This paper provides an empirical test of the predictions of a standard private
information model of labor strike activity using firm-specific sales data and wage and strike
data from 1124 collective bargaining agreements made in the U.S. manufacturing sector.
The implications of the private information model are tested directly by using the error in
forecasted sales as a proxy for the private information on profitability held by the firm, and
the variance of the forecast error in sales as a proxy for the union’s ex ante uncertainty. In
support of the model, strike incidence and unconditional duration are negatively correlated
with the forecast error and positively correlated with the variance of the forecast error.
Furthermore, the variance of the forecast error is systematically related to the stage of
settlement of collective bargaining agreements. Contrary to the predictions of the model,
there is no evidence that wages are correlated with the forecast error in sales in those
bargaining agreements which involve a strike.

This paper uses a unique data set of contracts and strikes to address the old question of the
relationship between the business cycle and strike activity. It also examines the factors that determine when
strikes occur and proposes a new test of the recent private information models of strikes. The data set
contains over 7000 contracts covering both manufacturing and non-manufacturing industries for the twelve
year period 1970 to 1981. Contrary to earlier findings there is no simple correlation between fluctuations
in the business cycle and either the number of strikes, strike incidence or strike duration. However, once
specific industry effects are controlled for then strike incidence but not strike duration varies procylically.
As suggested by the Total Cost theory of strikes both demand conditions by industry and labor market
conditions are important factors in determining strike activity. The level of demand by industry as proxied
by the industry producer price index has a negative effect on strike incidence but a positive effect on strike
duration. This opposite movement of strike incidence and strike duration which appears in this paper is
not predicted by any of the theoretical models of strikes. Higher national unemployment reduces the
probability of a strike, but it is regional unemployment and industry specific unemployment that have the
greater negative effect on strike incidence.
‘The recent private information models of strikes suggest that an important determinant of strikes is
a variable observable to the finn but unobservable to the union. In this paper I assume that the firm can
predict future demand conditions better than the union. In this case the difference between actual realized
future prices and the forecasts of those prices made today can be used as a proxy for this unobservable
variable. This paper shows that neither the level nor the variance of this residual has any significant effect
on strike activity. However, both the variance of past prices and the variance of past unemployment have a
significant positive effect on strike incidence. This suggests that some form of uncertainty is an important
determinant of strikes.

This paper presents evidence on two aspects of strike activity
associated with the renegotiation of union contracts: the effects of
contract characteristics on dispute probabilities; and the variation in
strike activity over tine within bargaining pairs. Cross-sectional and
longitudinal estimation techniques show that strike probabilities are
higher in summer and fall than winter and spring. Strike probabilities
are also increased by increasing the length of time between negotiations, and reduced in limited wage reopening negotiations. Finally,
strike probabilities are significantly affected by lagged strike out-
comes. Relative to a peaceful settlement of the previous contract,
strike probabilities are l0 percentage points higher following a strike
of two weeks or less, and 5 to 7 percentage points lower following a
strike of longer than two weeks.

Is there a systematic relation between wage rates and strike out-
cones? This paper addresses the question using a panel of over 2,000
collective bargaining agreements from the Canadian manufacturing sector.
Contrary to the implications of recent signalling and sequential
bargaining models, there is no correlation between contract real wage
rates and strike duration or incidence. Furthermore, lagged strike out-
comes do not affect future wage settlements. On the other hand, there
is some evidence that contract wages affect future strike outcomes, and
also that lagged strike outcomes affect future strike probabilities.

We study strike durations and outcomes for some 2000 disputes that occurred
between 1881 and 1886. Most post-strike bargaining settlements in the 1880s fell
into one of two categories: either a union "victory", characterized by a
significant wage gain or hours cut, or a union "defeat", characterized by the
resumption of work at the previous terms of employment. We find a strong
negative relation between strike duration and the value of the settlement to
workers, reflecting the declining probability of a union victory among longer
strikes. For the subset of strikes over wage increases we estimate a structural
model that includes equations for the capitulation times of the two parties and
a specification of the wage increase conditional on a union victory. We find
strong support for a relative bargaining power hypothesis: factors that enhance
the workers’ ability to withstand a strike tend to raise the wage increase in the
event of a successful strike, while factors that enhance the employer's ability
to withstand a strike tend to lower the wage increase in the event of a union
victory.

A common defnition of a labor union is that it is an association of workers who
bargain collectively with their employer regarding the terms and conditions of
employment. Economic analysis of labor unions falls largely into four related
categories: 1) unions as maximizing agents, 2) dispute resolution: strikes and
arbitration, 3) the effect of unions on wages, and 4) the determination of union
membership. This essay is organized around discussions of these topics.

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 concession schedule which maps
out a negative relationship between wages and strikes. In this
paper a concession schedule is estimated using a unique micro
data set of about 3000 contracts over the period 1970-1981.
Unlike previous wage determination studies, which use the
percentage change in nominal wages as the dependent variable,
this study uses the average expected real wage over the length of
the contract as the dependent variable as this is the wage that
is of interest to the negotiating parties. In order to estimate
the concession schedule it is necessary to control for all
observable variables which effect the level of wages and strike
activity. The most important determinants of the real wage are
found to be bargaining pair specific fixed effects and a general
time trend. Wage settlements at other firms in the sane industry
prior to the negotiations were also important. The estimated
concession schedule has a negative slope as predicted by the
private information models. The concession schedule is fairly
flat -— the real wage decreases by only 3% after a strike lasting
100 days.