American Economic Association
Strikes, Wages, and Private Information
Author(s): Sheena McConnell
Source: The American Economic Review, Vol. 79, No. 4 (Sep., 1989), pp. 801-815
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/1827933
Accessed: 29-03-2018 08:55 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
American Economic Association is collaborating with JSTOR to digitize, preserve and extend
access to The American Economic Review
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
Strikes, Wages, and Private Information
By SHEENA MCCONNELL*
Private information models of strikes suggest that the strike is used as an
information-revealing device by the union in the presence of asymmetrical information.
A testable prediction of these models is that there is a negative relationship
between strikes and the unpredicted component of the wage. This paper
shows that in a large sample of U.S. labor contracts the real wage falls by about 3
percent after a strike lasting 100 days.
This paper examines the relationship between
wages and strikes in order to shed
light on the empirical relevance of some
strike models. There are basically only two
major theories to explain why strikes occur.
One argues that strikes are accidents or mistakes
that occur during negotiations. The
only prediction of this theory is that strikes
should be less frequent and shorter when the
cost of the strike is higher. The other suggests
that the strike is used as an information-revealing
device in the presence of
asymmetrical information. The major assumption
in these models is that the union
uses the strike to gain information from the
firm's management on the size of the economic
rent that accrues from the firm's activity.
These models predict that strikes should
be more frequent and longer when the component
of the rent that is unobservable to
the union is lower than anticipated.
A major problem with testing this second
class of models is that the important determining
variable is by definition unobservable
to the union and hence is presumably also
unobservable to the econometrician. These
models also predict, however, that there
should be a negative correlation between the
final negotiated wage and strike activity. This
relationship should map out a negatively
sloped "concession schedule." This is a directly
testable hypothesis which does not
rely on any assumption about the form of
the private information held by the firm.
Moreover, it is unlikely that wages and
strikes would be correlated if strikes were
just mistakes.
The wage-strike relationship is examined
here using a unique data set of U.S. labor
contracts which includes information on both
the negotiated wage and any work stoppages.
It has two major advantages over
previously existing U.S. contract data sets.'
First, it contains a very broad cross section
of contracts, including contracts from both
manufacturing and nonmanufacturing industries.
Second, it uses the most comprehensive
listing of strikes available to identify
whether a strike occurred at a contract nego-
tiation.
In order to estimate the concession schedule,
it is necessary to control for all observable
variables which affect the level of wages
or strike activity. In particular, as the outcome
of previous wage negotiations is an
important source of information to the union,
previous wage settlements in similar firms
are included in the wage equation. The estimated
wage equation differs from those estimated
in previous studies of wage determination
in that the average expected real wage
over the term of the contract is used as the
dependent variable rather than the change in
the nominal wage.
*Department of Economics, London School of Economics,
Houghton Street, London WC2A 2AE, UK.
This research was partially funded by an Olin Fellowship.
I would like to thank David Card, John Abowd,
Janet Neelin, Scott Newlon, and two anonymous referees
for helpful comments. Debbie Nelson provided
valuable research assistance.
'For example, the data used by Wayne Vroman
(1984).
801
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
802 THE AMERICAN ECONOMIC RE VIEW SEPTEMBER 1989
The rest of the paper is organized as follows.
Section I presents a brief discussion of
the theoretical models of strikes. A description
of the data used in the paper is given in
Section II and Section III reports the results
of estimating the real wage model. Section
IV presents evidence on the negative correlation
between negotiated real wages and
strikes. Section V discusses the role of previous
wage settlements in the determination of
the real wage, and Section VI contains some
concluding remarks.
I. Background
The first strand of the theoretical strike
literature, which originated in John Hicks
(1963), suggests that strikes are mistakes or
accidents and does not explain why these
strikes occur. Hicks also suggested that the
union will be more prepared to accept a
lower wage after a long strike than after only
a short strike. John Kennan (1980) and
Melvin Reder and George Neumann (1980)
argue that even if strikes are mistakes they
should occur less often and for a shorter
duration the higher the total cost of the
strike to the union and the firm. While this
theory can be tested empirically,2 the cost of
the strike would presumably matter in almost
any economic theory of strikes.
The second strand of the strike literature
suggests that strikes are a result of imperfect
information on the workers' part as to the
firm's ability to pay higher wages. The absence
of complete information is an essential
feature of the strike model constructed
by Orley Ashenfelter and George Johnson
(1969). They maintained that there are three,
not two, parties involved in labor negotiations:
the firm's management, the union
leaders, and the union rank and file. The
union leaders and the firm's management
share all information relevant to the bargaining,
and so the union leaders know the maximum
wage the firm's management will agree
to. The union rank and file do not share this
information and may have unrealistically
high expectations about the wage the firm is
willing to grant. It is assumed that during a
strike these expectations are revised downward
and the minimum acceptable wage increase
for the rank and file falls along a
negatively sloped concession schedule. The
union leaders agree to take strike action to
avoid the politically damaging consequences
of accepting a wage increase much lower
than that expected by the rank and file. The
firm maximizes its expected profit taking the
union's concession schedule as given. The
equilibrium wage increase and strike length
occurs at the point along the concession
schedule at which there is a tangency with an
iso-profit curve or at a corner solution with
no strike.3 An oft-cited criticism of the
Ashenfelter-Johnson model is that it does
not explain why the wage expectations of the
union rank and file should fall as the strike
progresses.
The recently developed private information
models of strikes4 provide the first fully
consistent models of why strikes occur during
labor negotiations. The important contribution
of these models is that they recognize
that a strike can be used as a screening
device when there is an asymmetry in the
information held by the negotiating parties.
As such, the strike may be an ex ante-efficient
bargaining tool. These models provide
a rationalization for the downward-sloping
concession schedule of Hicks and Ashenfelter
and Johnson.
Suppose that the firm has information on
the size of the rent which is not shared by
the union, and the cost of the strike increases
with the size of the rent. A strike can
then be used as a screening device by the
union to infer information about the size of
the rent in the following way. The union
strikes more often and for longer when the
firm offers a low wage claiming that the
value of the rent is low. There is obviously
2See Sheena McConnell (1988) for evidence that
strikes do vary inversely with the cost of the strike to
the negotiating parties.
The concession schedule must be convex over some
strike lengths for there to be an interior solution.
4See, for example, Peter Crampton (1984), Drew
Fudenberg, David Levine, and Paul Ruud (1985),
Sanford Morton (1983), Beth Hayes (1984), Joseph
Tracy (1987a), and John Kennan (1986).
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WAGES, AND PRIVATEINFORMATION 803
no incentive for the firm to offer a high wage
and pretend that the value of the rent is high
when it is not, therefore there is no need for
the union to penalize the firm for offering a
high wage. On the other hand, if the union
does not penalize the firm for offering a low
wage, the firm is always better off offering a
low wage irrespective of the true value of the
rent. Hence, while these models predict that
strikes should vary inversely with the unobservable
size of the rent, the models also
predict that strikes should be inversely correlated
with the observable negotiated wage. If
there is a continuum of possible states of
nature, the possible wage and strike outcomes
map out a negatively sloped concession
schedule.5 This is a directly testable
prediction of the private information models.
Furthermore, there is no reason to expect
a negative correlation between wages
and strikes in a simple model in which strikes
are mistakes.
There have been some attempts directly to
test the private information models by assuming
a particular form of private information.
The fundamental problem in testing
this type of model is that the important
determining variable is by assumption unobservable.
Any measure of the unobservable
component of the rent must be based on the
assumption that the econometrician can
measure a variable which the union cannot.
Unions are presumably sufficiently sophisticated
to infer the value of the unobservable
variable in the same manner as the researcher.
Tracy (1987a) and Beverly Hirtle
(1985) attempt to proxy the unobservable
component of the rent using the residual
from a CAPM securities-pricing equation.
Although Tracy found some evidence in a
sample of contracts covering the years 1973
to 1977 that the probability of a strike is
positively correlated with the variance of this
residual, Hirtle, using a sample covering a
longer period of time, 1957 to 1980, found
little evidence to support this relationship.
These previous attempts to test the private
information strike models relied on essentially
arbitrary assumptions about the form
of the private information held by the firm.
The advantage of the approach adopted in
this paper is that it relies only on the relationship
between two readily observable
phenomena: wages and strikes. If the negative
relationship between wages and strikes
does not hold, then a prima facie case can
be made against these models without resorting
to any assumptions about the form of
the private information.
It is important to note that it must be the
firm that possesses the private information
for this relationship between wages and
strikes to hold. If the union possesses information
not shared by the firm, and the work
stoppage is used by the firm as a screening
device then there would be a positive correlation
between wages and strikes.
II. The Data
While there are many studies of the wage
determination process under collective bargaining
in Canada, there is a dearth of similar
studies for the United States.6 This is
because in the United States there is no
single source of contract data which includes
data on work stoppages and negotiated
wages. To construct a micro data set for the
United States it is necessary to collect contract
data, strike data, and wage data separately,
and match the three sets of information
together. The data used in this study
consist of 3,001 contracts negotiated between
1970 and 1981. These contracts cover
883 employer-union bargaining pairs in 20
manufacturing and 25 nonmanufacturing in-
dustries.7
The starting point of the data set is a
listing of contracts to which all data on
5See Morton for the derivation of a wage-strike
schedule in a private information model with a continuum
of states of nature. Unlike the concession schedule
in the Ashenfelter and Johnson model, this curve need
not be convex at any point.
6Wayne Vroman, Wallace Hendricks, and Lawrence
Kahn (1985) and John Abowd (1987) used large micro
contract and wage settlement data sets for the United
States.
A detailed description of the construction of the
data set used in this study is given in a data appendix
which is available on request from the author.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
804 THE A MERICA N ECONOMIC RE VIE W SEPTEMBER 1989
strikes and wages are matched.8 This contract
listing included most contracts which
covered 1000 or more workers and included
information on the name of the company or
association, the union(s) involved in the negotiations,
the date the contract was ratified,
the effective date of the contract, and its
expiration date. Unfortunately, it did not
provide any information on strikes at contract
negotiations or details of the negotiated
wage.
McConnell (1988) and Tracy (1987b) used
available sources of strike data to construct
a " master" listing of all strikes that occurred
in the United States during the period between
1970 and 1981 and involved 1000 or
more workers. The three principal sources of
strike data were the Industrial Relations
Facts, a weekly publication of the Bureau of
Labor Statistics (BLS); a tape provided by
the Bureau of National Affairs (BNA); and a
tape from the BLS which for reasons of
confidentiality did not include the name of
the firm or the name of the union. Using this
" master" strike listing, it was possible to
identify whether a strike occurred at the
beginning of a contract and, if so, the length
of the strike. Matching the strikes to the
contracts was aided considerably by the detailed
information on the expiration date of
the last contract and the negotiated date of
the new contract provided by the contract
data. As few of the theoretical models of
strikes discuss why strikes occur while a contract
is in effect (as, for example, in the case
of grievance strikes), I took care to include
only those strikes which occur at contract
negotiations over the terms of the new agreement.9
There were 518 strikes that matched
to a contract in the sample, yielding a strike
rate of 17 percent.
Wage changes for most of the contracts
listed on the situation tape are published by
the BLS in the Current Wage Developments
(CWD). A typical contract specifies an initial
wage increase effective within a few
months from the date the contract is negotiated
and non-contingent deferred wage increases
effective later in the contract. Many
contracts (38 percent in this sample) also
specify wage increases contingent on price
increases using a cost-of-living adjustment
(COLA) formula. The CWD lists not only
initial wage changes but also non-contingent
deferred increases and the realized value of
COLA increases.
While wage changes for most of the contracts
on the contract listing are published in
the CWD, the BLS does not publish data on
wage levels for individual firms. Fortunately,
at least one "base" wage level could be
found for most bargaining pairs from wage
data collected by the BNA'0 or from the
hourly wage in their respective four-digit
SIC industry. The complete history of wage
levels for each bargaining pair was then calculated
by adding (or subtracting) the wage
changes to the base wage level if the change
occurred after (or before) the effective date
of the base wage level.
In most of the existing wage determination
studies the change in the nominal wage
rate between the end of the last contract and
the end of the current contract is compared
with the expected change in the price level
over the contract." However, the wage rate
that should matter to the parties in a contract
negotiation is the expected real wage
level. In this study, the negotiated wage is
calculated as the average,'2 over the length
of the contract, 3 of the real wage levels
expected at the time of the negotiations.
If there is no COLA clause, the expected
real wage rate is just the nominal wage level
deflated by the expected price level. However,
if there is a COLA clause then the
expected nominal wage will depend on the
8The Current Wage Developments Situation Tape is
provided by the Bureau of Labor Statistics.
9Strikes over local issues during general multiplant
negotiations are also excluded.
0 I would like to thank John Abowd for making
these data available to me.
" See, for example, Louis Christofides, Robert
Swidinsky, and David Wilton (1980a), Craig Riddell
(1979), and Wayne Vroman (1984).
12This implicitly assumes that the union has a discount
rate of zero.
13The length of a contract is defined as the number
of months between the effective date and the expiration
date of the contract.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WAGES, AND PRIVATE INFORMATION 805
TABLE 1-FREQUENCIES OF CONTRACTS, STRIKES, AND AVERAGE WAGES BY INDUSTRY
Average Average
Number Number Strike Expected
of of Strike Durationa Real Wage
Industry Contracts Strikes Probability (days) (1967$)
Food and Allied products 284 46 0.16 44.8 3.10
Tobacco 24 3 0.13 26.7 3.34
Textile Products 62 6 0.10 20.3 2.38
Men's and Women's Apparel 150 3 0.02 16.8 2.14
Lumber and Wood Products 60 7 0.12 40.7 3.00
Furniture and Fixtures 20 4 0.20 16.8 2.86
Paper and Allied Products 187 21 0.11 59.9 3.39
Printing and Publishing 62 7 0.11 9.7 3.63
Chemicals 96 17 0.18 52.4 3.57
Petroleum Products 50 11 0.22 51.9 4.39
Rubber and Plastics 36 16 0.44 54.3 3.23
Leather 37 5 0.14 39.0 2.11
Stone, Clay, and Glass 86 10 0.12 51.2 3.39
Primary Metals 201 26 0.13 28.2 4.13
Fabricated Metal Products 112 36 0.32 41.2 3.54
Nonelectrical Machinery 124 55 0.44 43.8 3.61
Electrical Machinery 182 43 0.24 34.9 3.09
Transportation Equipment 242 82 0.34 46.5 3.75
Scientific Instruments,
Photo. Goods, and Watches 32 6 0.19 49.3 2.72
Miscellaneous Manufacturing 26 8 0.31 24.1 2.72
Railways 66 2 0.03 1.0 4.00
Public Transportation 4 0 0 - 3.51
Motor Freight 26 6 0.23 12.3 4.30
Water Transportation 23 5 0.22 6.8 3.21
Airlines 73 9 0.12 24.1 9.81
Communications 136 17 0.13 26.9 3.76
Utilities 233 24 0.10 75.8 4.03
Wholesale Trade-Durable Goods 12 2 0.17 58.5 3.23
Wholesale Trade-Nondurable Goods 24 6 0.25 18.8 3.62
Department Stores 54 2 0.04 4.5 2.18
Auto Dealers and Service Stations 23 6 0.26 45.3 3.57
Apparel Stores 7 0 0.00 - 2.04
Restaurants 32 0 0.00 - 1.91
Retail Trade-Miscellaneousb 13 1 0.08 19.0 2.19
Security and Commodity Brokers 3 0 0.00 - 3.87
Insurance Carriers 17 4 0.24 76.5 2.81
Real Estate 21 4 0.19 31.5 2.63
Lodging 42 3 0.07 20.0 1.79
Personal Servicesc 12 0 0.00 - 1.97
Business Services 27 4 0.15 9.8 2.41
Auto Repair 4 1 0.25 11.0 2.62
Miscellaneous Repair Services 4 1 0.25 11.0 3.66
Entertainment 6 0 0.00 - 2.33
Health Services 56 7 0.13 24.3 2.80
Motion Pictures 10 2 0.20 47.5 4.33
Total 3001 518 0.17 41.4 3.48
Source: As described in text.
aStrike duration is measured conditional on a strike's occurrence.
bPrimarily Drugstores and Liquor Stores.
'Primarily Dry-cleaning and laundry establishments.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
806 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 1989
TABLE 2-FREQUENCIES OF CONTRACTS, STRIKES, AND AVERAGE WAGES BY YEAR
(4) (5)
(1) (2) Average Average
Number Number (3) Strike Expected
of of Strike Durationa Real Wage
Year Contracts Strikes Probability (days) (1967$)
1970 64 22 0.35 24.3 3.15
1971 226 53 0.24 46.8 3.45
1972 181 27 0.15 38.5 3.29
1973 300 42 0.14 33.3 3.63
1974 381 78 0.21 36.3 3.52
1975 248 43 0.17 54.1 3.57
1976 251 55 0.22 40.1 3.51
1977 359 60 0.17 33.0 3.99
1978 209 42 0.20 63.9 3.50
1979 239 40 0.17 42.3 3.24
1980 307 25 0.08 35.3 3.33
1981 236 31 0.13 36.1 3.05
Source: As described in text.
a Strike duration measured conditional on a strike's occurring.
COLA rule and expected prices. As the CWD
rarely publishes exact COLA formulas, a
COLA rule was estimated using the change
in the price level and the subsequent actual
COLA payment. To find the expected COLA
payments, the expected change in the price
index was substituted for the actual change
in the price index in the estimated COLA
rule."4
Table 1 shows the frequency and duration
of strikes together with the average expected
real wage rate by industry. Strikes are most
frequent in durable-goods manufacturing industries,
where a strike occurs during 25
percent of contract negotiations, and are least
frequent in nonmanufacturing industries,
where strikes occur during only 11 percent
of negotiations. The average strike length is
41 days, the median strike length is 22 days,
and the modal strike length is only one day.
The longest strikes are in the nondurable
goods manufacturing industries, and strikes
are the shortest, in addition to being the
least frequent, in nonmanufacturing indus-
tries.
Table 2 gives strike frequencies, average
strike duration, and average wages by year.
There is a general decline in the frequency of
strikes over the period considered, with strike
frequency peaking in 1970 at 35 percent and
reaching a low of 8 percent in 1980. Strike
duration, however, peaks in 1978 and is lowest
in 1970, when strikes were most frequent.
The correlation between average strike duration
and average strike probability by industry
or by year is not significantly different
from zero. Furthermore, there is no significant
correlation between either strike frequency
or strike duration and the negotiated
wage by two-digit SIC industry or by year.
III. Estimates of the Real Wage Equation
According to the private information theory
of strikes, a work stoppage is used to
screen low rent firms from high rent firms,
and, hence, there should be a negative correlation
between wages and strikes. The absence
in this sample of a correlation across
industries or time periods between strikes
and wages should not be interpreted as refuting
the private information models. While
the private information models stress that a
variable unobservable to the union is important
in determining the frequency and duration
of strikes, they do not suggest that
14Expected future prices were estimated on the assumptions
that the price level can be explained by
twelve lags plus monthly dummies, this relationship is
used to forecast future prices, and expectations are
formed at the time the contract is negotiated.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL:STRIKES, WAGES,ANDPRIVATEINFORMATION 807
variables observable to both parties are
unimportant. Indeed, in common with the
"total cost" model of strikes, the private
information models predict that strike activity
should be inversely related to factors
which affect the cost of a strike.
The observable variables which influence
strike activity determine the position of the
wage-strike concession schedule but not the
equilibrium point on the concession schedule.
It is very important when attempting to
test the private information models to control
for all other variables which are observable
by both parties. If these are not adequately
controlled for, there is a danger that
we may be estimating a curve mapped out
by shifts in the concession schedule rather
than movements along it. For example,
higher unemployment may shift the concession
schedule downward because workers are
less likely to strike when there are fewer
opportunities for alternative work, and the
firm may be able to hire workers at a lower
wage. The resulting spurious positive correlation
between wages and strikes has nothing
to do with the existence of private informa-
tion.
For expository purposes we can think of
dividing the wage rate into two components.
The first component is a function of all the
variables which determine wages or strikes
and which are observable to both parties.
The second component of the wage is a
function of only that part of the rent that is
unobservable to the union. If strikes are to
be used as a screening mechanism they need
only be correlated with this second component
of the wage. In an empirical context, we
can think of the first component of the wage
as the predicted wage in a wage determination
equation and the second component of
the wage as the residual of the equation. The
private information models predict a negative
correlation between strikes and the
residual from the wage equation and not
between strikes and the predicted wage.
The following wage determination equation
is estimated:
log(wt) = ai + f(lt + e2t2
+ ,BXjt + ei,
where wit is the average expected real wage
negotiated by bargaining pair i for a contract
negotiated at time t, ai is a bargaining
pair-specific effect, t is a time trend, Xi, is a
column vector of independent variables, /B is
the row vector of coefficients on the independent
variables, and eit is an independently
and identically distributed error term. Table
3 presents estimates of this equation.15
Columns 3, 4, and 5 of Table 3 show the
results of estimating the model including a
fixed effect for each bargaining pair to control
for any specific company or union effects
which we do not observe. For comparison
with previous studies I report the results
of regressions with no fixed effects and with
only industry dummies in columns 1 and 2
of Table 3, respectively.'6 Nearly all the
variation in the negotiated real wage can be
explained by the bargaining pair effects and
the time trend. Inclusion of the bargaining
pair effects increases the total sum of squares
of the real wage explained by the model
from 20 percent to 95 percent. Column 6
presents the estimates of a model including
both fixed effects for each bargaining pair
and a dummy for each year.17
The model also includes as independent
variables the national unemployment rate
and the unemployment rate by state as measures
of the state of the labor market, the
annual rate of employment growth by industry
as a measure of the state of the product
market, and the average real wage rate over
all industries as a measure of the alternative
wage available to workers. There is some
evidence to suggest that strike activity decreased
during the Nixon Wage and Price
Controls and so dummy variables are included
for both the Nixon Controls and for
the existence of Carter's Council of Wage
and Price Stability (COWPS).
15The means of the explanatory variables are presented
in Appendix Table 1 for reference.
16Christofides, Swidinsky, and Wilton (1980a), Vroman
(1984), and Riddell (1979) use the differenced
nominal wage rate but do not difference the independent
variables.
17While there are 3,001 contracts in the complete
data set, 1,015 contracts are lost due to the absence of a
preceding contract in the sample. This leaves 1986 observations
for the estimation.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
808 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 1989
TABLE 3-OLS ESTIMATES OF THE REAL WAGE EQUATION,
DEPENDENT VARIABLE: LOGARITHM OF THE EXPECTED REAL WAGE RATE
(6)
Bargaining
(1) (2) (3) (4) (5) Pair Effects
No Fixed Industry Bargaining and Year
Effects Fixed Effects Pair Effects Effects
Included Included Included Included
Time Trend (x 100) 0.02 0.56 0.79 0.44 0.69
(0.26)a (0.17) (0.08) (0.12) (0.10)
Time Trend sqd (x 100) - 0.002 - 0.004 - 0.005 - 0.003 - 0.004 (0.001)
(0.001) (0.001) (0.001) (0.001)
(Log) of Average Price Level - 0.52 - 0.02 - 0.13 - 0.02 - 0.07 - 0.03
Over Previous Contract (0.42.) (0.23) (0.16) (0.17) (0.18) (0.18)
(Log) of Average Expected 1.17 0.23 0.35 0.19 0.15 0.03
Price Level Over Previous (0.37) (0.25) (0.13) (0.15) (0.15) (0.16)
Contract
Average Wage- 0.45 0.70 - 1.06 0.34 - 0.05
All Industries (Log)b (0.55) (0.38) (0.27) (0.22) (0.35)
National Unemployment - - - 0.01 - 0.01
Rate (Percent)c (0.01) (0.01)
State Unemployment -1.19 - 0.66 - - -0.14 Rate
(Percent) (X 100)d (0.47) (0.32) (0.24)
Employment Growth -0.13 0.01 - - 0.09 0.07
by Industry (Percent) ( x 100)' (0.13) (0.10) (0.06) (0.06)
Dummy =1 If Contract -0.16 - 0.02 - 0.01 - 0.01 - 0.01 - 0.01
Negotiated at Scheduled (0.03) (0.02) (0.01) (0.01) (0.01) (0.01)
Reopening
Dummy = 1 If Negotiated - 0.09 0.04 0.09 0.06 0.07 0.06
During 1st Stage of Nixon (0.09) (0.06) (0.03) (0.03) (0.03) (0.05)
Wage and Price Controlsf
Dummy =1 If Negotiated -0.06 0.02 0.07 0.06 0.06 0.06
During 2nd Stage of Nixon (0.03) (0.02) (0.01) (0.01) (0.01) (0.01)
Wage and Price Controlsg
Dummy = 1 If Negotiated - 0.06 - 0.06 0.07 0.06 0.06 0.06
During 1st Stage of (0.02) (0.02) (0.01) (0.01) (0.01) (0.01)
Carter's COWPSh
Dummy = 1 If Negotiated 0.01 0.01 --0.01 - 0.00 -0.00 0.01
During 2nd Stage of (0.03) (0.02) (0.01) (0.01) (0.01) (0.02)
Carter's COWPS'
Duration of Contract - 0.25 -0.10 - 0.13 - 0.13 -0.13 - 0.13
(Months) (x 100) (0.09) (0.06) (0.04) (0.04) (0.04) (0.04)
Dummy =1 If COLA 0.12 -0.01 0.03 0.03 0.03 0.03
Clause Present (0.01) (0.01) (0.01) (0.01) (0.01) (0.01)
Dummy for Type of
Bargaining StructureJ
1. Single Firm- 0.05 0.02
Multiple Plant (0.01) (0.01)
2. Industry Association -0.15 -0.01
(0.17) (0.03)
3. Multiple Union 0.08 0.02 - - - (0.05)
(0.01)
R-Squared 0.20 0.65 0.95 0.95 0.95 0.95
a Standard errors are in parentheses.
bThe hourly wage rate for producti
Earnings.
cThe monthly unemployment rate of the civilian labor force over 16-years-old. Source: Monthlv Labor Review.
dThe annual unemployment rate by state. Source: Geographic Profile of Employment and Unemployment.
'The growth rate of employment over the previous year by the two-digit SIC industry. Source: Emplovment an
Earnings.
fPolicy in effect August 1971 to December 1972.
'Policy in effect January 1973 to April 1974.
hPolicy in effect November 1978 to December 1979.
'Policy in effect January 1980 to August 1980.
JThe reference bargaining structure is a single plant company and a single union.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WAGES, AND PRIVATE INFORMA TION 809
As contracts are rarely fully indexed, unexpected
inflation over the previous contract
may affect wage negotiations. To allow for
this "price catch-up" effect, the average and
expected price level over the previous contract
are included in the model. Neither price
term is precisely estimated suggesting that
past uncompensated changes in the price
level have an insignificant impact on the
wage determination process.18
Some wage negotiations take place at a
scheduled reopening of a contract at which
only wages and salaries are discussed. The
incidence of strikes at contract reopenings is
only 5 percent, significantly lower than the
average for the whole sample, but as shown
in Table 3, the wage rate negotiated at a
contract reopening is not significantly different
from a wage negotiated at a contract
expiration. Both shorter contracts and the
inclusion of a COLA clause are associated
with higher negotiated wages. Columns 1
and 2 of Table 3 show that real wages are
highest for firms bargaining with more than
one union and lowest for an industry association
bargaining with a single union.
IV. Estimates of the Effect of Strike Activity
on the Real Wage
There is no clear consensus in the wage
determination literature on the relationship
between strikes and wages in either the
United States or Canada. Henry Farber
(1978), in the first empirical study of wages
and strikes using U.S. micro-contract data,
estimated the Ashenfelter and Johnson model
for ten firms and found that although the
rate of concession varied from firm to firm,
it was always positive. Fudenberg, Levine,
and Ruud (1985), using a similar small data
set of 159 contracts for 15 firms, estimated a
system of wage and strike equations suggested
by a private information model of
strikes. The level of sales was used as a
measure of the variable unobservable to the
union. Their estimated rate of concession is
positive and about 6.9 percent per year.
This negative correlation between wages
and strikes does not hold up when tested
using larger micro data sets. Riddell (1980),
using a data set of 2,360 Canadian labour
contracts from the period covering 1953
through 1973, found that the occurrence of a
strike had a positive effect on the nominal
wage change. On the other hand, David Card
(1988) found no effect of strike activity on
the real wage rate for a data set of 2,258
Canadian labour contracts from the years
1964 through 1985. Similarly, Vroman found
neither the occurrence of a strike nor the
length of a strike had any significant effect
on the nominal wage in the United States.
Estimates of the effect of the probability
and duration of a strike on the negotiated
real wage rate are presented in Table 4. The
models estimated in columns 1, 2, 3-4, and 5
of Table 4 differ from the models estimated
in columns 1, 2, 5, and 6, respectively, of
Table 3 only by the addition of strike variables.
The measure of strike length included
in the model is the unconditional duration of
the strike. This measure counts the absence
of a strike as a zero-length strike.
Column 1 of Table 4 shows the estimates
of a real wage equation which does not include
any bargaining pair or year-fixed effects
for comparison with the Vroman and
Riddell studies. As Vroman found with U.S.
contract data, there is no significant relationship
between strikes and wages. When industry-fixed
effects are included in column 2
there is actually a significant positive relationship
between strike duration and wages.
However, as shown in columns 4 and 5, in
this sample there is a significant negative
relationship between unconditional strike
duration and the negotiated real wage when
bargaining-pair effects are included. The real
wage falls by about 0.03 percent for each
day of the strike. The occurrence of a strike
is also negatively related to the real wage,
although the coefficient on the strike dummy
in column 3 is imprecisely estimated. The
fact that the estimated coefficients of the
strike variables change from positive to negative
in sign when bargaining-pair effects are
added suggests that there is some omitted
variable bias when only industry-fixed effects
are included. The addition of year effects to
18For a fuller discussion of price catchup see
Christofides, Swidinsky, and Wilton (1980a).
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
810 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 1989
TABLE 4-ESTIMATES OF THE EFFECT OF STRIKE ACTIVITY ON THE REAL WAGE RATE,
DEPENDENT VARIABLE: LOGARITHM OF THE EXPECTED REAL WAGE RATEa
(5)
(1) (2) (3) (4) ~~~Bargaining(1) (2) (3) (4) Pair Effects
No Fixed Industry Bargaining and Year
Effects Fixed Effects Pair Effects Effects
Included Included Included Included
Dummy =1 If Strike 0.38 -0.01 -0.82 0.21 0.28
Occurred (X 100) (2.07)b (0.01) (0.73) (0.89) (0.87)
Unconditional Strike 0.04 3.82 - - 2.35 - 3.12
Duration (Days)C (0.03) (1.96) (1.21) (1.20)
(X104)
R-Squared 0.20 0.65 0.95 0.95 0.95
a Regressions also include all independent
of Table 3.
bStandard errors are in parentheses.
'If no strike occurred, unconditional strike duration equals zero.
3.50 -
3.45-
3.40 -
3.35,,
3.30x
3.25-
3 3.20*
3.15-
xC
WU 3.10- _ I
0 50 100 150 200 250 300
Strike duration in days
FIGURE 1. THE ESTIMATED CONCESSION SCHEDULE
the model in column 6 has little impact on
the coefficients, but the model is estimated
more precisely.19
Figure 1 illustrates the concession schedule
estimated in column 4 of Table 4. After
the average strike length of 41 days the
negotiated wage is only about 1 percent, or
about 3.5 1967 cents, lower than it would
have been in the absence of a strike. The
annual rate of concession is approximately
8.6 percent.
If strike duration is included in the model
with a strike dummy, as in columns 4-5, the
effect of the strike dummy becomes much
smaller and insignificant. Thus, as predicted
by the private information models, in these
data the absence of a strike is equivalent to a
strike of zero days in duration.
V. The Effect of Previous Wage
Negotiations on the Real Wage
It has long been recognized that the outcomes
of wage negotiations are highly correlated
with previous wage settlements at other
firms.20 Failure to control for the effect of
prior wage settlements, or "wage spillover,"
will at best lead to inefficient estimates of
the predicted wage and, if strikes are correlated
with previous wage settlements, could
lead to biased estimates of the relationship
between strikes and real wages.
There are two principal rationales for including
previous wage settlements in a wage
9The model was also estimated using the complete
sample of 3,001 contracts excluding the price catch-up
terms from the model. The concession rate fell slightly
to 0.02 percent per day, but was still significantly different
from zero.
20See John Burton and John Addison (1977) for a
review of the empirical studies of the correlation between
wage settlements.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WA GES, A ND PRIVATE INFORMATION 811
determination equation. First, there may be
a variable omitted from the regression which
is also correlated with wages at other firms.
This may be a variable that is not observable
to the union or the econometrician. If the
value of this unobservable variable is correlated
across firms, the results of other wage
negotiations will provide valuable information
to unions at firms in the same industry.
The second rationale is that wage settlements
at other firms may enter directly into
wage negotiations. The reservation wage is a
determinant of the negotiated wage in nearly
all theories of wage determination and is
itself a function of wages offered at other
firms. Moreover, if turnover costs are high, a
profit-maximizing firm will set an optimal
wage differential between its wage and the
wage set by other firms competing for the
same workers.2' Also, for institutional or political
reasons, workers may care about the
level of their wage rates relative to those
paid at other firms.
In this study five different categories are
used to define the set of reference wages. In
the first two categories it is assumed that the
reference wages are those negotiated in the
same two- and three-digit SIC industry, respectively.
The third category assumes that
the important reference wages are those in
the same broad regional area irrespective of
the industry.22 The fourth category is defined
as the set of previous wage settlements
in the same two-digit SIC industry and region.
Finally, I have also defined a set of
reference wages consisting of previous wage
settlements by the same union regardless of
the industry in which they occurred.
All previous wage settlements at other
firms which occurred after the ratification of
the previous contract between the workers
and the firm are included in the reference
wage set. It is unlikely that any wages negotiated
prior to this date will enter directly
into the bargaining process, and, because all
information is revealed at the end of negotiations,23
no new information can be gained
from settlements at other firms prior to the
negotiated date of the last contract.
The simplest way to model the effect of
previous wage negotiations would be to take
a straight average of all the wages in the
reference wage set. This would give the same
weight to those wages negotiated one month
ago as to those wages negotiated three
years ago. To avoid this, following Louis
Christofides, Robert Swidinsky, and David
Wilton (1980b), a weighted average of the
reference wages is used where the weight
given the j th reference wage, Vj is the following
quadratic function:
(2) Vj = 0 + ?O1mj + 02mj,
where m is the number of months between
the date the reference wage was negotiated
and the date the current contract was negotiated.24
Constraining the sum of the weights
for each observation to equal one, E vj = 1, it
can easily be shown that
(3) vjr log( wj) =Elog( wj)/J
j j)log(W
?+- 2m)log(w1)
i
where Wj is the wage for the jth contract
settlement in the reference set, J is the total
number of settlements in the reference set,
mj = Emj/J and m= im/J. Thus the real
21See, for example, John Dunlop (1944).
22The ten regions are New England, Mid-Atlantic,
East-North Central, West-North Central, South Atlantic,
East-South Central, West-South Central, Mountain,
Pacific, and a category which includes bargaining
groups from more than one region.
Although there may be a pooling equilibrium in
which the wage is independent of the size of the rent. In
these cases, the negotiated wage will contain no information
on the size of the rent.
24This is just a Shirley Almon (1965) lag of the
reference wages.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
812 THE AMERICAN ECONOMIC RE VIEW SEPTEMBER 1989
TABLE 5-ESTIMATES OF THE EFFECT OF STRIKE ACTIVITY AND PREVIOUS WAGE SETrLEMENTS ON THE
REAL WAGE RATE, DEPENDENT VARIABLE: LOGARITHM OF THE AVERAGE EXPECTED
REAL WAGE MODEL INCLUDES BARGAINING PAIR SPECIFIC-FIXED EFFECTSa
Model: log(wit) = a i J8Xi,+ + aog(wi,)/J+ a01X(ni1j-mj)log(wj)+ a02 (mj - i )log(w.)? e,
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Previous Wage Two-digit SIC Three-digit SIC Region 2-Digit SIC Industry Union
Settlements by: Industry Industry and Region
d 0.17 0.18 0.17 0.17 0.08 0.10 0.10 0.12 0.08 0.08
(0.05)b (0.03) (0.03) (0.03) (0.06) (0.09) (0.03) (0.03) (0.04) (0.04)
01( X 100) -0.39 -0.42 -0.44 -0.55 0.43 0.41 0.90 -0.78 0.80 0.95
(0.51)c (0.48) (0.79) (0.77) (0.41) (0.38) (1.41) (1.27) (0.95) (0.94)
02( X 100) 0.01 0.01 0.02 0.02 -0.02 -0.02 0.03 0.02 -0.02 -0.02
(0.01)c (0.01) (0.02) (0.02) (0.02) (0.02) (0.03) (0.03) (0.02) (0.02)
Dummy = 1 If Strike -1.07 0.30 -1.16 0.26 -1.15 0.18 -1.08 0.46 -1.18 0.13
Occurs (X 100) (0.72) (0.86) (0.71) (0.86) (0.72) (0.87) (0.72) (0.87) (0.72) (0.87)
Unconditional - -3.38 - -3.47 - -3.23 - -3.50 - -3.15
Strike Duration (1.20) (1.19) (1.20) (1.21) (1.20)
(Days)( X 104)
Average Value of J 68 68 6 6 26 26 28 28 10 10
R-Squared 0.96 0.96 0.95 0.95 0.95 0.95 0.95 0.96 0.96 0.96
aRegressions also include all independent variables included in column 4 of Table 4.
bStandard errors are in parentheses.
cLarge sample standard errors estimated from the approximation: V(0 ) = [1/c9]2V(iT) + T-/d ]2V( a - 2[1/8 ]
[/8 ]cov(t, d), where T = d 0.
dJ is the number of previous contracts included in the respective category.
wage model reduces to
(4) log(wi,-) =ai+?fX,+?Elog(wj)/J
+ T1E(m1-rj)log(wj)
+ 2 m2 - j2)log(Wj)+et+ T2E (Mi - i j5>lgw) ? eit,
i
where T1 = 618 and T2 = 628The
results of estimating equation (4) are
shown in Table 5.25 The inclusion of wage
settlements for previous contracts improves
the fit of the model but has little effect on
the estimates of the other coefficients. The
estimates of the coefficient 8 show that the
average of the previous wage settlements has
a positive and sometimes fairly large effect
on the negotiated wage. The largest effects
are from other wage settlements negotiated
in the same industry. The elasticity of the
real wage with respect to prior wage settlements
in the same two-digit industry and
three-digit industry is 0.17. The effect of
previous wage settlements in the same geographical
region is insignificantly different
from zero. Interestingly, the reference wages
defined by two-digit industry and region have
a significant effect on the negotiated wage,
but it is smaller than the effect of reference
wages defined by two-digit industry alone.26
There is a small positive correlation between
the outcomes of successive wage negotiations
by the same union.
The odd-numbered columns of Table 5
show that after controlling for the reference
wages, the decline in the real wage associ-25Estimates of the coefficients 01 and 02 were found
by dividing the estimates of Tr and T2, respectively, by
the estimate of 8. As the estimates of 01 and 02 are
nonlinear functions of the OLS estimates, they are not
unbiased in small samples but are still asymptotically
unbiased. The large sample variances of these estimates
were approximated by the delta method. See Jan
Kmenta (1971, 442-44).
26fThe importance of reference wages by industry but
not by geographic region was also found by Christofides,
Swidinsky, and Wilton (1980b) in their study of wage
spillovers in Canada.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WA GES, AND PRI VA TE INFORMA TION 813
TABLE 6-THE EFFECT OF STRIKF ACTIVITY ON THE REAL WAGE BY
BROAD INDUSTRY GROUPING, DEPENDENT VARIABLE:
LOGARITHM OF THE AVERAGE EXPECTED REAL WAGE
Model Includes Bargaining Pair Specific-Fixed Effects and
Previous Wage Settlements by Two-Digit SIC industrya (1) (2)
Dummy = 1 If Strike Occurs (x 100) - 0.66 0.26
Manufacturing-Durable-Good Industries (1.06)b (1.27)
Dummy = I If Strike Occurs (x 100) - 0.57 2.05
Manufacturing-Nondurable Good Industries (1.22) (1.52)
Dummy = 1 If Strike Occurs (x 100) - 3.33 - 2.81
Nonmanufacturing Industries (1.57) (1.91)
Unconditional Strike Duration (Days) (X 104) - -2.08
Manufacturing-Durable-Good Industries (1.52)
Unconditional Strike Duration (Days)( x 104) - -6.66
Manufacturing-Nondurable Good Industries (2.41)
Unconditional Strike Duration (Days)( X 104) - -1.54
Nonmanufacturing Industries (3.17)
R-Squared 0.96 0.96
aRegressions also inclu
bStandard errors are in
ated with the occurrence of a strike increases
for each category of reference wages, yet it is
still not significantly different from zero.27
Similarly, comparing the even-numbered columns
of Table 5 to column 4 of Table 4
shows that there is a small increase in the
rate of concession when previous wage settlements
from each of the five categories are
included. The rate of concession is largest
3.5 percent per 100 days of strike-when
previous wage settlements by industry and
region are included in the model. Although
there is some evidence of omitted variable
bias when previous wage settlements are not
included, this bias is not large enough to
change the qualitative nature of the results.28
The negative coefficient on unconditional
strike duration implies that strikes occur
more frequently and last longer when the
wage offered by a firm is lower than the
"going" rate for that industry and region.
The strike ensures that the management is
not bluffing when it claims that there are
firm-specific problems which prevent it from
offering a higher wage.
Table 6 provides a breakdown of the relationship
between strikes and wages by broad
industry category. The occurrence of a strike
and the length of a strike have a negative
effect on the real wage in each of the industry
groups. In nonmanufacturing industries
it is the occurrence of a strike rather than the
length of the strike that matters. A wage
negotiated after a strike of any length is
about 3 percent lower than a wage negotiated
without a strike. The relative unimportance
of the effect of the duration of a strike
may be because in nonmanufacturing industries
there is less variation in the length of
the strike. The effect of strike duration on
the real wage is largest in nondurable-goods
manufacturing industries. The rate of concession
is twice as fast in these industries as
it is for all industries together. Neither the
occurrence of a strike nor strike length is
significantly correlated with the real wage in
durable-goods manufacturing industries.
27The inclusion of the strike variables has no effect
on the coefficients of the reference wage variables.
28The bias may be the result of omitting a suitable
measure of the workers' alternative wage from the wage
determination equation. Strike models suggest that, since
the cost of a strike to the workers falls with a rise in the
alternative wage, there should be a positive correlation
between strikes and the alternative wage. If reference
wages are a good proxy for the alternative wage, omitting
them from the wage determination equation could
impart the positive bias to the coefficient on the strike
variable that was found.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
814 THE AMERICAN ECONOMIC REVIEW SEPTEMBER 1989
TABLE Al-MEANS OF VARIABLES USED
IN THE REGRESSIONS
Average Expected Real Wage $3.48
Average Price Level Over the Past
Contract (1967 = 100) 167
Average Expected Price Level Over the
Past Contract (1967 = 100) 166
Average Real Wage for All Industries $2.84
National Unemployment Rate 6.5 Percent
Unemployment Rate by State 6.7 Percent
VI. Conclusion
The major finding of this paper is that
both the occurrence of a strike and the length
of a strike are negatively correlated with the
unpredicted component of the real wage. A
strike of 100 days is associated with a decrease
in the real wage of about 3 percent.
This result is robust to the inclusion in the
model of wages previously negotiated at
other firms. This finding provides support
for the theory that strikes are used as information
revealing tools rather than occurring
as a result of mistakes during negotiations.
REFERENCES
Abowd, John M., "Collective Bargaining and
the Division of the Enterprise," NBER
Working Paper No. 2137, January 1987.
Almon, Shirley, " The Distributed Lag between
Capital Appropriations and Net
Expenditures," Econometrica, January
1965, 33, 178-96.
Ashenfelter, Orley and Johnson, George E.,
" Bargaining Theory, Trade Unions, and
Industrial Strike Activity," American Economic
Review, March 1969, 59, 35-49.
Burton, John and Addison, John T., "The Institutionalist
Analysis of Wage Inflation: A
Critical Appraisal," Research in Labor
Economics, 1977, 1, 333-76.
Card, David, "Strikes and Wages: A Test
of a Signalling Model," unpublished research
paper, Princeton University, February
1988.
Christofides, Louis N., Swidinsky, Robert and
Wilton, David A., (1980a) "A Microeconometric
Analysis of the Canadian Wage
Determination Process," Economica, May
1980, 47, 165-78.
_ and , (1980b) "A
Microeconometric Analysis of Spillovers
within the Canadian Wage Determination
Process," Review of Economics and Statistics,
May 1980, 62, 213-21.
Crampton, Peter C., "Bargaining with Incomplete
Information: An Infinite Horizon
Model with Two-Sided Uncertainty," Review
of Economic Studies, October 1984,
51, 579-93.
Dunlop, John T., Wage Determination under
Trade Unions, New York: Augustus M.
Kelley, 1944.
Farber, Henry S., "Bargaining Theory, Wage
Outcomes, and the Occurrence of Strikes:
An Econometric Analysis," American Economic
Review, June 1978, 68, 262-71.
Fudenberg, Drew, Levine, David and Ruud,
Paul A., "Strike Activity and Wage Settlements,"
unpublished research paper, University
of California at Berkeley, May
1985.
Hayes, Beth, "Unions and Strikes with Asymmetric
Information," Journal of Labor
Economics, January 1984, 2, 57-83.
Hendricks, Wallace E. and Kahn, Lawrence, M.,
"Wage Indexation in the United States,
Cola or Uncola?," Cambridge, MA: Ballinger,
1985.
Hicks, John R., The Theory of Wages, New
York: Macmillan, 2nd ed., 1963.
Hirtle, Beverly J., "Interfirm Information,
Strikes, and Union Bargaining Behavior,"
unpublished research paper, MIT, October
1985.
Kennan, John, "Pareto Optimality and the
Economics of Strike Duration," Journal of
Labor Research, Spring 1980, 1, 77-94.
, "The Economics of Strikes," in Orley
Ashenfelter and Richard Layard, eds.,
Handbook of Labor Economics, Amsterdam:
North-Holland, 1986, ch. 19.
Kmenta, Jan, Elements of Econometrics, New
York: Macmillan, 1971, 442-44.
McConnell, Sheena, "An Empirical Study of
Strike Activity in the United States," unpublished
doctoral dissertation, Princeton
University, January 1988.
Morton, Sanford, "The Optimality of Strikes
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms
VOL. 79 NO. 4 MCCONNELL: STRIKES, WAGES, AND PRIVATE INFORMATION 815
in Labor Negotiations," Discussion Paper
No. 83/7, Tulane University, April 1983.
Reder, Melvin N. and Neumann, George R.,
"Conflict and Contract: The Case of
Strikes," Journal of Political Economy, October
1980, 88, 867-86.
Riddell, W. Craig, "The Empirical Foundations
of the Phillips Curve: Evidence from
Canadian Wage Contract Data," Econometrica,
January 1979, 47, 1-24.
_ "The Effects of Strikes and Strike
Length on Negotiated Wage Settlements,"
University of British Columbia Discussion
Paper No. 80-2, January 1980.
Tracy, Joseph S., "An Investigation into the
Determinants of U.S. Strike Activity,"
American Economic Review, June 1986, 76,
423-36.
, (1987a) "An Empirical Test of
an Asymmetric Information Model of
Strikes," Journal of Labor Economics, April
1987, 5, 149-73.
, (1987b) "Testing a Strategic Model
of Bargaining Using Stock Market Data,"
unpublished research paper, Yale University,
May 1987.
Vroman, Wayne, "Wage Contract Settlements
in U.S. Manufacturing," Review of Economics
and Statistics, November 1984, 66,
661-65.
U.S. Department of Labor, Bureau of Labor
Statistics, Washington: Current Wage Developments,
various issues.
.Washington: Employment and Earnings,
various issues.
Washington: Geographic Profile of
Employment and Unemployment, various
issues.
Washington: Monthly Labor Review,
various issues.
This content downloaded from 147.251.55.15 on Thu, 29 Mar 2018 08:55:08 UTC
All use subject to http://about.jstor.org/terms