Employer Behavior in the Face of Union Organizing Drives
Author(s): Richard B. Freeman and Morris M. Kleiner
Source: ILR Review, Vol. 43, No. 4 (Apr., 1990), pp. 351-365
Published by: Sage Publications, Inc.
Stable URL: http://www.jstor.org/stable/2524126
Accessed: 26-03-2018 13:45 UTC
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide
range of content in a trusted digital archive. We use information technology and tools to increase productivity and
facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
Sage Publications, Inc. is collaborating with JSTOR to digitize, preserve and extend access to
ILR Review
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR IN THE FACE OF UNION
ORGANIZING DRIVES
RICHARD B. FREEMAN and MORRIS M. KLEINER*
Using data from a 1986 survey of employers and a 1982-83 survey of
union organizers, the authors investigate the determinants and
consequences of employer opposition to union organizing drives. They
find that strong management opposition, as evidenced by, for example,
the filing of formal charges of unfair labor practices against
management, was most likely when the firm had relatively low wages,
poor working conditions, and supervisory problems; when the likelihood
of union victory was uncertain; and when the potential union
compensation differential-and thus the potential effect on firm
profits-was high. Opposition by supervisors was particularly effective
in defeating union drives. The authors conclude that firms' responses to
organizing drives were consistent with the motive of profit maximization,
and that management opposition has been an important
determinant of the decline of unionization.
IN the 1970s and 1980s management
became increasingly hostile to trade
unions. The goal of a union-free environment,
once espoused only by fringes of the
management community, spread until by
1983 45% of the firms in the BNA's
Personnel Practices Forum declared that
being nonunion was their major labor
* Richard Freeman is Professor of Economics at
Harvard University and Director, Labor Studies
Program at the National Bureau of Economic
Research. Morris Kleiner is Professor of Public
Affairs and Industrial Relations at the Humphrey
Institute and Industrial Relations Center, University
of Minnesota, and Research Economist at the
National Bureau of Economic Research. This study
was funded by the National Science Foundation. The
authors thank Gary Chamberlain, William Dickens,
and participants in seminars at Bradley University,
NBER, and the University of Minnesota for helpful
comments. They also thank Janet Bull, Alida
Castillo-Freeman, and Thomas Clifton for research
assistance.
Data and programs can be obtained from Morris
M. Kleiner, Humphrey Institute, University of
Minnesota, Minneapolis, MN 55455.
relations goal (Kochan, McKersie, and
Chalykoff 1986:487). Unfair labor practices
committed by management skyrocketed
despite a decline in NLRB representation
elections, and approximately onethird
of firms in which workers voted to
unionize failed to sign a collective contract,
effectively reversing the election
results. '
Management opposition, of one form or
another, has been found to be a key
determinant of NLRB representation election
outcomes (Dickens 1983; Freeman
1988), and many believe that increased
opposition has been a major cause of the
precipitous fall in private sector union
density over the past two decades (Freeman
and Medoff 1984, Chap. 15; Dickens
and Leonard 1985; Farber 1987). Despite
' The unfair practices data are from the NLRB.
The percentage of cases in which a contract was
signed is from Cooke (1985), McDonald (1983), and
Freeman and Kleiner (1986).
Industrial and Labor Relations Review, Vol. 43, No. 4 (April 1990). ( by Cornell University.
0019-7939/90/4304 $01.00
351
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
352 INDUSTRIAL AND LABOR RELATIONS REVIEW
the important part management opposition
has undoubtedly played in the deunionization
of the U.S. private sector,
however, relatively little is known about
what determines management behavior
during an organizing drive. Does the
extent of management opposition depend
on the likely costs of unionism to the firm,
as the profit-maximizing calculus would
lead one to expect? Are higher wages and
good working conditions-so-called positive
industrial relations-substitutes or
complements for hiring consultants, committing
unfair practices, or otherwise
campaigning intensely against unionization?
Which management tactics have the
greatest impact on representation election
outcomes?
In this paper we use two data sets
covering 1980s organizing drives to examine
these questions. The first data set is
based on a 1986 survey of 202 establishments
that had representation elections in
the Boston and Kansas City NLRB districts.
The survey asked managers about
the number of days consultants were hired
to oppose unions in the election drive,
perceived causes of the drive, wages and
benefits paid to workers, and the impact
of union victories on the careers of the
managers. We supplement these survey
data with information from the NLRB
about unfair labor practices and representation
election outcomes. The second data
set was obtained for 1982-83 from an
AFL-CIO survey of union organizers in
274 NLRB representation elections. The
organizer survey contains data on benefits
available at the firm, characteristics of
workers, management anti-union tactics,
organizers' perceptions of the issues that
mattered most in the representation campaign,
and organizers' perceptions of the
role of supervisors in campaigning against
unions.
Both surveys have weaknesses, in part
because they lack information on some
variables (the organizer survey contains no
data on consultant days used, and the
employer survey contains no data on
supervisors' activity in the drive) and in
part because they are obtained from
participants with definite biases. We believe,
however, that by examining the two
surveys together we obtain a more complete
and accurate picture of management
opposition than could be obtained using
each survey separately.
Theory of Management Opposition
How does management react to a union
organizing drive at one of its plants? What
influences management policies toward
unionism once management knows it is
headed for an NLRB representation election?
We postulate that the reaction to an
organizing drive depends on three factors:
the expected effect of management opposition
on the probability that the union will
win the election; the costs of opposition;
and the prospective loss of profits due to
unionization, which itself depends on the
union wage differential. Our model predicts
a nonlinear relation between management
opposition and the "innate" probability
of a union win, with firms opposing
unions more strongly when the election
outcome is uncertain than when the union
is especially likely either to win or to lose;
and predicts that opposition will increase
nonlinearly as the effect of unionism on
the firm's labor compensation grows.
We assume at the outset that principalagent
issues between management and
shareholders are of negligible importance
and thus that management's actions are
determined by perceptions of what unionization
will do to profits. This is not an
innocuous assumption. Union-induced
changes in shareholder returns are unlikely
to affect significantly the economic
position of most managers, particularly
the foreman and other lower-level supervisors
whose behavior is critical to any
management campaign to defeat unions.
(Our evidence suggests that firms' managerial
personnel policy substitutes for the
incentive of ownership in motivating plantlevel
management to oppose unionization
of work forces.) Conversely, lower-level
management may devote resources to
opposing unions when doing so is not in
the shareholders' interest, sacrificing profits
for control and greater flexibility at the
workplace. Further, as union corporate
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 353
campaigns have shown, unions can pressure
shareholders or management into
neutrality in representation elections by
imbedding the union drive into a broader
problem of, for example, obtaining assets
from a major financial institution (Pruitt,
Wei, and White 1988).
In its simplest form, our model contains
three basic equations:
(1) Probability of a union victory in the
campaign: P(X, Mo) with P' = dP/dMo <0,
with MO = management opposition and X
= other factors that determine worker
propensity to support a union, such as the
composition of the work force and special
conditions in the firm. We call the
probability that the union will win the
representation election absent any management
opposition the "innate probability"
of a union win and represent it as P(XO).
Because P is bounded between 0 and 1, we
postulate a logistic form, with dP =
- B( 1- P) (P) where B is the impact parameter
of management opposition in the
logistic equation.
(2) Cost of management opposition: C(Mo)
with C'>0, where examples of opposition
are committing unfair labor practices,
hiring specialized consultants to direct an
anti-union campaign, and directing supervisors
to try to convince workers to oppose
the union.
(3) Loss of profits due to unionization:
L(DW), where DW is the likely wage (or
cost) difference due to unionization of the
firm. Price theoretic considerations imply
that L'<O and L"<0. Moreover, because
the standard "welfare triangle" loss due to
monopoly wage gains depends on the
square of the wage differential, the relation
between the differential and loss of
profits is likely to have a parabolic form
(Freeman 1986).2
2 Formally, the loss in profits to management from
a given wage differential DW is L DW + 1/2 (DW) (DL),
where L is the final level of employment and DL is
the change in employment due to the differential.
Here, LDW is the transfer of profits from the firm to
workers absent any labor demand response; and 1/2
(DW)(DL) is the triangle loss in welfare (paid from
profits) due to the demand response. If the elasticity
of demand for labor is constant, h, the loss in profits
is L DW + 1/2 h (DW)2. This is a parabolic function in
We assume that the firm seeks to
minimize the expected loss from unionization,
subject to the cost and probability
functions. The solution yields an inverted
U-shaped relation between the extent of
opposition and the innate probability of a
union victory. One way to see the reason
for the inverted U is to consider management's
decision whether or not to oppose
an organizing drive at all. If the firm does
not oppose the drive it has an expected
loss of L(DW) P(XO). If it opposes the
drive and chooses the optimum level of
opposition MO* with cost C*, it has an
expected loss of C* + L(DW) P(X, MO*)which
is the sum of the probabilityweighted
loss if it fails to defeat the
organizing drive, (L + C*) P(X, MO*), and
the probability-weighted loss if it succeeds
in defeating the drive, [1 - P(X, MO*)] C*.
Given these costs, the firm will oppose the
union drive if the expected savings from
opposition exceed the cost of the management
campaign:
(4) - L(DW) [P(X, MO*) - P(X,O)]
> C*
or, letting P' = (B(1-P)P) approximate
the difference in probability, if B(1-P)P
> C*/L.
Because the derivative of the logistic (or
any similarly shaped probability function)
varies with the level of probability, equation
(4) implies a nonlinear relation between
company opposition and the (innate)
probability of a union victory: when
P is large and the union nearly certain to
win, the firm will forgo campaigning
against the union; similarly, when P is
small and the union nearly certain to lose,
the firm will not spend many resources
fighting the union. Only when there is
serious doubt about the likely victor will
management work hard to oppose the
union.
Turning to the optimum level of management
opposition, the interior solution
requires that the firm equate the marginal
the union wage differential. More generally, the
convexity of the profits function with respect to
wages implies that the loss in profits grows nonlinearly
as the wage differential grows.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
354 INDUSTRIAL AND LABOR RELATIONS REVIEW
benefit of opposition and the marginal
cost of opposition:
(5) P' L = B P(1-P) L = C'
Again we get a nonlinear relation between
the innate probability of a union win and
management opposition. The closer P is to
one-half, the greater is the marginal
benefit from opposition, and thus the
greater the likely opposition.
Solving the model for management
opposition yields the basic equation of
concern in this study-the dependence of
opposition on the likely union wage gain
and consequent losses in profits, the
innate probability that the union will win,
and the costs of opposition:
(6) MO = F(P(XO), C, DW),
where dF/dP rises and then falls with the
level of P; dF/dC < 0; and dF/dDw> 0 and
rising because of the concavity of the
profits function with respect to wages. As
the explanatory variables in (6) are exogenous,
the equation can be estimated by
single-equation methods.
The endogeneity of management opposition
creates problems, however, in estimating
the other key relation in the
model, the impact of opposition on outcomes
(equation [1]). When the innate
probability of a union win is high and
poorly specified in our equation, we
should observe little opposition, and might
incorrectly infer that the lack of opposition
caused the win. By the nonlinearity of
the opposition equation, however, when
the innate probability of a union win is low
we should also observe little opposition,
and might incorrectly infer that lack of
opposition caused the loss. The net result
of these effects is that single equation
estimates of the effect of opposition on
representation outcomes are likely to be
biased toward zero. We deal with this
problem by using instrumental variables to
estimate equation (1), replacing the actual
level of opposition by predicted levels
from our management opposition equation.
The key variables in this procedure
are the postulated nonlinear relations
between the propensity to unionize and
the union wage premium and management
opposition; in our model these
nonlinear terms do not enter the representation
election equation.
Finally, note that our data are limited to
organizing drives that proceed to an
NLRB representation election. We do not
deal with management opposition that
deters unions from a drive or from
carrying a drive to an election, nor do we
deal with management recognition of a
union without a representation election.
The absence of information on these cases
suggests that our estimates will understate
the impact of management opposition on
unionization in the economy as a whole.
Survey Data on
Management Opposition
The first data set that we use to assess
management opposition to union organizing
drives is based on interviews done in
1986 with firms in the Boston and Kansas
City National Labor Relations Board districts
that had elections during 1979 and
the 1980s.3 We contacted 243 firms that
had elections with over 20 employees in
the potential bargaining unit and obtained
on-site interviews with management in
202, for a response rate of 83.1 percent.
One hundred of the 202 firms were in the
Boston region and 102 were in the Kansas
City region; 5% had elections in 1985,
31% in 1984, 12% in 1983, 10% in 1982,
16% in 1981, 16% in 1980, and 10% in
1979. The states covered by the two
districts are generally reflective of the
national labor relations environment,4 and
the win rate of unions in our sample was
similar to the national average: unions
won 39% of the elections in our sample
compared to 38% for all elections conducted
in 1981 in firms with over 20
employees (Medoff 1984). In addition, the
proportion of firms that lost elections and
3 The Boston district encompasses Connecticut,
Maine, Massachusetts, New Hampshire, and Vermont;
the Kansas City district encompasses all or part
of Arkansas, Iowa, Kansas, Missouri, and Nebraska.
4 A composite ranking of private sector union
density in the states in our sample was 29th out of 51
(D.C. included).
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 355
signed collective contracts was close to the
national average: 64% of the elections won
by unions in our sample produced signed
collective contracts, compared with 63% in
the nation (McDonald 1983).
Our second data set is based on an
AFL-CIO Department of Organization
and Field Services survey of the organizers
involved in 274 NLRB election drives.
The sample covers 15 AFL-CIO affiliates
and consists almost entirely of units with at
least 50 eligible voters. The sample is
geographically dispersed, with observations
from all regions of the country.
Both data sets suffer from problems of
missing information for various observations.
To maximize the usable samples, we
assigned the mean value of a variable to
missing independent variables, and added
a dummy variable that took the value one
for that observation and variable (Little
and Rubin 1987). We report results for
samples limited solely by missing observations
on dependent variables or critical
independent variables; our analysis of
samples for which no independent variables
were missing, however, yields comparable
results.
Table 1 summarizes what the two
surveys tell us about the methods management
used to oppose union organizing
drives from 1979 to early 1986. Panel A
shows the frequency of use of consultants
and of filed and upheld unfair labor
practice charges in the employer survey.
Panel B records similar information for
the organizer survey and additional information
on the specific tactics used by the
companies to deter unionization. Because
the organizer survey coding does not
distinguish between a firm's use of lawyers
and its use of consultants in its effort to
defeat a union, the 70% consultants/
lawyers figure reported in the survey
overstates the use of consultants in running
the company's anti-union drive. We
believe the reported use of consultants by
41 % of firms in the employer survey is
more accurate, as it is based on the specific
question: "Was an outside consultant,
beyond normal legal assistance, hired to
assist you during the organizing camTable
1. Percentage of Firms Using Various
Methods to Oppose Union Organizing Drives
in the 1980s.
A. Employer Survey, 1986
Consultants Used (yes): 41%
< month 13%
1 month or more 28%
Unfair Labor Practices: Charges Filed 24%
Guilty 15%
B. Organizer Survey, 1982-83
Consultants/Lawyers Used (yes): 70%
Unfair Labor Practices:
Filed 36%
Discharges or Discriminatory Layoffs 42%
Tactics:
Company Leaflets 80%
-s 4 33%
5 or more 47%
Company Letters 91%
c< 4 41%
5 or more 50%
Captive Audience Speech 91%
62%
5 or more 29%
Supervisory Small Mtgs.
per/employee 92%
s 4 46%
5-8 12%
9 or more 33%
Supervisor Intensity in Opposing Uniona
Low 14%
Moderate 34%
High 51%
Source: Employer Survey: NBER survey of 202
establishments that faced organizing drives between
1979 and 1986. Organizer Survey: AFL-CIO Department
of Organization & Field Services, survey of 274
organizers in 1982-83 representation elections.
a Asked of only half the organizers. The specific
question was, "Describe on a 1 to 5 scale the
frequency, intensity and sophistication of the supervisor's
campaign, with 5 being very intense, etc."
paign?"5 The proportion of campaigns in
which unions charged the firms with unfair
labor practices was 24% in the employer
survey and 36% in the organizer
survey.6 Note, however, that organizers re-
5 The AFL-CIO questionnaire contained information
that differentiates between lawyers and consultants,
but that information was not coded and placed
on the computer file that we were given by the
AFL-CIO.
6The AFL-CIO questionnaire contained one
question on firings and a separate question on "other
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
356 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 2. Employers' Perceptions of the Causes of Organizing Drives and Organizers' Perceptions
of the Most Important Issues in Organizing Drives.
Employer Survey, 1986
Either
Primary Primary or
Cause Secondary Cause
Economic
Wages & Benefits 17% 27%
Job Security 10% 15%
"Voice"
Lack of Fairness 7% 13%
Dissatisfied Employee(s) 14% 18%
Lack of Communication about Company 12% 19%
Union Pressure
Pressure from Other Unions in Establishment or Area 11% 18%
Union Organizer 7% 10%
Organizer Survey, 1982-83
Most Important Issue
Beneficial for Union Issue
Supervisory 52%
Wages 27%
Fringes 6%
Safety 7%
Discrimination 3%
Pressure 5%
Beneficial for
Strikes 36%
Layoffs & Closing 16%
Dues 15%
Wages 2%
Other 31%
Source: See Table 1.
port many more discharges and discriminatory
layoffs than indicated in filed
charges. With respect to specific tactics, the
organizers reported that most firms used a
spectrum of tactics ranging from leaflets to
captive audience speeches to small meetings
between workers and supervisors. Of
the organizers asked to evaluate the intensity
of the supervisors' efforts to oppose
unions, half rated it as high.7
Table 2 provides information on the issues
that managers and the organizers saw
as important in the organizing campaign.8
unfair labor practices" that were filed with the
NLRB. Here, we assume firings were filed charges.
7 This question was added to the survey when half
the responses had already been gathered. It asked
organizers to rate intensity of opposition from 1 to 5.
We treat 5 as high, 3-4 as moderate, and 1-2 as low.
8 The surveys asked for somewhat different
The figures reveal a striking difference between
the perceptions of these groups.
Managers recognized that unfairness to
workers, inadequate communication with
workers, and poor handling of employee
dissatisfaction were important causes of
union campaigns, but regarded purely economic
issues-wages and benefits and job
security-as the key factors. In addition,
some managers saw the primary or secondary
impetus for the union drive as coming
from outside the work group, in the
form of union organizers or pressures from
other unions in the firm or in the local area
trying to unionize their workers.
Organizers, in contrast, saw workerinformation.
The employer survey asked for the
causes of the election drive, whereas the organizer
survey asked for the most important campaign issue.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 357
management relations or "voice" issues as
critical in organizing campaigns: they viewed
supervision as the most important issue in
the drive, with wages and fringes next in
importance, and regarded the risk of strikes,
layoffs, or plant closing and the onus of
union dues as the most important issues that
the company used to discourage employees
from voting for the union. Furthermore, in
responses not summarized in Table 2, organizers
reported that 23% of the drives
were initiated because the organizer or the
international targeted the firm (comparable
in magnitude to the percentage of firms citing
union pressure as a primary cause) but
that 57% began because workers in the firm
called the union.9
Determinants of Opposition
The results of our analysis of the
determinants of management opposition
in the employer survey are given in Table
3. Here we measure opposition with three
variables: 0-1 dummy variables for whether
the firm was charged with committing an
unfair labor practice in the election and
for whether the NLRB upheld the charge,
and a continuous variable for the number
of days (including 0) that the firm used a
management consultant. We used a logistic
form to estimate the equations with the
0-1 dependent variables and used a Tobit
analysis to estimate the equation for
consultant days.
The key independent variable in the analysis
is the likely loss of profits due to unionization,
which we proxy by the difference
between the log of the firm's own compensation
(obtained from our survey) and the
log of hourly wages of unionized production
or nonproduction workers in the onedigit
industry in its Census Region (estimated
from the Current Population Survey
Tapes for 1984), adjusted for the impact of
9 In the employer survey the relevant question
referred to unionized workers in the plant or at the
job site who were pressuring unorganized workers to
join or form a bargaining unit; in the organizer
survey the question referred to a paid official from a
local or national union who targeted and attempted
to organize the establishment.
unions on fringes.'0 To test for the hypothesized
positive relationship between differentials
and managerial opposition, we enter
this compensation differential in the equation
for opposition linearly and as a squared
term. Finally, we include controls for various
other differences between firms, such
as number of employees in the establishment,
whether the unit consists of production
or nonproduction workers, an index of
personnel practices (defines as the sum of
dummy variables for whether a firm had
any of the five practices listed in the table
note), whether the firm is in the Boston or
Kansas City NLRB district, the firm's onedigit
SIC industry, and whether the firm
deals with unions in other locations."1 Because
the employer survey lacks data that
might be used to indicate the innate probability
of a union victory, we are unable to
10 The question in our survey from which we obtained
the wages and benefits the firm paid was, "What
were the a) average wages per hour and b) total compensation
(wages and fringes) of the workers in the
unit which voted in the N.L.R.B. election 3 years prior
to the election, 1 year prior to the election, 6 months
prior to the election, 3 months prior to the election, at
the time of the election, and 1 year after the election,
and what are current wages?" Since managers obtained
this information from firm records, we believe
it is reasonably accurate. We compare it to wages of
unionized production or nonproduction workers in the
relevant one-digit industry in the Census region using
the CPS tapes and unionized fringe benefit data from
the BNA for 1984, as we believe that those wages provide
managers with a plausible indication of how much
unions might raise the wages in their firm. We estimated
variants of the equations in Table 3 by entering
separately the log of the firms' compensation and the
log of our estimated union wage and fringes in the
area, and obtained results comparable to those in the
table.
1 Based on the occupational categories for persons
in the bargaining unit, which were available to
us in the NLRB records, we classified the workers as
production or nonproduction workers. For industry
categories, we used the one-digit SIC categories (for
example, construction, manufacturing, business services).
In order to obtain enough observations from
the CPS for our regions, we had to use broad
occupation and industry categories. Further, the use
of a region variable serves as a control for the
different regional attitudes and cultures that may
affect employer behavior. Finally, our index of
experience with unions had five potential choices: no
experience, little experience, some experience, moderate
experience, and considerable experience.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
358 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 3. Determinants of Management Opposition to Union Organizing Drives:
1986 Employer Survey.
(Standard Errors in Parentheses)
Explanatory Variables Charge of Unfair Guilty of Unfair Consultant
and Summaiy Statisticsa Practice Practice Days Used
Union Compensation 2.26 -1.53 2.51 -2.77 3.24
Difference (1.16) (2.26) (1.30) (2.57) (11.53)
Union Compensation - 5.04 - 8.02 Difference
Squared (3.67) (4.12)
Sample Size 188 188 184 184 184
Technique Logistic Logistic Logistic Logistic Tobit
-2 Log Likelihood 193.0 191.8 146.8 144.0 236.0
a The other controls used in this study were number of employees in election unit; whether unit is
production/nonproduction workers, K.C./Boston, mfg./other; dummies for missing independent variables and
whether firm has organized units elsewhere; and an index of personnel practice (how many of the following
practices the firm uses: formal written grievance procedure, formal written seniority system for promotions,
formal written sick leave policy, written posting of training opportunities, and formal written policy for layoff
and recalls).
examine the impact of that variable in the
analysis.
The estimates in Table 3 show that our
model does a good job of explaining
unfair labor practices but fails to explain
the use of management consultants. Columns
1 and 3 show that the potential
union wage differential has the expected
positive impact on the probability that a
firm will be charged with or be found
guilty of an unfair labor practice. 12 In
columns 2 and 4 the differential is
estimated to have a nonlinear parabolic
form, with the square of the differential
having a positive substantial impact on the
chances of committing an unfair practice
whereas the linear differential has a
negative impact. The insignificant coefficient
on the union differential in the
equation for consultant days in column 5
shows, by contrast, that our model does
not explain the use of consultants. The
implication is that higher wages substitute
for the commission of unfair labor practices
but not for consultant days used.
The evidence on the determinants of
management opposition in the organizer
survey in Table 4 provides stronger
12 This finding is not inconsistent with Kleiner's
(1984) results showing that current profitability is
unrelated to committing an unfair practice. In the
model we investigate, it is the potential effect of
unions on future profits that is the key determinant
of unfair practices, not current profitability.
support for our model. Here we measure
management opposition with charges of
unfair labor practices,'3 as in Table 3, and
with a 0-1 dummy variable for whether
supervisors campaigned against the union
(which was unavailable in our data set for
employers). We do not look at the lawyers/
consultants measure in the organizer
survey, since (as discussed above) it does
not measure the use of consultants in the
election drive. With respect to independent
variables, the organizer survey lacks
information on wages paid but includes
questions about fringe benefits, from
which we have created an index of the
fringe benefits provided by the firm (the
number of fringes as reported in the list
given in the table note). Since unions have
a sizable impact on the provision of fringes
(Freeman 1981), this variable provides
another indicator of the likely impact of
unionization on costs: firms with few
fringes are likely to see sizable increases in
their fringe benefit costs if they become
organized. We also include an index of
good work conditions and supervisory
practices (obtained as minus the sum of
organizers' ratings of conditions/practices
on a 1-5 scale, with the highest ratings
13 When we experimented with a variable for
whether or not a firm fired or discriminatorily laid
off workers in place of the unfair labor charge
variable, we obtained results similar to those in the
table.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 359
Table 4. Determinants of Management Opposition to Union Representation Drives:
1982-83 Organizer Survey.
(Logistic Equation Estimates with Standard Errors in Parentheses)
Charged with Supervisors
Explanatory Variables Unfair Labor Campaigned
and Summary Practices Against Union
Statisticsa (1) (2)
Index of Benefits -0.220 0.410
(0.100) (0.170)
Index of Good Work Conditions -0.140 - 0.140
and Supervisory Practices (0.040) (0.050)
>70% of Workers Signed -0.810 - 1.510
Authorization Cards (0.450) (0.650)
<40% of Workers Signed 0.640 - 0.870
Authorization Cards (0.470) (0.730)
Union Propensity Index 0.370 -0.280
(0.240) (0.500)
(Union Propensity Index)2 -0.009 0.010
(.005) (.010)
Sample Size 232 245
-2 Log Likelihood 264.4 102.2
a The other controls used in t
previous experience with unions.
Index of Benefits sums 0-1 variables for medical insurance, dental insurance, sick pay, pension, grievance
procedure, arbitration, employee handbook, open door policy, and quality of work life.
Index of Good Work Conditions and Supervisory Practices sums answers to: Is supervision unpredictable,
inconsistent or autocratic? Is the work of a routine nature where employees receive little recognition from
supervisors or little personal job satisfaction? Are employees stuck in "dead-end" jobs with little chance for
promotion? Have there been major accidents in the plant, or are there serious threats to the health and safety
of workers? Are severe pressures, including mandatory overtime, placed on employees to maintain or increase
production? Is there evidence of discrimination or favoritism of any kind? Do the employees feel that the
company has little personal concern for them? The variables are coded from 1 to 5, with 5 being the worst
conditions (those best for the union organizing drive). We multiply the sum by minus one for ease of exposition.
Index of Union Propensity sums answers to: Was the workplace subjected to substantial anti-union propaganda
or attitude surveys (or both) before the organizing campaign started? Does the company have a
pre-employment screening process to weed out potential union sympathizers? Do a substantial number of
employees belong to civil rights groups, tenant associations, social advocacy groups, or church organizations
that are active in community affairs? Do organizers have access to employees, either on or off the job? Are labor
unions generally well accepted within the community? Have there been any recent shutdowns of large union
plants or establishments in the community? Have there been bitter, highly publicized strikes in the community
in the past few years? Is employee turnover relatively low?
The variables in the union propensity indices are scaled from 1 to 5, as described in the text, with 5 being the
best from the union's perspective.
indicating worse conditions, using the
specific questions given in the table note).
We view both the fringe and conditions/
practices variables as indicators of "positive
industrial relations" practices.
Finally, the organizer survey contains
two variables that are potentially good
measures of the innate propensity of
workplaces to vote for the union: the
percentage of workers signing authorization
cards, and an index of "union
propensity" created from questions regarding
the attitudes of workers and community
toward unionism and of employer
anti-union activity prior to the election, as
listed in the table note.
In Table 4 we test for the hypothesized
nonlinear impact of the innate probability
of a union victory on management opposition
in two ways. To test for nonlinearity
in the effect of the percentage of workers
signing authorization cards on management
opposition, we created two dummy
variables to reflect the extremes of the
distribution: "percent cards > 70%" takes
the value 1 in cases where more than 70%
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
360 INDUSTRIAL AND LABOR RELATIONS REVIEW
signed cards, and "percent cards < 40%"
takes the value 1 in cases where fewer than
40% signed cards. The hypothesized nonlinearity
should produce negative coefficients
on both dummy variables, distinguishing
the extremes from cases in which 40-
70% of the workers signed cards. To test
for nonlinearity in the effect of our union
propensity index, we enter the variable and
the square of the variable. Here the hypothesized
nonlinearity should yield a negative
squared term and positive linear term.
Three findings are evident in the table.
First, consistent with our model, firms
with fewer fringe benefits and with poor
work conditions and supervisory practices
are more likely to commit unfair practices
than those with more fringes and better
conditions and practices. Second, poor
work conditions and supervisory practices
have a similar impact on supervisors'
campaigning against the union: the worse
those conditions, the more likely the firm
will get supervisors to campaign against
the union. On the other hand, good
benefit packages also increase the probability
that supervisors campaign against
the union, perhaps because the company's
provision of good benefits gives supervisors
a strong argument against the union.
Third, there is some evidence of the
predicted U-shaped relation between our
indicators of the innate probability to
unionize and management opposition. In
column 1 the nonlinearity shows up in the
squared union propensity index, which
takes a parabolic form so that a greater
union propensity has first an increasing
then a decreasing effect on management
opposition. The estimated coefficients for
the percentage who sign cards, by contrast,
show a monotonic negative impact of
card-signing on unfair practices. This
result provides evidence that management
campaigns less virulently when a union
has strong support but not that it also
campaigns less when a union has weak
support. Perhaps management is sufficiently
risk-averse that it spends resources
even when it seems likely to defeat the
union with only limited opposition.
Nonlinearity shows up in a different
way in the column 2 calculations for the
determinants of whether or not supervisors
campaign against the union. Here,
the percentage who sign cards has the
predicted nonlinear effect, with both high
and low percentages reducing the probability
that a company will direct supervisors
to campaign against the union. Supervisors
campaigned against the union in
86% of companies in which more than
70% of the workers signed cards, in 95%
of companies in which 40-70% signed,
and in 84% of companies in which fewer
than 40% signed. By contrast, the union
propensity index has no such nonlinear
effect. Our finding that different indicators
evince nonlinear patterns in the
different equations suggests that the nonlinear
implication of our model should be
viewed with caution. A strong result,
however, is that firms with work forces
having the greatest innate probability to
vote union devote fewer resources to
opposing unionization than other firms.
Determinants of Representation
Election Outcomes
Because management opposition is endogenous
in our model, it is inappropriate
to use ordinary least squares to estimate
how such opposition affects the probability
that a union wins a representation
election. The error term in an equation
for a union victory (our C1) is likely to be
correlated with management opposition,
biasing the coefficient on opposition. To
deal with this problem, we use a limited
information instrumental variables analysis.
Specifically, we estimate a linear
approximation to the election outcome
equation and replace the endogenous
opposition variables by their estimated
values from our model:
(7) W= a + bX + cPr(opp) + U,
where W = 1 if the union wins the
election and 0 if it loses; X stands for
control variables; and Pr(opp) is the predicted
probability of opposition obtained
from the relevant logistic equation for
opposition. The instrumental variables
procedure should make Pr(opp) uncorrelated
with the error term U.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 361
In this analysis we use a linear probability
approximation to estimate the win
equation because linearity allows us to
apply readily the instrumental variables
technique, whereas there is no easy way to
apply the technique when the outcome
function is nonlinear. 14 Since the mean
win rate for unions is about 0.5 in both
our data sets, the linear approximation
should be a good one.
Table 5 presents the results of our
linear probability analysis of the determinants
of a union victory. Columns 1 and 3
record the coefficients and standard errors
for ordinary least squares estimates of
the linear probability equation in our two
data sets. We present the least squares
estimates as a description of the relation
among the variables in the data set and as
a measuring rod for assessing what the
instrumental variables do to the estimates.
Columns 2, 4, and 5 give our instrumental
variable estimates.
Since for the employer survey our
model worked reasonably well in explaining
unfair practices but failed to explain
consultant days used, in column 2 we
replace the variable for being found guilty
of an unfair practice by its predicted value
(from the model estimated in column 4 of
Table 3) but leave the consultant days
used variable as is. Both column 1 and
column 2 show that consultant days used
had a significant negative impact on the
probability of a union win (see Lawler
1984 for a similar result); and both show
that the compensation difference has a
positive effect but with a sizable standard
error. In the ordinary least squares equation
of column 1 the estimated effect of
unfair practice charges on a union win is
positive with a standard error of approxi-
14 This model gives the correct conditional mean
estimates as long as E(oPP I exogenous variables) =
Pr(oPP), where E(oPP I exogenous variables) is the
expected value of opposition conditional on the
exogenous variables; that is, the conditional mean
estimates are correct if we have correctly modeled
the form of the opposition equation. We prefer this
methodology to the maximum likelihood bivariate
logistic approach used by Lawler (1984) because it
does not require correct specification of an entire
system of equations.
mately the same size. In the instrumental
variable estimate of column 2 the estimated
effect remains positive with a larger
standard error.
Three factors stand out in the ordinary
least squares estimates of the probability
of a union victory in our analysis of the
organizer survey in column 3. First is the
massive statistically significant adverse effect
of supervisory opposition on the
probability of a union win: in our analysis
supervisory opposition is the most important
management action to deter unionization.
Second is the sizable impact of the
percentage of workers who sign cards on
the probability of a union win, and the
positive but more modest impact of the
index of union propensity, supporting our
use of these variables as indicators of the
innate probability of a union win. Third is
the positive coefficient on unfair practices,
counter to our model and to rational
behavior. (Why should management commit
unfair practices if doing so increases
the union's chance of winning?)
In the instrumental variable estimates in
column 4 we replace the estimated probability
of the employer committing an
unfair practice with the predicted value of
that probability from the model in column
1 of Table 4 and replace the 0-1 dummy
variable for supervisors campaigning
against the union with its predicted value
from column 2 of Table 4. The results are
striking: the coefficient on unfair practices
changes from positive to negative while
the coefficient on the effect of supervisors'
campaigning against the union increases.'5
Apparently, the endogeneity of management
opposition greatly biased the estimated
ordinary least squares estimates of
15 Even fairly weak results for the impact of unfair
practices on election outcomes are not at variance
with time series or industry/state calculations that
find that the proportion of workers organized
through NLRB elections is greatly reduced by the
frequency of unfair practices, since the pattern
found in those studies largely reflects variation in the
number of NLRB elections rather than in win rates. In
any case, unfair practices in aggregate studies should
be viewed largely as an indicator of management
opposition that has its primary effect on the number
of organizing campaigns.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
362 INDUSTRIAL AND LABOR RELATIONS REVIEW
Table 5. Determinants of Union Victory in NLRB Representation Elections:
Linear Probability Model.
(Standard Errors in Parentheses)
1986 Employer Survey 1982-83 Organizer Survey
Estimating Techniques and Instrumental Instrumental
Summary Statistics OLS Variables OLS Variables
Union Compensation Difference 0.270 0.22
(0.210) (0.240)
Consultant Days Used -0.007 -0.007
(0.003) (0.003)
Unfair Practicesa b 0.12 0.30 0.090 -0.620 -0.530
(0.10) (0.36) (0.070) (0.400) (0.380)
Supervisors' Campaign - 0.370 - 0.780 - 1.050
Against Unionsb (0.120) (0.420) (0.410)
Intensity of Supervisors' Campaignc - - - 0.160
(0.040)
% of Workers Signing 0.830 0.540 0.570
Authorization Cards (0.200) (0.240) (0.230)
Index of Union Propensity 0.012 0.007 0.002
(0.007) (0.007) (0.007)
Other Controls A A B B B
N 184.000 184.000 232.000 232.000 232.000
R 2 0.11 0.10 0.180 0.180 0.250
Note. A: See Table 3 controls. B: See Table 4 controls.
a In columns 1-2, "guilty of unfair labor practices"; in columns 3-5, "charged with unfair labor
(The organizer survey provides no data on whether the charges were upheld.) The instrumental variable
estimate in column 2 is obtained by replacing the unfair practice variable by its predicted value from the logistic
equation in column 4 of Table 3. The instrumental variable estimates in columns 4-5 are obtained by replacing
the unfair practice variable by its predicted value from the logistic equation in column 1 of Table 3.
b The instrumental variable estimates in columns 4-5 are obtained by replacing whether supervisors
campaigned against the union with the predicted value from the logistic equation in column 2 of Table 3.
c See note a to Table 1.
the effect of this opposition on election
outcomes in this data set.
Finally, in column 5 we add a variable
for which the organizer survey has information
in only about half of the cases -an
index of the intensity of supervisory
opposition, scaled from 1 to 5. (The
survey question we used to derive this
index is described at the bottom of Table
1.) With the mean entered for missing
values and a dummy variable to flag those
cases, the intensity of supervisors' opposition
turns out to be a major determinant
of outcomes, reinforcing our conclusion
that supervisors are critical in management
campaigns against unions.
Effect of Organizing Drives on
Management Careers
Because most union studies are concerned
with what unions do for their
members, they rarely investigate how
organizing drives affect management. (See,
however, Clark 1980 for intriguing evidence
on a small sample of firms.) An
analysis of management opposition to
organizing drives cannot, however, neglect
this issue. From the perspective of
our model it is important to delineate the
incentives managers, as opposed to stockholders,
have to oppose unionism. Accordingly,
in our employers survey we asked
management whether plant managers were
promoted, fired, sent for retraining, or
reassigned in the wake of the NLRB
representation campaign. In the larger
project of which this study is a part, we
asked firms to identify their closest competitors
(see Freeman and Kleiner 1990)
and we asked management in 33 of those
firms (none of whom faced an organizing
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 363
drive during the 1979-86 period) about
career outcomes for their managers.
Table 6 tabulates the pattern of management
career outcomes by the organizing
status of the firm. Line 1 shows the
distribution of changes in manager status
for 202 establishments in our sample in
which elections took place. Lines 2 through
4 give the distributions for varying outcomes
of the organizing drive: a union win
and a contract; a union win and no
contract; and a union loss in the election.
Line 5 gives the distribution for the
"control" subsample of nonunion establishments
that did not face an organizing
drive, and line 6 shows the distribution for
the matched pairs of these establishments.
The results show that an organizing
drive and a union victory have a definite
impact on the careers of managers. Eight
percent of managers in the sample of
establishments with organizing drives were
fired, and 10% in the subsample who lost
the election to the union and ended up
with a collective bargaining contract were
fired, compared to 2% in our control
sample. At the other end of the spectrum,
just 3% of managers in the sample facing
an organizing drive and none in the
sample who ended up with a union
contract were promoted, compared to
21 % of managers in our control group.
These findings (consistent with Clark's
earlier results) have two implications for
understanding the management response
to organizing drives. They suggest that
managers have a strong incentive to
engage in labor relations practices that
deter union organizing activity and, faced
with a drive, to fight fiercely to defeat the
union. They also suggest that a union
organizing drive signals stockholders and
top management that the plant management
is poor and should be replaced.
Table 6. Percentage of Firms Changing Manager Status, by Organizing Drive,
1986 Employer Survey.
Manager
Manager Reassigned
Description of No Manager Sent for to Another Manager No
Establishments Change Promoted Retraining Location Fired Other Answer
(1) All establishments
with organizing drives
(Percent) N = 202 76 3 1 3 8 8 1
(2) Establishments
where the union won
and a contract was
reached
(Percent) N = 50 82 0 2 2 10 4 2
(3) Establishments
where the union won
the election, but no
contract was reached
(Percent) N = 28 61 7 0 4 7 18 4
(4) Establishments
where the union lost
the election
(Percent) N= 124 77 3 0 3 7 8 1
(5) Sub-sample of firms
without union and no
election N = 33a 41 21 3 4 2 9 (6)
Paired Sub-sample
of firms [with those in
(5)] with organizing
drives N=33 70 0 0 3 15 6 6.1
Percentages sum to less than 100 due to rounding.
a These values are the means of the percentages of the change in manager status over the same time period
as firms experiencing an organizing drive.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
364 INDUSTRIAL AND LABOR RELATIONS REVIEW
Conclusions
In this study we have examined the
determinants and consequences of employer
opposition to union organizing
drives. Our results reveal some substitution
between "positive industrial relations"-as
indicated by wages, benefits,
and work conditions -and more adversarial
management tactics toward unions.
Firms that had higher wages and better
work conditions and supervisory practices
and more benefits were less likely to
commit unfair labor practices than firms
with lower wages and less favorable conditions.
In our organizer survey data set, we
found that in firms with poor work
conditions and supervisory problems, supervisors
were especially likely to campaign
against the union; that firms with a
large percentage of workers signing authorization
cards were less likely to commit
unfair practices; and that the most effective
"hardnosed" company tactic was to
have supervisors campaign intensely
against the union. In our employer survey
data set we found that companies that
brought in consultants were more likely to
defeat unions than other firms, and that
managers whose establishments faced or
lost organizing drives were more likely
than other managers to suffer setbacks to
their careers (firing, reassignment, retraining,
or failure to be promoted).
We found some support for the nonlinearities
predicted by our model. First, in
our employer survey, the potential union
compensation differential had a parabolicshaped
effect on management opposition
as proxied by unfair labor practices.
Second, in our organizers' survey, management
opposition tended to be stronger
when the likelihood of a union victory was
in serious doubt than when it was either
very high or very low. In that data set,
instrumental variable estimates that take
account of the endogeneity of management
opposition yielded more sensible
estimates of the impact of committing
unfair practices on the union's likelihood
of winning than did single equations
estimates.
We interpret our results as consistent
with the hypothesis that firms behave in a
profit-maximizing manner in opposing an
organizing drive. We also find that management
opposition, reflected particularly
in the actions of supervisors, is a key
component in union inability to organize
workers in the United States.
REFERENCES
Clark, Kim B. 1980. "The Impact of Unionization on
Productivity: A Case Study." Industrial and Labor
Relations Review, Vol. 34, No. 4 (July), pp. 451-68.
Cooke, William N. 1985. "The Failure to Negotiate
First Contracts: Determinants and Policy Implications."
Industrial and Labor Relations Review, Vol. 38,
No. 2 (January), pp. 163-78.
Dickens, William T. 1983. "The Effects of Company
Campaigns on Certification Elections: Law and
Reality Once Again." Industrial and Labor Relations
Review, Vol. 36, No. 4 (July), pp. 560-75.
Dickens, William T., and Jonathan S. Leonard. 1985.
"Accounting for the Decline in Union Membership,
1950-1980." Industrial and Labor Relations
Review, Vol. 38, No. 3 (April), pp. 323-34.
Farber, Henry S. 1987. "The Recent Decline of
Unionization in the United States." Science, Vol.
238, No. 11 (November), pp. 915-20.
Freeman, Richard B. 1981. "The Effect of Trade
Unionism on Fringe Benefits." Industrial and Labor
Relations Review, Vol. 34, No. 4 (July), pp.
489-509.
_ 1986. "The Effect of the Union Wage
Differential on Management Opposition and Union
Organizing Success." American Economic Review,
Vol. 76, No. 2 (May), pp. 92-96.
. 1988. "Contraction and Expansion: The
Divergence of Private Sector and Public Sector
Unionism in the U.S." Journal of Economic Perspectives,
Vol. 2, No. 2 (Spring), pp. 63-88.
Freeman, Richard B., and Morris M. Kleiner. 1990.
"The Impact of New Unionization on Wages and
Working Conditions: A Longitudinal Study of
Establishments." Journal of Labor Economics, Vol. 8,
No. 1 (January), pp. S-8-S-25
__ 1987. "Union Organizing Drive Outcomes
from NLRB Elections During a Period of Economic
Concessions." Proceedings of the 39th Annual
Meeting of the Industrial Relations Research Association
(New Orleans, Dec. 28-30, 1986). Madison, Wis.:
IRRA, pp. 41-47.
Freeman, Richard B., and James Medoff. 1984. What
Do Unions Do? New York: Basic Books.
Kleiner, Morris M. 1984. "Unionism and Employer
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms
EMPLOYER BEHAVIOR 365
Discrimination: Analysis of 8(a)(3) Violations."
Industrial Relations, Vol. 23, No. 2 (Spring), pp.
234-43.
Kochan, Thomas A., Robert B. McKersie, and John
Chalykoff. 1986. "The Effects of Corporate Strategy
and Workplace Innovations on Union Representation."
Industrial and Labor Relations Review,
Vol. 39, No. 4 (July), pp. 487-501.
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.
Little, Roderick J. A., and Donald Rubin. 1987.
Statistical Analysis with Missing Data. New York:
John Wiley & Sons.
McDonald, Charles. 1983. "Memorandum to the
National Organizing Committee." Washington,
D.C.: AFL-CIO.
Medoff, James. 1984. "Study for AFL-CIO on the
Public's Image of Unions." Daily Labor Report,
Bureau of National Affairs, Dec. 24, pp. D1-D23.
Pruitt, Stephen W., John Wei, and Richard White.
1988. "The Impact of Union-Sponsored Boycotts
on the Stock Prices of Target Firms." Journal of
Labor Research, Vol. 9, No. 3 (Summer), pp.
285-90.
Stephan, Paula E., and Bruce E. Kaufman. 1987.
"Factors Leading to a Decline in Union Win Rates:
1973-1981." Proceedings of the 39th Annual Meeting
of the Industrial Relations Research Association (New
Orleans, Dec. 28-30, 1986). Madison, Wis.: IRRA,
pp. 296-305.
This content downloaded from 147.251.185.127 on Mon, 26 Mar 2018 13:45:59 UTC
All use subject to http://about.jstor.org/terms