Union Certification Success Under Voting Versus Card-Check Procedures: Evidence from British Columbia, 1978–1998

Abstract

The author estimates the impact of compulsory election laws on certification success using data on over 6,500 private sector certifications from British Columbia over the years 1978-98. A unique quasi-experimental design is used by exploiting two changes in the union recognition law: first, in 1984, the introduction of mandatory elections; and second, in 1993, the repeal of elections and their replacement by the original card-check procedure. The author also estimates the effectiveness of management opposition tactics across union recognition regimes. Success rates declined by an average of 19 percentage points during the voting regime, and then increased by about the same amount when card-checks were re-instituted. The results indicate that the mandatory election law can account for virtually the entire decline. In addition, the findings suggest that management opposition was twice as effective under elections as under card-checks. (Author's abstract.)

Full text

Industrial & Labor Relations Review
|
Number 4Volume 57 Article 1
2004
Union Certification Success under Voting Versus
Card-Check Procedures: Evidence from British
Columbia, 1978–1998
Chris Riddell
Riddell, Chris (2004) "Union Certification Success under Voting Versus Card-Check Procedures:
Evidence from British Columbia, 1978–1998," Industrial & Labor Relations Review, Vol. 57, No. 4,
article 1.
Available at: http://digitalcommons.ilr.cornell.edu/ilrreview/vol57/iss4/1

Union Certification Success under Voting Versus Card-Check
Procedures: Evidence from British Columbia, 1978–1998
The author estimates the impact of compulsory election laws on certification success using data on over 6,500
private sector certifications from British Columbia over the years 1978–98. A unique quasi-experimental
design is used by exploiting two changes in the union recognition law: first, in 1984, the introduction of
mandatory elections; and second, in 1993, the repeal of elections and their replacement by the original card-
check procedure. The author also estimates the effectiveness of management opposition tactics across union
recognition regimes. Success rates declined by an average of 19 percentage points during the voting regime,
and then increased by about the same amount when card-checks were re-instituted. The results indicate that
the mandatory election law can account for virtually the entire decline. In addition, the findings suggest that
management opposition was twice as effective under elections as under card-checks.
The author thanks Michele Campolieti, Morley Gunderson, Felice Martinello, and Craig Riddell for useful
suggestions.
This article is available in Industrial & Labor Relations Review: http://digitalcommons.ilr.cornell.edu/ilrreview/vol57/iss4/1

493
Industrial and Labor Relations Review, Vol. 57, No. 4 (July 2004). © by Cornell University.
0019-7939/00/5704 $01.00
O
UNION CERTIFICATION SUCCESS UNDER
VOTING VERSUS CARD-CHECK PROCEDURES:
EVIDENCE FROM BRITISH COLUMBIA, 1978–1998
CHRIS RIDDELL*
The author estimates the impact of compulsory election laws on certification
success using data on over 6,500 private sector certifications from British
Columbia over the years 1978–98. A unique quasi-experimental design is used
by exploiting two changes in the union recognition law: first, in 1984, the
introduction of mandatory elections; and second, in 1993, the repeal of elec-
tions and their replacement by the original card-check procedure. The author
also estimates the effectiveness of management opposition tactics across union
recognition regimes. Success rates declined by an average of 19 percentage
points during the voting regime, and then increased by about the same amount
when card-checks were re-instituted. The results indicate that the mandatory
election law can account for virtually the entire decline. In addition, the
findings suggest that management opposition was twice as effective under
elections as under card-checks.
*Chris Riddell is a Post-Doctoral Fellow with the
Centre for Labour and Empirical Economic Research,
Department of Economics, University of British Co-
lumbia. He thanks Michele Campolieti, Morley
Gunderson, Felice Martinello, and Craig Riddell for
useful suggestions.
The data used in this paper may be obtained from
the British Columbia Labour Relations Board, Suite
600, Oceanic Plaza, 1066 West Hastings Street,
Vancouver, BC, CANADA V6E 3X1; 604-660-1300. A
data appendix with additional results, and copies of
the computer programs used to generate the results,
are available from the author at University of British
Columbia, Department of Economics, #997-1873 East
Mall, Vancouver, BC, CANADA V6T 1Z1;
chrisrid@interchange.ubc.ca.
ne of the most important public poli-
cies affecting the union organizing
process is the recognition procedure, that
is, the mechanism by which the union at-
tempts to demonstrate that it has sufficient
support within the proposed bargaining
unit to be certified. Up to the late 1980s, all
Canadian provinces (except Nova Scotia
after 1977) had “card-check” procedures,
whereby the union could be certified with-
out an election if it signed up a sufficient
proportion of the proposed bargaining unit.
The proportion required to avoid an elec-
tion varied by province, but generally
ranged between 50% and 55%. While it was
possible for a certification bid to require a
vote in a card-check province, few organiz-
ing drives actually went to an election. The
Canadian system in those years contrasts
starkly with the system in the United States
and, until the 1999 Employment Relations
Bill, with that in the United Kingdom—
although for very different reasons. As a
result, the union recognition procedure

494 INDUSTRIAL AND LABOR RELATIONS REVIEW
became a key component in theories ex-
plaining the three countries’ divergence in
union density (Weiler 1983, 1990; Freeman
1985, 1989; Disney, Gosling, and Machin
1995; Johnson 2002).
Since the late 1980s, the certification
process has changed in Canada. Now six
provinces, covering about 75% of the labor
force, have mandatory elections as the union
recognition procedure. In addition, over
the 1984–98 period, union density fell by
about 7 percentage points in Canada—
similar to the 8 percentage point decline in
the United States during the same years
(Riddell and Riddell 2004). Many observ-
ers believe it is no coincidence that unions
declined concurrently with the abandon-
ment of the card-check system.
While there is a literature in both Canada
and the United States that emphasizes the
importance of the certification process, only
a handful of studies have empirically inves-
tigated the impact of the union recognition
procedure. Those studies suggest that
mandatory elections, or legislation institut-
ing them, has a strong negative effect on
union organizing (Martinello 2000;
Johnson 2002). However, no study has
investigated why voting deters union orga-
nizing, which is perhaps the most impor-
tant question of all.
This paper estimates the impact of the
recognition procedure, and addresses the
question of why elections reduce certifica-
tion success. The analysis is based on over
6,500 private sector certifications in British
Columbia, covering the years 1978 to 1998.
A mandatory election law was adopted by
British Columbia in 1984, and was then
repealed in 1993. We thus have data from
six years of card-check laws, followed by
nine years of a voting era, followed by six
years of card-check laws—a regime succes-
sion that allows for a unique quasi-experi-
mental study. British Columbia is the only
province in Canada that has changed union
recognition regimes more than once, with
each regime lasting multiple years. More-
over, the timing of the various legislative
changes makes it possible to isolate the
effect of voting from the effects of other
amendments.
Previous Studies of the
Union Recognition Procedure
Many observers have noted that a pos-
sible explanation for the large gap in union
density between Canada and the United
States, which emerged in the mid-1960s, is
the union recognition procedure. In a
seminal paper, Paul Weiler argued that the
election system in the United States allows
employers greater opportunity to oppose
the organizing drive (Weiler 1983). Both
that study and Meltz (1985) provided some
evidence to suggest that elections reduce
union organizing success.
While the potential importance of the
union recognition law has not been over-
looked, the empirical literature investigat-
ing it is very limited and largely restricted
to Canada and the United Kingdom. One
Canadian study, Martinello and Meng
(1992), estimated the impact of labor legis-
lation (including elections, dues check-offs,
and strike-replacement worker laws) on the
likelihood that an individual will be cov-
ered by a collective agreement. No statisti-
cally significant effects were found with
respect to voting, but, as the authors noted,
there is simply not enough variation across
provinces in their single cross-section to
identify such an effect. In an analysis that
exploited variation over time and across
provinces, Johnson (2002) estimated the
impact of voting laws over the 1978–96
period using aggregate data from nine Ca-
nadian provinces. She concluded that elec-
tions reduced union organizing success
rates by 9 percentage points.
Following the Employment Relations Bill
in 1999, the union recognition procedure
in the United Kingdom became more simi-
lar to the Canadian card-check system
(Wood and Godard 1999). In the two de-
cades prior to the Bill, union density had
fallen substantially, a change that was largely
attributed to a series of legislative changes
in the 1970s and 1980s (Freeman and
Pelletier 1990). Overall, U.K. union den-
sity fell by 20 percentage points over the
period from the late 1970s to 1992, a time
when union recognition and decertifica-
tion were voluntary and not covered by

VOTING VERSUS CARD-CHECK PROCEDURES 495
legislation (Beaumont and Harris 1995).
Disney, Gosling, and Machin (1995) showed
that difficulty in achieving union recogni-
tion in new establishments was the key fac-
tor in the 1980s decline, and that an in-
crease in management opposition can ex-
plain some of the decline.
In the United States, the near-hegemony
of elections has discouraged systematic
study of union organizing under different
union recognition procedures. One study
analyzing the recognition procedure is
Eaton and Kriesky (2001). The authors
concluded that success rates are higher in
cases decided by card-check than in those
decided by elections, although their find-
ing is based on a sample size of only around
100 certifications. They also provided evi-
dence that the incidence of management
opposition is reduced in a card-check sys-
tem. However, the latter conclusion ap-
pears to be primarily based on a compari-
son of U.S. evidence with evidence from
two Canadian studies—Thomason (1994)
and Thomason and Pozzebon (1998)—nei-
ther of which had variation in the union
recognition law. Comparing results from
the two countries may not yield appropri-
ate conclusions.
While there has not been systematic
study of the union recognition proce-
dure in the United States, there is a grow-
ing U.S. literature on the importance of
the union recognition procedure that
draws on qualitative and case study evi-
dence. For example, Budd and Heinz
(1996) found that successful certifica-
tion at a Minneapolis hotel occurred in
only 32 days under a card-check arrange-
ment. Benz (1998) found similar evi-
dence in favor of card-checks, but con-
cluded that there is insufficient data to
determine whether the results can be
generalized. Additional studies were re-
viewed by Eaton and Kriesky (2001).
In summary, there is a literature suggest-
ing that the union recognition procedure
is related to certification success. However,
only one paper has estimated the impact of
voting alone (Johnson 2002), and none of
these studies examined why mandatory elec-
tions reduce union organizing success.
Legislative Background
Three legislative changes were made to
the British Columbia Labour Code over the
sample period of October 1978 to May 1998.
The first change, which introduced manda-
tory elections, was in 1984. Second, in
1987, sweeping changes were made to the
Code when the Industrial Relations Re-
form Act was introduced. Finally, in 1993,
mandatory elections were repealed. How-
ever, many of the “voting era” amendments
were maintained. Appendix 1 presents an
analysis of the various legislative changes,
which are discussed in more detail below,
under “Framework for Analysis.”
1
The key amendment made in May 1984
required the British Columbia Labour Re-
lations Board (BCLRB) to order a repre-
sentation vote on all certification applica-
tions for which at least 45% of the proposed
unit was “signed up” through the card-
check process. Prior to the 1984 Act, a vote
was required if the union was only able to
sign up between 45% and 55% of the pro-
posed unit, which rarely occurred (see Ap-
pendix 2). Automatic certification without
a vote was granted if 55% or more of the
proposed unit was signed up. The 1984 Act
also introduced expedited certification pro-
cedures as a regulation (that is, not formal
law), which dictated that the vote was to be
held within “approximately” 10 days of the
application filing date. The 10-day rule was
formalized in the 1987 Act, but the data
examined in this paper indicate that the
1984 Regulation was followed (see Appen-
dix 2). It is useful to stress from the outset
that employers had only a short time pe-
riod to launch a campaign against the union.
Two other substantive changes were made
1
The following laws are not discussed because no
amendment (or only re-labeling or very minor amend-
ment) was made over the sample period: first con-
tract arbitration, professional strikebreakers, unfair
labor practices relating to union certification/decer-
tification, duty to bargain in good faith, and fair
representation; union successorship; grievance me-
diation and arbitration; and Board review of arbitra-
tion awards.

496 INDUSTRIAL AND LABOR RELATIONS REVIEW
in 1984: the decertification rules were
amended to parallel the new certification
procedures, and unions were no longer
automatically allowed to picket an ally of the
employer or another location of the em-
ployer.
In 1987, several additional amendments
were made.
2
The most substantive change
was the creation of the new Dispute Resolu-
tion Division, which placed several restric-
tions on strikes, and introduced mediation
and fact-finding. Two other changes made
in 1987 were the introduction of final offer
voting and the broadening of the essential
services designation. In 1993, the voting
law was repealed, and replaced with the
pre–1984 card-check system whereby a
union could avoid an election if 55% of the
unit was signed up. However, the other key
1984 and 1987 amendments were retained.
3
Figure 1 shows success rates for the pri-
vate and public sectors, with raids and with-
drawn certifications excluded.
4
The pat-
tern in Figure 1 is striking. Success rates for
private sector organizing drives fell by nearly
20 percentage points following the intro-
duction of mandatory elections in May 1984.
Moreover, in 1993, success rates appear to
have recovered to their pre-voting levels,
where they held fairly steady through the
remainder of the sample period.
Interestingly, public sector success rates
did not systematically decline. This finding
is suggestive of an explanation based on
management opposition, since employer
resistance to certification is likely to be
negligible in the public sector.
5
However,
Figure 1 does not provide a sufficient basis
for concluding that mandatory elections
alone caused the decline in success rates,
because other potentially important legis-
lative changes were made over the sample
period. In addition, Figure 1 does not
reveal anything conclusive about the role
of management opposition. The task for
the remainder of the paper is therefore
twofold: first, to use the legislative analysis
to isolate the effect of voting from the
effects of other amendments; and second,
to use data on unfair labor practices (ULPs)
to examine the role of management oppo-
sition.
Framework for Analysis
I begin by outlining a simple conceptual
framework that hypothesizes why the union
recognition procedure could affect certifi-
cation success and why employer behavior
might change with a change in the union
recognition law. In general, mandatory
election laws are believed to lead to lower
certification success rates for two reasons:
management has more opportunity to op-
pose the bid in a voting system; and peer
2
A careful reading of the legislation indicates that
many of the 1987 changes were procedural in nature.
For instance, all applications had to be filed in a
timely manner (that is, a strict number of days was
explicitly given for virtually all applications) and in
writing, including appropriate documentation. Many
types of applications were made on a more informal
basis prior to 1987, and supporting documentation
was rarely required (unless requested by the Board at
a later date). These various procedural changes were
maintained in 1993.
3
In addition, any pure effect of the government on
union organizing is “controlled for” since there was
only one change in government over the entire sample
period: the election of the New Democratic Party in
1991.
4
A large majority of the social services certifica-
tions are public sector, but a few are private sector,
and cannot be isolated. Withdrawn certifications are
excluded for three reasons: such cases are often
withdrawn almost immediately after the filing date,
and thus important variables used in the analysis are
missing; classifying withdrawn applications as a fail-
ure may be inappropriate; and including withdrawn
applications could result in double-counting. With
respect to double-counting, I have found that in a
majority of withdrawn cases, the union re-applied
shortly after (in fact, the Board eventually instituted
a six-month waiting policy for this reason). In results
available upon request, it is shown that withdrawn
certifications simply shift the series downward in a
parallel fashion, and that the trend in Figure 1 cannot
be explained by a Canada-wide shock, as certification
success rates in most other provinces were largely
constant over the 1984–92 period (also see Martinello
1996).
5
Due to concerns over public image and the lack of
profit maximization objectives, management opposi-
tion (particularly egregious management opposition)
is likely to be negligible in the public sector.

VOTING VERSUS CARD-CHECK PROCEDURES 497
pressure from pro-union colleagues and
union organizers may artificially inflate the
“true” level of support in a card-check sys-
tem. With respect to management opposi-
tion, a change in the union recognition law
may change both the effectiveness and the
incidence of management opposition.
A voting regime may increase the effec-
tiveness of employer tactics for two reasons.
The amount of time employers have to
influence the organizing drive is unam-
biguously greater in the voting system than
in the card-check system. In the card-check
system in British Columbia prior to May
1984 and from 1993 to 1998, union orga-
nizers collected signatures from members
of the proposed bargaining unit, and then
made an application to the BCLRB. The
employer was contacted when the BCLRB
conducted its investigation to determine
whether the petition requirements (which
are discussed below) were satisfied. As
shown in Appendix 2, it was rare for a union
in one of the card-check regimes in British
Columbia to fail to gather 55% support—
sufficient for certification without a vote.
This was also the case in Ontario over the
1980 to 1988 period (Thomason 1994).
Thus, in the card-check years, many orga-
nizing drives were virtually completed be-
fore the employer was aware of the drive.
While Riddell (2001) showed that some
employers find out about the bid and en-
gage in an anti-union campaign prior to
the application, in general, management
does not have sufficient time to launch a
meaningful campaign. Under voting,
unions still had to gather cards and make
an application, but then a secret ballot vote
was held 10 days later. This extra step in the
certification process gave the employer
greater opportunity to launch a campaign.
The secret ballot vote itself is a second
reason management opposition is likely to
100
95
90
85
80
75
70
65
60
Figure 1. Certification Success Rates, British Columbia, 1978-1998.
1978
1979
1980
1981
1982
1983
1984
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
Private Sector
Jan.-May
June-Dec.
Public Sector

498 INDUSTRIAL AND LABOR RELATIONS REVIEW
be more effective in a voting regime. Un-
der card-checks, if the employer coerces
employees into refusing to sign cards, union
organizers and pro-union colleagues can
counteract the coercion tactics. In a secret
ballot vote, the opportunity to counteract
employer threats is likely diminished.
Elections may also increase the incidence
of management opposition. If the effec-
tiveness of management opposition is
greater in a voting regime—and assuming
employers know this—then a cost-benefit
calculation implies that employers will
adopt such tactics more frequently, given
that the chances of defeating the bid are
greater while the costs remain the same. As
well, there is simply more time available to
the employer under elections. Thus,
whereas employers in a card-check system
may not realize organizing activity is occur-
ring until after the cards had been col-
lected, under a voting system they will have
an opportunity to oppose the union.
The empirical analysis proceeds in two
stages. First, I use data on certifications
from 1978 to 1998 and changes in legisla-
tion over that period to estimate the impact
of mandatory elections on certification suc-
cess. Second, I merge data on unfair labor
practice complaints, which are available
only for the years 1987 to 1998, to examine
whether the incidence and effectiveness of
management opposition are greater in a
voting regime. The second part of the
analysis also estimates what part of the total
change in success rates across the two union
recognition regimes—from voting (1987–
92) to cards (1993–98)—can be attributed
to management opposition.
To analyze the impact of elections on
certification success, I estimate the probit
model
(1) SUCCESS
i
= f (α
0
+ α
1
LEG84
i
+ α
2
LEG87
i
+ α
3
LEG93
i
+ α
4
X
i
+ α
5
IND
i
+ α
6
REG
i
+ α
7
YEAR
i
+ ε
i
),
where LEG84 = 1 if the i
th
certification was
applied for under the 1984 Labour Code
Amendment Act; LEG87 = 1 if the certifica-
tion was applied for under the 1987 Indus-
trial Relations Reform Act; LEG93 = 1 if the
certification was applied for under the 1993
Labour Code Amendment Act; X is a vector
of other explanatory variables (discussed
below); IND is a vector of industry dummies
to capture unobserved factors associated
with the industry of the unit; REG is a vector
of regional dummies to capture unobserved
factors associated with the region within
the province;
6
and YEAR is a vector of year
dummies to capture unobserved changes
over time, such as changes in economic
conditions.
The coefficients on the three legislative
variables are evaluated relative to the omit-
ted reference group, which is the pre–1984
legislation (the 1973 Labour Code Amend-
ment Act) that was in effect from October
1978 until May 1984. A hypothesis test of
the equality of α
1
and α
2
will indicate
whether there was any difference between
the 1984 and 1987 reforms, both of which
had mandatory elections. If α
1
and α
2
are
statistically different, then α
2
– α
1
yields the
impact of the additional reforms made in
1987. Note that it is not possible to isolate
any of the individual law changes made in
1987.
The 1984 legislation made relatively few
changes to the Labour Code, so the coeffi-
cient α
1
mainly reflects the impact of com-
pulsory elections. However, α
1
does not
fully isolate the effect of elections, since
amendments to decertification and picket-
ing provisions may have reduced the ben-
efits to unionization, and thus may have
affected certification success.
7
The decerti-
fication and picketing rules were both main-
tained in 1993 along with the key 1987
6
Statistics Canada’s UI Regions are used to create
the region dummies, as well as to define the unem-
ployment rate. These are the only regional defini-
tions available over the sample period that reflect
economic and social differences within a province.
7
My reading of the legislation is that decertifica-
tion became much easier following the 1984 law, and
in fact the number of decertifications increased mark-
edly (although this may not be a causal relationship).
Regulations governing picketing unambiguously be-
came much more restrictive following the 1984 law, as
indicated in Appendix 1.

VOTING VERSUS CARD-CHECK PROCEDURES 499
changes. Thus, if α
1
and α
2
are not statisti-
cally different (that is, there is no addi-
tional impact of the 1987 reforms), α
1
– α
3
yields the impact of mandatory elections
alone. If α
1
and α
2
are statistically different,
then the voting effect is α
1
– [(α
2
– α
1
) + α
3
].
To examine whether management op-
position is more effective under voting than
under card-checks, I use data on ULPs—
available only since 1987—and estimate the
following probit model:
(2) SUCCESS
i
= f(γ
0
+ γ
1
LEG87
i
+ γ
2
ULP
i
+ γ
3
LEG87
i
∗ULP
i
+ γ
4
X
i
+ γ
5
IND
i
+ γ
6
REG
i
+ γ
7
YEAR
i
+ ν
i
)
The inclusion of the interaction term
between the voting regime and ULPs allows
for testing the hypothesis that management
opposition is greater under elections. That
is, γ
1
indicates the impact on certification
success rates of the 1987 voting legislative
package relative to the omitted legislative
regime (card-check, 1993–98), γ
2
indicates
the average effect of ULPs, and γ
3
indicates
the additional impact, if any, that ULPs had
when voting legislation was in effect.
A further question is: how much of the
fall in success rates associated with the vot-
ing regime can be attributed to a change in
the incidence of management opposition,
and how much can be attributed to a change
in the effectiveness of management opposi-
tion? To address this question, I estimate
regressions separately for each legislative
regime in years for which ULPs are avail-
able—one regression for certifications dis-
posed under the voting regime (1987–92)
and another for those handled under the
card-check regime (1993–98). Applying
the standard Oaxaca-Blinder technique,
modified for a probit (see Appendix 3), I
can then use the estimated coefficients from
the two regressions to decompose the
change in success rates into two compo-
nents: the part due to changes in the
means of the X’s across legislative regimes
(capturing changes in the incidence of man-
agement opposition), and the part due to
changes in the coefficients across legisla-
tive regimes (capturing changes in the effec-
tiveness of management opposition). It is
not straightforward to sub-decompose the
total coefficient component into each indi-
vidual coefficient’s contribution, which is
necessary in order to estimate how much
the change in the effectiveness of manage-
ment opposition contributed to the change
in success rates. I use the method devel-
oped by Nielson (1999) to provide an esti-
mate of the contribution of the change in
effectiveness of management opposition
(see Appendix 3 for further details).
Data and Methodological Issues
The principal data were collected from
the records of the British Columbia Labour
Relations Board. A number of sample re-
strictions were imposed. For reasons dis-
cussed above, public sector certifications
were dropped from the sample. Also
dropped were observations involving raids,
since such cases have always required a
vote, and involve workers choosing between
the incumbent union and the raiding union.
Certain construction applications were
dropped because a representation vote has
never been required in temporary construc-
tion projects. Certifications were also omit-
ted if they were dismissed as a result of
petition violations, either because of juris-
dictional issues (deemed to be under fed-
eral jurisdiction) or because the bargain-
ing unit was deemed inappropriate for col-
lective bargaining. These restrictions re-
moved 992 certifications from the original
private-sector sample, resulting in a final
sample of 6,650 certifications.
8
As noted, the measure of management
opposition is whether a ULP was filed
against the employer for illegally attempt-
ing to coerce employees from supporting
the union.
9
An additional advantage of
8
In some cases it was difficult to determine whether
a given construction certification was exempted from
the vote; however, the results are not sensitive to the
exclusion of all construction certifications.
9
Penalties for ULPs are largely compensatory, and
there was no change in these penalties over the sample
period. For intimidation, the BCLRB would post a
cease and desist order. In cases of dismissal, the
BCLRB could order reinstatement and payment of

500 INDUSTRIAL AND LABOR RELATIONS REVIEW
examining British Columbia is that the
BCLRB main database, unlike data pro-
vided by many of the provincial labor rela-
tions boards, separates ULPs by the section
of the legislation under which they were
filed. Thus, only ULPs that were filed where
it was indicated that management had ille-
gally attempted to influence one or more
persons’ decision regarding certification
are included in the sample.
10
It is appropriate to consider the advan-
tages and disadvantages of using ULPs,
which are the only available measure of
management opposition with which to ex-
ploit the unique quasi-experimental design
of the study. A key issue is that ULPs
capture illegal employer tactics, and thus I
am not able to examine the effectiveness of
legal employer tactics across union recog-
nition regimes.
11
In addition, I cannot
distinguish between different types of em-
ployer tactics. It may be the case that
certain types of illegal employer tactics
(such as dismissals) have varying effects
across union recognition regimes while oth-
ers (such as broad threats) do not.
Some light can be shed on the tactics the
ULP variable may be capturing by examin-
ing the results reported by Martinello and
Yates (2003). The authors identified five
categories of employer strategies (where
one category is “do nothing”). Their re-
sults indicated that ULPs tend to be filed in
cases where management goes full out and
adopts legal (or largely legal) strategies—
challenging the bargaining unit or holding
captive-audience meetings—and more mar-
ginally legal/outright illegal strategies such
as threatening plant closure, improving
working conditions during the campaign,
and dismissing employees for union in-
volvement. If their results for Ontario can
be generalized to British Columbia, our
ULP variable is likely capturing anti-union
campaigns in which the employer is using
the full array of tactics.
Appendix 2 shows the number of dis-
missal and non-dismissal ULP cases. Two
broad types of ULPs are included in the
non-dismissals: general coercive activities,
such as threats of certain consequences in
the event of successful certification; and
non-discharge actions taken against an
employee (or group of employees), includ-
ing demotions, denial of overtime pay, and
changing working conditions. Historically,
dismissals dominated the ULP filings in the
province; unfortunately, the non-dismissal
ULP numbers are not reliable after 1988.
12
Riddell (2001) showed that non-dismissal
ULPs tend to be threats made against the
workplace in general, and also found that,
for British Columbia, it is unusual for a
non-dismissal ULP to be filed in isolation
lost wages under Subsection 8(4)(c). Appendix 2
shows the proportion of dismissal ULPs in which
some form of compensation was granted. The final
power available, Subsection 8(4)(e), allowed auto-
matic certification to be granted where the Board
deemed that the true wishes of the workers were not
expressed and could not be expressed through a
second vote. As indicated in Appendix 2, the Subsec-
tion 8(4)(e) power was rarely used.
10
This includes all ULPs falling under Section 3 of
the Industrial Relations Act for 1987 to 1992 and
Section 6 of the Labour Code for 1993 to 1998.
11
One piece of evidence that increases my confi-
dence in the reliability of ULPs as a measure is the
rarity of applications that are rejected (that is, not
found to be meritorious; this can be seen in the
BCLRB annual reports over the 1978 to 1992 years
only). A strong majority of applications either are
found to be meritorious or are withdrawn. Moreover,
Riddell (2001) found that in a majority of the with-
drawn cases in British Columbia a settlement was
reached.
12
There has been a dramatic inflation of ULP
charges in British Columbia, especially since 1993.
However, much of this is artificial, due to the creation
of a new ULP section in 1993 that was already covered
under previous legislation. Also, unions filed more
ULP charges (that is, under different subsections)
per case. Riddell (2001) found no support for the
notion that a second ULP has a cumulative effect on
certification success. Given all of this, the results
presented in the paper use a ULP variable defined as
being at least one ULP filed. In this way, the mean of
the ULP variable will not be affected by any artificial
inflation. Finally, the BCLRB changed its system of
counting ULPs in 1989. Readers should therefore be
wary of the ULP numbers in the BCLRB annual
reports.

VOTING VERSUS CARD-CHECK PROCEDURES 501
(that is, without a dismissal ULP also being
filed). The raw data on ULPs from British
Columbia thus appear consistent with the
Martinello and Yates (2003) finding that
ULPs tend to be filed in cases where man-
agement “goes all out.”
13
There are advantages of using ULPs
rather than survey-based measures of em-
ployer tactics. Administrative data are much
less prone than survey data to measure-
ment problems such as non-response bias
and recall error (see, for instance, Pierret
2001). In addition, a comparison of two
Canadian surveys—Martinello and Yates
(2003), which surveyed unions about em-
ployer tactics, and Bentham (2002), which
surveyed employers about employer tactics—
illustrates a potential pitfall of self-reported
measures. A cross-survey comparison of
the reported incidence of most employer
tactics shows dramatically lower reported
incidences in the employer-based survey
than in the union-based survey, with gaps
on the order of 20 percentage points for
particularly egregious practices such as
threats of relocation.
Another issue is the potential
endogeneity of ULPs.
14
Employers who
believe success is contestable and can be
influenced are more likely to pursue an
anti-union campaign than employers who
see success as either very likely or very
unlikely. Standard regression techniques
may thus provide estimates that underesti-
mate or overestimate the true effect of
employer resistance. Lawler (1984), Free-
man and Kleiner (1990), and Riddell (2001)
are the only studies I know of to have esti-
mated equations that correct for
endogeneity.
15
To address the possible endogeneity of
management opposition, I estimate a two-
stage least squares model. For an instru-
ment to be valid it must be correlated with
the presence of an unfair labor practice,
but must not directly influence organizing
success. There are two main factors con-
tributing to the presence of a ULP: factors
that would explain why employers resist
unions, and factors that affect the union’s
(or employees’) decision to file a ULP. For
instance, there may be an organizing drive
in which the employer is engaging in “sup-
pression” tactics but the union does not
feel it is worthwhile to file a complaint.
I can think of no credible instrument
that would explain why employers resist
unions, since any obvious factor could be
argued to explain certification success as
well. Thus, I focus on factors that may
affect the decision to file a ULP. The first
instrument used is the processing time of
ULPs lagged by one month.
16
The ratio-
nale is that if it takes a long time for the
BCLRB to render a decision on ULP cases,
the union may feel it is not worthwhile to
file. This is particularly the case where ULP
processing times are substantially longer
than certification processing times. The
second instrument is a dichotomous vari-
able equaling one if ULP processing time
in the month prior to the certification ap-
plication was greater than the processing
13
I asked the Board for their opinion of the nature
of non-dismissal ULPs, and also examined the Board’s
Decisions for ULP cases (unfortunately, the decisions
published in the Board’s annual review are not ran-
domly selected). Both pieces of evidence suggest that
the non-dismissal ULPs tend to be broad, serious
threats against the workplace, with plant closure/
relocation and (negative) changes in working condi-
tions being the two most common threats.
14
There is also a sample selection bias present,
since management opposition may have prevented
some certification bids from being launched; how-
ever, such a bias is likely minimal in the case of British
Columbia (and Canada, for that matter) relative to
the United States due to the high probability of
winning the bid.
15
Canadian studies of employer tactics include
Thomason (1994), Thomason and Pozzebon (1998),
Riddell (2001), Bentham (2002), and Martinello and
Yates (2003). U.S. studies include Dickens (1983),
Lawler (1984), Lawler and West (1985), Cooke (1985),
Freeman and Kleiner (1990), and Bronfenbrenner
(1997), among others.
16
ULP processing time is the number of days be-
tween the ULP filing date and the decision date. The
results (particularly with respect to the difference in
the ULP coefficient across union recognition regimes)
are quite robust with respect to the choice of lag.

502 INDUSTRIAL AND LABOR RELATIONS REVIEW
time of the current organizing drive. Use
of this instrument, like the first, is based on
the reasoning that ULP processing times in
excess of certification processing times re-
duce the incentive to file a ULP. In this
case, it is implicitly assumed that the union’s
expectation of the length of the organizing
drive equals the ex post processing time.
While this is obviously not entirely realistic,
it is a reasonable assumption, since orga-
nizers, at the time the union is considering
filing a ULP, will likely have a good feeling
for how long the BCLRB will take to process
the application.
17
The final instrument is
the lagged success rate of compensation
claims (for three months prior to bid) for
dismissal-related ULPs under subsection
8(4)(e). The intuition in this case is that
unions may be more likely to file a ULP if
the likelihood of receiving compensation is
higher.
This paper exploits two legislative
changes, occurring seven years apart, to
examine the impact of the mandatory elec-
tion law and the effectiveness of manage-
ment opposition across legal regimes. Thus,
the appropriate data to use are from a large
administrative dataset on certification at-
tempts that covers the necessary years. Al-
though reliance on this dataset constrains
my ability to control for other factors that
may influence certification success, I am
able to construct some controls for union
characteristics.
I derive two variables meant to capture
the decision-making power of the proposed
bargaining unit if certified. The first vari-
able, called unit representation, is the size
of the unit divided by the size of the union
local to which the bargaining unit belongs.
This variable may indicate how “important”
the unit will be relative to the union local,
and thus the unit’s perceived decision-mak-
ing power. The second variable, local rep-
resentation, is the size of the local to which
the unit belongs divided by the union’s
total membership in the province. A vari-
able indicating the share of the unit’s local
that is female is also included. Bronfen-
brenner (1997) found that successful cer-
tification was more likely in units that
were 60% or more female than in other
units.
An industrial specialization index, which
equals the percentage of a given union’s
certifications that occur in the union’s pri-
mary jurisdiction industry, is also computed
for each individual union. This is identical
to the “representational specialization” vari-
able used in Fiorito, Jarley, and Delaney
(1995), which the authors found to have an
important positive influence on organizing
effectiveness. However, given that there
are around 150 unions in the data, we have
much more variation in this variable than
did Fiorito, Jarley, and Delaney.
I also create a set of controls for the
affiliation of the union. In Canada, there is
considerably more variation in affiliation
than in the United States, and to some
extent, affiliation can control for union
characteristics.
18
Moreover, the role of
union affiliation has not, to date, been
17
The BCLRB has ordered a second vote in cases
involving a ULP, but if the delay in the second vote is
quite long (due to submissions being made) employ-
ees may lose interest in the organizing drive.
18
The affiliations are the following: an interna-
tional union affiliated with the AFL-CIO and the
Canadian Labour Congress (“AFL-CIO/CLC”); an
international union not affiliated with the CLC (AFL-
CIO only), which includes all trades unions (the
Electrical Workers, for example) that formed the
Canadian Federation of Labour (“CFL”) in 1982; a
CLC national union (“CLC”); a Confederation of
Canadian Unions (“CCU”) national union, which
consists of a group of left-wing, former national-
independent unions that merged in 1983; a national
independent union (“NIU”); a provincial indepen-
dent union (“PIU”), which is a multi-employer inde-
pendent union operating only in British Columbia
and likely to be a specialist union, such as the Health
Science Association; and an independent local union
(“ILU”). The Teamsters are separated because they
were a U.S. independent for most of sample period,
then became an AFL-CIO only (but should not be
grouped with the CFL trade unions), and then joined
the CLC in 1993. All union-related variables were
created using the annual British Columbia Directory
of Labour Unions.

VOTING VERSUS CARD-CHECK PROCEDURES 503
examined in the context of certification
success.
19
Results
Table 1 presents estimates of the impact
of mandatory elections on certification suc-
cess rates where I attempt to isolate the
voting effect by exploiting the fact that
different laws were changed at different
times over the 1984 to 1993 period. To
simplify the table, I show only marginal
effects.
20
Summary statistics for all vari-
ables are listed in Appendix 4.
Column (1) shows that certification suc-
cess rates were, on average, 24 percentage
points lower under the 1984 voting rules,
and 22 percentage points lower under the
1987 voting rules, than under the pre–1984
rules. A hypothesis test of the null that the
1984 coefficient equals the 1987 coefficient
cannot be rejected (χ
2
= 0.02). As regres-
sors are added, the gap between the 1984
and 1987 estimates grows somewhat, but
the 1984 effect is systematically larger than
the 1987 estimate. In the preferred specifi-
cation, which includes year effects, the 1984
impact is 23 percentage points, but the
1987 effect has fallen to 17 percentage
points. Not surprisingly, given the stan-
dard error on the 1984 legislation coeffi-
cient, an equality test still cannot be re-
jected (χ
2
= 0.34). Thus, I conclude that
there is no discernible difference between
the 1984–86 voting regime and the 1987–
92 voting regime. Essentially, the set of
additional reforms made in 1987 had no
additional impact on certification success
rates.
While the 1987 reforms had no addi-
tional impact on certification success, the
1984 estimate alone still does not isolate
the effect of voting, given the decertifica-
tion and picketing changes that were also
made in 1984. As discussed, the effect of
these other factors can be estimated using
the 1993 legislation, since the other 1984
amendments were maintained in 1993.
Column (1) shows that only three percent-
age points separate the certification suc-
cess rates of the 1993 legislative regime and
the pre-1984 regime, both of which had a
card-check union recognition procedure.
This effect holds up in the second specifica-
tion, where a set of controls is added. At
this point the 1993 estimate would suggest
that we should deduct 3 percentage points
from the 1984 estimate, to net out the
effect of the decertification and picketing
rules. However, after I add year effects,
which absorb unobserved factors correlated
with time, the 1993 estimate is not statisti-
cally different from zero. Column (4) adds
another set of controls—industry effects
interacted with a voting (1984–92)
dummy—to test for any differences in the
voting effect across industries, which the
industry effects alone would miss. How-
ever, the legislation estimates remain virtu-
ally the same. Thus, I conclude that voting
accounts for virtually the entire fall in suc-
cess rates that occurred over the 1984 to
1992 period, and that the various other
reforms had no impact on certification suc-
cess. Of course, it must be emphasized that
these other reforms may have had impor-
tant effects along other dimensions.
One concern is that the unit characteris-
tics, union affiliation, unemployment rate,
industry, region, and year dummies are not
adequate controls, and that unobserved
factors are contaminating the results. That
is, something else may be explaining the
dramatic decline in certification success
rates that we observe over the June to De-
cember 1984 period when voting was intro-
duced (Figure 1), and the subsequent in-
19
Two striking features of Canadian unionization
are the decline of U.S. internationals and the rise of
the independent. In 1962, two of every three union
members belonged to an international union, but by
1995 this ratio had fallen to about 20% (CALURA,
various years). Conversely, while there were very few
independent unions in Canada in 1962, by 1995 over
20% of the some 4 million Canadian union members
belonged to an independent.
20
The reported estimates are the change in the
probability of certification for an infinitesimal change
in each continuous independent variable and, for
each dummy independent variable, the discrete
change in the probability of certification when the
dummy variable takes on a value of 1 compared to 0.
See the dprobit routine in Stata version 8.0, which was
the statistical package used in the study.

504 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 1. Estimates of the Change in Probability of
Certification Success: Isolating the Impact of Mandatory Elections, 1978–1998.
(Standard Errors in Parentheses)
Specification
Variable (1) (2) (3) (4) (5)
1984 LCA Act –.241*** –.226*** –.229*** –.229*** –.234***
(.023) (.023) (.076) (.081) (.051)
1987 IRR Act –.218*** –.200*** –.167*** –.153*** –.177***
(.015) (.015) (.047) (.052) (.044)
1993 LCA Act –.031*** –.033*** –.001 –.023 –.011
(.012) (.012) (.037) (.040) (.038)
Industrial Concentration — .071*** .071*** .067*** —
(.026) (.026) (.026)
Unit Representation — –.283*** –.281*** –.280*** —
(.047) (.047) (.047)
Local Representation — .002 .002 .006 —
(.018) (.018) (.019)
% Female — –.019 –.016 –.018 —
(.030) (.030) (.030)
AFL-CIO/CLC — –.035** –.037*** –.037*** —
(.016) (.016) (.016)
CFL — –.013 –.015 –.016 —
(.015) (.015) (.015)
CCU — –.016 –.017 –.018 —
(.030) (.030) (.030)
NIU — .042** .041** .042* —
(.017) (.017) (.020)
PIU — –.023 –.022 –.016 —
(.046) (.046) (.045)
ILU — .075** .075** .072** —
(.024) (.024) (.025)
Teamsters — –.064*** –.066*** –.068*** —
(.022) (.022) (.023)
Unemployment Rate — –.000 –.013** –.013** –.008
(.002) (.006) (.006) (.005)
Industry Dummies No Yes Yes Yes Yes
Regional Dummies No Yes Yes Yes Yes
Year Dummies No No Yes Yes Yes
Industry * 1984–92 Dummies No No No Yes Yes
394 Union Local Dummies No No No No Yes
Log Likelihood –2,639.7 –2,553.1 –2,547.8 –2,537.9 —
R Squared — — — — .38
Sample Size 6,650
Notes: The omitted union category is national unions affiliated with the CLC. The calculation of the marginal
effect is discussed in the text.
*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level.
crease in success rates in 1993 when voting
was repealed. Two final strategies are
adopted to address this concern. First, I
examine all other legislative changes that
were made in 1984 and in 1992/early 1993.
No other single piece of legislation was
changed at both times. In addition, I can
find no other legislation in 1984 or 1992

VOTING VERSUS CARD-CHECK PROCEDURES 505
that could have plausibly affected organiz-
ing success. Second, the final column in
Table 1 displays the results from a specifica-
tion that includes dummy variables for
union locals—a total of 394 dummy vari-
ables. This specification absorbs all varia-
tion across union locals, controlling for
unobserved factors such as union strate-
gies, union staff, union finances, and union
mergers. The coefficients on the legisla-
tion variables remain the same.
21
The union affiliation estimates merit dis-
cussion. The three sets of international
unions are all less likely to certify than were
the CLC national unions (the omitted cat-
egory), although the estimate is not statisti-
cally significant in the case of the CFL
unions. Second, the NIUs and ILUs were
much more likely to successfully certify than
were the CLC unions. The above results are
of interest because, as noted, two striking
features in Canadian union incidence are
the dramatic decline of international (U.S.-
based) unions and increase of indepen-
dent unions. The results suggest that union
organizing effectiveness may be one reason
for these changes. Finally, the industrial
concentration variable is statistically sig-
nificant and positively related to certifica-
tion success, supporting the finding of
Fiorito, Jarley, and Delaney (1995), while
the unit representation variable is statisti-
cally significant and negatively related to
certification success.
22
I now turn to the issue of whether man-
agement opposition underlies the effect of
elections. Column (1) in Table 2 presents
the estimated marginal effects from the
first regression in the second stage of the
analysis, wherein I test the hypothesis that
management opposition is greater under
elections. The average negative impact of
ULPs on certification success is estimated
to be 5 percentage points. Note that this is
a larger impact than that found by
Thomason (1994) in the case of Ontario
under card-check rules. However, the esti-
mated marginal effect on the interaction
term (1987 legislation * ULP) is very large,
and implies that ULPs reduced certifica-
tion success rates by an additional 8 per-
centage points in the voting regime—when
workers voted 10 days following the appli-
cation. That is, the full effect of ULPs over
the 1987 to 1992 period is estimated to be
a 13 percentage point reduction in success
rates. In percentage terms, this represents
an increase in the effectiveness of manage-
ment opposition of about 160%.
The remaining results in Table 2 are
from regressions estimated separately by
legislative regime. Columns (2)–(5) indi-
cate that the OLS and probit marginal ef-
fects for the impact of a ULP are virtually
identical. The estimates imply that a ULP
under card-checks reduces success rates by
nearly 6 percentage points, while the im-
pact for the voting years is 12 percentage
points. This suggests that management
opposition is twice as effective under vot-
ing—a smaller increase than is found
using the interaction approach. The re-
sults for the remaining variables are very
similar to those presented in Table 1
and, to simplify the exposition, are not
presented in Table 2.
The two-stage least squares estimates sug-
gest that OLS and probit estimates of the
21
The union “fixed effects” specification is based
on a linear probability model. However, OLS esti-
mates of models (1)–(4) are very similar to the mar-
ginal effects predicted from the probit models (for
example, OLS estimates for the column 3 specifica-
tion are –.215 and –.164 for the 1984 and 1987 legis-
lation variables, respectively, and –.229 and –.161 for
the column 4 specification).
22
The magnitude of these results is small. A one
standard deviation increase in industrial concentra-
tion increases the likelihood of success by 1.5 per-
centage points, while an analogous increase in unit
representation decreases the probability of success by
3 percentage points. The negative relationship be-
tween unit size and certification success may be driv-
ing the unit representation estimate. However, in
other analysis (not reported here) I examined the
distribution of the unit representation variable (with
and without unit size included as a covariate), and the
results indicate that those certifications with very
high values (above the 90th percentile) of the unit
representation index are driving the results, even
after unit size is included.

506 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 2. Estimates of the Change in Probability of
Certification Success: The Impact of Management Opposition, 1987–1998.
(Standard Errors in Parentheses)
Specification
Voting Card-Check Voting Card-Check Voting Card-Check
Interaction Regime Regime Regime Regime Regime Regime
Model: Only: Only: Only: Only: Only: Only:
Variable Probit Probit Probit OLS OLS 2SLS 2SLS
1987 IRR Act –.173*** — — — — — —
(.015)
Unfair Labor Practice –.051*** –.118*** –.057*** –.120*** –.059*** –.353*** –.281***
(.019) (.031) (.016) (.028) (.013) (.098) (.069)
ULP * Voting Regime –.084*** — — — — — —
(.028)
Log Likelihood –1040.9 –673.1 –356.2 — — — —
R Squared — — — .17 .14 — —
Sample Size 3,023 1,325 1,698 1,325 1,698 1,325 1,698
Notes: For probit models, only the estimated marginal effect is shown. The calculation of the marginal effect
is discussed in the text. All specifications include controls for industrial concentration, unit and local
representation, proportion female, union affiliation, and the unemployment rate, as well as region, industry,
and year effects. The 2SLS estimates are computed using the regression results from a first stage, where ULP
is instrumented with three instruments (in addition to the other above-noted variables): lagged ULP processing
time, a dummy equal to one if lagged ULP processing time is greater than certification processing time, and the
success rate for ULP-compensation claims under subsection 8(4)(e) from the previous three months.
*Statistically significant at the .10 level; **at the .05 level; ***at the .01 level.
impact of ULPs understate the true impact,
a finding that is consistent with previous
research. The results from the first stage
are as anticipated, with the two ULP pro-
cessing time–related variables being nega-
tively associated with the likelihood of a
ULP and statistically significant, and the
subsection 8(4)(c) success rate being posi-
tively related but not statistically signifi-
cant. Industry and the unemployment rate
are also key factors in the first stage. The F-
statistics from the first stage are 30.99 and
42.84 for the card-check and voting regime
regressions, respectively. However, cau-
tion is required in interpreting the 2SLS
estimates, since the standard errors increase
dramatically, and the magnitudes of the
estimated effects appear too large to be
credible. Nevertheless, the voting regime
estimate is still larger than the card-check
regime estimate.
Table 3 presents the decompositions of
the differences in certification success rates
that occurred between the card-check and
voting regimes, with the decomposition
computed so as to yield a positive gap (card-
check success minus voting success). The
decompositions are an attempt to address
the question of how much of the change in
success rates across union recognition re-
gimes can be attributed to changes in the
incidence and effectiveness of management
opposition. The actual difference in suc-
cess rates was 18.6 percentage points and
the predicted difference, based on a probit
model, is somewhat lower at 17.4 percent-
age points, while the predicted difference
from OLS is 18.8 percentage points.
23
The
main result from the second panel of Table
3 is that changes in the means of the X’s are
unimportant. Essentially all of the gap in
success rates can be attributed to differ-
23
The predicted success rates are higher than those
in Figure 1 because those certifications rejected due
to petition violations have been excluded from the
sample.

VOTING VERSUS CARD-CHECK PROCEDURES 507
ences in the estimated coefficients across
the two regimes. This result is identical for
both the probit and the standard Oaxaca-
Blinder decomposition (based on a linear
probability model). In addition, decompo-
sitions that switch the weights (that is, hold-
ing coefficients constant at voting levels
instead of card-check, and vice-versa for
means) yield virtually the same results.
The finding that the incidence of man-
agement opposition did not change across
regimes is contrary to the conventional wis-
dom that management opposition is higher
in a voting regime. For instance, many
authors have noted that there appears to be
more employer opposition to unions in the
United States—where elections are typi-
cally required—than in Canada—where,
until recently, elections have typically not
been required. The results from British
Columbia, which exploit variation in the
union recognition law, demonstrate that
caution is required when comparing re-
sults from different countries.
However, there is reason to expect simi-
lar incidence of management opposition
across the two union recognition proce-
dures if the inherent level of support in
attempted organizing drives is higher in a vot-
ing regime. This possibility is an implica-
tion of management opposition being en-
dogenous. Specifically, given that success-
fully certifying a unit is more uncertain
under elections, unions may respond in
either of two ways when a voting regime is
introduced: by increasing the number of
attempted certifications, or by making fewer
attempts but focusing on campaigns in
which they believe success is very likely—
that is, on campaigns in units with a higher
inherent level of support. As seen in Ap-
pendix 2, the number of certification at-
tempts fell by around 50% in British Co-
lumbia following the introduction of man-
datory elections. This may indicate that
unions focused their organizing efforts on
units with a high inherent level of support.
Of course, unions may simply have become
Table 3. Decomposition of the Difference in
Certification Success across Legislative Regimes, 1987–1998.
Actual Probit OLS
Predicted Certification Success Rates
Card-Check Regime 0.931 .939 .931
Mandatory Elections Regime 0.745 .765 .743
Difference in Success Rate 0.186 .174 .188
Standard Decompositions
Total Difference Due to Means — –.003 –.001
Sub-Decomposition of Means — — —
ULP — –.002 –.001
Union Characteristics — –.006 –.001
Union Affiliation — .001 .000
Industry — .003 .001
Other Controls — .001 .000
Total Difference Due to Coefficients — .183 .189
Approximation Residuals — –.006 —
Total — .174 .188
Nielson Decomposition
Difference Due to Change in Mean of ULP — — –.002
Difference Due to Change in Coefficient of ULP — — .045
Difference Due to Change in Coefficient of Constant Term — — .139
Predicted Gap — — .182
Notes: See Appendix 3 for details on the decompositions and Appendix 4 for means.

508 INDUSTRIAL AND LABOR RELATIONS REVIEW
discouraged during the voting years, in
which case it is unclear whether the inher-
ent level of support would be higher or
lower in the voting regime.
If unions are more selective in a voting
regime, then we may expect fewer ULPs
under a voting regime than under a card-
check regime, because employers will rec-
ognize the greater inherent level of sup-
port and decide that an anti-union cam-
paign is not worthwhile. Thus, while there
are reasons to anticipate a higher incidence
of ULPs in a voting regime—higher effec-
tiveness of anti-certification tactics, and
more time to employ them—there is also
reason to expect fewer ULPs under voting.
Given the fact that most structural fac-
tors (such as shifts in the industrial and
occupational distribution of employment)
do not change abruptly, it seems unlikely
that structural factors explain the striking
observed changes in certification success
shown in Figure 1. Moreover, if structural
factors were an important determinant of
the decline in success rates during the vot-
ing era, we should see some change in the
industrial or regional distribution of certi-
fication attempts. On the contrary, as shown
in Table 3, changes in the means of the
explanatory variables contributed nothing
to the overall change in success rates. As
noted above, it is possible that unobserved
structural change in the economy affected
the incidence of management opposition.
As a check on the finding that structural
factors are unimportant, I estimated sepa-
rate regressions of the probability of union-
ization for each time period (1984 versus
1998) using data on non-agricultural, paid
workers from the 1984 Survey of Union
Membership—a supplement to the Labour
Force Survey—and the 1998 Labour Force
Survey.
24
Essentially, I replicated the analy-
sis of Riddell and Riddell (2004), but re-
stricted the sample to British Columbia
workers. This approach allows us to take
into account various demographic and la-
bor market characteristics that are unavail-
able in the certifications data. Using the
same procedures discussed in Appendix 3,
I decomposed the change in probability of
unionization over the 1984 to 1998 period
into two components: the part due to
changes in the coefficients (changes in the
propensity to be unionized) and the part
due to changes in the means (changes in
the proportion of the labor force that have
given characteristics, such as gender, edu-
cation, part-time status, industry, occupa-
tion, and job tenure).
The results from these nationally repre-
sentative surveys indicate that the decline
in the probability of unionization in British
Columbia over the 1984 to 1998 period was
10 percentage points, well above the 7 per-
centage point decline for Canada. The
decomposition results for British Colum-
bia are quite similar to the results for
Canada, as estimated by Riddell and Riddell,
with only 5% of the change in the probabil-
ity of a non-agricultural, paid worker being
unionized over the 1984 to 1998 period
being attributable to changes in the means
of personal characteristics.
25
It therefore
appears unlikely that structural factors have
played even a modest role in the decline of
unionization in British Columbia.
The final panel in Table 3 uses the ap-
proach suggested by Nielson (1999) to sub-
decompose the total coefficient compo-
nent. The estimated impact of a ULP is
only about one percentage point higher in
a regression that includes only a control for
ULP (and a constant term). Given the
huge volume of statistics produced by this
approach and the fact that the omission of
the other control variables does not change
the ULP estimate, I only present the results
from the parsimonious regression, as sug-
gested by Nielson under such circum-
stances. The part of the difference in suc-
cess rates that can be attributed to a change
24
The 1984 Survey of Union Membership was the
first Canadian survey comparable to the 1998 Labour
Force Survey that included a question on coverage by
collective agreement.
25
These results are available upon request to the
author.

VOTING VERSUS CARD-CHECK PROCEDURES 509
in the mean is essentially zero with an esti-
mate similar to the other decompositions
(–.002), while the part of the gap that can
be attributed to a change in the effective-
ness of ULPs (that is, a change in coeffi-
cients) is estimated to be 4.5 percentage
points, or about 25% of the total gap. I
emphasize that while the Nielson method is
intuitively appealing, it is only one possible
approach to sub-decomposing the coeffi-
cient component, and so some caution is
required in interpreting the 25% estimate.
Overall, the results in Tables 2 and 3
illustrate that even in an environment with
very strict time limits on the election pro-
cess—time limits that were complied with—
management opposition is highly effective
when workers must vote.
Conclusions
It has been argued that mandatory elec-
tions reduce certification success. A key
factor believed to underlie the effect of
voting is management opposition to the
certification bid. In particular, it has been
hypothesized that the incidence and effective-
ness of management opposition tactics are
greater when workers’ preferences are ex-
pressed via a secret ballot vote. This paper
exploits variation in the union recognition
law to test the management opposition hy-
pothesis using data on over 6,500 private
sector certifications from British Colum-
bia. The union recognition law changed
from card-check to mandatory elections,
then back to card-check, providing a unique
opportunity to identify the impact of the
legal regime on union organizing.
My analysis uses the fact that different
laws were changed at different times in the
province, allowing for the impact of elec-
tions to be isolated from other legislative
changes such as decertification, picketing,
first contract arbitration, essential services,
strike and lockout, dues check-off, dispute
resolution, and last offer laws. The results
indicate that mandatory elections can ac-
count for virtually the entire 19 percentage
point decline in private sector certification
success rates that occurred over the 1984 to
1992 period.
Management opposition—as measured
by unfair labor practices—was at least twice
as effective in the voting regime as in the
card-check regime. The additional impact
of ULPs in the voting regime ranged from
6 to 8 percentage points. However, the
incidence of management opposition was
slightly lower in the voting regime. This
apparent anomaly may have arisen because
unions, perceiving a reduced likelihood of
success, reduced their number of certifica-
tion attempts and concentrated only on
units with high levels of inherent support.
Faced with high union support in those
settings, employers may have been less likely
to oppose the bid even though their oppor-
tunity to do so was greater. The decompo-
sitions indicate that management opposi-
tion can account for 25% of the overall
decline in certification success rates. How-
ever, other evidence suggests that ULPs
may only be capturing the more egregious
cases of management opposition. Thus,
there is reason to believe that the 25%
contribution is understated.
To conclude, I wish to discuss some limi-
tations of this study, some caveats, and some
suggestions for future research on this im-
portant topic. Perhaps most important,
caution should be exercised in extending
the results in this paper to the decline of
union organizing success in the United
States, and to the differing fortunes of
unionization in the United States and
Canada. Recall that elections in British
Columbia included a 10-day expedited vot-
ing procedure—a law that was strictly en-
forced. At no time did British Columbia
truly have a U.S.-style system in which there
was voting and no time limit on when the
vote occurred. That said, given that man-
agement opposition was twice as effec-
tive under elections and that organizing
drives have become increasingly long in
the United States, I feel that the results
are consistent with the “management
opposition” story of the decline in U.S.
certification success rates as told by
Weiler, Freeman, and others.
Although restraint is called for in using
the results reported here as a lens for inter-
preting union certifications in the United

510 INDUSTRIAL AND LABOR RELATIONS REVIEW
States, I believe the results have clear policy
implications for U.S. labor law reform.
In particular, many authors have recently
argued that card-check procedures—
through negotiated agreements—show
considerable promise for reducing ille-
gal employer tactics, and generally creat-
ing a fairer playing field in the union
organizing process (Budd and Heinz
1996; Benz 1998; Eaton and Kriesky 2001).
Other authors argue that card-check pro-
cedures artificially inflate support levels
(Yager, Bartl, and LoBue 1998). An obvi-
ous compromise is so-called “instant” elec-
tions, as proposed by Weiler (1983, 1990),
whereby the vote takes places a few days
following the application. The case of Brit-
ish Columbia—with a strictly enforced 10-
day time limit during the voting years—
illustrates that employers can still influ-
ence the election outcome even with a
short time limit.
Finally, the analysis in this paper demon-
strates that elections appear to account for
the entire decline in certification success
rates. However, as noted, while success
rates remained high—relative to those in
other countries—the number of certifica-
tion attempts fell by around 50% in 1984,
and even after the 1993 legislation restor-
ing card-check rules it never recovered to
the pre–1984 levels. While a number of
explanations could exist for the decline in
organizing activity, it may be the case that
other industrial relations legislation had
an impact on the number of certification
attempts. Future research should investi-
gate the effect of both union recognition
laws and other industrial relations laws on
overall union organizing.

VOTING VERSUS CARD-CHECK PROCEDURES 511
Appendix 1
Labour Code in British Columbia, 1978 to 1998
Labour Code in 1984 Labour Code 1987 Industrial 1993 Labour Code
Law Effect as of 1978 Amendment Act Relations Reform Act Amendment Act
Union Section 39: Section 39 Section 43 replaces: Section 18 replaces:
Recognition Automatic certifica- repealed and Vote held if 45%+ sign automatic certification
tion if 55%+ sign replaced: Vote cards; expedited voting if 55%+ sign cards; vote
cards, vote held if held if 45%+ of 10 days. held if between 45%
between 45% and sign cards; expe- and 55%; expedited
55%. dited voting regu- voting of 10 days (if vote
lation of “approx- required).
imately” 10 days.
Decertifica- Section 52: Upon Section 52 Section 52 amended: Section 33 replaces:
tion application by some repealed and Vote held if 45%+ sign Vote held if 45%+ sign
party, Board rules replaced: Vote cards; expedited voting cards; expedited voting
on whether union held if 45%+ of 10 days. of 10 days.
has ceased to be sign cards; expe-
agent of employees dited voting regu-
or employer has lation of “approx-
ceased to employ imately” 10 days.
workers.
Strike and Section 79: No Section 79: No Part 8.1 introduced: Part 7 replaces Part 8.1
Lockout strike or lockout trike or lockout Written notice first (see below): Time
Regulations can take place while scan take place required plus minimum limits modified; Section
a collective agree- while a collective 72-hour wait, potential 70 introduced; Section
ment is in force agreement is in 48-hour wait following 57 and 59 replaces
(subject to essential force (subject to mediator review if medi- Section 79: Temporary
service designation), essential service ator appointed, poten- strike replacement
or before bargaining designation), or tial 48-hour wait follow- workers restricted. No
and strike vote. before bargaining ing fact-finding report if strike or lockout can
and strike vote. fact-finder appointed; take place while a
Section 79 in effect as collective agreement is
well. in force (subject to
essential service
designation), or before
bargaining and strike
vote.
Picketing Sections 85 and 86: Sections 85 and Section 85 amended: Section 65 replaces:
Employees can 86 repealed and After application and After application and
picket at employer, replaced: After review or by its own review or by its own
any other location application and motion, the Board may motion, the Board may
of the employer, a review, Board allow employees to allow employees to
common site, or any may allow em- picket at another location picket at another
location of an ally ployees to picket of the employer or a location of the em-
of the employer. at another loca- location of an ally of the ployer or a location of
tion of the em- employer; cannot picket an ally of the employer;
ployer or a loca- a common site unless cannot picket a
tion of an ally of common site is an ally common site unless
the employer; (and the application is common site is an ally
cannot picket a approved). (and the application is
common site unless approved).
common site is an
ally (and the appli-
cation is approved).
Last Offer No law applicable. No law applicable. Section 137.7 intro- Section 78 replaces:
Received duced: Employer may Employer may request
Voting request that, before a that, before a strike
strike vote, employees vote, employees vote on
vote on last offer last offer received.
received.
continued

512 INDUSTRIAL AND LABOR RELATIONS REVIEW
Appendix 1 Continued.
Labour Code in 1984 Labour Code 1987 Industrial 1993 Labour Code
Law Effect as of 1978 Amendment Act Relations Reform Act Amendment Act
Essential Section 73: Fire- Section 73: Fire- Section 137.8 replaces: Section 72 replaces:
Services fighters, police, and fighters, police, Employer can apply for Employer can apply for
hospital unions can- and hospital “cooling-off period” essential service
not strike and can- unions cannot and/or essential service designation if dispute
not be locked out; strike and cannot designation if dispute threatens “health/
no other essential be locked out; no threatens “economy of safety/welfare of
service designation. other essential province, health/safety/ province.”
service designation. welfare of province, or
education.”
Collective Section 69: Medi- Section 69: Medi- Part 8.1 (sections 81 and Part 7 (sections 76, 77,
Bargaining ation services ation services 82) replaces: Mediation and 79) replaces:
Dispute (Minister may upon (Minister may introduced by applica- Mediation by applica-
Resolution application appoint upon application tion,at the direction of tion, or at the direction
Procedures a mediation officer appointa media- the Commissioner or by of the Board (non-
to confer with both tion officer to IRC’s own motion (non- binding; full review and
parties and assist in confer with both binding; full review and report conducted, can
reaching an agree- parties and assist report conducted, can be be made public); fact-
ment). First contract in reaching an made public); fact- finding as above (non-
arbitration available agreement). First finding introduced as binding; full review and
under section 70. contract arbitra- above (non-binding; full report conducted, can
tion available review and report con- be made public); public
under section 70. ducted, can be made interest inquiry board
public); public interest relabeled as Industrial
inquiry board introduced Inquiry Commission
(employees can be made (employees can be
to vote on decision); in- made to vote on
tervention power (by decision). First contract
resolution of Legislative arbitration available
Assembly) to end a dis- under section 55.
pute. First contract
arbitration available
under section 1375.5.
Additional No law applicable Section 5 intro- Section 5.1 introduced: Section 5 replaces: Job
Union (mandatory dues duced: Job loss or Job loss or job-related loss or job-related
Membership check-off in effect, job-related discrim- discrimination prohib- discrimination prohib-
and Dues section 10). ination prohibited ited for not being a ited for being member
Regulations for not participa- union member unless of different union;
ting in unlawful individual failed to pay mandatory dues check-
industrial action dues; mandatory dues off removed in cases of
(mandatory dues check-off removed in religious convictions, in
check-off in effect, cases of religious convic- effect in all other cases.
section 10). tions, in effect in all
other cases.
Notes: The 1973 LCA (including 1974 and 1976 amendments) was the legislation in effect as of 1978. To
simplify the exposition, the following terms are adopted: introduced means that the section is making its first
appearance in the Code, replaces means that the section replaces the existing section (often the section number
remains the same). Note that in many cases the language of the law was identical even if a new section number
was introduced.

VOTING VERSUS CARD-CHECK PROCEDURES 513
Appendix 2
Selected Statistics from British Columbia, 1978 to 1998
Cases Filed Cases Filed Average % of Cases
Certifications Sub-Section Sub-Section Total in Micro-
Filed, Certifications 8(4)(E) 8(4)(C) Processing Data Non-
Annual Filed, [Number [Number Time for Vote Going to a Dismissal Dismissal
Year Report Micro-Data Successful] Successful] Certification Vote ULPs ULPs
1978 672 54
a
17 [1] 55 [37] 145.60 4.00 71 46
1979 662 500 25 [1] 59 [40] 167.27 3.24 69 34
1980 790 556 22 [0] 53 [38] 116.83 4.67 77 39
1981 848 659 34 [2] 69 [58] 135.24 3.79 99 49
1982 916 593 15 [2] 74 [65] 126.15 4.60 107 54
1983 678 410 18 [0] 56 [47] 141.95 5.31 81 67
1984 376 221 21 [3] 62 [52] 39.03 55.45 80 56
1985 333 197 16 [2] 42 [36] 21.94 99.00 62 69
1986 385 249 18 [2] 56 [39] 25.70 99.02 74 84
1987 343 164 17 [0] 41 [30] 23.15 100.00 77 73
1988 509
b
238 10 [0] 37 [25] 14.75 100.00 53 55
1989 424 253 10 [0] 23 [13] 18.39 100.00 37 97
1990 437 287 18 [3] 34 [26] 17.84 99.01 64 250
c
1991 414 268 20 [1] 35 [34] 15.83 99.72 73 232
1992 340 207 32 [6] 31 [28] 24.44 97.11 56 220
1993 683 424 31 [2] NA 40.43 14.31 NA NA
1994 646 336 31 [2] NA 43.56 8.50 NA NA
1995 606 298 35 [0] NA 35.93 10.58 NA NA
1996 611 348 41 [1] NA 37.33 9.54 NA NA
1997 589 279 52 [3] NA 37.55 7.61 NA NA
1998 528 109
a
40 [0] NA 39.25 9.49 NA NA
Notes: Column (1) is the total number of certifications filed, as given in the BCLRB annual report. Column
(2) is the total number of certifications based on the micro-data used in the analysis, and thus excludes raids,
withdrawn and “public-sector” certifications, as well as petition violation cases (except <45% signed up).
Column (3) refers to the power—subsection 8(4)(e), relabeled subsection 14(4)(f) in 1993—of the BCLRB to
impose automatic certification without a vote in cases of a ULP. Subsection 8(4)(c) in column (4) refers to
compensation claims for dismissal ULPs. With respect to the number of successful 8(4)(c) cases in parentheses,
only those dismissal ULPs that resulted in a decision are included (that is, withdrawn cases excluded, which
includes some settlements). I consider a successful case as being an individual who received some form of
compensation (that is, re-instatement, back-pay, or both). Column (5) is the time from the application date to
the decision date, since the date of the vote is not always recorded in the data; the 1984 cell consists of both voting
and card-check laws. Column (6) is the percentage of column (2) cases (that is, excluding raids, withdrawn, and
public sector certifications as well as petition violations) that went to a vote. Columns (7) and (8) show the
number of ULP cases by the nature of the charge. The source for columns (2) and (6) is the micro-data on
certification applications collected from the BCLRB, while all other statistics are from the Annual Report of the
BCLRB.
a
For the micro-data, the 1978 cell consists of October to December only and the 1998 cell consists of January
to May only.
b
Includes 75 certifications of teachers (over 31,000 employees—all successful), who were given the right to
unionize in 1988.
c
The non-dismissal ULP numbers for post–1989 are not comparable to the numbers for earlier years (see the
text for further discussion).

514 INDUSTRIAL AND LABOR RELATIONS REVIEW
Appendix 3
Decomposition Techniques
If we rewrite (1) from the text as
(2a) Prob(SUCCESS =1) = F(X
ij
'_
β
j
),
then the predicted probability of certification success is
(2b) P
ˆ
(X
ij
β
j
) = 1/N
j
∗ Σ Φ (X
ij
'
β
ˆ
j
),
where X
ij
is a vector of the covariates of the i
th
bargaining unit in legislative regime j; β
ˆ
j
is a vector of probit
parameters estimated from legislative regime j; N
j
is the number of bargaining units in regime j; and Φ is the
standard normal cumulative distribution function.
If a jurisdiction changes union recognition procedures, then the change in predicted certification success
rates can be written as
P
ˆ
(X
CARD
, β
CARD
) – P
ˆ
(X
VOTE
, β
VOTE
),
which yields a positive gap. This can be decomposed as
(2c) P
ˆ
(X
CARD
, β
CARD
) – P
ˆ
(X
VOTE
, β
VOTE
) = [P
ˆ
(X
CARD
, β
CARD
) – P
ˆ
(X
VOTE
, β
CARD
)]
+ [P
ˆ
(X
VOTE
, β
CARD
) – P
ˆ
(X
VOTE
, β
VOTE
)].
The first term on the right-hand side of (2c) indicates what part of the decline in success rates is due to a
change in the means of the covariates holding the coefficients constant at their card-signing levels. The second
term indicates what part of the decline in success rates is due to a change in the coefficients holding the means
constant at their voting levels. An alternative to (2c) reverses the comparison group (that is, difference in means
holding the coefficients constant at voting levels and difference in coefficients holding the means constant at
card-check levels).
Due to Φ being non-linear, there is no unique way to decompose the decline in success rates. Following
Doiron and Riddell (1994) and Yun (2000), I adopt a modification of (2c) that allows the decline in success rates
to be decomposed despite the non-linear specification. The approach adopted is based on a linear approxima-
tion of the probability of certification and proceeds in two stages. In the first stage, the sample average of the
standard normal CDF is approximated by the standard normal CDF at the sample mean of the explanatory
variables. In the second stage, the differences between the two standard normal CDFs from stage one are
approximated by a first-order Taylor expansion about X
–
j
β
ˆ
j'
= X
–
j
β
ˆ
j
, where j ≠ j'. For more details, see Doiron and
Riddell (1994) and Yun (2000). The final decomposition formula is
(2d) P
ˆ
(X
CARD
, β
CARD
) – P
ˆ
(X
VOTE
, β
VOTE
) = (X
–
CARD
– X
–
VOTE
) β
ˆ
CARD
φ(X
–
CARD
β
ˆ
CARD
)
+ (β
ˆ
CARD
– β
ˆ
VOTE
)X
–
VOTE
φ(X
–
VOTE
β
ˆ
VOTE
) + R
1
+ R
2
,
where R
1
and R
2
are residual terms from each approximation stage (see Yun 2000 for the R
1
and R
2
formulas)
and φ is the standard normal probability density function.
Essentially the only difference between (2d) and a standard Oaxaca-Blinder decomposition is the approxi-
mation error and the fact that the coefficients are multiplied by a standard normal PDF. Given that nonlinear
decompositions remain an uncertain area, in the sense that there are a number of methods available with no
consensus on which method is more reliable, I also estimate the certification success equations by union
recognition regime using a linear probability model and decompose the change in success rates using a standard
Oaxaca-Blinder decomposition.
As is well known (Jones 1983), it is not straightforward to further sub-decompose the total coefficient
component—that is, (β
ˆ
CARD
– β
ˆ
VOTE
)X
–
VOTE
φ(X
–
VOTE
β
ˆ
VOTE
)—into each individual coefficient contribution, which is
required to identify how much the change in the effectiveness of management opposition contributed to the
change in success rates. Using the method in Nielson (1999), I provide one possible estimate of the contribution
of the change in effectiveness of management opposition. The drawback of Nielson’s approach is that even with
a relatively small number of covariates, a huge number of computations are required. The Nielson method
essentially uses the implicit constant term for each “group of people” (in the present case, each “legislative
regime”) as a basis for the decomposition, since these parameters are the same regardless of what omitted
categories are used, and is computed as follows (in the case of only a single dichotomous regressor, ULP, and
simplifying to a linear probability model):
(2e) (β
ˆ
CARD
– β
ˆ
VOTE
)X
–
VOTE
= ζ
VOTE1
[(β
ˆ
0,CARD
– β
ˆ
0,VOTE
)
+ (β
ˆ
ULP,CARD
– β
ˆ
ULP,VOTE
)] + ζ
VOTE2
[β
ˆ
0,CARD
– β
ˆ
0,VOTE
],
where ζ
VOTE1
and ζ
VOTE2
are the proportion of voting regime certifications with and without a ULP, respectively
(Nielson 1999).
i=1
n
j

VOTING VERSUS CARD-CHECK PROCEDURES 515
Appendix 4
Summary Statistics
Card-check/ Voting/ Card-check/ Voting/
Full ULP ULP Full ULP ULP
Sample, Sample, Sample, Sample, Sample, Sample,
1978– 1993– 1987– 1978– 1993– 1987–
Variable 1998 1998 1992 Variable 1998 1998 1992
Unfair Labor Practice — .231 .213 NIU .041 .077 .070
(.422) (.409) (.198) (.266) (.256)
1976 LCA Act .432 PIU .010 .020 .007
(.495) — — (.101) (.140) (.082)
1984 LCA Act .084 — — ILU .013 .009 .017
(.279) (.114) (.097) (.131)
1987 IRR Act .213 — — Teamsters .079 .065 .091
(.410) (.269) (.246) (.288)
1993 LCA Act .270 — — Manufacturing .185 .191 .227
(.444) (.388) (.394) (.419)
Industrial Concentration .625 .565 .606 Construction .381 .304 .297
(.212) (.210) (.213) (.486) (.460) (.457)
Unit Representation .038 .048 .041 Transportation .068 .079 .090
(.121) (.126) (.126) and Storage (.252) (.270) (.286)
Local Representation .261 .236 .310 Primary Industries .043 .058 .046
(.268) (.271) (.309) (.202) (.234) (.210)
% Female .165 .176 .191 Trade .120 .114 .120
(.204) (.199) (.211) (.325) (.317) (.325)
CLC .188 .346 .222 Hotel and .064 .075 .077
(.391) (.476) (.416) Restaurant (.245) (.264) (.267)
AFL-CIO/CLC .192 .162 .251 Business Services .027 .029 .026
(.394) (.368) (.434) (.163) (.167) (.160)
CFL .456 .312 .314 Other Services .112 .149 .116
(.498) (.463) (.464) (.315) (.356) (.321)
CCU .021 .010 .028 Unemployment 9.705 8.907 9.924
(.143) (.099) (.165) Rate (2.557) (0.635) (1.222)
Sample Size 6,650 1,698 1,325
Notes: Means (and standard deviation in parentheses) are based on the certifications micro-data used in the
regression analysis (that is, raids, petition violations, “public” sector certifications excluded). To simplify the
table, summary statistics for region (UI region within British Columbia) and year are not shown.

516 INDUSTRIAL AND LABOR RELATIONS REVIEW
REFERENCES
Beaumont, Phillip B., and Richard I. Harris. 1995.
“Union De-Recognition and Declining Union Den-
sity in Britain.” Industrial and Labor Relations Review,
Vol. 48, No. 3 (April), pp. 389–402.
Bentham, Karen. 2002. “Employer Resistance to
Union Certification: A Study of Eight Canadian
Jurisdictions.” Relations Industrielles/Industrial Rela-
tions, Vol. 57, No. 1 (Winter), pp. 159–85.
Benz, Dorothee. 1998. “Scaling the Wall of Employer
Resistance: The Case for Card Check Campaigns.”
New Labor Forum, Fall/Winter, pp. 118–28.
Bronfenbrenner, Kate. 1997. “The Role of Union
Strategies in NLRB Certification Elections.” Indus-
trial and Labor Relations Review, Vol. 50, No. 2 (Janu-
ary), pp. 195–212.
Budd, John W., and Paul K. Heinz. 1996. “Union
Representation Elections and Labor Law Reform:
Lessons from the Minneapolis Hilton.” Labor Stud-
ies Journal, Vol. 20, No. 4 (Winter), pp. 3–20.
Cooke, William N. 1985. “The Rising Toll of Dis-
crimination against Union Activists.” Industrial Re-
lations, Vol. 24, No. 3 (Fall), pp. 421–42.
Dickens, William. 1983. “The Effect of Company
Campaigns on Certification Elections: Law and
Reality Once Again.” Industrial and Labor Relations
Review, Vol. 36, No. 4 (July), pp. 560–75.
Disney, Richard, Amanda Gosling, and Stephen
Machin. 1995. “British Unions in Decline: Deter-
minants of the 1980s Fall in Union Recognition.”
Industrial and Labor Relations Review, Vol. 48, No. 3
(April), pp. 403–19.
Doiron, Denise J., and W. Craig Riddell. 1994. “The
Impact of Unionization on Male-Female Earnings
Differences in Canada.” Journal of Human Resources,
Vol. 29, No. 2 (Spring), pp. 504–34.
Eaton, Adrienne E., and Jill Kiesky. 2001. “Union
Organizing under Neutrality and Card Check Agree-
ments.” Industrial and Labor Relations Review, Vol.
55, No. 1 (October), pp. 42–59.
Fiorito, Jack, Paul Jarley, and John Thomas Delaney.
1995. “National Union Effectiveness in Organizing:
Measures and Influences.” Industrial and Labor Rela-
tions Review, Vol. 48, No. 4 (July), pp. 613–35.
Freeman, Richard B. 1985. “Why Are Unions
Faring Poorly in NLRB Representation Elections?”
In Thomas Kochan, ed., Challenges and Choices
Facing American Labor. Cambridge, Mass.: MIT
Press, pp. 45–64.
____. 1989. “The Changing Status of Unionism
around the World: Some Emerging Patterns.” In
Wei-Chiao Huang, ed., Organized Labor at the Cross-
roads. Kalamazoo, Mich.: Upjohn Institute for
Employment Research, pp. 111–38.
Freeman, Richard B., and Morris M. Kleiner. 1990.
“Employer Behavior in the Face of Union Organiz-
ing Drives.” Industrial and Labor Relations Review,
Vol. 43, No. 4 (July), pp. 351–65.
Freeman, Richard B., and Jeffrey Pelletier. 1990.
“The Impact of Industrial Relations Legislation on
British Union Density.” British Journal of Industrial
Relations, Vol. 28, No. 2 (July), pp. 141–64.
Johnson, Susan. 2002. “Voting or Card-Check: How
the Union Recognition Procedure Affects Organiz-
ing Success.” Economic Journal, Vol. 112, No. 479,
pp. 344–61.
Jones, F. L. 1983. “On Decomposing the Wage Gap:
A Critical Comment on Blinder’s Method.” Journal
of Human Resources, Vol. 18, No. 1 (Winter), pp. 126–
30.
Lawler, John J. 1984. “The Influence of Management
Consultants on the Outcome of Union Certification
Elections.” Industrial and Labor Relations Review,
Vol. 38, No. 1 (October), pp. 38–51.
Lawler, John J., and Robin West. 1985. “The Impact
of Union-Avoidance Strategy in Representation Elec-
tions.” Industrial Relations, Vol. 24, No. 3 (Fall), pp.
406–20.
Martinello, Felice. 1996. “Certification and Decerti-
fication Activity in Canadian Jurisdictions.” Current
Issues Series. Kingston, Ont.: IRC Press.
____. 2000. “Union Activity in Ontario.” Canadian
Public Policy, Vol. 26, No. 1 (March), pp. 17–34.
Martinello, Felice, and Ronald Meng. 1992. “Effects
of Labor Legislation and Industry Characteristics
on Union Coverage in Canada.” Industrial and Labor
Relations Review, Vol. 46, No. 1 (October), pp. 176–
90.
Martinello, Felice, and Charlotte Yates. 2003. “Union
and Employer Tactics in Ontario Certification Ap-
plications.” Mimeo, Brock University.
Meltz, Noah. 1985. “Labor Movements in Canada
and the United States.” In Thomas A. Kochan, ed.,
Challenges and Choices Facing American Labor. Cam-
bridge, Mass.: MIT Press, pp. 315–34.
Nielson, Helena S. 1999. “Discrimination and De-
tailed Decomposition in Logit Models.” Economic
Letters, Vol. 61, No. 1 (October), pp. 115–20.
Pierret, Charles R. 2001. “Event History Data and
Survey Recall.” Journal of Human Resources, Vol. 36,
No. 3 (Summer), pp. 439–66.
Riddell, Chris. 2001. “Union Suppression and Certi-
fication Success.” Canadian Journal of Economics,
Vol. 34, No. 2 (May), pp. 396–410.
Riddell, Chris, and W. Craig Riddell. 2004. “Chang-
ing Patterns of Unionization: The North American
Experience.” In Thomas A. Kochan and Anil Verma,
eds., Unions in the 21st Century: An International
Perspective. London: Palgrave Macmillan.
Thomason, Terry. 1994. “The Effect of Accelerated
Certification Procedures on Union Organizing Suc-
cess in Ontario.” Industrial and Labor Relations Re-
view, Vol. 47, No. 2 (January), pp. 207–25.
Thomason, Terry, and Silvana Pozzebon. 1998.
“Managerial Opposition to Union Certification in
Quebec and Ontario.” Industrial Relations/Relations
Industrielles, Vol. 53, No. 4 (Fall), pp. 750–69.
Weiler, Paul. 1983. “Promises to Keep: Securing
Workers’ Rights to Self-Organization under the
NLRA.” Harvard Law Review, Vol. 96, pp. 1769–
1827.
____. 1990. Governing the Workplace: The Future of
Labor and Employment Law. Cambridge, Mass.:

VOTING VERSUS CARD-CHECK PROCEDURES 517
Harvard University Press.
Wood, Stephen, and John Godard. 1999. “The
Statutory Union Recognition Procedure in the
Employment Relations Bill: A Comparative Analy-
sis.” British Journal of Industrial Relations, Vol. 37,
No. 2 (June), pp. 203–44.
Yager, Daniel V., Timothy J. Bartl, and Joseph J.
Lobue. 1998. Employee Free Choice: It’s Not in the
Cards. Washington, D.C.: LPA.
Yun, Myeong-Su. 2000. “Decomposition Analysis for
a Binary Choice Model.” Discussion Paper #145,
Institute for the Study of Labor.