*2 AGEE, Circuit Judges.
Vacated and remanded with instructions by published opinion. Judge GREGORY wrote opinion, Judge which into plant organized produc- wrote a six joined. Judge AGEE MICHAEL mill, mill, departments: tion beam hot cold part concurring separate opinion maintenance, mill, melting, shipping. dissenting part. job opening at the becomes When *3 available, position is advertised over a OPINION bidding system plant-wide posting and GREGORY, Judge: Circuit by personnel depart- controlled the central Employees ment. are allowed to bid on racial allegations of case involves This any department. Although, by in positions manufacturing at a steel discrimination general manager policy, plant’s ap- Carolina, by owned Huger, in South proves promotions all and handles discrim- and Nucor Steel Corporation Nucor investigations, ination and harassment “Nucor”). We find Berkeley (collectively, department that each suggests record its discretion the district court abused manager has unbridled discretion to make denying law in as a matter of and erred department utilizing within his promotions plaintiffs-appel- to the class certification objective subjective whatever or factors he vacate the order and We therefore lants. supervisors There were no black wishes. court for the case to the district remand Equal until after the institution of the certification. Employment Commission Opportunity In- charges preceded litigation. I. deed, supervisor a white testified that his appellants pres- allegations The department manager wore a Confed- —who in of their racial discrimination support ent flag erate emblem on his hardhat —told speak claims and hostile work environment promote him that he would never a black (J.A. and em- supervisors white for themselves: employee supervisor. 1885— referred to black em- ployees frequently 86.) “yard “nigger,” “bologna lips,” ployees as present litigation August arose on The “porch monkey.” employ- White ape,” and 25, 2004, employees when seven at black the black em- frequently referred to ees plant, along employee-plaintiffs with “DAN,” which stood for “dumb ployees as states, by in plants owned other nigger.” epithets These racial were ass (2000) § brought suit under U.S.C. plant-wide sys- over the radio
broadcast
Act
Rights
and Title VII of the Civil
tem, along
“High
with “Dixie” and
Cot-
in the
District Court for the
Monkey noises were also broadcast
ton.”
Arkansas on behalf of
Western District of
system response
the radio
to the
over
approximately
themselves and
one-hun-
employees.
of black
communications
past
present
employ-
dred other
and
black
flag
perva-
display of the Confederate
plant.
litigation
ees at the
At the time the
and items con-
throughout
plant,
sive
commenced,
employees
there
were
logo alongside the Confed-
taining Nucor’s
plant,
at Nucor’s South Carolina
working
flag
plant’s gift
in the
erate
sold
seventy-one
of whom
were black. The
shop. Additionally, several e-mails
District of Arkansas severed the
Western
racially
offensive
depicted
people
by
black
brought
case and transferred the claims
ways,
by showing
as
them with noos-
plaintiffs
such
the seven
South Carolina to
necks,
appel-
were circulated
es around them
the District
South Carolina.
Once,
employee
permanent
injunction,
back
employees.
various
lants seek
compensatory
punitive damages,
and
up
pay,
a noose and told black co-worker
held
attorney’s
and
fees.
him.
that it was for
(4)
7, 2007,
May
representative parties
fairly
On
filed
will
alleging:
adequately protect
motion for class certification
the interests of
the class.
(1)
A pattern
practice
against African-American
treatment
factors,
In our
review these
we bear
respect
op-
with
in mind that
(2)
plant;
pro-
at the
portunities
Nucor’s
requirements
the final three
of Rule
procedure,
motion
which allows white
23(a)
merge,”
“tend to
managers
supervisors
subjec-
to use
typicality “serving
guideposts
to promote employees,
tive criteria
has a
...
determining whether
maintenance of
disparate impact on African-American
*4
a class action is economical and whether
apply for
employees
promotions,
who
the named plaintiffs claim and the class
(3)
requires
and
African-Ameri-
claims are so interrelated that the inter-
employees
plant-wide
can
to work
ests of the class
fairly
members will be
hostile work environment.
adequately protected
and
in their ab-
(J.A. 8980.) The district court denied class
sence.”
certification, and the
plain-
would-be class
Broussard v. Meineke Discount Muffler
appeal.
tiffs now
(4th
331,
Cir.1998)
Shops, 155 F.3d
337
Falcon,
13,
(quoting
457
U.S.
157 n.
102
II.
2364) (alteration
original).
S.Ct.
In-
review
We
the district court’s certifica
deed,
only
“[certification is
concerned with
tion decision for abuse of discretion. Doe
(not
merit)
commonality
apparent
Chao,
(4th
170,
Cir.2002),
306 F.3d
of the claims and the existence of a suffi-
grounds,
on other
aff'd
ciently
group
persons
numerous
who
(2004).
S.Ct.
L.Ed.2d
“[A] may assert
Lilly
those claims.”
v. Har-
action,
any
Title VII class
like
other class
Supermarket,
ris-Teeter
action, may only be certified if the trial
(4th Cir.1983).
332-33
satisfied,
court is
rigorous analysis,
after a
The district court below considered each
23(a)
prerequisites
of Rule
have
23(a)
turn,
of the Rule
requirements in
and
been satisfied.” Gen. Tel. Co.
the Sw. v.
we shall do the same.
Falcon,
147, 161,
(1982).
III. the district court found that satisfied 23(a) Rule of the Federal Rules of Civil requirement because there were nine- Procedure provides following: ty-four employees black who worked at the One or more may members of class through from 2001 2004. Nucor ar- representative sue or be sued as parties gued only that because nineteen of these on behalf all only of members if: employees positions bid on from 2001 to (1) the joinder class is so numerous that numerosity requirement had not impracticable, all members is been met. the district court (2) questions there are or law fact held that the other black should class, common to the be counted in “potential the class because (3) the claims or repre- defenses of the applicants eligible prove are parties sentative typical are of the would have applied for but class, (J.A. 8994.) claims or defenses of the discriminatory practice.” finding, commonality typicality requirements challenge do Appellees 23(a). it to correct. of Rule presume therefore and we (internal omitted). Lilly, citation we B. further concluded that statistical evidence unnecessary commonality: to establish Second, commonality, the district appellants. against ruled court [Pjlaintiffs not, need at the time of the subjectivity in deci- reasoned certification, demon- motion insufficient to es- sion-making alone was strate statistical evidence blacks claim, impact discred- disparate tablish a promoted] at a [or have been terminated presented evidence ited the statistical whites, any than higher rate claim, impact support other differential. Certification is work rejected the hostile environment (not concerned with the court im- claim. find that We merit) apparent of the claims and the direct discounted the properly sufficiently existence of a numerous alone, evidence, which, was sufficient to may group persons who assert those Moreover, commonality. the dis- establish *5 claims.... appel- excluded the improperly
trict court
Lilly,
332-33.
missing
alternative benchmark for
lants’
The district court determined that
in
employment data and therefore erred
statistics were
the
insufficient
to
finding their statistical calculations
Yet,
commonality.
appel
establish
Fi-
commonality.
to establish
insufficient
certainly presented compelling
lants have
finding
court erred in
nally,
discrimination,
direct evidence of
such as
for
on the
insufficient basis
junior
promotions
denials of
when more
environment claim.
hostile work
granted promotions
white
(J.A.
1017),
ability
denial of the
1.
during regular
cross-train
shifts like their
allega-
that
This Court has noted before
(J.A. 1000, 1023),
counterparts
white
practice
tions of “a
treatment
aby
supervisor
statement
white
that he
...
in the exercise of unbridled discretion
promote
employee
would never
black
questions of law and fact common
rais[es]
(J.A. 1885-86).
alone
This evidence
estab
[subject]
employees.” Lilly,
to all
black
common claims of discrimination
lishes
F.2d at
The Fifth Circuit rea-
720
333.
Stastny
of class certification. See
worthy
Trinity
in
similarly Shipes
soned
Indus-
Co.,
F.2d
v. S. Bell Tel. & Tel.
(5th Cir.1993):
tries,
F.2d
(4th Cir.1980) (noting that class certifica
requirements
The threshold
of common-
claim
pattern-or-practice
might
tion for a
ality
typicality
high;
are not
Rule
inferential, statistical,
be based on
or di
23(a) requires only that resolution of the
discrimination).
rect evidence of
That the
questions affect all or a sub-
common
presented
appellants also
statistical data
of the class members.
stantial number
only strengthens their case.
discriminatory em-
Allegations of similar
presented
five
practices,
appellants
such as the use of
ployment
subjective
processes
support
statistical calculations
of their
entirely
personnel
claims,1
discriminate,
we need
operate
satisfy the
discrimination
but
focus
that
First,
average
by comparing
percentage
appellants presented a calculation
positions
plant from 2000
disparity
rates at Nucor
blacks in skill
at the
of the
in utilization
weighted
the same
average
one.
In order to demonstrate
was
as the
on
composition
bidding pools
the racial
of the
job promotions,
disparity
subsequent, post-2000
period.
for
percentage
compared the estimated
sought promotions
blacks who
between
provided
promotions
selected
December 2003
December 1999 and
through
argues
data for 2001
and it
of blacks who
percentage
the estimated
this data should have been
in-
used
promotions during
period.
received
extrapolated
stead of the
1999-2000 data
19.24%,
figure
The former
was
and the
appellants gathered
the extra-
because
figure
ap-
latter
was 7.94%.
polated
change
data from the
of status
rely entirely
actual
pellants could not
forms, which do not
indicate whether a
applicant
making
data
this calculation
position
open
bidding.
for
The dis-
destroyed
Nucor has
such data for
appel-
trict court decided to exclude the
Instead,
promotion periods prior to 2001.2
pre-2001
post-2003
lants’
data and Nucor’s
in an effort
to establish an alternative
rely solely
data and
on the data that Nucor
benchmark,
appellants, using change-
provided
January
through
Decem-
forms,
twenty-seven
identified
of-status
Although
2003.3
argued
ber
during
positions filled
the 1999-2000 time
that the 80% rule was the proper standard
period
extrapolated
from that data
by which to evaluate their statistical evi-
using
assumption
dence,4
the racial com-
the district court determined that
position
bidding pool
jobs
proper
those
standard deviation
was the
percentage
filing
to 2004 with the
of blacks in the
after
of their suit in 2004. Howev-
surrounding population
er,
*6
in 2000 who were
post-lawsuit promotions would not en-
qualified
positions
plant.
for the skill
at the
accuracy
hance the
of the data on Nucor’s
area,
surrounding
up
In the
blacks made
Indeed,
promotion practices.
pro-
such
population qualified
to hold
38.2%
likely
motions would
detract from their accu-
positions.
only
skilled
of
13.42%
racy, since it is no secret that the institution
plant
through
at
the
the
from 2000
litigation
prompts companies
of
often
2004 were black. The second calculation was
change
practices.
post-suit
their
These
first, except
appellants
the same as the
the
changes
only
weight
minimal
with re-
period
extended the time
to 2006 to demon-
gard
plaintiffs’
to the merits of the
claims.
disparity
strate that the
continued.
Co.,
Holsey
See
v. Armour &
disparity
To demonstrate a
in the number of
(4th Cir.1984) ("We
214 n. 5
find no error in
job applications by
applicants,
black
the third
the fact that the district court minimized the
compared
percentage
calculation
the
quali-
of
significance
post-
of evidence of Armour's
surrounding population
fied blacks in the
in
complaint hiring
of black em-
percentage
2000 with the
of blacks who bid
ployees.”). The district court found in the
plant promotions
between 2001 and De-
issue,
appellants'
why
favor on this
which is
percentage
cember 2003. The former
was
data,
post-2003
did not utilize Nucor’s
38.2, and the latter was 16. The fourth calcu-
doing
we find no abuse of discretion in its
so.
Finally,
lation is discussed
the fifth
infra.
compared
percentage
quali-
calculation
the
80%,
four-fifths,
4.
surrounding community
fied blacks in the
rule
"is an adminis-
percentage
2000 with the estimated
by agencies
of blacks
trative rule of thumb used
con-
promotions
who received
from December
cerned with
VII
Title
cases. It offers a defini-
figure
1999 to December 2003. The former
tion of what is a serious difference in the
38.2%,
figure
and the latter
was 7.9%.
passing
protected
rates for
classes.
If the
protected
selection rate for a
class is less than
judgment regarding
2. We
propri-
make no
the
group
of the selection rate for the
80%
select-
ety of Nucor's destruction of the data.
rate,
highest
ed
the
that constitutes adverse
complains
appellants
Serv.,
Nucor
that the
them-
impact.” Chisholm v. U.S. Postal
unduly
selves have created an
restricted data
(4th Cir.1981).
F.2d
495 n. 22
by excluding
promotions
set
the
that occurred
by assuming
applicant
pool
In
cant
to do so.5
its subse-
by which
method
pro-
contained the same
pool for 1974-77
statistics,
the dis-
quent assessment
of blacks and women as did the
portions
had
appellants
court found
trict
pool.” Id. at 940. The
applicant
dispar-
statistically significant
proven
the use of this alternative
approved
Court
against them on
it therefore ruled
ity, and
concluded that it was “the
benchmark and
their claims.
commonality prong of
County’s
proof
most salient
labor
argue that
the district
appellants
case,
Nucor
market.”
Idf
this
since
nega-
improper
court’s decisions
job
destroyed
pre-2001
promotions
their statistics.
tively affected
data,
appellants
likewise were free to
an alternative bench-
attempt
to utilize
involving de
In another Title VII suit
calculations,
mark in order to form their
data,
County
stroyed
United States
its discretion
and the district court abused
(4th Cir.1980), Fair
Fairfax,
was
than two standard deviations.9 The
treatment
and a
itively proven disparate
presented
lants have therefore
valid statis
rather,
question
disparate impact;
independently
tical evidence that
indicates
appellants’
the basis of
discrimi-
whether
disparate impact
disparate
treat
support
nation claims was sufficient
Nucor,
job promotions
ment in
at
and we
Lilly,
See
720 F.2d
class certification.
in-depth
reiterate
that an
assessment of
data, the
excluding
appellants’
In
appellants’
the merits of
claims at
ruled,
assumptions
“These
district court
stage
improper.
would be
See Eisen v.
may
the 1999-2000
be
[regarding
data]
156, 177,
Jacquelin,
Carlisle &
reasonable and the statistics based thereon
(1974);
S.Ct.
L.Ed.2d 732
Thorn
may
prove
relevant to
discrimination at
be
Co.,
Ins.
Jefferson-Pilot Life
However,
necessity
plant.
(4th Cir.2006).
311, 319
We therefore find
their
assumptions
probative
diminishes
appellants’
calculations based
(J.A. 8985.) Yet,
value.”
evidence need
adequate
their alternative benchmark were
even
probative,
not be conclusive
commonality.10
to establish
relatively
proba-
evidence that
is of
weak
may
meeting
tive value
be useful
summary,
appellants’
because the
di-
commonality requirement.
rect evidence alone was sufficient to dem-
included,
onstrate common claims of
treat-
With the 1999-2000 data
the record indicates that
ment and disparate impact, their statistical
provide
during discovery,
purposes,
data
more
Nucor
For our
a threshold of two stan-
argues
appeal
corresponds roughly
that the
cannot
dard deviations
ato 95%
23(f).
discovery
agree
order under FRCP
We
significance,
confidence level or a .05 level of
i.e.,
"jurisdiction
with the Eleventh Circuit that
probability
there
5%
23(f)
granted by Rule
does not extend to [a]
result is due to chance. Three standard devi-
[discovery]
separate
order.” DeLeon-Grana-
equate
ations would
to a
confidence
99.7%
Trees,Inc.,
Chalmer,
dos v. Eller & Sons
F.3d
generally
level. See
Bruce J.
Under-
(11th Cir.2007).
We
n.
therefore ex-
(1987);
standing
Wikipedia,
Statistics 97-98
press
opinion
regarding
deviation,
no
at this time
http://en.wikipedia.org/
Standard
wiki/Standard_deviation
compel
not to
court's decision
(last
visited June
produce
data.
additional
2009).
Supreme
any-
9. Court has
crediting
indicated that
10. The dissent
issue
takes
with our
thing greater than
appellants'
two or three standard devi-
statistics.
we em-
suspi-
phasize
stage,
dealing only
ations
racial discrimination cases is
that at this
we are
cious.
School Dist. v. United
presented
with whether the
Hazelwood
*8
States,
299,
14,
2736,
433 U.S.
308 n.
97 S.Ct.
sufficient statistical information to establish
Partida,
(1977);
commonality.
found in the
favor
Caridad,
held,
Supreme
Court has
tion of their motion. See
v. Met
workplace
permeated
“When the
with
R.R.,
283,
ro-North Commuter
intimidation,
‘discriminatory
ridicule, and
(2d Cir.1999) (“More detailed statistics
insult,’
‘sufficiently
perva
that is
severe or
the Plaintiffs’
might
required
to sustain
to alter the
sive
conditions of the victim’s
statistics],
persuasion,
but [these
burden
employment and create an abusive work
evidence,
conjunction
the anecdotal
environment,’
ing
Title
is violated.”
VII
satisf[y] the Class Plaintiffs’ burden of
Inc.,
17,
Sys.,
Harris v.
510 U.S.
demonstrating commonality
purposes
Forklift
(1993)
114 S.Ct.
agement provided “clear evidence” of “the
The district court determined that
practices.”
decentralized nature of Nucor’s
prove
for the
typicality
order
18.) However,
(Appellees’ Br.
the “clear
claim,
disparate
they
for their
treatment
contradictory
evidence”
cite is itself
satisfy
would have to
the framework set
manager
on this matter. While the
of the
Green,
in McDonnell
Douglas
forth
mill
department
beam
indicated
his
159 em- “example” the district court previously, “note[d]” “[i]n noted As rulings in its on the litigation, abused its discretion conflicts ployment discrimination treatment, disparate disparate appellants’ arise, might example, employ- for between environment impact, and hostile work applicants ees and who were denied em- therefore, claims; of it was also an abuse will, relief, ployment granted and who if the district court to base its discretion for compete fringe for benefits rulings. conclusion on these typicality 23, seniority. Rule the same Under Moreover, supra, as we noted plaintiff represent could not these classes.” Broussard, merge,” typicality and “tend to 331, Setting 100 aside Id. S.Ct. F.3d at and the question language the of whether this was satisfy the presented sufficient evidence controlling, plainly meant to be does not requirement of Rule typicality threshold apply pro- to this which deals with 23(a). and not the competition motions for
“fringe seniority” posed by benefits or new D. hires.
Finally,
adequacy, the district
court can
regarding
plaintiffs’
address
plaintiffs
found the
not to be ade-
injunctive
claims for
or other relief after
disparate
for the
quate representatives
liability and other common issues are de-
disparate impact
treatment
claims.
23(c)(5);
termined. See Fed.R.Civ.P.
Int’l
However, the court could “discern no con-
States,
Bhd.
Teamsters v. United
of
among
plaintiffs
flicts of interest
the
324, 361-62,
S.Ct.
52 L.Ed.2d
regarding
members
hostile work en-
class
(1977) (“[A]s
typical
is
of Title VII
(J.A. 8993.)
vironment claims.”
Given
suits,
pattern-or-practice
question
above conclusions that we have reached
individual relief does not arise until it has
treat-
regarding
proved
employer
been
has fol-
claims,
disparate impact
ment and
we also
employment policy
lowed an
of unlawful
district comb’s assessment of
find
proof
discrimination. The force of that
adequacy
regard
factor with
to these
dissipate
stage
does not
at the remedial
claims was an abuse of discretion.
(“Bifur-
trial.”); Hill,
class. See INSTRUCTIONS. Cir.1986). (3d But this F.2d that the we find stage proceedings, AGEE, concurring in Judge, representatives Circuit adequate are part: dissenting part treatment impact and disparate putative class.
claims of
I. majority opinion in the as I concur the district court previously, As noted 111(A)(the determi- district court’s Section repre- adequate to be found the factor), por- numerosity nation on the work environ- for the hostile sentatives 111(B)(2) reversing the dis- tion of Section dispute do not appellees ment claim. The judgment as to the hostile trict court’s of dis- find no abuse finding, and we concerning claim work environment concluding. court’s so factor,1 cretion in the district commonality portion of Sec- III(C) factor for typicality
tion as to the only, environment claim the hostile work IV. III(D) reversing the district and Section adequacy to the judgment court’s court’s assess- review of the district Our I be- representation.2 23(a) Rule factors leads us ment of the majority fails to adhere to opinion lieve the its discre- that the court abused conclude usurps of review and the role the standard class certification. Given denying tion court, respectfully I dissent of the district above, that the we find our consideration III(C) 111(B)(1), (2), and as to sections requirements of Rule class certification typicality regarding 23(a) regard with have been satisfied disparate treatment and factors for disparate appellants’ disparate impact, impact following claims. For the treatment, environment and hostile work reasons, I would affirm the district court’s vacate the district claims. We therefore judgment grant not to class certification as appellants’ court’s denial of the motion to those claims. certification, and we remand the case class court with instructions to to the district II. action and to certify class consistent engage proceedings further review the district court’s certifica- “We of discretion.” tion decision for abuse opinion. with this facility. How- judgment environment across the whole 1. the extent the district court's To ever, evidence should Nucor adduce relevant regard upon a was based determination showing and uncon- on the merits distinct (C) 111(B)(2) appel- to Sections ap- nected work environments as affects claim did not lants' hostile work environment employees, the pellants and other questions common of law or fact or was may action as Rule take such further typical because the evidence 23(c)(1)(C)permits. separate, unconnected work environ- showed ments, supported that determination is not holdings represented by in the I also concur While Nucor's evidence shows its the record. the district court did not footnote plant departments operation of the was dis- excluding post-2003 discretion in abuse its data, separated among departments, tinct and that Rule and footnote 23(f) probative interlocutory appeal permit that evidence is not as to the effects did not discovery employees’ working alleged on the order. acts
161
jurist
442
Gregory
Capital Corp.,
v. Finova
F.3d
could have made the decision at
Cir.2006)
(4th
188,
issue based on the evidence in the
(citing
190
McClain v.
record.
(4th
Bank,
898,
Nat.
105 F.3d
902
S.C.
great deal of
must
[A]
deference
Cir.1997)).
party
It
seeking
is the
afforded to a District Court’s decision to
prov
who bears the burden of
deny
plaintiff
certification
certification of a
class for
Indeed,
of Rule 23. Lienhart
a class action.
even if
ing
requirements
reason-
(4th
Inc.,
138,
persons
disagree
able
can
as to whether
Dryvit Sys.,
v.
255 F.3d
146
Cir.2001);
proper,
the class certification was
Am. v.
Int’l Woodworkers of
certainly
District
decision
Court’s
does
Chesapeake Bay Plywood Corp., 659 F.2d
Cir.1981).
not amount to an abuse of discretion.
(4th
1259, 1267
“A court has
deciding
discretion in
whether to
broad
Poe,
1370,
47 F.3d
Simmons
1381-82
allow the maintenance of a class action.”
(4th Cir.1995).
cannot hold that the
“[W]e
ESB, Inc.,
1343,
Roman v.
550 F.2d
1348
expand
district court’s refusal to
the class
Cir.1976).
(4th
discretion,
... was an
though
abuse
clearly
taking
reasons
existed for
the other
A district court has abused its discre-
course,
taking
surely
would not ei-
guided by
if
tion
its decision “is
errone-
ther have abused discretion.” Lewis v.
legal principles”
upon
ous
or “rests
a
Mills, Inc.,
561,
Bloomsburg
773 F.2d
564
finding.”
factual
clearly
erroneous
(4th Cir.1985).
AB,
Westberry v. Gislaved Gummi
178
appeal
granting
On
from
order
or
(4th Cir.1999).
257,
F.3d
261
We do not
certification,
denying
appellate
court
ask whether we would have come to the
novo,
reweigh
does not
the evidence de
but
if
conclusion as the district court
same
is to accord the district court’s decision
the matter
examining
we were
de novo.
Lienhart,
“broad discretion.” See
255
Fellheimer,
Braverman,
Eichen &
P.C.
(Within
F.3d at 146
the framework of Rule
(3d
Techs.,
1215,
v. Charter
57 F.3d
district court
“[a]
has broad discretion
Cir.1995). Rather,
reviewing
after
deciding
certify
whether
a class.”
record and the reasons the district court
(internal
omitted)).
quotations
decision,
offered for its
we reverse for
core,
At its
the district court determined
of discretion if we form “a definite
abuse
could not meet either the
and firm conviction that the court below
typicality
factors for certi-
judgment
committed a clear error of
23(a)
fication under Rule
upon weigh-
conclusion it reached
a
failed
proof
to meet their burden of
as to
ing
Westberry,
of the relevant factors.”
the direct or statistical
evidence
dis-
While
Amchem,
matters,
is not a
a certification determination
521 U.S. at
relevant
underlying
the merits of the
judgment
L.Ed.2d
claims,
pleadings-based
it a
de-
neither is
conducting “rigorous analysis”
of such
*13
by any
that can be verified
termination
matters, Falcon,
161,
at
102
457 U.S.
evidence, no matter how deficient. Cer-
2364,
making “findings”
and for
S.Ct.
tainly
Supreme
the United States
since
requirements
that the
of Rule 23 have
Telephone
in General
Co.
Court’s decision
23(b)(3).
satisfied,
been
see Fed.R.Civ.P.
2364,
Falcon,
72
102 S.Ct.
Moreover,
only
if courts could
consider
(1982),
courts have
740
L.Ed.2d
“parties
then
would have
pleadings,
“rigorous
to undertake a
required
been
inject
latitude to
issues to
wide
frivolous
just
plaintiffs’
claims
analysis” finding
predo-
bolster or undermine
support
but of the evidence to
pled,
& David S.
minance.” Robert G. Bone
appropri-
to make an
those claims
order
Evans, Class
and the Sub-
Certification
on Rule 23 certification. 457
judgment
ate
Merits,
51 Duke L.J.
1269
stantive
(“[A]
161, 102
Title
U.S. at
S.Ct.
VII
(2002).
action,
action,
any
like
other class
may only be certified if the trial court is
23(c),
originally
Rule
which
re-
When
satisfied,
analysis,
after a
that the
rigorous
quired certification orders to be made
23(a)
prerequisites of Rule
have been satis-
practicable
“as soon as
after commence-
fied.”). “[Sjometimes may
necessary
it
be
action,”
ment of
was amended in
[the]
probe
pleadings
the court to
behind the
require
to
the court to determine
coming to rest on the certification
before
early practica-
“at an
class certifications
question.” Id. at
S.Ct.
time,”
Advisory
ble
Committee
bar,
In the case at
the district court
explained
preexisting
Civil Rules
duty
performed
rigor
followed its
and
longstanding practice
prompted
analysis, although with a result that
ous
change:
if
majority may
not have reached
may
gather
Time
needed to
infor-
be
sitting
were
as the trier of fact. This
necessary make the certifi-
mation
to
explaining
has been clear in
that an
Court
objection
examining
to
of a
Although
the merits
cation decision.
an evalua-
claim,
class certification
as reflected
probable
tion of the
outcome on the
majority opinion’s citation to Eisen v. Car
part of the cer-
properly
merits is not
Jacquelin,
lisle &
decision, discovery in
tification
aid of
(1974), supra
The Eisen relied, require rule, trict court does not findings and if made in connection accept plaintiffs’ pleadings court to when requirements overlap those find- assessing whether a class should be cer ings that will have to be made on the Eisen, Supreme tified. Court s, overlap merit such coinciden- preliminary held that the district court’s tal.
hearing on the merits
the case—con
365-66.
cluding
plaintiff
was “more than
likely”
prevail
inappropriate
majority’s
—was
citation to Thorn v. Jef
purpose
determining
whether
Co.,
Ins.
L.Ed.2d 351 class deter Thus, the district court’s examination generally mination involves consider and evaluation of plaintiffs’ evidence to ations that are enmeshed the factual support the class certification claim was legal comprising issues plain (internal only appropriate, required but under quotation tiffs cause of action” omitted)). Gariety. Falcon and That the majority marks conclusion, would have reached a different Thus, while an evaluation of the merits lacking albeit all advantages of the to determine strength plaintiffs’ district court’s knowledge first-hand part analysis, case is not of a Rule 23 nothing establishes because the stan spelled out in factors Rule 23 must goes dard of review awhat reasonable through findings, be addressed even if found, jurist could have may very which they overlap with issues on the merits. well lead to differing views on the same prohibition against Eisen’s set assessing Morris, (“We of facts. plaintiffs’ See 448 F.3d likelihood of success on the do not ask whether we part merits as of a Rule 23 certification would have come to conclusion does not mean that consideration of same as the district court if novo.”) necessary examining facts we were to Rule 23 determina- the matter de Fellheimer, Braverman, merely tion is (citing foreclosed Eichen & Techs., required are to be as proved part of the P.C. v. Charter (3rd Cir.1995)). merits. The under Rule 23 interview and selec- jective nature of the view, majority opinion steps my promotion process.”) improp- stage tion of Nucor’s of review and the standard beyond added); Appellant Br. of at 37 (emphasis to achieve erly reweighs the evidence so (“Plaintiffs rely on the mere exis- within the did not properly that was a certification subjective promotion procedure, deny. tence of a discretion district court’s statistical, presented a rule of man- a combination to sanction but majority appears showing that where a rea- anecdotal and direct evidence certification even dated class racially discriminatory jurist properly procedure find such sonable could Nonetheless, as plant.”). in- evidentiary throughout basis was plaintiffs’ named below, limited appellants’ a result is not consonant noted when sufficient. Such examined, a reasonable standard of review. direct evidence appellate with the failed to jurist could find that evidence III. proof as to appellants’ meet the burden typicality for the either majority opinion initially rests on of all African-Amer- certification of class direct the conclusion plant employees. ican require evidence alone was sufficient *15 dispa- majority appears certification of the claims for Similarly, opinion class the impact. Su- rate treatment and in to sustain reweigh to the evidence order majority I the pra agree at 152-53. As noted appellants’ the statistical model. below, could be suffi- opinion that direct evidence within its the district was cient, a proper support in a to certifi- proper exercise of discretion to accord the corollary statistical weight cation decision without statistical evidence little or no proof, although prac- that is not a common A the certification decision. reason- class disparate impact record, in at con- jurist, tice least this could find the able on Nonetheless, majority’s conclu- specula- text. evidence too appellants’ statistical the direct evidence alone was sion—that foundation so as lacking proper tive and a to merit certification —seems evidentiary meeting sufficient without value in to be argument that odd decisional basis because appellants’ proof. burden of in the by was not made A. The Direct Evidence appellants’ Neither
district court.
Issues,
argu-
or oral
Statement of
briefs
argue
did not
to the dis-
error,
raised,
failure
point
ment
as a
of
“direct evidence” was
trict court
their
grant
to
certification
of the district court
support
finding
alone sufficient to
a
of
on the
evidence alone.
basis
direct
certifi-
commonality
typicality
or
for class
They
argument
presented
below was
cation under Rule 23.
appellants’
direct evidence to “bolster” their statistical
direct evidence “bolstered” their
evidence,
claim,
proof
standalone
for Rule
evidentiary
main
which was their
as
23(a)
See,
However, the
analysis.
e.g.,
purposes.
appellants’
statistical
J.A. 965
(“The
methodology
disposi-
is not the
plaintiffs have also bolstered the sta-
choice
on
the standard of
pattern
practice
appeal.
evidence of a
or
tive issue
What
tistical
examine in
case is
subjective bias with
direct and anec-
review should
both
court, based on the
by
dotal evidence of actual racial bias
the whether the district
....”)
alone, could have reason-
(empha-
decision-makers
direct evidence
(“The
to
added);
commonality ably found that evidence insufficient
sis
J.A. 8473
proof
meet the
burden
typicality
plaintiffs’ pattern-or-prac-
pro-
for the
by
typicality
claim further
the sub-
tice
bolstered
jurist
department.
That a reasonable
in another
posed class.
Even the state-
alleging
supervisor
also found the same evidence ment
bias of a
a
could have
was
department
is irrelevant
statement
manager
sufficient for certification
charge of the beam mill
of discretion standard where the
about
the abuse
in the
mill.
support
opposite
could
beam
evidence also
record,
plain, on this
finding.
appears
It
existence of a valid individual
[T]he
found,
jurist could
that a reasonable
necessarily
claim does not
warrant
did, that the appel-
as the district court
conclusion that
the individual plaintiff
(though
lants’ direct evidence
considered
may successfully maintain a class action.
evidence)
only supplemental
simply
equally
It is
clear
plaintiffs
inadequate
proof.
to meet their burden of
attempt
prove
the existence of a com-
panywide policy, or
even
consistent
23(a)(2)
(3),
appel-
Under Rule
practice
given
within a
department, may
required
prove
lants were
there were
though
fail even
against
discrimination
of law or fact common to the
“questions
one or two
proved.
individuals has been
representa-
and “claims ... of the
class”
Cooper
Richmond,
v. Fed. Reserve Bank of
parties
typical
tive
are
... of the class.”
867, 877-78,
467 U.S.
104 S.Ct.
class,
purported
23. The
all
Fed.R.Civ.P.
(1984)
L.Ed.2d 718
(characterizing the
former and current African-American em-
Falcon);
holding
see also Int’l Bd. of
ployees
plant during
of the Nucor
the rele-
States,
Teamsters v. United
period,
plant’s
vant time
worked in all the
(1977)
al promotion-related consideration, Nucor excluded from February January 2001 to cant data from appellants’ experts calculated the actual to the own According 2006.4 expected versus standard deviation based re- analysis of this data expert, statistical job posting bidding on the actual data - - .84, in a sulted standard deviation from 2001 to 2003 as 1.48.6 In other statistically significant no indicating words, result pro- were unable to disparity expected in the actual versus statistically significant duce evidence of employ- African-American percentage promotions on race discrimination based successfully bidding jobs. on J.A. job posting ees actual data.7 See Castaneda v. Partida, however, object- appellants, 5872-73. The 496 n. S.Ct. (1977) period (recognizing time L.Ed.2d 498 ed to use of the 2001 to 2006 purported Complaint If the class was beam mill lants then filed a Third Amended may employees, January well have been abuse of 2005. J.A. 43. certify not to claims discretion factor). (ignoring numerosily 6. Our case law makes clear that standard proposed pur- that was not the class. The proper deviation is the method of ported plant's class in this is case all analysis. statistical See EEOC v. Federal Re- and, African-American rea- Richmond, serve Bank stated, simply sons the direct evidence in- (4th Cir.1983), grounds, rev'd on other sub sufficient to find an abuse of discretion *17 Cooper v. Fed. nom. Reserve Bank Rich- of court. 867, mond, 2794, 81 ("[W]e (1984) adopted L.Ed.2d 718 the solely 4. The case is concerned proper determining rule that the method for promotions with Nucor’s actions on within 'legal significance’ on the basis of statistical plant's the South Carolina workforce. internal through the evidence is use of the standard Hiring employment or termination not at of ....”) analysis (citing deviation Moultrie v. Moreover, pro- issue. it is uncontested that - Martin, (4th Cir.1982)). only motions at the are from Nucor I. 48 standard deviation resulted from data existing pro- within workforce and factors, employees’ that controlled for two subject motions at issue in this case are those training discipline. When those two fac- by employees. to bid controlled, tors were not the standard devia- - 1.53, ap- tion was calculated which the opinion 5. The district court’s states that suit pellants’ expert statistically testified was not a 25, originally August was filed on significant However, difference. J.A. 5857. court records indicate that suit was originally against filed Nucor Steel on a com- pany-wide 7. The actual data basis in the Western District of from 2001 to 2003 came jobs, portion posting Arkansas from "the notices of the on December 2003. That bids pertaining by employ- of the suit claims that were submitted the individual discrimination ees, bidder, Nucor-Berkeley plant prevailing personnel at the at issue here was and the everybody jobs.” transferred to the Carolina files and that bid on those District South August appel- J.A. 9024. on 2004. J.A. 9007. The expected discounting analysis that differences between val tion that greater ue and the observed number than the 2000 data was without a valid founda tion, significant);purely speculative, or three standard deviations are two and thus not enti Indus., Inc., tled to probative weight. v. Halstead Even the appel War ren (4th Cir.1986) lants’ own experts agreed F.2d 759 n. 17 the 2000 data significant suffered from (recognizing that standard deviation infirmities. greater than two or three excludes Dr. Bradley, Edwin one of the appel- of under-representa “chance” as the cause experts, lants’ statistical stated that “[a]ll tion). statistics involve a comparison between a benchmark of what was expected to occur appellants prevail Since the could not actually happened.” what J.A. 404 promotions based the actual data from added). Thus, (emphasis there were two through they sought to create appellants’ experts, variables Dr. additional data would dilute the actual Bradley Fox, and Dr. Liesl needed to for- - produce data the needed mulate as to what was expected to occur in greater 2.00 or standard deviation. The (1) actually 2000 and happened: what appellants by contended could do so job number of posted promotions available estimated comparing percentage “the for bidding by existing employ- sought promotions who between blacks (2) during ees and whether job those December 1999 and December 2003 with promotions “similarly situated” as percentage the estimated of blacks who (i.e., job defined the district court pro- promotions during period.” received motions for which at least one African- Supra proposed at 154. their American employee applied). The as- (the data”) only data substitute “2000 sumptions required for both un- variables for 2000 estimated because the bulk validity dermine the of the 2000 data. (from 2003), period time 2001 to reflected job for African- postings actual which variable, order to construct the first American were known to have promotion positions the number of avail- applied.8, 9 contended to bid, able in open employee for the district court that if their version twenty-seven relied on “change- promotion statistics the 2000 personnel of-status” forms culled from the (four through period years) was con produced by files Nucor. It is not clear (2001 years sidered instead three to from the twenty-seven record whether the 2003), they required could show the stan change-of-status all, represent forms prima selected, dard deviation to make a facie jobs open case *18 appellants’ of discrimination. It is the 2000 because these forms are not identi- creating means of the 2000 data that vali argues change-of- fied. Nucor dates the district court’s exercise of simply company discre status forms “are record detail, majority opinion explained infra, 8. The recites the additional 9. As in more the "2000 period time for statistical information as "change-of-status” data” of 27 consisted period "1999-2000.” The actual was Decem- employment by forms for taken actions 4, 2001, January quite ber 1999 to not January from December 1999 to 2001 and the months. J.A. 1162 n. 16-17. I refer to that resulting appellants’ projection percent- of the January December 1999 to 2001 time ages ap- of African-American who period simply accurately "2000” to more plied employment positions represent- for the single year appel- reflect which change-of-status during ed forms sought lants to create a statistical data substi- promoted. timeframe and were not tute. posted posi- any change of known whether these were employee’s which documents (i.e. status, promot- by employees). for bid employee whether the was tions” available demoted, record, in- ed, pay Thus, a standard received 5911 n. 1. based on the J.A. crease, Appellee’s it, or was transferred.” that appellants or lack of to conclude Br. at 36. necessary for established the first variable verify their statistical the 2000
Indeed, forms change-of-status highly data is dubious. in the record for 2000 fail to bolster found claim that those forms re- appellants’ variable, For the second whether to bid. only promotion positions open flect positions twenty-seven positions represent example, change-of-status nine forms For situated,” Bradley Fox “similarly Drs. whether the 2000 data based on the nebu- vant to the formulation of statistical evi- provided by positions that is one reason the district court found not that the 2000 data plaintiffs ings through three in probation promotions promotion. increase in forms, seven seven lous er Two 1999 and able dence for the 8406-07, 8416, 8673-74, 8701. Of these (J.A. 8673) appear cember 1999 and (J.A. 8407), difficult, if (J.A. 8397, change-of-status available for positions reflecting promotion open- positions one a bid employee bidding searched for an January January the record identifies of some pay (J.A. 8399) Nucor and located reflects a new hire and not a process. promotion positions 8406) filled for an existing employee. January remaining through personnel 2001.” J.A. 8984 bidding impossible, 2001). type, though only a promotion acquired reflect represented twenty- forms between December dated between De- existing employee. simply reflects an On this claims. and thus rele- but J.A. completion proves five reflect to discern tually (including record, 8397-99, Perhaps twenty- Anoth- those avail- (em- files “the one pra at volved African-American for those racial the same as the ployee actually its plicants) for the” bidding pool [applicants] those on which an African-American em- could seven Fox testified in her defined assumed Dr. Fox: At least one. Counsel: Dr. Fox: Counsel: Counsel: discovery what did lections involved African-American actually bidding lar? these selections involved—these 27 Court’s definition bidding on the composition “promotions” an 154; “similarly “that the racial confirm that one of the [27] Job Okay. Okay. Okay. J.A. 8984. orders the district court had you bid. title. job weighted average Other of the Now, How 2001-2003 situated” do to determine that on the job.10 Drs. deposition selections met many the 2000 data However, through what determine-— composition than the bidding pools Bradley job? J.A. 5853. jobs mean same or simi- employee period. of these as follows: one, jobs and Fox twenty- job of the how (ap- Su- Dr. job se- ac- in- many involved? added). phasis Dr. Fox: I don’t know. twenty- But even if one assumes the You don’t know. You have no Counsel: change-of-status represented forms seven *19 way knowing that? promotion openings actual in Nucor’s Welch, expert, Dr. Finis testified “it is not Dr. Fox: No. change-of-status
10. In addition to assertion that the asserts that even if each form Nucor’s twenty-seven change-of-status may represented promotion, way a forms not there is no represented job posted bidding. "promotion,” know if the was also That that any American bid. means Okay. Is not definition Counsel: situated, postings “similarly which were for situ- that not also similarly does jobs ated” but had no African-American an African- job on which include bidders, bidders, all and therefore white actually bid? American were not included in our calculation. order, That was the Court’s Dr. Fox: Thus, representa- the African-American yes. among “similarly tion bidders for situat- you any Okay. So have made Counsel: jobs provided by Nueor-Berkeley is ed” analysis 27 to your of these attempt necessarily inflated. within the Court’s defini- bring them Welch, According 583. to Dr. J.A. or similar? tion of same necessarily by that applying means destroyed. It’s Dr. Fox: Those were weighted average 2001 to 2003 of the bid that.11 possible to do data, applicants “they to the 2000 pool are Instead, Bradley Dr. 5852-54. J.A. overstating expected of Afri- number jobs “simi- Dr. Fox determined were can-American selections.” J.A. 5912. In similarity on the larly situated” based view, Dr. Welch’s “there is no reason to job title alone. J.A. 5854. applicant assume that the for each of pools assumptions Dr. criticized these Welch twenty-seven] positions necessarily in- [the ... any “it is not known whether because at least African cluded one American.” employees applied” for African-American Thus, n. 4. appellants’ sup- J.A. 5912 job twenty-seven selections.12 J.A. variable, port the record for the second also believed that 5911 n. 1. Dr. Welch twenty-seven positions that used to Bradley may and Fox have missed “Drs. create the 2000 data were positions for of African-Americans into a two selections legitimately it could which be assumed data,” posting in the job that is included applied, least one African-American is as Americans, “African and that there were dubious as the first variable. plaintiff, a named who moved including appellants certainly were entitled to job job to 2001 that is a held prior into a proffer valid statistical evidence for 2000 they posi- at the time won a by employees upon expand which to the standard devia- posting in the data.” J.A. 5911-12 n. tion said, however, analysis. being That tion 2, 3. precedent principle requires no or of law Bradley recognized and Fox also Drs. inherently given unreliable evidence evidentiary weight. The appellants chose proof, a method of statistical but based it Bidding records 2001 to [from 2006] “similarly a trier of fact could determine provided only situat- on evidence jobs inherently faulty.13 Consequently, African- ed” for which at least one was evidence, appellants proffered 13. The district court's observations that the 11. The no from any employee “assumptions may themselves or other Nucor be reasonable (past present) any African-American may be rele- and the statistics based thereon employee any positions bid on ostensi- says. than it J.A. vant” means no more 8985. bly represented the 2000 data. court, context, merely ac- The district possibility knowledging the theoretical Accordingly, 12. Dr. Welch limited his present could relevant evi- job postings containing appli- to "all of the disproved by proposition dence—a the evi- “[p]rior to 2001 cant information” present. dence did applicant regarding we know neither the mix anything regarding we race nor do know qualifications.” J.A. *20 (4th Bank, F.2d Cir. determined Nat’l jurist could have reasonable 1981). contrast, meet their burden the case at bar is the failed to 23(a) through factors proof years for the Rule of actual opposite, of exact with three evidence. Accord- proffered statistical precedent requires is no data. There court did not abuse its ingly, the district in all cases to add an additional a court failing give to by discretion data, in particularly lacking data period of data, resulting devi- and the standard Moreover, Fair- evidentiary credibility.15 evidentiary de- analysis, weight and ation a County, municipality, fax as a was under nying certification.14 statutory duty employment to maintain 1970s, duty in the which it failed to data
IV.
duty
to Nu-
applicable
fulfill. No such
if we
that the 2000 data
Even
assume
County
Fairfax, 629 F.2d at
cor. See
of
(and
resulting
analy
deviation
its
standard
(“In
keep-
n. 4
of the record
violation
sis)
evidentiary
to some
was entitled
Sharing
Revenue
ing regulations of the
by
majori
weight,
precedent
cited
Act,
Act
defendants
and the Crime Control
ty opinion
support
does not
the conclusion
destroyed pre-1978 applications for
had
its discretion
the district
abused
employment.”).
refusing
in
certification as to the
County
Another distinction is
of
disparate impact
treatment
claims.
appeal
an
from the dis-
involved
Fairfax
County
Our decision United States
judgment
trict court’s
on the merits of the
(4th Cir.1980),
Fairfax,
Moreover, significant there are distinc- of Fairfax “in recruit- volved claims discrimination County tions between the case at bar and ment, hiring, assignments, pro- particular A distinction of im- of Fairfax. motions,” sufficiency 629 F.2d at portance “missing” is that data applicant flow data in that case County involved three out of of Fairfax data, hiring pro- involved years the four from which the statistical Wrings, data. Unlike came motions which analysis was to be made. There was no surrounding community from the in Coun- credible in that case which to use basis ty Fairfax, only promotes from year the one of actual data to foster valid existing within its workforce. The data analytical sample. year’s One data was hiring County a universe for simply upon insufficient which to make of Fairfax analysis. larger statistical Accord v. Am. was thus much and without restriet- EEOC Appellants spo- appellants' direct available. have never made a Insomuch as neither the any evidence or statistical evidence met the Rule pointed liation of evidence claim or proof, 23 burden of it was not an abuse of legal duty part have main- on the of Nucor to discretion to determine the two deficient precise tained the records found absent. not, together, forms of evidence could taken certainly While entitled to appellants' proof. meet the burden of data for noth- seek to create substitute ing entitled them to a data substitute based on majority opinion imply 15. The seems to some assumptions. insufficient evidence and invalid purpose part bad on the of Nucor because the promotions 2000 data on actual bid was un- *21 on esti- direct projections variables that made and statistical evi- ing dence failed to show problematic. typi- and mated data cality purported for the bar, contained class sufficient to the data universe case meet the certification threshold. As the limiting extrapolations variables restrictive judgment district court’s was neither current Nucor “clearly based on a erroneous factual find- and who in fact bid on qualified who were ing” legal principles,” nor “erroneous it comprise group. could should be accorded the deference due un- for the County does not stand of Fairfax der the abuse of discretion standard of claim- that class certification proposition review. are entitled to a waiver of the rules of ants meeting for their purposes evidence view, my majority opinion fails to judg- court’s proof. burden of district proper accord the district court deference legal ment was not based erroneous and, instead, in this case reaches factual and was thus not an abuse of principles conclusions de novo that are not within the discretion. proper scope appellate of review an majority’s opinion
court. The could rea- V. sonably interpreted require class cer- a trial appellate long plaintiffs an court reviews tification so as future seek- When upon ing produce any determination an abuse of class certification can court’s data standard, exceeding accord that discretion it must set two standard deviations supported by regard factual unreliability court’s conclusions without to the of that proper process by See United data or the the record deference. which it was de- (4th Pittman, Equally v. 209 F.3d rived. troubling prospect States is the Cir.2000) (The abuse of discretion “stan- that certification is required no matter significant of review mandates a plaintiffs dard how attenuated and insufficient a appellate measure of deference to the direct evidence is shown be. Such an courts.”); calls of trial our judgment ignores pru- see also extension of case law Mason, judgment proper United States 52 F.3d dent discretion of a Cir.1995) (4th (“Under court, the abuse of district which has seen and heard standard, first-hand, may prolific discretion this Court not evidence and con- its that of the judgment merely substitute strains the district court to a minis- court; rather, evidentiary grounding. we must determine terial act devoid of discretion, incongruent whether the court’s exercise of That result is an appel- facts, considering the law and the late court’s adherence to the standard of arbitrary capricious.”). review for abuse of discretion. I would judgment thus affirm the district court’s recently This Court has stated that “[a]t denying class certification on the appel- core, its immovable the abuse of discretion disparate impact lants’ treat- requires reviewing standard ment claims the reasons set forth enough primary show deference to a deci- above, respectfully dissent from the judgment sion-maker’s that the court does majority opinion regard. in that merely reverse because would come to a different result in the first in- Long Evans v. Corp.
stance.” Eaton Plan, Disability
Term Cir.2008).
(4th Based on the record jurist plainly a reasonable could find