against Google are not before me. But unless OFCCP has considered these challenges and
decided that this is a case for modification of reversal or existing law, it might want to consider
further whether it should redirect its efforts while continuing this investigation.
Timeliness. Google argues that OFCCP may not pursue a continuing violation going back
beyond the two-year period that normally limits an OFCCP compliance review. OFCCP
counters that, under the so-called “paycheck rule,” earlier discriminatory compensation practices
present a new violation with each paycheck, if the current pay of protected persons is reduced on
account of those earlier unlawful policies. OFCCP relies principally on the Lilly Ledbetter Fair
Pay Act of 2009, 123 Stat 5; 42 U.S.C. § 2000e-5(e)(3)(A), superseding Ledbetter v. Goodyear
Tire & Rubber Co., Inc., 550 U.S. 618 (2007).
The Fair Pay Act amended the filing deadline (statute of limitations) provisions in Title VII and
certain other specified statutes to make the “paycheck rule” explicit.93 But, as Congress did not
extend the Act to any other provision of law, a district court declined to apply it to claims under
several other civil rights statutes. Russell v. County of Nassau, 696 F. Supp. 2d 213, 230 (E.D.
N.Y. 2010) (holding time-barred claims under 42 U.S.C. §§ 1981, 1983, 1985, and Title VI of
the Civil Rights Act). Google argues that, as the Fair Pay Act does not refer to Executive Order
11246, it does not apply here and that, if anything, the Supreme Court’s holding in Ledbetter
applies.

92

It is unclear on this record when Title VII first applied to Google. Title VII applies only to employers with 15
employees or more. On September 1, 2014, Google had over 19,000 employees; it had over 21,000 employees a
year later. The record thus establishes that Title VII applied in 2014 and 2015. Undoubtedly, it applied for a good
number of years before 2014 – starting whenever Google had 15 employees. But, as there is no evidence on the
record of when that first happened, it is entirely possible that Title VII did not apply to Google in the year it was
formed (1998) or perhaps a year or two after that.
93

The Fair Pay Act also applies to the Americans with Disabilities Act, the Age Discrimination in Employment Act,
and sections 501 and 504 of the Rehabilitation Act of 1973. EEOC is enforcement of all these statutory provisions,
not OFCCP.

35

 It appears to me, however, that the Fair Pay Act legislatively overrules the Ledbetter to the full
extent of that decision. Ledbetter would have applied to the Executive Order only by analogy;
the case does not directly concern the Executive Order. Similarly, the legislation addresses only
Title VII and statutes related to it. I infer that, to the extent the Court’s decision could by applied
by analogy to a case brought under the Executive Order, Congress’ legislation overruling
Ledbetter can as well.94 I conclude that case has no bearing on the present dispute.95
Nonetheless, if OFCCP eventually makes findings adverse to Google based on a theory of
discrimination in starting pay, it will confront this issue again.
Adverse impact. Of greater concern is the potential legal insufficiency of OFCCP’s theory.
Earlier this year, the Ninth Circuit rejected an assertion that an employer’s use of prior salary in
setting starting salaries at hire cannot be justified as a decision based on a factor other than sex
and motivated by a legitimate business purpose. Rizo v. Yovino, 854 F.3d 1161, 1164-66 (9th
Cir. 2017) (reversing and remanding denial of employer’s motion for summary judgment in a
Title VII and the Equal Pay Act case), citing Kouba v. Allstate Ins. Co., 691 F.2d 873, 876-77
(9th Cir. 1982)). The court overturned the district judge’s holding “that prior salary alone can
never be a ‘factor other than sex.’” 854 F.3d at 1165-66. It might be that OFCCP wants to
litigate the merits of this theory as a test case that it will pursue to the Supreme Court.
More likely, the Ninth Circuit’s repeated holding is the reason that OFCCP offers more than an
argument that reliance on pay at a prior job – without more – is adverse impact discrimination.
Rather, OFCCP asserts more: that Google negotiates starting salaries, that research shows that
women proportionately are not as effective negotiators as men (at least when negotiating on their
own behalf), and that this has resulted in the sex-based wage disparity.
The central difficulty with this theory is that OFCCP has spent two days with a group of
investigators interviewing Google’s relevant executives and managers, and it has not established
that Google engages in what most would consider negotiation when it sets starting salaries.
As to campus hires, the record shows – as OFCCP essentially concedes – that Google does not
negotiate starting salaries at all. Google does not consider prior job salaries for campus hires; for
many of these hires, the applicant never had prior job. Google will offer campus hires something
more only if the campus hire has a more favorable offer from a competing firm. But Google
does not offer increased salary; it offers only one-time “sweeteners,” either cash (a signing
bonus) or a one-time stock grant that could vest over time. OFCCP offers no explanation how
these one-time extra payments affect any ongoing wage disparity over the years.
There is a group of support staff with whom Google does not negotiate starting pay.

94

Although the Act specifies which statutes it is amending, it nowhere states that its principles cannot be applied by
analogy to other anti-discrimination provisions. See PL 111–2 [S 181] (Jan. 29, 2009).
95

Still, OFCCP misplaces its reliance on Bazemore v. Friday, 478 U.S. 385 (1986). That case found an employer
liable for pre-Title VII discriminatory policies, but only because the employer continued to apply those policies after
Congress enacted Title VII. 478 U.S. at 391-93. The Supreme Court thus distinguished Bazemore in AT&T Corp.
v. Hulteen, 556 U.S. 701 (2009), because in that pregnancy discrimination case (unlike Bazemore), the employer
withdrew the discriminatory policy as soon as Congress passed the Pregnancy Discrimination Act.

36

 Finally, a group somewhat short of 80 percent of Google’s hires are from other employers in the
industry (“industry hires”). If the applicant has a higher current wage than Google would
normally offer for the job under consideration, or if the applicant has a higher competing offer,
Google might increase its offer. Although Google attempts to bring in the applicant with onetime payments, there are hires whom it offers an increase in salary above the amount it would
generally offer a new hire in the particular job.
But two factors show how this does not support OFCCP’s theory about negotiating starting
salaries. First, the increased offer is determined by a “compensation team” that is not given any
identifying data about the applicant; no team member knows the applicant’s gender; and the team
does not communicate directly with the applicant. Second, this is not an open negotiation: the
only factors that make Google willing to offer something more are greater current pay or a more
favorable competing offer. That is not what most would consider a “negotiation.” Rather, it is
the compensation team’s unilateral response to one of the only two factors it will consider. It is
not a conversation between a hiring manager and an applicant, where the differences – whatever
they may be – between the negotiating skills on average of men and women are said to differ.
Again, the compensation team acts with no information about the applicant’s gender and no
contact with the applicant.
OFCCP’s case is even weaker on starting salaries for Google employees who are promoted. In
most instances, the salary is set by a mechanical calculation of a certain percentage of the market
reference point for the new job. Exceptions occur, but these result in mechanical calculations as
well. If the employee will receive no raise or a raise less than 5 percent in the promoted position,
Google will start the employee with a 5 percent raise. If the employee would get more than a 20
percent raise under the model calculation, the raise is limited to 20 percent. OFCCP’s assertion
that Google negotiates raises of 5 to 20 percent is mistaken; there is no credible evidence to
support it. In all, there is no negotiation when Google sets starting salaries for promoted
employees.
The parties also offered evidence on annual merit increases in salary that Google gives most
employees. OFCCP’s theory has to be that the merit increases are a percentage of existing pay
or otherwise are linked to existing pay in a way that perpetuates the initial sex-based disparities
over time, year-by-year. But that is not what the evidence showed.
On the contrary, as discussed in the findings of fact above, Google’s merit pay increases are
calculated to bring convergence to the pay for workers who are performing equally in the same
job. The starting point is a mechanical calculation based on current pay and the latest
performance rating. But, if two employees are performing equally and one is being paid more,
the policy generally will provide the employee being paid less with a larger increase. Over time,
the two will be paid the same if they continue to perform the same job with the same
performance rating. Although OFCCP protests that the system could take years the salaries
converge, this argument misses the mark. The point is that, over time, any initial salary disparity
resulting from, for example, a practice of negotiating starting salaries, will dissipate, not continue
year-by-year into the present. At the least, Google’s merit increase policy significantly detracts
from OFCCP’s theory that discrimination at hire persists because annual pay increases merely
perpetuate it.

37

 OFCCP showed that, in addition to these policies, Google’s managers have limited discretion to
adjust pay. Google offered testimony that the purpose of this is to enable managers to bring
employees performing equally on the same job into better pay alignment. Of course, once a
manager has discretion, there is always a potential that the discretion will be used for an
unauthorized purpose, including potentially an unlawful purpose. But none of this supports
OFCCP’s theory that a practice of negotiating starting salaries has disadvantaged women and
caused a sex-based wage disparity unfavorable to Google’s employees “pretty much across the
entire workforce.” For that result, it would take more than a few managers misusing their limited
discretion to adjust pay.
I am not deciding that OFCCP’s investigation is at a close or that OFCCP will not at some point
be entitled to information other than the data I discussed earlier in this decision. But OFCCP has
neither offered anything sufficient to refute Wagner’s testimony or shown how its theory has any
grounding in Google’s practices. Despite having several investigators interview more than 20
Google executives and managers over two days and having reviewed over a million
compensation-related data points and many hundreds of thousands of documents, OFCCP
offered nothing credible or reliable to show that its theory about negotiating starting salaries is
based in the Google context on anything more than speculation. It could also be that for
undisclosed reasons OFCCP does not believe Mr. Wagner, but OFCCP makes no argument
against Wagner’s credibility.
Even if OFCCP ultimately does not believe Google’s description of its compensation system, the
first step in determining what might be causing any sex-based disparity is for OFCCP to
understand fully what it is that it is deciding not to believe. The record shows that OFCCP has
not taken sufficient steps to learn how Google’s system works, identify actual policies and
practices that might cause the disparity, and then craft focused requests for information that bears
on these identified potential causes. Without this, the requests become unreasonable: unfocused,
irrelevant, and unduly burdensome.
As I discussed earlier, OFCCP is on a search for the cause of a disparity it has found on a
preliminary basis. Under Directive 307, OFCCP should engage in an iterative process, asking
Google for information, interviewing Google’s officials and managers, reviewing the
documentary materials and data Google has produced, considering information gathered from the
EEOC and the California Department of Fair Employment and Housing, and reviewing
information from any other source it has. Based on that, it should consider Google’s statements
of its policies and practices, such as the testimony Wagner gave. It should determine whether
Google’s representations are consistent with the data and other information obtained. It should
then adjust its models and request further information consistent with observable indicators in
the information it has. Without this process, as outlined in Directive 307, OFCCP’s requests for
information are untethered to any factual basis and are no more than speculation. As such, the
requests impose without rationale a burden on Google and the more than 25,000 involved
employees (whose private information OFCCP seeks). If OFCCP accurately understood
Google’s practices or had evidence to refute – or at least bring into question – Google’s
statement of what those practices are or have been, OFCCP could have made an entirely different
showing on relevance. Lacking that, OFCCP’s showing fails as ungrounded speculation.

38

 When OFCCP obtains the information that a final order in this matter will require Google to
provide, it will be able to consider that information and might decide to pursue again the salary
and job histories it currently seeks. It might be able to present a persuasive showing of
relevance. Nothing about this decision bars OFCCP from pursuing the same requests if OFCCP
can make a showing of relevance sufficient to make its requests reasonable. But it has not yet
done that.
Whatever speculative relevance there might be also fails to outweigh the burden this request
places on Google and its employees. OFCCP has conducted repeated compliance reviews of
Google. In 2007, it conducted a compliance review of Google’s Mountain View facility96 and
could then have sought starting salary information within the two-year review period applicable
to that investigation. This is the information it seeks now, ten years later. For all I know,
perhaps OFCCP did request that information during the 2007 investigation. Whatever OFCCP
did at that time, there seems to be little reason to revisit now what should have been
encompassed in that audit.
The burden on Google is considerable. Google intentionally makes it difficult to access the kind
of information OFCCP is requesting. It stores the information in different electronic locations,
not all together. For the information it has produced thus far, it essentially had to create “back
door” access to extract what was requested while keeping confidential other information. It had
to review every document and each data entry to be certain that it was not divulging confidential
employee information beyond what OFCCP was requesting. Although Google has a litigation
support staff who help respond to discovery demands, complying with OFCCP’s requests
exhausted those resources and required Google to hire an outside contractor. Similarly, although
Google has in-house attorneys, it retained outside counsel in addition. This is a real disruption of
Google’s ordinary, productive processes.
OFCCP offered evidence that it offered to do the work of collecting the information from
Google’s files if Google preferred. OFCCP also argued that Google expended resources
unnecessarily because it chose for its own reasons to review the materials before producing them
to OFCCP.
I reject these arguments. OFCCP could not have accessed the data. Google had to develop
scripts to access it, and OFCCP could not have done that unless Google had made significant
disclosures of proprietary and security information. Such a disclosure would have been an even
greater burden on Google. Even with security information from Google, it is not at all certain
that OFCCP could have developed the needed scripts.
OFCCP goes disturbingly far when it asserts that Google’s pre-disclosure review of the data was
unnecessary and simply Google’s voluntary choice. Nothing brings into question Google’s
testimony that it did this to avoid unnecessary disclosure of employees’ private information.
OFCCP’s advancing its argument is part of what appears to be its persistent neglect of Google’s
96

Tr. 65. There were also compliance reviews at Google facilities (not all Mountain View) in 2010, 2011 and 2012.
Tr. 66.

39

 employees’ privacy. Not everything the federal government might find of interest is something
private individuals want the federal government to see. This is true even when the federal
officials’ purpose is, as OFCCP states of itself, to protect the workers.
As I stated above, information in government files can be hacked, and government employees
can and do misuse it – though thankfully only rarely as far as I know. Google would likely have
legal exposure to its employees if it unnecessarily revealed their private information. At the
least, as I discussed above, it would damage Google’s relationships with employees, Google’s
reputation as an employer, and Google’s ability to recruit and retain the best employees.
Google’s review of materials before disclosing them to OFCCP was a legitimate and necessary
effort to protect its employees, not some detour for the company’s own purposes. Without
OFCCP’s demands – and the breadth and scope of those demands – Google would not have
untaken the document and data review.
If OFCCP wants to correct Google’s policies going forward while compensating adversely
affected employees who worked for Google during performance of the June 2, 2014 contract, it
need not look back 19 years to 1998. It can achieve the same ends going back far fewer years.
The provision in the Compliance Manual allowing OFCCP investigators to look back more than
two years only when a potential continuing violation is at issue does not imply that OFCCP
investigators can look back across decades.
If policies are adversely affecting current employees based on sex, OFCCP should be able to
establish that for a large proportion of all affected employees by looking back three or four years.
If, using its focused, iterative approve, OFCCP finds discrimination consistent with its theory,
OFCCP can look back somewhat further if it chooses, again consistent with the iterative
approach in Directive 307. Or perhaps, instead, it could conciliate with Google and arrive at a
resolution that will compensate adversely affected employees promptly and correct the unlawful
practices discovered. But it is doubtful that it could enforce a request for material going back 19
years under any circumstances in the present case.
OFCCP argues that Google is so large and profitable that no burden is too great. As I have
discussed, the burdens at issue are not merely financial. Even OFCCP’s expert conceded as
much.97 And in any event, Google’s compliance with OFCCP’s earlier demands hindered its
normal operations, and the additional steps that this Order will require imposes a greater burden.
Enlarging that burden further as OFCCP requests goes too far in the light of OFCCP’s
inadequate showing of focus and relevance.
If OFCCP is looking for a test case to develop the law of adverse impact sex discrimination to
include instances of negotiation and anchoring, it must find a vehicle that presents those facts. If
97

OFCCP also goes too far in its argument based on Google’s announcement that it was spending $115 million in
2014 and $150 in 2015 to promote diversity initiatives. 97 OFCCP argues this shows why it is not too great a burden
to impose a cost of compliance in this case that will exceed the $600,000 that GSA has paid in total on the AIMS
contract. This argument suggests an animus that is difficult to understand. What is the policy wisdom of using
Google’s diversity initiatives against it? I would think that the Department would laud government contractors that
spend hundreds of millions of dollars on diversity initiatives, not use those voluntary efforts against these
companies.

40

 OFCCP is looking to remedy conduct going back nearly a decade before Google had its first
government contract – and 17 years before OFCCP initiated the current audit, perhaps OFCCP
should consider referring the matter to the EEOC. If the EEOC agrees with OFCCP’s policy
purposes, it can issue a commissioner’s charge of adverse impact sex discrimination based on
negotiation and anchoring. The charge can be assigned to EEOC’s well-established and highly
successful systemic program.98 What OFCCP cannot do is obtain an order enforcing its demands
for information without a minimal showing of relevance.
As OFCCP’s salary history, job history, and related requests exceed even the considerable
deference owed OFCCP on a determination of relevance, and as they create an unreasonable
burden on Google and its employees, OFCCP will have to do more – along the lines discussed in
the text above – if Google is to be ordered to provide this data.
Recommended Order
1. Within 60 days after this Order becomes final, Google must provide OFCCP with a snapshot
for September 1, 2014, limited as described above.
2. After this Order becomes final, when OFCCP provides Google with a list of up to 5,000
employee99 names and a request for contact information, Google must within 30 days of
receiving this request provide for each named person, the personal address, personal
telephone number, and personal email address of that person if Google has that information
in its records. If Google does not have the information in its records, it need not ask the
employees for it, nor need Google ask third parties to supply the information.100 Google
should designate which employees are managers and which, if any, are in Google’s control
group. OFCCP may not contact or interview any designated managerial or control group
member except consistent with written advice from an attorney at the Solicitor’s office, who
will advise OFCCP about applicable ethical requirements.101
3. After considering information learned under this Order or from other sources, it may make a
request for contact information for up to 3,000 additional employees. The process in the
paragraph above applies.
98

EEOC would investigate the case under Title VII. Its jurisdiction would not face limitations based on when and
whether Google was a federal contractor. It could readily show that the Fair Pay Act applies to supersede the
Supreme Court’s Ledbetter decision because the application of the Act to Title VII is explicit in the Act.
99

These employees must be selected from those who appear on either of the two snapshots (September 1, 2014 or
September 1, 2015).
100

See Tr. 89-90.

101

Regional Director Wipper is an attorney admitted to the California bar. The Department has a number of
investigators who are attorneys. On investigations, attorneys often give specific direction to non-attorney
investigators about what questions to ask particular persons. Director Wipper testified that OFCCP interviews
managerial employees. She said that OFCCP “would only be talking to a manager about their specific experiences
with potential discrimination. We wouldn't be asking anything about confidential information. And we probably
would notify Google before we contacted the managers.” Tr. 42 (emphasis added). This could present issues under
attorney or governmental ethics requirements and should occur only under the direct guidance of attorneys at the
Solicitor’s Office.

41

 4. OFCCP’s request for salary and job histories and the related data listed above is denied
without prejudice to OFCCP’s renewing the request if it can show that the request is
reasonable, within its authority, relevant to the investigation, focused, and not unduly
burdensome. Before re-asserting this request or any similar request, OFCCP must offer to
engage with Google in meaningful, good faith conciliation to resolve any dispute, including
by showing why the information sought is reasonable, relevant, focused, and not unduly
burdensome. If Google offers information that tends to show the request is unreasonable,
irrelevant, unfocused, or unduly burdensome, OFCCP must consider Google’s information,
determine whether it is accurate, determine how it affects the relevance of the request or the
burden of compliance, modify the request if appropriate, and only then may OFCCP go
forward with the request.102 If Google chooses not conciliate or does not conciliate in good
faith, OFCCP has fulfilled its conciliation obligation under this Order.
5. This Recommended Decision and Order will be served on counsel for OFCCP and counsel
for Google by email or facsimile. All other service is by U.S. mail.
SO ORDERED.

Digitally signed by STEVEN BERLIN
DN: CN=STEVEN BERLIN, OU=ADMIN
LAW JUDGE, O=US DOL Office of
Administrative Law Judges, L=San
Francisco, S=CA, C=US
Location: San Francisco CA

STEVEN B. BERLIN
Administrative Law Judge

NOTICE OF APPEAL RIGHTS: To appeal, you must file exceptions (“Exceptions”) with the
Administrative Review Board (“Board”) within ten (10) days of the date of receipt of the
administrative law judge’s recommended decision.
The Board's address is: Administrative Review Board, U.S. Department of Labor, Suite S-5220,
200 Constitution Avenue, NW, Washington DC 20210, for traditional paper filing. Alternatively,
the Board offers an Electronic File and Service Request (EFSR) system. The EFSR for electronic
filing (eFile) permits the submission of forms and documents to the Board through the Internet
102

For example, Google might offer a report of its own multiple regression and other statistical studies that show no
sex-based pay disparity. OFCCP must consider the report and determine whether it is persuasive before OFCCP
presses requests for information based on its contrary studies. This might also give the parties a useful opportunity
to discuss issues such as which employees are similarly situated for comparison purposes. OFCCP is not required to
accept Google’s views on the question, but Google knows more about its organization than OFCCP does, and
OFCCP should evaluate Google’s views before imposing a burden on Google to produce more information. See
Directive 307.

42

 instead of using postal mail and fax. The EFSR portal allows parties to file new appeals
electronically, receive electronic service of Board issuances, file briefs and motions
electronically, and check the status of existing appeals via a web-based interface accessible 24
hours every day. No paper copies need be filed.
An e-Filer must register as a user, by filing an online registration form. To register, the e-Filer
must have a valid e-mail address. The Board must validate the e-Filer before he or she may file
any e-Filed document. After the Board has accepted an e-Filing, it is handled just as it would be
had it been filed in a more traditional manner. e-Filers will also have access to electronic service
(eService), which is simply a way to receive documents, issued by the Board, through the
Internet instead of mailing paper notices/documents.
Information regarding registration for access to the EFSR system, as well as a step by step user
guide and FAQs can be found at: https://dol-appeals.entellitrak.com. If you have any questions or
comments, please contact: Boards-EFSR-Help@dol.gov
If filing paper copies, you must file an original and four copies of the Exception with the Board,
together with one copy of this decision. If you e-File your Exception, only one copy need be
uploaded.
Exceptions may be responded to by other parties. Responses must be filed with the Board within
seven (7) days after receipt of the exceptions. Briefs or exceptions and responses shall be served
simultaneously on all other parties to the proceeding. See 41 C.F.R. § 60-30.36.
After expiration of the time for filing exceptions, the Board is to issue a final Administrative
order which shall be served on all parties. If the Board does not issue a final Administrative order
within thirty (30) days after the expiration of the time for filing exceptions, this recommended
decision shall become a final Administrative order which shall become effective on the thirtyfirst (31st) day after expiration of the time for filing exceptions. See 41 C.F.R. § 60-30.37; see
also 41 CFR 60–30.30 (which is applicable to the Board's review of this recommended decision,
except as to specific time periods).

43