The opinion
UNITED STATES DISTRICT COURT
SOUTHERN DISTRICT OF OHIO
WESTERN DIVISION
EQUAL EMPLOYMENT
OPPORTUNITY COMMISSION,
Plaintiff,
Case No. 1:17-cv-515
v. JUDGE DOUGLAS R. COLE
R&L CARRIERS, INC., et al.,
Defendants.
OPINION AND ORDER
“Facts are stubborn things, but statistics are pliable.” Mandala v. NTT Data,
Inc., 975 F.3d 202, 205 (2d Cir. 2020) (citing Mark Twain). This case turns on
statistical analysis that David Neumark, the Equal Employment Opportunity
Commission’s statistician, conducted. The EEOC says his analysis reflects stubborn
facts—that R&L Shared Services and R&L Carriers Inc. (individually “Shared
Services” and “Inc.” and collectively “R&L”) discriminated against women. But R&L
says Neumark’s analysis stretches well beyond the permissible bounds of statistical
pliability and is thus inadmissible. For the reasons discussed below, the Court
concludes that Neumark’s analysis may well be wrong, but that, even if so, it is not
so clearly wrong to preclude admissibility. Rather, its correctness is a matter for the
jury to decide. And that determination also ends up resolving many of the other
outstanding motions here as well.
The EEOC brought this case in 2017, alleging that R&L discriminated against
female applicants for dock loader positions at its Wilmington, Ohio, terminal for
years. Trial is now approaching. Currently before the Court are three motions for
summary judgment: one by Shared Services, another by Inc., and a final one—for
partial summary judgment—by the EEOC. Accompanying these are a host of motions
to exclude the testimony of various witnesses—several expert witnesses and one
deceased lay witness.
Of the witnesses, Neumark is key. Shared Services concedes that if the Court
admits Neumark’s testimony, there is a genuine dispute of material fact. Meaning, if
Neumark is in, Shared Services’ summary judgment motion is out. Likewise, though
it changed its tune in its post-hearing brief, the EEOC agreed at oral argument that,
should the Court exclude Dr. Neumark, there is little left to its case.
For the reasons discussed below, the Court concludes that Dr. Neumark’s
statistical analysis, while open to attack on various points, clears the bar for
admissibility. That means the Court DENIES R&L’s Motion to Exclude Neumark’s
Testimony and Opinions (Doc. 132) and DENIES R&L’s Motion to Strike Neumark’s
March 16, 2022, Declaration (Doc. 163). As a result, the Court also DENIES Shared
Services’ Motion for Summary Judgment (Doc. 130).
Inc.’s Motion for Summary Judgment (Doc. 131) turns on an unrelated issue.
Inc. claims that it is not part of an integrated enterprise with Shared Services. The
Court disagrees and concludes that it does constitute an integrated enterprise. So the
Court DENIES Inc.’s motion.
Then there is the EEOC’s Motion for Partial Summary Judgment (Doc. 134).
There, it claims that four of R&L’s affirmative defenses fail as a matter of law. At oral
argument, R&L seemingly withdrew one of those defenses—the EEOC’s alleged
failure to conciliate, and anyways, the parties have since had a go at mediation. So
the Court DENIES the EEOC’s motion on this front as moot. As to the other three,
the Court concludes that those affirmative defenses fail as a matter of law. The Court
thus GRANTS the EEOC’s motion in part.
As for the remaining motions to exclude, the Court will treat those as motions
in limine, and will thus reserve ruling on them until closer to trial, or even perhaps
until the Court has heard a voir dire of the relevant witness at trial.
BACKGROUND
The procedural history of this case is ponderous. Given the unwieldy record,
the Court limits its discussion of facts to those strictly necessary to this decision.
The EEOC sued R&L in 2017, alleging R&L discriminated against female
applicants for dock loader positions at its Wilmington terminal, violating Title VII of
the Civil Rights Act of 1964 and Title I of the Civil Rights Act of 1991. (Compl., Doc.
1, #3–4). It is bringing a Teamsters pattern-or-practice case—a type of discrimination
case hinged on statistical evidence of discrimination. So the EEOC hired David
Neumark, an economist, who offered such statistical analysis (specifically, a multiple
regression analysis) in expert reports (Docs. 132-1, 125-1). Not surprisingly, R&L
hired its own expert, Paul White, to refute his conclusions (Doc. 127-4), and R&L also
deposed Neumark (Docs. 120, 137).
R&L later moved to exclude Neumark, arguing that he omitted major variables
from his regression analysis and thus his testimony was inadmissible. (Doc. 132).
Shared Services moved for summary judgment, arguing that without Neumark’s
testimony, the EEOC had no case. (Doc. 130). Neumark offered a declaration
defending his initial report. (Doc. 153-7). R&L moved to strike that declaration. (Doc.
163).
Separately, Inc. moved for summary judgment, arguing that it was not liable
for any of Shared Services’ behavior, since it did not constitute an integrated
enterprise with Shared Services. (Doc 131). The EEOC, meanwhile, moved for partial
summary judgment in its favor on R&L’s integrated-enterprise affirmative defense,
along with three other affirmative defenses that R&L offered in its answer. (Doc. 134).
In November 2022, the Court heard oral argument about Neumark and the
summary judgment motions. There, R&L withdrew its conciliation defense after
being informed that the remedy would be to postpone trial and return to conciliation.
Given a rapidly approaching trial, it’s time for the Court to decide these motions.
LAW AND ANALYSIS
I. R&L’s Motion to Exclude Neumark’s Testimony and Opinions
R&L opens its motion by quoting another old saw attributed (some say
wrongly) to Twain—“There are three kinds of lies: lies, damned lies and statistics.”
(Doc. 132, #8972). That’s telling. A jury, not a judge, sorts lies from truths. The judge’s
job is to ensure that testimony, whether honest or a “damned lie,” meets the legal
requirements to present to a jury. See Counts v. Gen. Motors, LLC, 606 F. Supp. 3d
547, 583 (E.D. Mich. 2022) (citing Schechner v. Whirlpool Corp., No. 2:16-cv-12409,
2018 WL 6843305, at *7 (E.D. Mich. Oct. 30, 2018)). Here, Neumark’s testimony does.
A court will admit an expert’s opinion into evidence if it meets three
requirements:
1. [T]he witness must be qualified by knowledge, skill, experience,
training, or education.
2. Second, the testimony must be relevant, meaning that it will assist the
trier of fact to understand the evidence or to determine a fact in issue.
3. Third, the testimony must be reliable.
In re Scrap Metal Antitrust Litig., 527 F.3d 517, 528–29 (6th Cir. 2008) (citing Fed.
R. Evid. 702) (internal quotation marks omitted).
R&L confines its challenge to the last of those—it disputes the reliability of
Neumark’s testimony. The Federal Rules of Evidence provide trial courts “general
standards to assess reliability,” including:
• whether the testimony is based upon sufficient facts or data
• whether the testimony is the product of reliable principles and methods,
and
• whether the expert has applied the principles and methods reliably to
the facts of the case.
Id. at 529 (internal quotation marks omitted). And the Supreme Court, in Daubert v.
Merrell Dow Pharm., Inc., 509 U.S. 579 (1993), “provided a non-exclusive checklist
for trial courts to consult in evaluating the reliability of expert testimony,” including:
“testing, peer review, publication, error rates, the existence and maintenance of
standards controlling the technique's operation, and general acceptance in the
relevant scientific community.” In re Scrap Metal, 527 F.3d at 529 (citing United
States v. Langan, 263 F.3d 613, 621 (6th Cir. 2001)).
Neumark offers multiple regression analyses purporting to show that R&L
discriminated against women when hiring dock loaders. (See Resp., Doc. 162,
#11319–21). What is multiple regression analysis?
Multiple regression analysis is a statistical tool for understanding the
relationship between two or more variables. Multiple regression
involves a variable to be explained—called the dependent variable—and
additional explanatory variables that are thought to produce or be
associated with changes in the dependent variable.
…
In a case alleging sex discrimination in salaries, for example, a multiple
regression analysis would examine not only sex, but also other
explanatory variables of interest, such as education and experience.
Daniel L. Rubinfeld, Reference Guide on Multiple Regression, in Reference Manual on
Sci. Evid. 179, 181 (Michael J. Saks, David L. Faigman, David H. Kaye, Joseph
Sanders, eds., 2d ed. 2000).
Regression is an accepted form of statistical analysis, and courts regularly
admit regression models into evidence. Universal Coin & Bullion, Ltd. v. Fed. Express
Corp., No. 2:12-CV-2778, 2015 WL 12001264, at *5 (W.D. Tenn. June 30, 2015)
(collecting cases). As the EEOC notes, “[b]ecause, ‘as a scientific technique, the
validity and acceptability of a properly performed multiple regression analysis is
widely accepted,’ courts ‘need not examine the four factors set forth in Daubert,’ but
can focus instead on the specific challenges to admissibility raised by the opponent.”
(Doc. 162 at #11330–31 (quoting Est. of Hill v. ConAgra Poultry Co., No. 4:94-cv-198,
1997 WL 538887, at *4 (N.D. Ga. Aug. 25, 1997))). That said, as R&L rightly points
out, this doesn’t mean that courts should abdicate their roles as gatekeepers just
because the offered evidence is a regression analysis. (Reply, Doc. 167, #11537–38).
R&L’s chief complaint here is that Neumark’s regression analysis omits key
“explanatory variables.” (Doc. 132, #8977). “[N]ormally, failure to include variables
will affect the analysis’[s] probativeness, not its admissibility.” Bazemore v.
Friday, 478 U.S. 385, 400 (1986) (Brennan, J., concurring in part and joined by all
Justices)); Conwood Co., L.P. v. U.S. Tobacco Co., 290 F.3d 768, 794 (6th Cir. 2002)
(same). And “[v]igorous cross-examination, presentation of contrary evidence, and
careful instruction on the burden of proof are the traditional and appropriate means
of attacking shaky but admissible evidence.” Conwood, 290 F.3d at 794 (quoting
Daubert, 509 U.S. at 596) (internal quotation marks omitted).
But regression analysis becomes unreliable and therefore inadmissible when
it omits major explanatory variables. Bazemore, 478 U.S. at 400; see, e.g., Kentucky
v. Marathon Petroleum Co. LP, 464 F. Supp. 3d 880, 894–95 (W.D. Ky. 2020);
Universal Coin, 2015 WL 12001264, at *10–12; The Iams Co. v. Nutro Prod., Inc., No.
3:00-cv-566, 2004 WL 5496244, at *5–6 (S.D. Ohio June 30, 2004), on reconsideration
sub nom., 2004 WL 5831566 (S.D. Ohio Aug. 23, 2004).
Perhaps unsurprisingly, the parties disagree whether the variables Neumark
allegedly omitted here are major. The contested omitted variables are:
• whether an applicant applied for “any” position or for the loader
position in particular.
• whether an application is active.
• which Recruiting Specialist hired the applicant.
(Doc. 132, #8973–74). Beyond the omitted variables, R&L says that Neumark also
makes several material errors in the data underlying his analysis, which also make
that analysis inherently unreliable. (Id. at #9004–06). All these alleged flaws lead to
what R&L characterizes as an abysmal R2 value. (Id. at #8997–98).
R-square (R2) is a statistic that measures the percentage of variation in
the dependent variable that is accounted for by all the explanatory
variables. Thus, R2 provides a measure of the overall goodness-of-fit of
the multiple regression equation. Its value ranges from 0 to 1. An R2 of
0 means that the explanatory variables explain none of the variation of
the dependent variable; an R2 of 1 means that the explanatory variables
explain all of the variation.
Rubinfeld, Multiple Regression, in Sci. Evid. at 215–16. The R2 value in Neumark’s
most complete model—which R&L’s expert calculates in percentage form as ranging
from 5.16% to 13.16%—is low. (See Doc. 127-4, #8494). EEOC admits that. (See Doc.
162, #11349–56). R&L says this too renders Neumark’s testimony inherently
unreliable. (See Doc. 132, #8995–98). The Court will address each of these alleged
problems in turn.
a. Major omitted variables
R&L claims that Neumark omitted several major explanatory variables. (Doc.
132, #8984). First, the Court will address who has the burden to show that an omitted
variable is major. Then, the Court will determine whether to consider Neumark’s
March 16 declaration, where he tries to address some of these variables. Finally, the
Court will consider each of the variables to see if they qualify as omitted major
explanatory ones.
i. Burden
The Court begins with the parties’ dispute on the burden of proof. R&L says it
doesn’t have the burden to prove that Neumark’s testimony is inadmissible. Rather,
the EEOC must show it is admissible. (Doc. 167, #11549). This is true. And the EEOC
does address the Federal Rules of Evidence and the Daubert factors to do this. (Doc.
162, #11327–29).
But the EEOC also collects several persuasive cases from sister circuits to
support the notion that, when a party claims that an expert has omitted a major
explanatory variable in a regression analysis, the challenging party ought to provide
some evidence that the omitted variable is in fact major. (See id. at #11331–32 (citing
Hemmings v. Tidyman’s Inc., 285 F.3d 1174, 1188 (9th Cir. 2002); Sobel v. Yeshiva
Univ., 839 F.2d 18, 34 (2d Cir. 1988); Palmer v. Shultz, 815 F.2d 84, 101 (2d Cir.
1987))). This makes sense. Otherwise, the Supreme Court’s Bazemore rule—that
omitting variables in regression analysis seldom affects admissibility—is rendered
meaningless. Parties could avoid that presumption by claiming with impunity that
every omitted variable is major, forcing the proffering party to justify every one of
those variables.
R&L says these cases are inapt, because “each case considered the credibility
of an expert’s conclusion at trial.” (Doc. 167, #11550). That is a very different
procedural posture than considering the admissibility of an expert’s conclusion before
trial, the argument goes. R&L then offers a case that makes this exact point, Freeland
v. AT&T Corp., 238 F.R.D. 130 (S.D.N.Y. 2006), which characterizes Sobel as going
to probativeness rather than admissibility. (Doc. 167, #11550); Freeland, 238 F.R.D.
at 146. Ultimately, Freeland concludes that in proving admissibility, “it is the
proponent who must establish that the major factors have been accounted for in a
regression analysis.” Id. at 145.
This is a good argument and a close call. But ultimately, the Court concludes
R&L is mistaken. While Sobel and its predecessor Palmer both concern the
probativeness of an expert’s testimony, Hemmings explicitly addresses admissibility
and cites Sobel alongside Bazemore while doing so, suggesting that Sobel is in fact
applicable to the admissibility question as well. See Hemmings, 285 F.3d at 1188
(citing Sobel, 839 F.2d at 34) (“[Plaintiff] contends that failure to include [variables]
rendered the analysis inadmissible under [Bazemore]. We disagree. [Plaintiff] did not
prove at trial that any of these factors were important to the subjective and undefined
promotion process or compensation awards. We have recognized that a defendant
may not rest an attack on an ‘unsubstantiated assertion of error.’” (citation omitted)
(emphasis added)). While this was an appeal focusing on the plaintiff’s failure to
provide any assertion of error during trial rather than before trial, recall that the
court explicitly notes that it is analyzing admissibility. This is the standard facing
the Court here.
More troubling for the Court is Freeland, which seems to suggest that the
proponent of expert testimony must prove that it included all major factors. See
Freeland, 238 F.R.D. at 145. But ultimately Freedland is unpersuasive for two
reasons. First, of the cases discussed above, more courts seem to go the other way
(and at the circuit level too). And second, following Freeland would force the Court to
flout the Sixth Circuit’s commands. After all, the Sixth Circuit said: “[i]n order to be
admissible on the issue of causation, an expert’s testimony need not eliminate all
other possible causes of the injury.” Conwood, 290 F.3d at 794 (internal quotations
and citations omitted). If this Court required experts to prove that all variables they
omitted were not major, that would force them to eliminate all other possible causes
of injury, contradicting Conwood. So in considering each omitted variable, the Court
will look for some evidence R&L provides that shows the variable is major.
ii. Neumark’s March 16 declaration
Neumark argues that the omitted variables are not major ones in his March
16 declaration. (See generally Doc. 153-7). R&L says that the declaration should be
excluded, because it is an expert opinion offered years after the close of discovery, in
violation of Fed. R. Civ. P. 26. (See generally Doc. 163). The Court disagrees.
“Nothing in Rule 26 … precludes an expert from revising or further clarifying
opinions, particularly in response to points raised in the presentation of a case. …
Rule 26 must be read in light of its dual purposes of narrowing the issues and
eliminating surprise.” McHugh v. Olympia Ent., Inc., 37 F. App’x 730, 735 (6th Cir.
2002). The Sixth Circuit suggests that when new expert testimony “did not constitute
unfair surprise” and “was not a departure from the general scheme of the [expert’s]
report,” it is not a new expert opinion under Rule 26. See id. at 736.
McHugh “contemplate[s] that Rule 26 should not be read so narrowly to
prevent an expert witness at trial from (1) rebutting the analysis by another expert
or (2) clarifying his or her opinion.” Hochstein v. Microsoft Corp., No. 4-cv-73071, 2006
WL 8066573, at *5 (E.D. Mich. Oct. 13, 2006). And while McHugh “deal[s] with
purportedly ‘new’ expert opinions at trial rather than at the summary judgment
stage[,] … similar reasoning should apply to expert declarations attached to summary
judgment motions.” Id. (citation omitted).
Hence for the purposes of summary judgment, this Court holds that the
proper standard for evaluating Defendant’s Motion to Strike Portions of
the expert declaration in question is to determine initially what, if any,
of Plaintiff's expert's declarations are “new.” By “new,” the Court should
determine initially whether the expert is attempting to rebut the
analysis of another expert or to clarify his or her position, comporting
with the “general scheme” of the report.
Id. (citing McHugh, 37 F. App’x. at 735–36).
Neumark’s March 16 declaration constitutes a clarification “in response to
points raised in the presentation of a case.” McHugh, 37 F. App’x. at 735. R&L’s
expert and R&L during deposition both suggested that Neumark was missing several
variables. (Doc. 127-4, #8471; Doc. 120, #6639–40, 6644). Recall, though, that R&L
did not suggest that the variables were important enough to affect admissibility until
the summary judgment stage. (Resp., Doc. 171, #11765). At that point, Neumark
offered his analysis clarifying why he omitted those variables and rebutting the idea
that they were important or incompatible with his conclusions. (See generally Doc.
153-7). So not only does his declaration not depart from the “general scheme” of his
original opinion, but it doubles down on that scheme. And R&L cannot claim unfair
surprise about his analysis of these variables, because it raised these variables and
now says they are so important as to render Neumark’s expert opinion inadmissible.
All the declaration does is explain why that is not the case.
R&L calls McHugh an “outdated ruling,” because it cites cases which predate
the 1993 changes to the Federal Rules of Civil Procedure. (Reply, Doc. 172, #11787).
That makes no sense. The Sixth Circuit decided McHugh in 2002, well after those
rule changes. The McHugh court was aware of those changes and felt that its holding
was compatible with the new rules. R&L makes no suggestion that, since McHugh,
the rules have again changed.
And while R&L suggests that the Court should disregard McHugh’s “general
scheme” rule because neither the Sixth Circuit nor this Court has applied it since,
(Doc. 172, #11788), many other district courts in our circuit have. See, e.g., Hochstein
v. Microsoft Corp., No. 4-cv-73071, 2006 WL 8066573, at *5 (E.D. Mich. Oct. 13, 2006);
SCA Hygiene Prods. Aktiebolag v. First Quality Baby Prods., LLC, 250 F. Supp. 3d
244, 261 (W.D. Ky. 2017); Counts v. Gen. Motors LLC, No. 16-cv-12541, 2020 WL
6937937, at *9 (E.D. Mich. Nov. 25, 2020); Auto Prop. & Cas. Ins. Co. v. Abernathy
Motorcycle Sales, Inc., No. 118-cv-1077, 2022 WL 567052, at *3 (W.D. Tenn. Feb. 24,
2022). That’s persuasive. So McHugh is good law.
Thus, the Court DENIES R&L’s Motion to Strike Neumark’s March 16
Declaration (Doc. 163).
iii. “Any” position
The Sixth Circuit has suggested that where an expert “ruled out all plausible
alternatives for which he had data” and “accounted for all variables raised by [the
other party’s] own expert,” this is sufficient but not necessary to show that he did not
omit any major variables. Conwood, 290 F.3d at 794.
Consider the first omitted variable—whether applicants were seeking any
open position or only loader positions. As R&L rightly points out, the EEOC’s chief
psychologist himself admitted that female applicants who expressed a desire for “any”
position were more likely to choose non-loader positions. (Doc. 132, #8990). But R&L
does not offer any evidence that this omitted factor is a major factor in Neumark’s
regression analysis—for example, by showing that controlling for the variable
reduces or eliminates the sex differential in hiring. And though Neumark omitted the
variable, he also explained why in his rebuttal report: “there is absolutely no data or
statistical evidence that self-removal is important.” (Doc. 125-1, #7901). And in his
March 16 declaration responding to R&L’s motion to exclude his testimony, he more
directly addressed that variable. He says that “the loader hiring rate for female
applicants who applied for ‘any’ position at R&L is higher than the overall loader
hiring rate for female applicants.” (Doc. 153-7, #10627 (emphasis original)). Meaning,
if he controlled for the variable, he says it “would actually increase the female hiring
shortfall.” (Id. at #10627–68 (emphasis original)). So it cannot be a major omitted
factor.
R&L contests this analysis—and argues that Neumark has omitted some
applications which affect his analysis and conflated raw-data and regression analysis
(Doc. 172, #11790). This may be true, but like the expert in Conwood, Neumark
accounted for the variable, so this ought to be a question for the jury.
iv. Active applications
Turn now to the second omitted variable—whether an application is active.
Applications were active for six months after being submitted. R&L says this is also
a major explanatory variable. (Doc. 132, #8993). Here, R&L offers some evidence that
this is a major variable. In White’s expert report, he shows that splitting the analysis
into six-month pools results in a dramatic reduction in any statistically significant
gender differential in hiring. (Doc. 127-4, #8470).
But again, like Conwood suggests, Neumark accounts for this variable later
on. In his depositions (Doc. 120, #6840; Doc. 137, #9622), his rebuttal report (Doc.
125-1, #7892), and his March 16 declaration (Doc. 153-7, #10628–29), Neumark
explained that he intentionally decided against pooling the data into six-month
increments, citing common statistical practice. This strikes the Court as a reasonable
position—slicing the pool into six-month segments creates small sample sizes,
making statistical significance tougher to come by. The Court understands why R&L
would advocate for this approach, and the Court agrees with R&L that this may be
fertile grounds for cross-examination. But as Neumark has accounted for the
variable, and explained why he made the choice he did, his testimony is admissible.
Whether the jury will agree with that method is a separate issue, but not one for the
Court to decide.
v. Recruiter era
Finally, let’s consider the third omitted variable—which recruiter hired the
applicant. As R&L explains, each Recruiting Specialist had independent hiring
criteria, (Doc. 132, #9002–03), though looking at those criteria suggests that they
were not dramatically different (Doc 130-1, #8871–86). When controlling for who
hired the applicant, R&L’s expert, Paul White, says the gender hiring differential
dramatically reduces. (Doc. 132, #9002). But again, Neumark’s analysis accounts for
this variable as well—in his rebuttal report, he disputes White’s conclusions about
how this variable affects his conclusions. (Doc. 125-1, #7903). Given that he accounted
for the variable as Conwood asks, whether his explanation is meritorious is once
again for the jury to decide, after vigorous cross-examination by R&L.
b. Other material errors
R&L also claims that Neumark’s analysis contains serious flaws which make
it inherently unreliable. (Doc. 132, #9004). These flaws include things like: not
searching for typos of “forklift,” when looking for applications containing forklift
experience, and each of the following, which the Court takes from R&L’s motion:
• Systemic coding errors with respect to the “not signed” factor. It was
always set to 1 for the paper applications in Dr. Neumark’s dataset,
indicating that applicants did not sign their applications. However, the
paper applications were almost always signed.
• Dr. Neumark only utilized the conviction box (yes/no). An applicant did
not check the box but listed several actual convictions, which
erroneously resulted in no conviction in Dr. Neumark’s dataset.
• Dr. Neumark erroneously listed anyone with a flexible salary as an
actual salary of $0.
• There is an application that suggested that someone finished a year of
college, but wrote “none” into the degree field, resulting in no education.
• There is an application that contained “associate’s” in the major field,
but the applicant just wrote “degree” into the degree field, resulting in a
flag of bachelor’s rather than associate’s.
(Id. at 9005–06 (cleaned up)). The Court agrees that these appear to be errors. But
R&L offers no evidence that these errors affect Neumark’s conclusions in any way.
Meanwhile, Neumark offers affirmative evidence in his rebuttal report that
controlling for at least some of these errors would not affect his conclusions. (Doc.
125-1, #7919). Questions of accuracy and weight are left to the jury. R&L has not
shown that these errors make Neumark’s conclusions about discrimination
unreliable. The company will have its opportunity to question him about these
“mistakes” on cross-examination.
c. The low R2 value
R&L argues that the low R2 value here means that Neumark’s testimony must
be excluded, relying on Griffin v. Board of Regents of Regency Univ., 795 F.2d 1281
(7th Cir. 1986). (Doc. 132, #8978). But the district court affirmed by Griffin held that
the testimony there “would not support an inference of discrimination because the R2
was low and because there was a lack of individual examples of discrimination.”
Griffin, 795 F.2d at 1292 (emphasis added). Moreover, the Griffin court was
addressing a somewhat different issue there than presented here. In Griffin, the
question was whether the statistics sufficed to establish a prima facie case of
discrimination, or whether defendant had successfully rebutted any such inference.
So the district court was making “an informed decision about the probative value of
plaintiffs’ statistics.” Id. (emphasis added). Here, by contrast, the question is one of
admissibility at trial. Finally, the Griffin court explicitly noted that “the R2 alone
cannot determine the validity of a model. We recognize that sex discrimination may
be present even though R2 is low.” Id. at 1292 n.23.
This follows the best practices recommended to courts. True, “the R2 value is
appropriately considered in assessing statistical models.” Id. at 1291 (citations
omitted). But, “[a]s a general rule, courts should be reluctant to rely solely on a
statistic such as R2 to choose one model over another.” Rubinfeld, Multiple Regression,
in Sci. Evid. at 216–17. Unlike in Griffin, here there is anecdotal evidence, the Court
is considering admissibility rather than probative value, and Neumark performed
multiple regression analyses, not one regression. Yes, he concedes that the R2’s for
the models are low. But Neumark explains that “[f]airly low R-squared values are
very common when looking at data on individual decisions or outcomes,” such as in
the context of allegedly discriminatory hiring. (Doc. 125-1, #7923). And he also notes
that at least one “famous paper” in the field had R2 values like those he achieved here,
and that “[p]ublished research in economics commonly draws strong conclusions
based on size and significance of effects, without regard to explanatory power.” (Id.).
R&L and its expert may disagree, and R&L can probe this issue on cross-examination.
But, again, the probativeness of Neumark’s report is a question for the jury.
In sum, R&L has identified definite potential shortcomings in Neumark’s
regression analysis. None of them, though, whether considered alone or in
combination, are severe enough to prevent the EEOC from presenting the evidence
to the jury. It will then be up to the jury to decide whether the EEOC or R&L has the
better of the battle of the statistical experts. Thus, the Court DENIES R&L’s Motion
to Exclude Neumark’s Testimony and Opinions (Doc. 132).
II. Shared Services’ Motion for Summary Judgment
The party seeking summary judgment bears the initial burden to show the lack
of a genuine issue of material fact in the record. Celotex Corp. v. Catrett, 477 U.S. 317,
325 (1986). Afterward, the non-movant can avoid summary judgment only by pointing
to evidence sufficient to create a genuine issue of material fact. Anderson v. Liberty
Lobby, Inc., 477 U.S. 242, 248 (1986). In making that determination, though, the
Court must view the evidence in the light most favorable to the non-movant. See
Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 587 (1986).
Resolving Shared Services’ Motion ends up a straightforward task. Though
Shared Services raises many other arguments in the motion, Shared Services
conceded at oral argument that the motion fails if the Court admits Neumark’s
testimony, since that testimony would create a genuine dispute of material fact. And
Shared Services was right to make that admission—the EEOC says that Neumark’s
statistical analysis shows R&L had a pattern of discriminating against women, while
R&L’s statistics say otherwise. That creates a triable issue of fact, leaving to the jury
to weigh the competing evidence. Conwood, 290 F.3d at 794. As explained above, the
Court will admit Neumark’s testimony. Since that creates a contested issue of
material fact, the Court DENIES Shared Services’ Motion for Summary Judgment
(Doc. 130).
III. Inc.’s Motion for Summary Judgment
Inc. moves for summary judgment because it says it cannot be held responsible
for any alleged discrimination by Shared Services. (Doc. 131, #8894). The EEOC,
however, says that the two companies “are so interrelated that they constitute an
integrated enterprise and are thus liable under Title VII.” (Doc. 160, #11272 (citing
Armbruster v. Quinn, 711 F.2d 1332, 1338 (6th Cir. 1983) abrogated on other grounds
by Arbaugh v. Y & H Corp., 546 U.S. 500 (2006))). This motion, then, turns on whether
the two companies constitute an integrated enterprise. The Court finds that they do.
An integrated enterprise exists between companies when they share:
a. interrelation of operations, i.e., common offices, common record
keeping, shared bank accounts and equipment;
b. common management, common directors and boards;
c. centralized control of labor relations and personnel; and
d. common ownership and financial control.
Swallows v. Barnes & Noble Book Stores, Inc., 128 F.3d 990, 993–94 (6th Cir. 1997)
(numbering changed) (citing York v. Tenn. Crushed Stone Ass’n, 684 F.2d 360, 362
(6th Cir. 1984)). “All four criteria need not be present in all cases and, even when no
evidence of common control of labor relations policy is presented, the circumstances
may be such that the Title VII single-employer doctrine is applicable.” Armbruster,
711 F.2d at 1338. At least three of the four factors suggest the existence of an
integrated enterprise. The Court will address each in turn.
a. Interrelation of operations
Inc. and Shared Services share an interrelation of operations. When “two
entities set forth a shared principal address … [it] shows there may be common or
interrelated aspects to their operations.” EEOC v. Care Ctrs. Mgmt. Consulting, Inc.,
942 F. Supp. 2d 771, 779 (E.D. Tenn. 2013). Using the same logo and similar
advertising also suggests interrelation. See EEOC v. Indi’s Fast Food Rest. Inc., No.
3:15-cv-590, 2016 WL 7473130, at *3 (W.D. Ky. Dec. 28, 2016). Interrelation of
operations also exists when “two companies provided exclusive services to each other,
were marketed as twin operations, used each other’s logo and letterhead
interchangeably, issued checks on each other’s behalf, and kept business and
personnel records at the same office.” Swallows, 128 F.3d at 994 (citing EEOC v.
Dolphin Cruise Line, Inc., 945 F. Supp. 1550, 1553–54 (S.D. Fla. 1996)).
The two companies share a headquarters. (Doc 134-1, #9568). The companies
share the same name when advertising to prospective employees—“[t]he on-line
recruiting guide for ‘R&L Carriers’ for dockworkers, drivers, management employees,
mechanics, security, information technology, and hardware support makes no
distinction between R&L Carriers, Inc. and R&L Carriers Shared Services, LLC.” (Id.
at #9575). The service agreement between the two companies specifies that Shared
Services will provide Inc. with “accounting, tax and payroll services; maintenance of
files; legal services; business supplies and the provision of executive and necessary
management personnel.” (Doc. 160, #11274).
Meanwhile, though R&L notes that Inc. has not used the services agreement
since 2008 (Doc. 131, #8901–02), its subsidiaries use those services—including one
wholly owned subsidiary who leases all its employees from Shared Services. (Doc.
134, #9547). And while each characteristic discussed in the cases above was
determinative at the motion-to-dismiss stage, all the characteristics together seem
enough to establish interrelation at the summary-judgment stage. So this factor
favors a finding that the two companies constitute an integrated enterprise.
b. Common management
When two companies share corporate officers, this suggests an integrated
enterprise. See Care Ctrs. Mgmt., 942 F. Supp. 2d at 779. Inc. and Shared Services
shared 29 corporate officers, including their Chairman, CEO, President, CFO, and
many Vice-Presidents. (Doc. 134-1, #9566–67). So this factor also favors a finding that
the two companies constitute an integrated enterprise.
c. Centralized control of personnel
When employees in both companies “report to common authority figures, who
both appear authorized to take some employment action” this suggests that the
companies are an integrated enterprise. Indi’s Fast Food, 2016 WL 7473130, at *4.
While Inc. has no employees, as the previous factor suggests, it shares corporate
officers with Shared Services, including Vice-Presidents of both Operations and
Human Resources. (Doc. 113-7, #5524–25, 5527–28). So the Shared Services
employees reported to officers seemingly wielding hiring and firing power, who
occupied the same positions of authority in Inc. as well. This means Shared Services
employees report to common authority figures of both companies. So the third factor
favors a finding of an integrated enterprise.
d. Common ownership
The Sixth Circuit has suggested that the fourth factor is not met when neither
of the companies is a sham. Swallows, 128 F.3d at 995. The Court does find the
corporate structure a little suspect. Members of the Roberts family own Inc. and
indirectly own Shared Services. (Doc. 134-1, #9565). This fact is clear from deposition
testimony. (See generally Doc. 113-7). R&L repeatedly makes the misleading claim
that Inc. only owns a minor stake of Shared Services (Doc 131, #8898), without
addressing the fact that Inc.’s owners and wholly owned subsidiaries own all the rest
of Shared Services. The EEOC offers a useful chart to illustrate that fact:
gl
Lines, Inc.
R+L Transfer, Florida Transfer
R.L.R. Paramount
‘pa.
Entity Entity Percent of Entity Owned
23
(Doc. 134-1, #9565). That structure could easily be a tactic to put some distance
between Inc. and Shared Services without giving up any ownership interest. But the
Court is reluctant to inquire into motives, so this factor hangs in equipoise.
Still, the other factors suggest that the two companies constitute an integrated
enterprise and the fourth factor does not refute that, so the Court finds that they do.
Thus, the Court DENIES Inc.’s Motion for Summary Judgment (Doc. 131).
IV. The EEOC’s Motion for Partial Summary Judgment
The EEOC moves for partial summary judgment against four of R&L’s
affirmative defenses—whether:
• the EEOC satisfied its statutory obligation to try to resolve determined
violations through informal conciliation before it filed its lawsuit
(Fourth Affirmative Defense)
• the EEOC’s claim is barred by a statute of limitations (Fifth Affirmative
Defense)
• the EEOC’s claim is barred because rejected female applicants did not
exhaust administrative remedies (Eleventh Affirmative Defense); and
• Inc. is an employer because it is part of an integrated enterprise with
Shared Services (Eighteenth Affirmative Defense).
(Doc. 134, #9523–24 (cleaned up)). The Court DENIES AS MOOT the EEOC’s
Motion (Doc. 134) as to the Fourth Affirmative Defense, failure to engage in
conciliation. That is because R&L withdrew that defense at oral argument, after
being informed that the remedy would be to push back its trial date and recommence
conciliation. The Court GRANTS the Motion as to the Eighteenth Affirmative
Defense, given that the Court already determined above that Shared Services and
Inc. constitute an integrated employer. What remain are the Fifth and Eleventh
Affirmative Defenses. Those are fairly straightforward to resolve.
The statute of limitations to which the Fifth Affirmative Defense refers applies
to individual plaintiffs, but “imposes no limitation upon the power of the EEOC to file
suit in a federal court” itself. Occidental Life Ins. Co. v. EEOC, 432 U.S. 355, 366
(1977). As the EEOC brings this suit of its own volition as a pattern-or-practice case,
the statute of limitations does not apply. So the Court GRANTS the EEOC’s Motion
as to the Fifth Affirmative Defense.
Similar logic applies to the Eleventh Affirmative Defense. The Sixth Circuit
has held that the EEOC can bring a discrimination action on behalf of individuals
who did not themselves pursue any discrimination claims. See EEOC v. Keco Indus.,
748 F.2d 1097, 1101 (6th Cir. 1984). This means there cannot be an exhaustion
requirement when the EEOC itself brings the action. So the Court also GRANTS the
EEOC’s Motion as to the Eleventh Affirmative Defense.
CONCLUSION
For these reasons, the Court DENIES R&L’s Motion to Exclude Neumark’s
Testimony and Opinions (Doc. 132), R&L’s Motion to Strike Neumark’s March 16,
2022, Declaration (Doc. 163), Shared Services’ Motion for Summary Judgment (Doc.
130), and Inc.’s Motion for Summary Judgment (Doc. 131). The Court GRANTS in
part the EEOC’s Motion for Partial Summary Judgment (Doc. 134), as to R&L’s Fifth,
Eleventh, and Eighteenth Affirmative Defenses, but DENIES it as moot as to R&L’s
Fourth Affirmative Defense, given that R&L withdrew that defense at oral argument.
SO ORDERED.
March 27, 2023
DATE DOUGLAS R. COLE
UNITED STATES DISTRICT JUDGE
26