UNITED STATES COURT OF APPEALS
FOR THE FIFTH CIRCUIT
No. 97-60708
AVONDALE INDUSTRIES, INC.,
Petitioner/Cross-Respondent,
v.
NATIONAL LABOR RELATIONS BOARD,
Respondent/Cross-Petitioner,
Petition for Review for Enforcement of the
order of the National Labor Relations Board
July 7, 1999
Before REYNALDO G. GARZA, JONES, and DeMOSS, Circuit Judges:
EDITH H. JONES, Circuit Judge:
Avondale Industries, Inc., a shipbuilding company,
engaged in a short but hotly-contested union representation
election with the New Orleans Metal Trades Council, AFL-CIO (“the
Union”), which was vying for a bargaining unit of approximately
4,000 employees. The issue that is determinative on appeal is
whether the National Labor Relations Board enforced voter
identification procedures that sufficiently protected the integrity
of the election. Under the circumstances presented here -- a very
large work force, the NLRB’s foreknowledge that a substantial
number of the votes would be challenged, observers unable to be
personally acquainted with the voters, multiple voting zones, an 11
1/2-hour election day without rigid controls over employee access
to the polls, serious allegations of improper campaign tactics, and
a close result -- we conclude that the voting identification
procedures were fatally flawed. The NLRB’s bargaining order may
not be enforced, and a new election must be conducted.
I. FACTUAL AND PROCEDURAL HISTORY1
On May 11, 1993, the Union petitioned the NLRB for a
representation election at Avondale. Following a representation
hearing, Avondale and the Union executed a consent agreement for an
expedited election to be held on June 25, 1993 -- sixteen days
after the agreement was consummated.
Two weeks before the election, Avondale supplied the
Union with an Excelsior2 list of eligible voters. The Excelsior
list included the last name, first initial, and home address of the
eligible employees. Although the Union requested a more detailed
Excelsior list containing the first name and employee number of the
1
On appeal, Avondale has challenged several procedural aspects of the
representation election including the NLRB’s certification of the bargaining
unit, the challenge ballot process employed during the election, and the Union’s
allegedly racially discriminatory use of challenges. Avondale also asserted
several due process claims regarding the post-election hearing. Because we
decide this case on the issue of voter identification, a complete discussion of
the other issues underlying the appeal is unnecessary.
2
Excelsior Underwear, 156 N.L.R.B. 1236, 1239-40 (1966) (requiring
employer to turn over list of the names and addresses of all eligible voters in
certification election). The Excelsior list provided by Avondale listed all
eligible voters, including a group of employees whose unit status was not yet
determined -- the Not On List (“NOL”) employees.
2
eligible voters, Avondale initially refused to provide the
information.3 On the eve of the election, however, Avondale
revised its position and offered to place employees’ first name and
employee number on the identification lists to be used during
actual balloting. The Union rejected this offer.
Electioneering was contentious and bitter. Moreover,
because the company and union could not agree on the size and
composition of the bargaining unit, the NLRB was aware that
hundreds of NOL employees would have to cast challenged ballots.
As the election approached, the Union threatened to issue up to
1,000 challenges. Particularly troubling are allegations that the
Union made racial appeals to win votes.
The election was held at Avondale as scheduled. Polling
was conducted at five voting zones dispersed within the sprawling
industrial facility. Each of the zones was manned by three NLRB
officials and four election observers -- two from Avondale, and two
from the Union. Two lists of employees were furnished. A “Master
Voting List” enumerated all employees (except the NOL employees)
eligible to vote in the election, identifying them by their last
name, first and middle initial, and address. The “Zone Voting
3
When created, the Excelsior list provided by Avondale complied with
controlling NLRB precedent. In 1994, the NLRB revised the Excelsior list
requirements. Now, an employer is required to furnish the full name of all
eligible employees. See North Macon Health Care Facility, 315 N.L.R.B. 359, 361
(1994) (finding an employer’s refusal to supply only the last name and first
initial of eligible employees constituted objectionable conduct).
3
List” employees broke down the Master List to identify employees
who were assigned to vote in a particular zone.
Supervisors escorted most employees to the voting zones
during regular work times. When an employee presented himself to
vote, the employee identified himself by name at the check-in
table. If the employee could not be identified by name, the
observers were advised to ask the employee’s address or to identify
him by his identification badge. Avondale, a defense industry
contractor, enhanced plant security with employee identification
badges that contain the employee’s first name and employee number
and a thumbnail-size photograph. When the employee’s name appeared
on the Zone Voting List, the list was marked and the employee
allowed to vote. If the employee could not be identified, or was
challenged by the Union “for cause,” or was an NOL employee, the
employee voted subject to challenge. Eventually, 850 challenges
were issued. Given the 3,000 unchallenged voters, complaints were
made against approximately one out of every four voters.
While initial results suggested a 600-vote margin of
victory for the Union, Avondale’s defense of the challenged ballots
was surprisingly successful. The NLRB hearing officer counted over
70% of the disputed, counted ballots for the employer, reducing the
4
Union’s victory margin to about 250 votes. A swing of 130 votes
would reverse the election results.4
Pertinent to this appeal, Avondale challenges the
integrity of the election in two ways. First, the company
complains that NLRB refused to enforce any system of routine voter
identification beyond voluntary self-identification. In practical
effect, no one can be sure who voted in the representation
election. Second, the NLRB disabled Avondale’s post-election
investigation of this issue, as it steadfastly withheld copies of
the marked voter list that would have revealed who voted, how often
they may have voted, and the status of challenged voters. After
Avondale successfully prosecuted a Freedom of Information Act case
up to this court to obtain the voter lists, and then analyzed the
lists, uncovering potentially suspicious voting involving hundreds
of ballots, NLRB refused to reopen the certification hearing and
summarily dismissed this part of Avondale’s claim. The NLRB,
acknowledging that no routine voter identification system was in
place, responds that it used “standard voter identification
4
The original count was 1804 for the Union and 1263 for Avondale, with
850 challenged ballots. Five hundred fifteen challenged votes were counted, with
Avondale receiving 369 of these votes to 146 for the Union. See ibid. Fifty-
nine votes remained undetermined, which, under established Board precedent, are
assumed to be added to the lower total to determine whether the election can
stand. See, e.g., NLRB v. Pinkerton’s, Inc., 621 F.2d 1322, 1329-30 (6th Cir.
1980); Byers Eng’g Corp., 324 N.L.R.B. 125 (1997); Wolverine Dispatch, Inc., 321
NLRB 796 (1996). Accordingly, the final vote total was 1950 for the Union to
1691 (1632 plus the 59 undetermined voters) for Avondale, for a gross difference
of 259.
5
procedures,” that Avondale’s arguments are exaggerated, no evidence
of voter fraud exists, and the election should not be overturned on
speculation.
II. ANALYSIS
An order requiring an employer to negotiate with a union
will be enforced if the NLRB’s decision to certify the union is
“reasonable and based on substantial evidence in the record.”
NLRB v. McCarty Farms, Inc., 24 F.3d 725, 728 (5th Cir. 1994). The
certification order’s enforceability depends, in turn, on the
validity of the underlying election. See id. (citing NLRB v. Hood
Furniture Mfg. Co., 941 F.2d 325, 328 (5th Cir. 1991)). If the
union was not certified properly, this court may refuse to enforce
the unfair labor practice order and remand the proceedings to the
NLRB. See Deming Div., Crane Co., 225 N.L.R.B. 657, 657 n.3
(1976).
A representation election is presumed to be fair and
regular, unless proven otherwise. See NLRB v. Mattison Mach.
Works, 365 U.S. 123, 124, 81 S. Ct. 434, 435 (1961). In overseeing
a representation election, the NLRB aspires to “an ideal atmosphere
in which a free choice may be made by employees, protected from
interference by employer, union, Board agent or other parties.”
Home Town Foods, Inc. v. NLRB, 416 F.2d 392, 396 (5th Cir. 1969)
(citations omitted). While “laboratory conditions” represent the
ideal, “clinical asepsis is an unattainable goal in the real world
6
of union organizational efforts.” NLRB v. Sumter Plywood Corp.,
535 F.2d 917, 920 (5th Cir. 1976). Thus, a reviewing court must
compare the conduct of a challenged election to the laboratory
conditions paradigm, recognizing the contaminating influence of
“the realities of industrial life.” Exeter 1-A Ltd. Partnership v.
NLRB, 596 F.2d 1280, 1283 (5th Cir. 1979).
When examining the voter identification procedures
employed in a representation election, this court does not sit to
determine “whether optimum practices were followed, but whether on
all the facts the manner in which the election was held raises a
reasonable doubt as to its validity.” NLRB v. ARA Servs., Inc.,
717 F.2d 57, 68 (3d Cir. 1983); see also Polymers, Inc., 174
N.L.R.B. 282, 282-83 (1969). Even under this deferential standard,
however, reasonable doubt means “reasonable uncertainty,” not
“disbelief” or “conclusive proof”. Allentown Mack Sales and
Service, Inc. v. NLRB, 522 U.S. 359, ____, 118 S. Ct. 818, 823 &
n.2 (1998). Voter identification procedures appropriate for
representation elections in small units may be inadequate when the
eligible voting pool becomes very large. As the NLRB Casehandling
Manual suggests, “[Voters] may also be asked for other identifying
information, as appropriate and as formerly agreed on.” NLRB,
Casehandling Manual (Part Two), Representation Proceedings, §
11322.1 (emphasis added) (“NLRB Casehandling Manual”)
7
Thus, in Monfort, Inc., 318 N.L.R.B. 209, 209-10 (1995),
NLRB rejected challenges to individuals who claimed their names had
been crossed off the voter lists before they arrived at the polls,
where employee identification cards were used to assist in
identification in a 1,500 employee unit. More on point with this
case, in Newport News Shipbuilding and Dry Dock Co., 239 N.L.R.B.
82, 88 (1978), the election officials utilized the last four digits
of social security numbers to identify voters in a 19,000 employee
unit. NLRB relied heavily on this method of identification to
prove the overall accuracy of the election. As a result, the Board
emphasized that limited deviations from the procedure -- whereby
observers asked “only” for voters’ names for a few minutes at one
polling site when polling was very heavy -- could not prejudice the
overall election results. See id. NLRB described the Newport News
voting procedure as “certainly consistent with established Board
practices.” Id.
In the Avondale election, by contrast, only the voter’s
name was required to permit an employee to vote. There is no
dispute that this was the routine practice followed in the
election. During the post-election hearing, after Avondale had
introduced voluminous testimony on the practice, the NLRB hearing
officer entered an order precluding further similar evidence as
“cumulative or irrelevant.” NLRB justifies verbal self-
identification as its norm and standard procedure, evidenced by the
8
testimony of Assistant Regional Director (“ARD”) Joseph Norton, the
NLRB’s director for the Avondale election:
[N]ot only in this election, but in any
election, we assume that when John steps up to
the table and he is asked his name and he
says, I am John Doe. And [election observers]
say, where do you live? And he says, I live
at 102 Smith Street, that this is the John Doe
that lives at 102 Smith Street. What further
identification do you need. There is a
presumption, if you will, that [voters] are
not going to walk in to lie. Moreover, in
this particular situation, [voters] are
actually being controlled from the point of
where they worked to the voting area. Why
would [voters] not be telling the truth? Why
would voters lie as to their name and where
they live? (emphasis added).
Parenthetically, it must be added that asking employees to give
their addresses for identification purposes was not, contrary to
ARD Norton’s testimony, the routine procedure employed in the
Avondale election.
NLRB attempts to deflect its responsibility to protect
the integrity of the ballot by noting that the Union and Avondale
could not agree on a system of objective voter identification5 and
by implying that the ARD learned too late the inadequacy of
employee badges for this purpose. NLRB also charges the election
observers with dereliction for any deficiency in identification
5
Avondale offered to provide full first and last names and clock
numbers on the Master and Zone Voting Lists the night before the election.
Unfortunately, the Union, retaliating for what it believed was inadequate
information on the Excelsior list, rejected Avondale’s offer. At oral argument
in this court, the Union offered no logical explanation for this rejection.
9
procedures, because observers appointed by the parties “not only
represent their principals but also assist in the conduct of the
election.” NLRB Casehandling Manual, § 11310. But notwithstanding
the flexibility allowed for in-person voting procedures and the
incentives for the parties to agree and cooperate, the buck stops
with the NLRB. The first admonition in the Board’s manual for the
conduct of elections is:
The responsibility for the proper conduct of the election
is the Regional Director’s acting through the assigned
Board agent. Extreme care must be exercised both in the
preparation and in the conduct of elections.
NLRB Casehandling Manual, § 11300.
The integrity of an election cannot be maintained without
assurance that the voters who cast ballots were eligible to do so.
Thus, voters had to be employees whose names were on the lists
agreed to by the Union and Avondale, and they had to identify
themselves as the persons enumerated in the list. The importance
of reliable voter identification is reflected in NLRB’s elaborate
precautionary rules governing mail-in ballots. See NLRB
Casehandling Manual, § 11336. Obviously, more flexibility is
called for in developing identification procedures for in-person
balloting. Representation elections cover bargaining units that
may range from a few dozen employees at one worksite to thousands
of employees dispersed among multiple shifts at numerous worksites.
Verbal self-identification is appropriate when -- as is probably
10
true in a large portion of cases -- it is likely that the observers
are personally acquainted with the voters. It is wholly
inadequate, however, as the sole guide to identification, where a
very large bargaining unit is contemplated, and the voter lists
contain virtually the only information that will assure the
identity of the voters. The procedures used in Newport News and
Monfort, Inc. confirm this common sense notion and equally condemn
the unthinking adoption of “standard practice” for a multi-thousand
employer like Avondale.
The voter identification procedure in this case was
utterly insufficient. It is undisputed that most of the observers
did not know the hundreds of employees who appeared during their
stints at each of the voting zones. It is also undisputed that
employee badges, which stated the employee’s first name only and
had a tiny photograph, were inadequate to prove that the voter was
the person whose last name, first and middle initials, and address
appeared on the Zone and Master Voting Lists. According to the
procedure used, an employee whose first name was “Jane” could
identify herself as any voter on the Zone list with a first initial
“J” and could vote on no more sure proof of identity.6 Lest this
6
Not only could an employee without an identification badge often vote
unchallenged merely by giving another person’s name to the observers, but a voter
with an identification badge and the same first initial of a voter on the Zone
Voting List could also vote unchallenged. Thus, an employee with “John” on his
identification badge could vote for any employee identified on the Zone Voting
List whose first initial was “J.” In either situation, no observer would need
to ask for further identification, such as an address -- the fraudulent voter’s
11
be thought an unlikely possibility, the NLRB hearing officer
adjudicated fourteen claims regarding employees whose ballots were
challenged because they appeared to vote after their names had
(perhaps erroneously) already been crossed off the voting lists.7
The NLRB hearing officer confirmed the identification procedures as
a source of potential confusion:
[m]ost importantly, many of the employees had
corresponding initials which might have created some
confusion in regard to the checking off of the correct
names. I have also noted that the employees’ badges did
not contain the last names of employees which might have
exacerbated the confusion. . . .8
Had the voting list contained full names and employee clock numbers
or social security numbers, information which could quickly be
corroborated by the employee or a driver’s license, no uncertainty
would have arisen.
NLRB objects that most voters truthfully identify
themselves. This is undoubtedly true, and was undoubtedly true in
this election, but it does not detract from NLRB’s own reliance on
assumed name would have appeared on the list as an eligible voter. In his
testimony, ARD Norton admitted as much.
7
This being Louisiana, the duplicate names included Chauff, Boutain,
Thibodeaux, and Plaisance, as well as more common names like Taylor.
8
The hearing officer implied at this point that the deficient
information on the voting lists was attributable to Avondale, because it
“provid[ed] an Excelsior list with first initials of employees instead of
Christian names”. The officer erred in drawing this conclusion. Avondale
volunteered before the election to add full names to the voting lists, but the
Union refused. Moreover, ARD Norton testified that the Excelsior list was a
completely separate document that he never saw; consequently, any limitations on
the Excelsior list cannot be lumped with the inadequacy of the voting lists.
12
the last 4 digits of social security numbers in Newport News as a
visible assurance of election integrity. Moreover, the truism
lacks force in this election for two powerful reasons. First, the
election contest was bitter and hostile, sure to provoke suspicion
in whichever party lost. From the number of threatened challenges
and the parties’ inability to agree whether hundreds of employees
would be in the bargaining unit, to the parties’ other
disagreements every step of the way toward the election, ARD Norton
should have foreseen a prolonged administrative struggle. An
objective voter identification procedure would have belied
suspicions, discouraged attempts at vote fraud, and averted this
source of future litigation. Second, the election was close. The
Union finally achieved 54% of the votes, but a margin of only 130
votes out of nearly 4,000 cast and counted controlled the result.
The wisdom of hindsight cannot alone dictate rejection of the
inadequate voter identification procedure, but it confirms that
NLRB must deploy its flexibility with “extreme care”, especially
when conducting high profile, hotly-contested representation
elections.9
9
NLRB contends that requiring verification by means of employee driver
licenses, social security numbers or clock numbers is impractical. But social
security numbers were used successfully, indeed were relied on by NLRB, in
Newport News, where a 19,000 member bargaining unit was at issue, and, given the
65 polling places used there, an average of 1,250 voters were appearing at each
station. The common experience of any adult who has voted at the local precinct
and been required to produce a voter registration card conflicts with NLRB’s
complaint.
13
On a more fact-specific level, parroting the hearing
officer’s allocation of responsibility, the NLRB and the Union have
attempted to saddle Avondale with the blame for any failure in
voter identification. They contend that Avondale exercised
extensive control over when and where the employees voted, thus
limiting the potential for voter fraud. Although the company’s
escort procedures were of some benefit, they by no means furnished
the exclusive means for voting. No employer controls could be or
were applied during lunch, breaks, or shift changes, when employees
were allowed to vote without being escorted to the voting zone by
their supervisor. No one had to vote when taken to the polling
place by the supervisor. The supervisor did not oversee the name
under which an escorted employee purported to vote, and nothing
prevented unescorted fraudulent voting during lunch, breaks, and
shift changes. Employees absent from work could come in at any
time to vote. Staggered voting and zoned polling places can not
replace inadequate voter identification procedures.
The crux of the inadequate identification procedure is
this: no one knows exactly who voted in the Avondale election.
Moreover, the challenge procedure could not function properly,
where the initial and usual method for identifying a voter was an
identity between his first name and stated last name and the
initial and last name on the Zone or Master List. Election
observers would have had to interfere with the “routine” procedure
14
in nearly every case in order to compel objective identification or
insist upon challenges. The ultimate basis for approving the
outcome of this election is the NLRB’s hope that most employees
voted truthfully. Such a hope does not fulfill the standard of
“extreme care” that the NLRB itself sets for the conduct of
representation elections. The NLRB’s reliance on mere hope,
unsupported by objectively verifiable voter information, raises a
reasonable doubt as to the fairness and validity of the election.
While evidence of voter fraud could not be reliably
investigated based on the voter identification procedures used at
Avondale, the company nevertheless produced specific evidence of
potential voter fraud stemming directly from the failed
identification procedures. When Avondale finally received the
marked, unredacted voting lists,10 many employees allegedly absent
from work on the date of the election were shown as voting. The
gate logs at Avondale, a highly secure facility, failed to confirm
entry by many of these absent employees.11 In Voting Zone 3, many
10
Avondale was forced to file a Freedom of Information Act claim
against the NLRB in order to recover the lists. See generally Avondale Indus.,
Inc. v. NLRB, 90 F.3d 955 (5th Cir. 1996).
11
The marked, unredacted voting lists indicate that potentially 126
employees absent on the day of the election cast ballots in the representation
election. Of the 185 ballots cast by absent workers, Avondale is able to account
for only 59 of the “absentee” ballots through excused, election-related absentees
and by checking available gate logs for plant entry by otherwise absent employees
for election purposes. Although Avondale was unable to locate the gate log for
one gate, and employee movement during shift changes and lunch breaks is not
recorded on the logs, the evidence was sufficiently specific to raise
considerable doubt regarding the absentee employees’ participation in the
election.
15
employees were marked in a strange fashion, potentially indicating
that the employees cast more than one ballot. Other anomalies
appear in comparing the actual voting lists with extrinsic
identification methods.12 Although the NLRB offers plausible
explanations for some of these anomalies, no hearing was held on
Avondale’s late-obtained evidence, and we are thus required to view
it in the light most favorable to Avondale. See Trencor, Inc. v.
NLRB, 110 F.3d 268, 270 (5th Cir. 1997). At the very least,
Avondale’s evidence raises serious questions concerning the
possible occurrence of vote fraud.
III. CONCLUSION
The NLRB’s failure to implement more extensive
identification procedures for this large-scale representation
election, combined with evidence of potential voter fraud, raises
serious questions regarding the validity of the representation
election conducted at Avondale. We do not minimize the
difficulties faced by ARD Norton in overseeing the contentious
election and mediating between two difficult parties, nor do we
condemn the herculean efforts of the hearing officer in post-
12
According to the affidavit of Rhonda W. James, a paralegal employed
by Avondale’s attorneys, the marked, unredacted voting lists showed at least 13
“phantom” ballots (i.e., while 3265 votes were cast from “listed voters” in the
election, only 3252 employees were marked as having voted) and 100 multiple-vote
voters. In fact, the NLRB sustained a challenge to the ballots of seven voters
because, “Although . . . observers might have inadvertently checked off the wrong
names . . . this evidence was insufficient to prove that any of the challenged
ballots at issue were the first and only ballots cast by the voters in question.”
NLRB Decision and Direction, at 6-7 (Feb. 5, 1997).
16
election proceedings. Unfortunately, based on the above-discussed
characteristics and systemic failure in the Avondale electoral
process, the mistakes cannot be remedied by remand for additional
post-election hearings. Accordingly, we REMAND this action to the
NLRB with instructions to set aside the representation election
held at Avondale on June 25, 1993, and to conduct further
proceedings consistent with this opinion. The NLRB’s order
requiring Avondale to bargain with the Union is VACATED.
Avondale’s motion to supplement the record is DENIED.
VACATED and REMANDED.
17