MSU RETURNING MATERIAL“:
PEace in book drop to

LIBRAIUES remove this checkout from
‘3“ your record. E3353 W111

 

be charged if hoof—1's

 

returned after the date
stamped be10w.

 

.‘—. Mm..." u—uno—‘u— al-au-u———-—

 

 

 

 

 

 

PUBLIC POLICY AND THE DURATION
OF

PUBLIC EMPLOYEE STRIKE ACTIVITY

By

Robert Charles Rodgers

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements

for the degree of

DOCTOR OF PHILOSOPHY

Labor and Industrial Relations School

1981

ABSTRACT

PUBLIC POLICY AND THE DURATION OF PUBLIC EMPLOYEE

STRIKE ACTIVITY

by

Robert C. Rodgers

Long strikes by public employees arouse public
concern because of a potential threat to the health and
safety of a community. High political (and possibly
economic) costs accrue to employers when essential services
are interrupted. Costs also are incurred by workers as a
result of foregone income. The determinants of the duration
of all nonrecognition strikes occurring in local government
bargaining units in the United States between January,

1977 and October, 1978 were thus examined in this study.
Factors expected to affect strike duration included economic
considerations (the expectations of the union membership
and the economic constraints on the public employer's
ability to make concessions) as well as noneconomic
considerations (bargaining unit structures and the local
labor relations environment). In addition, the impact on

duration of various impasse resolution mechanisms were

Robert C. Rodgers

ABSTRACT: Public Policy and the Duration of Public
Employee Strike Activity

examined.

Results for the impasse resolution factors,
controlling for other relevant considerations in a regression,
indicated that states with compulsory mediation provisions
tended to have slightly shorter strikes. Mandatory penalty
requirements were not, however, related to duration. An
injunction reduced the length of strikes in the smaller
bargaining units, but in the larger units, this effect
was reversed. An unexpected and also inconclusive result
was a positive association between mediation and strike
duration. Results for other determinants indicated that
strikes were longer when (l) the labor costs of the
striking bargaining unit were a large component of total
labor costs, (2) the union supported the strike, (3) the
issue concerned economic matters, and (4) noneducational

personnel were involved.

Oct

.nu

.fiu

b.

v
«A

~
I

~.

.
I“.

ACKNOWLE DGMENTS

This study benefited greatly from the many comments
and suggestions from Professors Richard Block, Jack Stieber
and Cheryl MarantO‘ from the School of Labor and
Industrial Relations, Jack Hunter from the Psychology
department, and Terry Moe from the Political Science
department. I am deeply indebted to each of these
individuals for their many hours of thoughtful review
and evaluation of many drafts of this manuscript. As
Chairman of the Dissertation Committee, I am also deeply
indebted to Rich Block for his work with me on this and
many previous research projects.

Acquisition of the data required the c00peration
of many individuals at the federal level of government.
Alan Stevens, Chief of the Governments Division,Bureau of
the Census, made it possible to match the Bureau of Labor
Statistics (BLS) case numbers identifying each public
employee strike with the correspondent Census jurisdiction
identification number. This enabled an identification of
the characteristics of the labor organization calling the
strike with the actual jurisdiction taking the strike.
Assistance from Larry Curran of the Governments Division

was also particularly helpful in this regard.

ii

From the Department of Labor, Division of
Industrial Relations, Gene Becker assisted with the
acquisition of machine readable data files containing
responses to the 1975-1978 Work StOppage Questionnaire.
Evelyn Traylor from this Division was responsible for
constructing a coding book for this data. In addition, she
spent many hours clarifying for me the meaning and source
of these data. John Bonner, Chief, Division of Public
Sector Labor Relations, Department of Labor, also provided
considerable insight into relevant issues from a public
policy point of view. The direction of this project was
influenced greatly by these recommendations.

This study required the acquisition of many large
machine readable data sets. Generous research support
from the National Association of Schools of Public Affairs
and Administration/National Association of Counties
Personnel Dissertation Fellowship made acquisition of these
data possible. Alma Beals from NASPAA was particularly
instrumental in facilitating completion of this project.

In exchange for a c0py of some of the data purchased from
this fellowship grant, Craig Olson from Purdue University
kindly provided a copy of the 1977 annual employment data.

At the state level of government, helpful comments
and reactions were received from Robert Pejorsky from the
Michigan Employment Relations Commission, Mr. Queen from

Pennsylvania's Bureau of Mediation, and Steve Hunter from

iii

’1“

f

(I)

(U

'1

()

f)

New Jersey's Public Employment Relations Commission.

Endless hours of complicated computer programming
assistance far beyond the technical capability of the
principal investigator were provided by Christopher Brown
and Harriet Dhanak of the MSU Politometric Laboratory.
Merger of a multitude of large data sets would not have
been possible had it not been for their skilled and
relentless assistance. Einar Hardin also provided helpful
suggestions which facilitated construction of the master
data set.

I am deeply indebted also to Henry Ng of the
Univeristy of Texas, Austin for typing the final manuscript
and for his many hours of additional work preparing the
tables and figures for inclusion in the final manuscript.

Calculation of the number of days idled as a
result of a strike was made possible by countless hours
of meticulous calculation by Nanette Rodgers who also
commiserated with me after each and every computer run
inhich was unsuccessful. Her encouragement, and the
inspiration from our children, Adrienne and Jonathan, were

instrumental to the ultimate completion of this project.

iv

TABLE OF CONTENTS

Chapter 1 Toward a Recognition of the Importance of
Studying Strike Duration . . . . . . . page

The Enhanced Visibility of the Public
Employee Strike . . . . . . . . . . page

Chapter 2 The Literature on Public Employee
strikes 0 O C I O O O O O O O I C O 0 page

An Overview of the Research on Public
Employee Strike Activity . . . . . . page

Public Employee Strike and Public
Policy . . . . . . . . . . . . . . . page

What Does the Research Tell Us? . . . page

Chapter 3 The Model . . . . . . . . . . . . . . . page

The Prediction of Strike Activity Using
Cost Benefit Theory . . . . . . . . page

A Framework for Predicting the Duration
of Work StOppages . . . . . . . . . page

The General Factors . . . . . . . . . page
The Dependent Variable (Sit) . . . . . page
The Independent Variables . . . . . . page

Expectations of the Union Membership:
The Bit Vector . . . . . . . . . . page

Relative Wage Changes (CE) . . . . page
Relative Wage Levels (RE) . . . . page
Internal Wage Comparisons (EV) . . page

Constraints on the Employer's Ability
To Pay the Concessions Demanded by
the Union Membership:
The Cit Vector . . . . . . . . . . page
Relative Importance of the Wages of
the Striking Workers to Total
Personnel Costs (WE) . . . . . . page

11

ll

14

18

24

24

28
29
33

34

34
34
36

37

39

39

TABLE OF CONTENTS (continued)

The Number of Bargaining Units

(NU) . . . . . . . . . . . . . . page
Fiscal Constraints . . . . . . . . . page
Fiscal Slack (SL) . . . . . . . . page
Local Tax Burden (TB) . . . . . . page

The Bargaining Relationship:

The Bit Vector . . . . . . . . . . page

Number of Employees Involved in the
Strike (82) . . . . . . . . . . page

Combined Occupational Composition of
the Bargaining Unit (CM) . . . . page

Firefighter Strikes (FIR) . . . . page
Police (POL) . . . . . . . . . . . page
Sanitation Strikes (SAN) . . . . . page
Health Care Workers (HOS) . . . . page
Other Employee Types (0TH) . . . . page
Degree of Inessentiality (NS) . . page

The Bargaining Environment:

The Pit Vector . . . . . . . . . . page

Negotiation of the First Agreement
(FN) . O O O O O O O 0 O O O O 0 page

Union Support for the Work StOppage
(US) . . . . . . . . . . . . . . page

The Issues Under Dispute . . . . . . page

Additional Considerations: Where and
When - The Ki Vector . . . . . . page

t
Location . . . . . . . . . . . . . page
Season . . . . . . . . . . . . . . page

vi

41

42

44

46

46

47
48
48
49
50
51
51

53

53

54

54

57
57

58

 

TABLE OF CONTENTS (continued)

The Public Policy Vector (Lit) . . . page
The Court Injunction (IJ) . . . . page
Mediation (ME) . . . . . . . . . . page
Permissive Strike Law (PL) . . . . page

The Expanded Estimating Model . . . page

Chapter 4 The Analysis . . . . . . . . . . . . . . page

The Method of Analysis and the Data . page

The Results . . . . . . . . . . . . . page
The Bit Variables . . . . . . . . . . page
The Pit Variables . . . . . . . . . . page
The Kit Variables . . . . . . . . . . page
The Cit Variables . . . . . . . . . . page

The Insignificance of the Number of
Units Variable . . . . . . . . . . page

The Duration of Strikes in Towns,
Counties and Cities . . . . . . . page

The Performance of Other Variables in
the Cit Vector . . . . . . . . . . page

The Eit Variables . . . . . . . . . . page

Some Concluding Considerations . . . . page

Chapter 5 Analysis of Public Policy Variables . . page

Do Court Injunctions Enhance or Deter
the Duration of Strikes? . . . . . . page

Do State Compulsory Penalty Laws Make
a Difference? . . . . . . . . . . . page

The Relationship Between the Injunction
and the Number of Strikers . . . . . page

vii

6O
61
61

63

72
72
76
79
91
93

96

96

97

100
102

108

114

114

115

118

TABLE OF CONTENTS (continued)

The Impact of the Injunction Depending
on the Type of Employee . . . . . . page

The Permissive Strike Law . . . . . . page
The Relationship Between Mediation and
Strike Duration . . . . . . . . . . page
Chapter 6 Summary of Results and Conclusions . . . page
Summary of Significant Findings . . . page

Summary and Discussion of Insignificant
Results . . . . . . . . . . . . . . page

Rural (RU) . . . . . . . . . . . . . page
First Negotiation (FN) . . . . . . . page
Winter (WR) . . . . . . . . . . . . page
Supplemental Benefit Issue (SB) . . page

Compulsory Penalty Laws (PE) . . . . page

Injunctions (IJ) . . . . . . . . . . page -

The Contributions and Limitations of
the Study . . . . . . . . . . . . . page

Appendices
Appendix A: Variable Descriptions and
Predictions . . . . . . . . page

Appendix B: Means, Standard Deviations
and Data Sources . . . . . page

Appendix C: The Correlation Matrix . . page

viii

121

124

125

134

134

145
146
147
147
147

148

148

153

156

158

Vbl~
\aq

Table

Table

Table

Table

Table

Table

Table

Table

Table

LIST OF TABLES

. . . . . . . . . . . . . . . . . . . . . page 77
The prediction of strike duration (Sit) with and
without "employee type" variables

(The 412 Sample)

. . . . . . . . . . . . . . . . . . . . . page 82
The prediction of strike duration (S. ) when
excluding variables drawn from the merged Census
data source (The 558 Sample)

0 O O O O I O O O O O O O O O O O O O 0 page 89
Regression results for S. which consider
a curvilinear specificatign of size

0 O O O O O O O 0 O O O O O O O O O O O 0 page 99
The prediction of the duration of strikes
occurring in cities, towns and counties

. . . . . . . . . . . . . . . . . . . . page 105
The prediction of strike duration when considering
the independent effects of CE and EV

. . . . . . . . . . . . . . . . . . . . . page 106
The independent effect of RE

. . . . . . . . . . . . . . . . . . . . page 117
States with and without compulsory penalty laws
(PE), by the number of strike occurrences and the
number of injunctions issued (1977-1978)

. . . . . . . . . . . . . . . . . . . . . page 120
The Effect of the Court Injunction on the Duration
of the Strike (1977-1978)

. . . . . . . .‘. . . . . . . . . . page 123
The prediction of strike duration when considering
the impact of the injunction (IJ) on teacher
strikes (EDU)

ix

LIST OF TABLES (continued)

Table 10 O I O O O O O O O O O O O O O I O O O O 0 page 126
A comparison of mean values for significant

predictors when considering Pennsylvania and
Ohio strikes

Table 11 . . . . . . . . . . . . . . . . . . . . . page 130
The prediction of strike duration when considering
the impact of compulsory mediation provisions
in state collective bargaining statutes

Table 12 . . . . . . . . . . . . . . . . . . . . . page 135
A comparison of the mean beta weights on variables
which were significant in Tables 1 through 6 for
strikes occurring in all jurisdiction types and
strikes occurring in cities, towns, and counties

LIST OF FIGURES

. . . . . . . . . . . . . . . page 6

Mean Duration of Strikes by Local Government

Employees Compared to the Mean Duration of Strikes

Figure 1 .
(1960-1977)

in the Service Industry

xi

Chapter 1
Toward a Rec0gnition of the Importance of

Studying Strike Duration

Over the past decade, industrial relations scholars

have just begun to examine the determinants of public

employee strike activity. Adopting the methodology of the

(earlier private sector strike studies, research on public

.sector strikes has taken into consideration the multi-

ciimensional character of the work stOppage. These dimensions

include frequency (number of strikes occurring), breadth

(the number of persons involved), and impact (the duration

c>r length of the strike). In view of the legal sanctions

vvhich prohibit or limit strikes in public employment, the
:identification of the determinants of strike frequency has

lJeen of particular interest not only to researchers but to

public officials as well.

One dimension of the strike, its duration, has

litrgely been neglected, however. Certainly fifty strikes

‘vfiich.last one day would be less likely to cause concern
from a public policy point of view than fifty strikes of
fifty days duration, especially if the strikes are by
essential employee groups such as the firefighters or the
police. The economic costs incurred by workers which result
from foregone income can be substantial.

High political costs can also be incurred by both

marlagement and labor. Long strikes arouse public concern

2
over a potential threat to the health and safety of the
community. With protracted strikes, day to day accommodation
between the parties which is critical to the successful
administration of a contract is temporarily interrupted.
This can result in a weakening of the political support held
by the union leadership, especially when little progress is
being made at the negotiating table. Thus, an enhanced
understanding of the causal factors accociated with the
duration of the public employee strike would be welcomed by
public administrators, unions and scholars alike.

Prior efforts to explain the duration of public
employee strikes have, unfortunately, been inconclusive.
{This has been attributed by some to the use of state level
:strike data.2 Problems of interpretation can result when
asize deflators are used in studies having states as the
Iprimary unit of analysis.3 This technique is designed to
(zontrol for the fact that a greater number of strikes (and
Ipossibly longer strikes) occur in states which have more
Ipublic employees. However, when any of the three strike
nneasures (frequency, breadth or duration) is adjusted by
Ciividing it by a size proxy such as employment, the perform-
Eince of the independent variables in a regression are
jJnfluenced both by their relationship with the size proxy
fits well as with the strike measure. This procedure can thus
IDesult, to the dismay of the researcher, in the discovery of

Siigns on coefficients which were unexpected.

Cost benefit frameworks are probably better suited

(1.:

f

(Y

(I)

(3

V‘.
“4..

a".

‘4

3
to an explanation of strike occurrences than of strike
duration. This is because predictions for many of the
variables are a function of whether the vested interests of
labor or of management are considered in the assessment of
the factor's influence. Theoretically, it remains indeter-
minate in many instances whose influence would be expected
to dominate.

In recognition of these caveats, this study is
based on the notion that the duration of the strike is
affected by considerations other than environmental condi-
tions such as unemployment levels or inflation rates. Four
basic factors are considered. These include: (1) the
earnings advantage (or disadvantage) of the employee group
'which voted to strike; (2) the fiscal and political charac-
teristics of the corresponding jurisdiction taking the
strike; (3) the characteristics of the bargaining unit
calling the strike; and (4) a vector of variables which
characterize the local labor relations environment.

In studying duration, it has been observed that the
identification of factors which characterize differences
across local strike events is essential.5 The bargaining
unit initiating the strike will thus be the basic unit of
analysis. This orientation is useful because of the emphasis
on the duration of specific strike events rather than the
mean duration of all strikes occuring within broad geo-
graphical areas.

Using the 558 strikes in local government

4
bargaining units which occurred during the negotiation or
renegotiation of a collective bargaining agreement between
January 1, 1977 and October 15, 1978, this study examines
the determinants of their duration. In addition to economic,
structural and environmental factors, this research considers
whether mediation and the court injunction act to reduce
(or possibly increase) strike duration. These findings,
moreover, will have important, albeit qualified, implications
of the establishment of public policies which encourage the

involvement of third parties at impasse.
The Enhanced Visibility of the Public Employee Strike

A central premise of this analysis is that a better
understanding of the determinants of strike duration is just
as critical at this stage in the emerging deve10pment of a
public sector strike theory as are studies emphasizing their
occurrence or breadth. Aside from the fact that strikes by
public employees are a new phenomenon historically, the
scarcity of work on duration is due partly to the fact that
less working time has been lost overall than has been
observed in private sector employment. Of the 11,109 agree-
ments which became effective at all levels of local govern-
ment in 1977, 95.9% were settled without a strike. Among
the 452 bargaining units which elected to withhold services
for at least one day, an average of only 7.7 working days

were subsequently lost.6 This is considerably less than the

5
duration of strikes in private sector industries. During
calender year 1977, the mean duration of strikes in manu-
facturing was 29.4 days, 21.3 days in contract construction,
21.2 days in the services industry, and 30.7 days in finance,
insurance and real estate.7

For comparative purposes, it is instructive to
compare the strike experience of local government employees
with their closest private sector counterparts--the services
industry. The secular trend over the past twenty years shows
that the mean duration of strikes in the services industry
exhibited a slight upward trend. As shown in Figure l, the
range in duration was bounded by 11 calender days in 1965
and 33 calender days in 1971. Local government employee
strikes, on the other hand, became noticeably longer over
this same period. As also shown in Figure l, strikes
initiated by local government employees in the early 1960's
lasted on the average only two (2) calender days. By the
late 1970's, the mean duration of public employee strikes
averaged eight (8) calender days. This is less than the
overall strike experience in the services industry during
1960 - 1977. Public employee strikes, on the other hand,
have become noticeably larger.

In addition to evidence that supports the prOpo-
sition that the public employee strike is lengthening, there
are also reasons for believing that strikes will become more
frequent in the 1980's. Consider, for instance, recent

developments which provide added incentives to public

Figure 1

Mean Duration of Strikes by Local Government Employees
Compared to the Mean Duration of Strikes in the Service Industry
1 960-1 977

Mean Duration
in Calendar Days

33

2 1
20 ‘
a"
1 9 ,o
.0
.0
.0

18

'60 '61 ’62 '63 '64 '65 '86 '67 '68 '69 '70 '71 '72 '73 '74 '75 '76 ‘77

Year
Data Sources. BLS, Work Stoppages in Government, 1977 (Report 554): Work Stoppages in Government. 1975 (Report 483); Work Stoppages rn

Government. 1974 (Report 453); Work Stoppages In Government. 1973 (Report 437); Work Stoppages in Government. 1972 (Report 434), Work

Stoppdgns Government Employees, 1942 1961 (Report 348); Work Stoppagos by Industry Group. 1950-1977. BLS Bulletin 2032, Analysts 01
Work Sloppages. 1977.

Mean Duration of Mean Duration of
u—_—Q Strikes by Loca| raga-canon... Strikes in the
Government Employees Services Industry

7
employers to take strikes. Tightening fiscal constraints
have begun to plague many local governments in the 1970's.
As a result, employers are now more willing to accept strikes
because they result in a savings of expenditures without a
commensurate reduction in revenues.

Further, a once pOpular alternative to the strike
in public employment, compulsory arbitration, has recently
been criticized.9 This is primarily because arbitration
awards have been viewed by some employers as too generous.10
Anticipation of an arbitration settlement which could be
unacceptable would be expected to enhance the attractiveness
of taking a strike. The strike during the decade of the
1980's is thus eXpected to become better accepted as a viable
Ineans of settling public employee disputes among less essen-
tial employee groups.ll

Public employee strikes may thus become not only
longer, but also more frequent. The potential contribution
of studies which predict the frequency of the strike can
thus be bolstered by a better understanding of the determi-
nants of the duration of the strike. As frequency studies
continue to be of interest to public policy officials, it
is hoped this study will spark interest in further studying
duration.

As previously observed, one concern of this study
is to evaluate the impact on strike duration, if any, of

12

mediation and the court injunction. It is important,

therefore, to review the research not only on public employee

8
strike activity, but also the research which evaluates the
impact of public policy on strikes. This review is presented
in Chapter 2. Chapter 3 presents a discussion of a general
model which is used to identify factors expected to contrib-
ute to an explanation of duration. Chapter 4 discusses the
results when the variables defined in Chapter 3 are entered
in multiple regressions. Chapter 5 further considers and
analyzes the empirical relationship between the public policy
factors selected for examination and strike duration. Con-

clusions of the study are presented in Chapter 6.

9
Footnotes (Chapter 1)

1

The seminal work on public employee strike activity was
by John F. Burton & Charles E. Krider, "The Incidence of
Strikes in Public Employment," in Daniel S. Hamermesh,
Labor in the Public and Nonprofit Sectors (Princeton,
N.J.: Princeton University Press, 1975), pp. 135-185.
For a more thorough discussion of the literature, see
Chapter 2.

 

As noted by one set of researchers, "In some cases, state
boundaries may be artificial delimiters of the Operation
of metrOpolitan and regional influences." James L. Perry
& Leslie J. Berkes, "Predicting Local Government Strike
Activity: An Exploratory Analysis," Western Political
Science Quarterly, Vol. 30 (December, 1977), p. 527.

 

 

Robert N. Stern, "MethodolOgical Issues in Quantitative
Strike Analysis," Industrial Relations, Vol. 17, No. 1
(February, 1978). For a general discussion of this
problem, see Malcolm Fisher, Measurement of Labour Disputes
and Their Economic Effects (Orgafiization for Economic
Co-Operation and Development, (1973).

 

 

 

An alternative approach is to enter the size proxy as an
independent (rather than a divisor of the dependent)
variable. This tends to raise the explanatory power of
the regression at the expense of reducing the significance
of the other independent variables in the equation due to
high multicolinearity.

5

Robert N. Stern, "Intermetropolitan Patterns of Strike
Frequency," Industrial and Labor Relations Review,
Vol. 29, No. 2 (January, 1976), p. 233.

6

U.S. Department of Commerce & U.S. Department of Labor,
Labor Management Relations in State and Local Government,
1977 Census of Governments, Vol. 3, No. 3 (Washington,
D.C., 1979).

7
Bureau of Labor Statistics, Analysis of Work StOppages,
1977, Bulletin 2032, Table 14 (G.P.O.: 1977), pp. 24—28.

10

8
Theodore W. Kheel, "Resolving Deadlocks without Banning
Strikes," in Joseph Loewenberg & Michael H. Moskow (eds.).
Collective Bargaining in Government (Englewood Cliffs:
Prentice Hall, 1972), p. 262. See also Robert C. Rodgers,
"A Replication of the Burton Krider Model of Public
Employee Strike Activity," Industrial Relations Research
Association Series, (1981), pp. 241-251. The following
remark is also reflective of this emerging point of View:
"Until only recently the public supported a variety of
anti-strike measures in the governmental sector, but lately
they have shown an increased willingness to take at least
some strikes." from David Lewin, "Collective Bargaining
and the Right to Strike," in A. Lawrence Chickering (ed.),
Public Emplgyee Unions: A Study of the Crisis in Public
Sector Labor Relations (San Francisco: Institute for
Contemporary StudIes, 1976), PP. 145-146; for a report of

a cost savings which accrued to a city government as a
result of a strike, see George A. Sipel, Public Management,
V01. 58, No. 2 (February, 1976), pp. 4-5.

 

 

 

 

9

See J. Joseph Loewenberg, "Compulsory Arbitration for

Police and Firefighters in Pennsylvania in 1968," Industrial
and Labor Relations Review, Vol. 23 (April, 1970), pp. 367-
379.

 

 

10
Compulsory arbitration, for instance, has been cited by the
Mayor of Detroit, Michigan as a primary cause for this
city's recent financial difficulties.

11
This does not mean, however, that the use of compulsory
arbitration schemes, whether conventional or any one of
the many forms of final offer arbitration, will necessarily
subside. Some experts in the field have suggested that
they are likely to become even more frequently used
mechanisms for resolving public employee disputes in the
1980's. Graig Overton, "The Climate for Collective
Bargaining in General Purpose Local Government in the
1980's," Industrial Relations Research Association Series
(September, 1981), pp. 290-295.

 

12
A research agenda in the 1980's included the need to learn
more about the effectiveness impact of third parties in the
resolution of labor management disputes. U.S. Department
of Labor, Labor Management Relations Research Priorities
for the 1980's, Final Report to the Secretary of Labor,
(GPO: I980), p. 46.

 

Chapter 2

The Literature on Public Employee Strikes

An Overview of the Research on Public Employee

Strike Activity

As observed, we have little understanding of the
determinants of the duration of strikes in the public sector.
As will be evident in this discussion of the literature, the
number of studies are limited and the evidence inconclusive.
While time series analyses of private sector strike
activity have been common since 1952,1 our brief experience
with public sector strikes for the most part has precluded
similar analyses due to a lack of a sufficient number of
observations.2 Cross sectional analyses of interstate
differences in public employee strikes have, on the other
hand, not been precluded.

In their cross sectional study, Burton and Krider
found the performance of the fifteen variables selected
for examination across multiple strike measures, including
duration, to be inconsistent and disappointing. "Most
of the variation among states in strike activity in a
particular year," they reported, "cannot be explained by
our variables, and those variables that appear important in
one year often are unimportant or have an Opposite
effect in other years."3 The explanatory power of their

multiple regressions was low, especially considering the

11

12
large number of predictors which were used. Inconsistent
signs were also found on many coefficients across the four
years of their analysis (1968 through 1971). As noted by
the researchers, this may have been due to the high multi-
colinearity among the variables selected for examination.

Results when using duration as the dependent
strike measure were especially unilluminating. Referring
to the inability of previous research to explain duration,
the authors subsequently questioned the merit of future
attempts to study strike duration. Duration may, in fact,
be "a particularly erratic aSpect of strike activity."4
An important benefit of this study is that it will provide
an empirical basis for evaluating this concern.

In a subsequent cross sectional study by Perry
and Berkes,5 the develOpment of a theory of public sector
strikes was avoided in light of "the paucity of previous
research and the general lack of agreement about deter-
minants of local employee strike activity." Rather than
making predictions about the relationship of specific
variables with multiple measures of strike activity, they
factor analyzed 40 variables which were hypothesized to
influence public employee strikes.6 These variables were
divided into four general categories: macro-environmental
variables, governmental sector variables, public employment
variables and legal policy variables. Factor loadings
suggested to the authors ten plausible groupings.

These were identified as follows, with examples

13

of the variables associated with each factor in paren-
theses: (l) urbanization—industrialization (percent of
the pOpulation which is urban, percent in non—agricultural
establishments), (2) union influence in the private sector
(union membership as a percent of non-agriculture employ-
ment), (3) race-poverty (percent of the pOpulation
below a "low income level"), (4) fiscal effort (govern-
ment expenditures per capita, government expenditures per
$1,000 personal income), (5) prOperty tax reliance (per
capita prOperty tax revenues, prOperty tax revenues as a
percent of expenditures), (6) local employee status (ratio
of government employee earnings to private sector employee
earnings, local government employment as a percent of
non-agriculture employment), (7) past strike activity
(number of public employees involved in work stOppages
(1958-1968, logged), (8) local reliance (percentage of
government revenue raised by local governments, percent
of state and local government expenditures), (9) pro-
fessionalism (prOportion of cities with a city manager
form of government, percent of the pOpulation within city
manager cities), (10) facilitation of bargaining,
consisting largely of provisions in the law (sc0pe of
bargaining, unfair labor practice provisions, union
security provisions).

After generating "factor variables" for each
of the above ten groupings, the reconstructed variables

were entered into multiple regressions for each of the

14
strike measures selected for examination. For one measure,
the number of days idle per non-education, local government
employee, more than 50% of the variance was explained by
the entry of the factor variables listed above.7 The one
and only significant predictor, however, was the local
employee status factor. A predicted negative sign for
this factor turned out to be positive.

The absence of a control for the size of the
bargaining unit makes interpretation of these findings
subject to considerable qualification. As observed, the
frequency of strikes is theoretically and empirically
related to the number of employees negotiating or rene-
gotiating agreements and also to the number seeking
reCOgnition status. It is thus not surprising to find that
there are more strikes in states which have more public
employees. If it is also true that strikes in the larger
states tend, on the average, to be longer, then the local
employee status variable used by Perry and Berkes may
actually have been proxying the number of local public
employees in the state. The author's description of and
subsequent prediction for this factor may thus have been

misspecified.

Public Employee Strike and Public Policy

In a third study, Perry considered the direct

impact of collective bargaining laws on public employee

15
strike activity for three categories of employee groups:
teachers, local government employees and state government
employees.8 Positive, although generally insignificant,
correlations were found between multiple measures of
strikes and union security provisions in state public
employment bargaining laws.9 Perry expected positive
coefficients, since union bargaining power and union
expectations should be greater when union security
provisions are in effect.

One plausible explanation for the finding of
insignificant coefficients is the fact that bargaining
power may not be linearly related to strike activity. If
unions with a high degree of bargaining power are able to
obtain concessions without striking, while unions with
relatively low levels of bargaining power seldom strike
because they lack the necessary strength, then the
relationship between strikes and bargaining power would
be curvilinear. A positive and significant relationship
between strikes and union security provisions would thus
be seen only when considering labor organizations which
were weak or "moderately strong."10

A central conclusion of the Perry study was that
state enacted legislation has had little effect on public
employee strike activity.‘ This observation is consistent
with similarly insignificant results obtained by Burton
and Krider on a set of legal variables which were also

expected to be positively associated with strike activity.

16

In the Burton and Krider, the Perry and Berkes
and the Perry studies, a dummy variable was created to
represent the presence (or absence) of a provision in
state legislation covering public employees. In the factor
analytic study by Perry and Berkes, eleven provisions found
in state laws were considered. These consisted of the
existence of an impasse procedure, a grievance procedure,
provisions pertaining to the legality of strikes, manage-
ment rights, the sc0pe of bargaining, unfair labor
practices, union security, bargaining rights, unit
determination, rules of procedure and the establishment of
a labor relations administrative policy. Burton & Krider
considered, among other factors, the existence of a third
party procedure law and a strike prohibition law. Perry
considered the presence (or absence) of strike policies
and, as noted, union security provisions. He also used
a comprehensive index which was constructed by summing
across twelve selected collective bargaining provisions
in the state law.11 In none of these studies was a
consistently significant relationship reported between the
legal variables and strikes.

Legal indexes used in prior research, regardless
of whether they were constructed to capture a particular
provision in a collective bargaining law or the compre-
hensiveness of such a law, are not factors which necessarily
have a direct impact on strike activity. A provision in

a state's collective bargaining law may require the

l7

mediation of all disputes unsettled by the termination
date of the contract. Nonetheless, many disputes which
reach impasse at this stage are settled without the
assistance of mediators.12 Courts may be empowered, under
state law, to deliver cease and desist injunctions.
Injunctions can not be issued by the courts in most
states, however, unless the employer is willing to file a
formal complaint. The courts do, of course, deliver cease
and desist injunctions, but these orders, eSpecially
without effective enforcement, may be ingored more often
than obeyed. This response, if it exists, would, over
time, reduce the effectiveness of penalty provisions as a
strike deterrent.

Two explanations are possible for the absence
of a significant relationship between laws and strikes.
As observed, there is reason to believe that laws do not
necessarily have a direct impact on the actual strike
behavior of the parties. Alternatively, the direction of
the effect, if and when it is realized, may simply be
equivocal. Strike prohibition laws may serve their
intended purpose by significantly reducing strike
occurrences, eSpecially in states which complement this
law with compulsory arbitration. In states which have no
compulsory arbitration Option available to the parties, the
strike prohibition law may actually encourage militant
behavior.

Public policy may thus act both to encourage

18
and discourage the duration of strikes.13 The theoretical
possibility of this two-edged effect is, similarly, a
limitation of the basic frameworks used in previous
empirical studies.14 An attempt has been made in this
study to consider factors, such as the involvement of
mediators and the issuance of court injunctions, which are
used to limit the duration of actual strike occurrences.
This eliminates the problems associated with considering the
impact Of legal variables which may not have a direct
impact on strikes. This approach does not, however, avoid
the theoretical ambiguity inherent in the direction of
their impact. The empirical relationship between selected

Iaublic policy factors and strike duration will be considered

:in.greater detail in Chapter 5.
What Does the Research Tell Us?

A recurring theme in these studies is that little
ies actually known about the determinants of public sector
eunployee strike activity. Perry and Berkes suggested that
"Emalitical and organizational models" ought to be
illcorporated in the construction of a public sector strike
tflaeory.15 Their findings, however, precluded a precise
iJiterpretation of what is meant by political or even
<3rganizational considerations. While Perry and Berkes
exPlained more of the variation in duration than the other

researchers, the sign on the coefficient of their only

19
significant predictor was inconsistent with their
expectation. Perry expected positive and significant
coefficients on his public policy variables, but discovered
that correlations with his strike measures were not
significant.

Unlike the Perry and Berkes and the Perry studies,
Burton and Krider specified a model which provided a
themaretical basis for making predictions. Like Perry,
they considered the impact of state policy indicators on
the: incidence of strikes. Environmental and organizational
factumrs, however, were also taken into consideration. Even
wit}: this more fully specified model, their overall results,
as 111 the Perry study, were inconclusive.

Multiple strike measures were used in these
studies. Across measures, no attempt was made to determine
Statxistically whether or not these strike measures were
captulring similar or different dimensions of the strike.
Wit}; an implicit assumption that the strike measures were
homogeneous, no differences were taken into consideration in
what; was expected in the performance of the coefficients on
the independent variables when using duration as Opposed to
frequency or breadth.:L6

If there are basic differences across the various
dimensions of the strike, including its occurrence, breadth
and duration, the develOpment of a different model and
the selection of different independent predictors for each

Of the strike measures would have been preferable.17 The

20
lack of an understanding of the general characteristics
which differentiate these measures has thus far precluded
this approach, however.

Treating all of the dimensions as equivalent,
then, identical independent variables were used in prior
studies to predict all dimensions of the strike. It was
assumed that the framework develOped to predict strike
frequency was theoretically relevant to examinations of the
other strike dimensions, including duration. This study
takes a different approach. A model is prOposed which has
theoretical relevance for one and only one of the strike
dimensions, duration. Factors which would be expected to
influence the decision to engage in or take a strike may
be different, presumably, from the determinants of the
strike's duration. An important implication from the
empirical results of prior studies is that strike frequency
mode ls have not been particularly useful frameworks for
Predicting strike duration. The model developed in this
Study, to be discussed in Chapter 3, thus modifies the
frameworks which have been the cornerstone of work in this

area .

21

Footnotes (Chapter 2)

This research, an often cited work in the literature, is
that of Albert Rees, "Industrial Conflict and Business
Fluctuations," Journal of Political Economy, Vol. 60,

No. 5 (October, 1952) . This line of empirical investigation
began with Alvin Hansen's publication in 1921 of "Cycles

of Strikes," American Economic Review, Vol. 11, No. 4
(December, 192D: pp. 616-621.

One time series study has been reported by Andrew R.
Weintraub and Robert J. Thornton, "Why Teachers Strike:

The Economic and Legal Determinants," Journal of Collective
Negotiations, Vol. 5, No. 3 (1976), pp. 193-206.

PublIc employee strike activity has been worthy of note
since 1965 when the Bureau of Labor Statistics first began
disaggregating strikes by public employees at the state

level. Before this time, there were relatively few strike
occurrences.

 

John F. Burton & Charles E. Krider, "The Incidence of
Strikes in Public Employment," in Dan S. Hamermesh,
ijor in the Public and Nonprofit Sectors (Princeton,
N.J.: Princeton University Press, 1975), p. 170.

Ibid., p. 149, p. 156. The difficulty inherent in predict-
lng uration was further delineated during a personal

conversation with John Burton in Atlanta, Georgia,
December 28, 1979.

James L. Perry & Leslie J. Berkes, "Predicting Local
Government Strike Activity: An Exploratory Analysis,"

Western Political Science Quarterly, Vol. 30 (December,
1977), p. 5121.

6

The use of a factor analytic technique, a priori, assumes
tfhat the factor components are orthogonal. No theoretical
Justification is provided for why one might expect this

to be the case and there is no reason to believe that

Good predictors of strikes would necessarily be uncorrelated.

7

Perry and Berkes, unfortunately, fail to specify if their

strike measures pertain to education employees, to non-
education employees, or to both.

22

8
James L. Perry, "Public Policy and Public Employee

Strikes," Industrial Relations, Vol. 16, No. 3 (October,
1977)! Pp. 27§-282.

9

This study is not clear on which laws were actually
examined, i.e., state laws covering local employees,
state laws covering teachers, state laws covering state
employees, or some combination thereof.

10
Burton and Krider, Op. cit., p. 153.

11

This comprehensive index was initially proposed by Thomas
Kochan, "Correlates of State Public Employee Bargaining
Laws," Industrial Relations, Vol. 12, No. 3 (October,
1973). PP. 322-327. Kochan finds positive correlations
between a comprehensive index of collective bargaining
laws and the breadth of strikes by police, firefighters,
teachers, local government employees and state government

employees. Only the correlation with teacher strikes was,
however, significant.

12

This conclusion is evident from a comparison of state laws
which require mediation and the actual use of mediation
across all states. It is just as common for states
Without such laws to involve mediators at impasse as it
is for states with these laws. A more detailed discussion
of this anomaly can be found in Chapter 5.

l3

Kochan and Wheeler report that the militant tactics of
firefighters, as captured by such considerations as slow-
downs, sickouts and picketing, were not related to three
legal provisions in the state bargaining law or to an index
constructed to represent the comprehensiveness of the law.
See Thomas A. Kochan & Hoyt Wheeler, "Municipal Collective
Bargaining: A Model and Analysis of Bargaining Outcomes,"
Edustrial and Labor Relations Review, Vol. 29, No. 1
(October, 1975).

14

For a reaffirmation of this theme, see Robert C. Rodgers,
"Replication of the Burton-Krider Model of Public Employee
Strike Activity," IRRA Series, 1981. Strike activity (or
the "militant tactic") has, on the other hand, been
recently used as a predictor of collective bargaining out—
comes. See Thomas A. Kochan & Hoyt Wheeler, "Municipal
Collective Bargaining: A Model and Analysis of Bargaining
Outcomes," Op. cit.; Paul F. Gerhart, "Determinants of

23

Bargaining Outcomes in Local Government Labor
Negotiations," Industrial and Labor Relations Review,
Vol. 29, No. 3 (March, 1976), pp. 331-351.

15
Perry & Berkes, op. cit., p. 527.

16
It has been reported in a study of strikes in the private
sector that the same predictors perform very differently,
depending on the particular strike measure being examined.
For a comparison of the performance of the same predictor
when using nine different strike specifications, see Jack
Skeels, "Measures of U.S. Strike Activity," Industrial and
Labor Relations Review, Vol. 24, No. 4 (July, 1971),
pp. 515—525.

 

17
Robert N. Stern, "Toward an Empirical Merger: SociolOgical
and Economic Conceptions of Strike Activity," Twenty-Eighth
Annual IRRA Meeting (1975), p. 66.

 

Chapter 3

The Model

The Prediction of Strike Activity Using

Cost Benefit Theory

It is unfortunate that, as noted by McClean,
"there is not a sufficiently well-develOped theory of the
role and determinants of the strike in industry specific
terms from which one can develop a genuinely satisfactory
model."1 As noted, this caveat is eSpecially evident for
the one dimension of the strike under consideration here,
its duration.

One approach, develOped by Ashenfelter and Johnson,
has been used with some success in a number of previous
studies of public and private sector strike activity.2 A
guiding premise of their analysis is that labor and manage-
ment, having complete and undistorted information, are
rational decision makers. It is then hypothesized that the
strike is continued as long as the present value of "holding
out" exceeds the estimated cost of "giving in." The impact
of the strike event itself on the decision making of both
parties is thus taken into consideration.

If, for instance, unemployment insurance or
Umion strike benefits are received, the propensity to
"hold out" is enhanced because the net cost to the workers

<>f remaining on strike is reduced. On the management side,.

24

25
political costs are incurred as the strike interrupts
essential services. Alternatively, economic benefits can
accrue to the employer, since expenditures are averted
without a commensurate reduction in revenues. These
simultaneous effects should offset one another.

This model is based on the assumption that
decision making by labor and management will be both rational
and purposeful. It is assumed, moreover, that during the
give and take of negotiations, the union leadership and
the employer should be able to identify the terms of an
economic settlement which are mutually advantageous. Strikes
subsequently occur, they hypothesize, to square the
expectations of the union membership with what the
employer is able and willing to pay.

The behavior of union members is thereby viewed
as irrational.3 A decision to strike is not in their best
long term interests, but workers are believed nonetheless
to be willing to strike after the expiration of the previous
contract when the offer falls short of the expectation.4

The model of strike duration developed for
purposes of this analysis modifies the Ashenfelter and
Johnson framework as follows. All parties which have a
vested interest in the successful negotiation of a contract
(management, union leadership, the union membership, and the
public at large) are assumed to be fully capable of
rational decision making. Thus, the behavior of the union

nmmbership is believed to be no less rational or purposeful

26
than the behavior of the union leadership who are elected
to represent their interests or the public employer from
whom they are seeking concessions.

This strike duration model, developed in the
following section of this chapter, is similar to the
framework adopted by Ashenfelter and Johnson to the extent
that the decision to abort or continue the strike is
assumed to be a rational process. Both models take into
consideration the interests of more than one party and
thereby recognize that something can be gained (or lost)
by each as a result of an additional day of idleness. As
observed, however, this approach lends itself to the
generation of a variety of competing hypotheses.

"It is difficult to generate unambiguous
predictions regarding the effects of
environmental variables on the dispute
resolution process. This is especially
problematic...since those variables that
increase union militancy or resistance

in bargaining may have the offsetting
effect of reducing management resistance."5

A consideration such as the ability of the
employer to meet union demands, a factor of central impor—
tance to this type of theory, has an ambiguous effect. On
the one hand, a greater ability to pay enhances the
prOpensity of the employer to "give in." This should

lead to a reduction in the duration of the strike.

27

Conversely, a greater ability to pay also
enhances labor's incentive to "hold out" for concessions
that otherwise might not have been expected. This lends
support to the expectation of an increase in duration. It
is thus not surprising that when using this model,
insignificant coefficients on many of the variables
selected for examination have been found.6

A further limitation of this approach is the
implicit assumption that the decision to end the strike
is tied to a contract settlement which is acceptable to a
majority of the voting membership. In the empirical
application of their model, Ashenfelter and Johnson adjust
their dependent variable, the number of strikes beginning
in a quarter, by the number of contract expirations during
this same period. There are many instances, especially in
the public sector, when workers return to their jobs without
a contract.7 Moreover, the extent to which the decision
to end a strike is tied to a rational consideration of the
costs of agreeing versus the costs of disagreeing is
questionable,8 given that the actual costs involved are
often discernable only after the strike has ended and after
a final settlement has been reached.

In studies which adOpted the Ashenfelter and
Johnson framework, cost-benefit analysis thus became a
useful tool for explaining strike frequency.9 There are
compelling reasons for believing that cost-benefit analysis

should also be relevant to the prediction of strike duration.

28

The costs to each of the parties of engaging in long
strikes should, after all, be a relevant consideration.

Ashenfelter and Johnson skirt the indeterminant
predictions suggested by a cost-benefit framework by
assuming away the rationality of the union membership and
by considering primarily the net costs of the strike to
the employer. Moreover, their primary interest was in
being able to predict the frequency of strikes in private,
rather than public, employment. When considering the
impact of cost factors to labor or to the public, important
considerations for a study of public sector strikes, the net

impact of cost considerations becomes unavoidably problematic.

A Framework for Predicting the Duration of Work

StOppages

In recoqnition of the problems associated with
adopting a cost-benefit model for a study of public employee
strike duration, the Ashenfelter and Johnson framework has
been modified and extended. Additional considerations
which rational actors would be expected to take into
account, including some that are Specifically relevant
to the public sector, are incorporated into the model.

As noted in Chapter 1, it is important to
consider factors which are Specifically relevant to the
local bargaining situation. Because the basic unit of

analysis is the bargaining unit which elected to initiate

29
a strike, Spurious statistical relationships which result
from using deflated strike measures will not be a problem.
As in previous strike studies, multiple regression
will be the primary statistical technique. Based on the
foregoing review of the literature, five general factors
(including a control vector) have been identified for

examination. These are discussed below.

The General Factors

In this model, the duration of the strike is, in
part, a positive function of the difference between the
union membership's expectation of the attainable wage
concession and the final offer of the last, pre-strike
negotiation round. Other considerations equal, enhanced

expectations should lead to longer strikes.10

It is thus
assumed that final pre-strike offers analyzed across strikes
do not have significant moderating affects on expectation
levels.

In collective bargaining, expectations are
primarily influenced by the earnings of comparable employees
working in neighboring or contiguous jurisdictions. The
greater the discrepancy between the perception of what the
union membership believes constitutes an acceptable
contract offer (as influenced by what comparable employees

have been able to obtain) and the terms offered, the greater

is the overall level of discontent.11 This level of

30

discontent, in turn, Should directly affect the duration of
the strike. A vector (Bit) of three variables has thus
been selected to capture the overall expectation level of
bargaining unit i employees who elected to initiate a strike
during time period t.

The construction of the variables associated
with this vector is guided by the belief that the behavior
of the union membership and the union leadership is
significantly affected by comparisons with the earnings
obtained by other, similar employee groups.12 Data
constraints in prior studies precluded the possibility of
considering any factor other than the impact on strikes of
mean earnings. A typical measure, for instance, was the
mean earnings of all non-education employees in a state.
As reported in Chapter 2, however, inconsistent or unexpected
Signs on all variables have been the rule, rather than the
exception, in these studies.

This study builds on these results by relating
the earnings of the employee group which elected to strike
to the earnings of other, similar employee groups working
for other employers as well as to the mean earnings of
other employee groups working for the same employer. It
is thus not the absolute level of earnings, but the earnings
differential which may, it is believed, be the relevant
factor. Variable Specifications for the Bit vector and
for the other vectors to be considered in this study will

be discussed in the following section of this chapter.

31

Economic and political constraints also have an
unmistakable impact on the ability and willingness of an
employer to make concessions during the course of a strike.
In a. Jrecent study of the collective bargaining experiences
across nine school districts in California, Charles Perry
concluded that the economic hardship experienced by six
of tiles districts studied (as measured by a marked decline
in the tax base, an absolute drOp in tax support of varying
typess, and the failure of the community to adjust prOperty
tax JJBVieS upward) was associated with the severity of the
strikes in those communities.13 Strikes in these districts,
he suggested, were "both prolonged and bitter."

Expectations thus affect the magnitude of the
concessions sought by labor, while restrictions on the
employer affect the magnitude of the concessions offered
durirng the course of the strike. A vector (Cit) of four
variables has been selected to represent the economic and
pOlitical constraints confronted by the bargaining unit i
emPlOyer during time period t.

These variables are not proxies for the financial
capacity of the employer to make concessions. Financial
cOnditions of public entities are particularly difficult to
aScertain, given the political nature of the budget process
and the complexity of public accounting procedures.

Further, even if accurate measures of an employer's ability
to pay were attainable, the net effect of these factors

remains indeterminant. As observed, there are reasons for

32

believing that strikes may be longer, just as there are

reasons for thinking that they may be shorter, when

considering the financial capacity of an employer to award

concessions. Some of the variables to be considered in the

Cit vector do not have Signs which can be predicted a

pric>ri. This does not mean, however, that their ultimate

imPEiCt will necessarily be inconsequential. This is an

empnirical, rather than a theoretical question at this point
in 1:he early develOpment of a theory of strike duration.
The recognition of economic factors alone,

however, is not expected to be particularly successful in

eXplaining strike duration. Existing evidence suggests that

economic considerations are actually better determinants of

strike occurrence than of strike duration.

"Efforts to explain duration with
an economic model have invariably
explained less variance and iden-

tified fewer significant variables
than similar models of frequency

and breadth."14

Moreover, it was suggested in Chapter 1 that one

Of the problems with the methodology used in prior strike

Studies is a failure to take into consideration the overall

Context in which the labor relations system functions.15

Accordingly, noneconomic considerations have been identified

and subsequently grouped into three vectors. These are Bit

(the bargaining relationship in unit i during time period

33
t), Pit (the bargaining environment in unit i during time
period t), and K121; (other relevant factors).

Having discussed the general purposes served by

vectors Ei't' Cit, Bit’

discrussion of the dependent variable, as well as the

Pit' and Kit' we now turn to a

inderpendent variables which are associated with each vector,

respectively.

The Dependent Variable

(Sit)

The Bureau of Labor Statistics (BLS) calculates
the number of idle days which resulted from a strike by
counting the total number of days lost from the beginning of
the strike to the end and by subtracting all weekends
and holidays which fall within this time period. Accordingly,
Strikes beginning on Monday and terminating on the following
Sunday are reported as 5 days of idleness for all employee
grcups.

Firefighters, police and hospitals, on the other
hand, are required to provide service on a continuous, 24
hour a day basis. Sit' the number of working days lost as
a result of a strike, is thus the equivalent of the calender
duration of the strike for nurses, firefighters and police.
The coding scheme used to estimate the number of work days
idled for the remaining employee groups is identical to

the procedure employed by the BLS to calculate working days

34
idles- Log (Sit) will be used for purposes of the
statistical analysis, since its statistical parameters
represent a more normalized distribution. A description

of these parameters can be found in Appendix B.
The Independent Variables

EXpectations of the Union Membership

The E. Vector
zt

Relative wage Changes (CE)

This variable measures the historical gain (or
loss) of the earnings received by the striking bargaining
unit vis-a-vis the wages of comparable employee groups in
other jurisdictions within the same state. It is assumed
that comparisons are made by public employees with their
occupational counterparts, i.e. firefighters compare their
wage position relative to that of other firefighters,
teachers compare their earnings with the wages of other
teachers working in neighboring school districts. One
would expect a decline in strike duration when expectations
for an acceptable contract award (based on what comparable
employee groups were able to Obtain) rose less rapidly than
the rate of increase in previous contracts awarded to the
Strikers.

One variable used in this study to represent

GXpectations measures the expansion (or contraction) of the

35

wages; awarded to other, comparable employee groups.16 Over
the iiiscal period prior to the strike, this is defined as
the change in the mean, monthly earnings of the comparable
employee group in the same state. This measure, as one
indicator of expectations, takes into account the extent

to veriCh wage concessions historically obtained by other
emplrayees may have generated differing levels of discontent.

Thus ,

CE=(ME -ME )/ME

7‘; t-J 75—]

Where: MB is the mean monthly earnings
of comparable, full time
employee groups working in
the same state;

t is October, 1977 for strikes
initiated between October 15,
1977 and October 15, 1978

t is October 15, 1976 for all
strikes initiated between
January 1, 1977 and October

15, 1977.17

AS noted previously, this is a refinement of earnings
measures used in earlier studies in that it takes into
consideration the fact that worker expectations Should

be affected by changes in the compensation awards received

36
18
by other, comparable employees. CE should thus have a
positaive sign. The operational definitions for CE and the
othexr' variables to be used in this analysis are presented

in Appendix A .

Relative Wage Levels (RE)

Although bargaining history Should directly
affect the extent to which workers are dissatisfied, a
compuarison of existing, pre-strike wage levels obtained by
comparable employee groups should also be a relevant factor.
EXpecrtations of the membership are thus affected not only
by past, but by current circumstances as well. The ratio
of the mean monthly earnings of comparable employee groups
in the same state (RE) to the mean monthly earnings of the
bargéiining unit which elected to strike should also be

posiizively associated with strike duration. Thus:
RE= (MEt/Et)

Where: MB is the mean monthly earnings
of comparable, full time
employee groups employed by
jurisdictions located in the
same state;

E is the mean monthly earnings
of the bargaining unit which

elected to initiate a strike

37
during time period t;

t is October, 1977 for strikes
initiated between October 15,
1977 and October 15, 1978;

t is October 15, 1976 for all
strikes initiated between

January 1, 1977 and October
14, 1977.

Internal Wage Comparisons (EV)

The maintenance of the differential in earnings
across employee groups with differing skill levels is also

affected by the perceived equity of pay within the same

jurisdiction. It is important to the higher skilled

emPlCJyee groups to maintain differentials. The variation

in compensation levels, however, can be narrowed by awarding

incrreases in pay for all employee groups by fixed absolute,

ratller than fixed percentage increments. In so doing,

employers are able to offer higher, and therefore more

attractive initial salaries for entry level positions. A

loss in the relative earnings advantage among skilled

emPloyees Should enhance expectations for this group. The

less skilled employee groups would be expected to respond
Similarly to the award of greater relative increases to the

higher skilled employees.

The variable EV, the percent change over the

fiscal year period before the strike in the mean monthly

38

earnings of all other employee groups working in the

jurisdiction which took the strike to the mean monthly

earnings of the bargaining unit which elected to initiate

the strike, should thus have a positive sign. Therefore,

EV

Where:

MJ

is the mean monthly earnings of
all full time employee groups
working for the same employer;
is the pre-strike, mean monthly
earnings of full time bargaining
unit employees who elected to
strike;

is October, 1977 for strikes
initiated between October 15,
1977 and October 15, 1978;

is October 15, 1976 for all
strikes initiated between
January 1, 1977 and October

14, 1977.

39
Constraints on the Employer's Ability to
Pay the Concessions Demanded by the Union Membership
The Cit Vector
Relative Importance of the Wages of the Striking Workers to
Total Personnel Costs (WB)

Considerable emphasis has been placed in the
collective bargaining literature on the economic bargaining
power derived by a small number of organized workers whose
salaries are a small component of total Operating costs.19
Fewer strikes are expected when the wage bill of the
striking employees is low relative to total costs because
of the greater relative bargaining power attributed to
organized labor in this instance.20

"Where the prOportion of labor cost is
small, it may not seem worth while for
the management to refuse to co-Operate
with the employees unless there are
irritating union rules limiting the
freedom of management."21

As an argument which is relevant to explanations Of
strike frequency, how is the prOportion of labor cost to
total cost related to strike duration? If this ratio is
small, the employer Should have greater financial capacity
to sponsor the employment of temporary strike replacements.
The successful recruitment of replacements may require wage

offers which exceed, by a considerable margin, the wages

40
paid to regular employees. When few workers are involved,
the employer may have the financial capacity to pay high
enough wages to attract temporary replacements, hOping that
this will force the strikers back to their jobs.

If, on the other hand, the component of labor
costs of the striking employee group to total labor costs
is high, as is the case with teachers, hiring strike
replacements becomes more difficult. Unlike the former
situation, the additional economic resources required to
recruit replacement personnel can be substantial. Also,

a sufficient number of qualified personnel, especially if
the striking group is large, may Simply be unavailable.

Employees comprising a relatively small component
of labor costs who are threatened with strike replacements
may be cajoled into "giving in" earlier. Further, it may
be more rational for the employer to settle rather than
hire temporary replacements. If a heightened level of
conflict results during the give and take of negotiations,
resistance to "giving in" could, on the other hand, be
bolstered. The interplay among these considerations: the
greater bargaining power of labor when labor costs are
minimal, the greater capacity of the employer to recruit
strike replacements, and the recognition by labor that the
employer may have the capability to make concessions, is
indeed complex.

If the prOportion of labor costs of the strikers

to total labor costs acts primarily as a constraint on the

41

employer, strikes should be longer. But this variable
may also proxy the ability of an employer to grant the
concessions demanded. As such, its effect is indeterminant.

WB, the ratio of the wage bill of the striking
bargaining unit to total wage and salary expenditures of
the jurisdiction taking the strike, is the measure selected
to represent the component of labor costs to the overall
labor costs incurred by the employer. Since public
employers are labor intensive, the denominator of W8 is
total labor costs, not total costs.22 The construction
of this variable is intended to represent a financial
constraint on the employer. However, to the extent that
WB also represents the ability of the employer to make

concessions, its Sign must remain indeterminant.

The Number of Bargaining Units (NU)

While the short run costs of granting a
favorable contract to organized employees who account for
a small component of overall costs is minimal, long termt
costs may be formidable. Public administrators who are
required to negotiate with more than one bargaining unit
have reason to be hesitant to award favorable contracts to
smaller units.23 Agreements which favor one employee group
can become the contract pattern during subsequent negotiation
rounds. There also exists the possibility that a contract
could be used by the larger and more powerful labor unions

as a bargaining strategy to obtain more favorable contract

42
terms for themselves.

The number of bargaining units in a jurisdiction
should therefore be a practical constraint on an employer's
willingness to make concessions. This consideration is
captured by the variable (NU), the number of recognized
bargaining units in the jurisdiction taking the strike. NU

is expected to be positively associated with duration.

Fiscal Constraints

Highly constrained financial conditions may give
strength to the employer to "hold out." This would suggest
longer strikes, ceteris paribus. If the financial resources
required to meet demands are simply unavailable, labor may
be convinced early in the course of the strike that their
cause is futile. Two variables have been Operationalized
to capture the extent to which an employer may be constrained
financially and politically to make concessions. These are
fiscal slack (SL) and the local tax burden (TB). A

discussion and Operationalization of these variables follow.

Fiscal Slack (SL)
One bargaining strategy available to organized
public employee groups is to target unused surplus funds
as a financial source for the award of pay concessions.
Since labor organizations have a vested interest in protecting

the job security of their respective memberships, they

43
would be expected to View favorably the commitment of a
dollar to pay personnel over the commitment of a dollar to
cover the wide cadre of Operating expenses such as supplies,
materials, equipment, utility services, etc. The greater
the portion of the budget which is used to support these
Operating needs, the greater also should be the determination
among the union leadership and membership to obtain the
concessions sought.

An accurate accounting of the availability of
these funds would require extensive examination of the audit
statements of each jurisdiction taking a strike. Although
this is beyond the sc0pe of this study, one proxy is
available--the prOportion of total revenues which are not

consumed by direct personnel costs (SL).

SL = (RV - PE) / RV

Where: RV = total revenues received
by the jurisdiction taking
the strike during the
fiscal year 1975;
RE = total expenditures for
personnel during the

fiscal year 1975.

In one sense, SL is a crude measure for the

availability of excess funds which might conceivably be

44
used to justify concessions. It does nevertheless represent,
to a limited degree, the extent to which labor could be
eXpected to know if the employer does (or does not) have
the ability to make concessions.

The greater is SL, the greater the tension one
would expect between labor and management, especially since
a strike has occurred. In effect, there is a larger share
of the pie to fight over, even though management may have
already committed the bulk of these funds to support
non-personnel expenses or possibly capital expansion
programs. This suggests there should be longer strikes.

It remains unclear, however, whether an employer might be
willing to concede earlier. The Sign on SL, as with other

variables in the Ci vector, thus remains indeterminant.

t
Local Tax Burden (TB)
The extent to which local citizens must bear the
costs of government Should be a factor in determining the
length of a strike. The greater the local tax burden
relative to the services received, the greater should
be the degree of community resistance toward awarding
concessions to public employees which have elected to
strike. This represents a constraint to the extent that
additional concessions can be financed only through
increased tax assessments. If the strikers respond to
public pressure, shorter strikes would be expected.

It is also conceivable that the local tax burden

45
and lack of community support for an additional tax assess-
ment reduces the employer's willingness to make concessions.
A high tax burden, relatively Speaking, should become a
formidable obstacle for an employer interested in identifying
the source of future revenues. Under these circumstances,
TB would be positively associated with strike duration.
Little is known about labor's response across various
strike situations to financial constraints faced by the
employer. Accordingly, the eXpectation for TB, a variable
constructed to measure the local tax burden, is indetermi-
nant. This variable is an index which is composed of two

variables:

TB

(PT / PE) + (PT / RV)

Where: PT = total property taxes paid
by local citizens (including
businesses) during fiscal

year 1975;

PE total expenditures for

personnel, 1975;

RV = total revenues reviewed by
the jurisdiction from all

sources during fiscal year

1975.

The first component of this index, PT/PE, captures

46
the extent to which voters perceive a return in terms of
the delivery of government services relative to their tax
contributions. The second component of the TB index,
PT/RV, captures the extent to which local taxes support
government services relative to the revenues received from

other governmental sources.

The Bargaining Relationship
The Bit Vector

Researchers have been interested for many years
in examining the relationship between the size of the
work group and strike prOpensity.25 Regardless of the
terminology (plant size, firm size, size of community or
Size of work group), it has been hypothesized that small
work units are less dehumanizing than large units.26 Given
the greater degree of informality and personalization
associated with small work groups, acceptable settlements

Should be easier to identify.

Number of Employees Involved in the Strike (SZ)

Previous work on strikes has, as noted in
Chapter 2, been oriented toward the macro level of analysis.
This explains why measures of work group size have been so
indirect. Burton and Krider used, as a proxy, a variable
which measured the level of urbanization in their cross

section strike study.27 Average firm size and average

47
plant size in an industry have been used in a recent study
of strikes in private sector employment.28 This study
considers a more direct measure, the number of employees
who were involved in the strike (SZ). In using size, the
breadth of the strike (one of the three dependent measures
often used in strike studies) is taken as a positive predictor

of duration.

Combined Occupational Composition of the Bargaining Unit
(CM)

Labor organizations of course do not necessarily
speak with a unified voice at the bargaining table. When
units are composed of diverse occupational groups, highly
diversified needs can exist with respect to work rules, job
security and equity in pay. As an "unstable amalgam of
interests," one WOUld expect that longer periods of time
might be required to discuss, and subsequently to resolve
these differences.

More homogenous units, with clearly formulated
demands, should be able to engage in a unified negotiation
strategy which is credible and convincing. An employer,
recognizing that there is little dissension among the
membership over what constitutes an acceptable settlement,
may "give in" more readily to demands than when competing
interests must be served.29 The existence of combined,
heterogeneous, units (CM) should thus be positively

associated with duration.

48
It is also reasonable to believe that the duration
of the strike is related to the employee type involved.
Justifications for why some employee groups would be
eXpected to have longer strikes relative to other groups,
and why others might be expected to have shorter strikes,

are discussed below.

Firefighter Strikes (FIR)

From a public policy point of view, considerable
importance has been attributed to maintaining uninterrupted
firefighting services. Replacement of firefighters is
particularly difficult, however, in view of the SOphisticated
nature of modern fighting equipment. Neither the state
police nor the national guard are, as a general rule,
proficient in using this equipment. Firefighters have also
long been recognized as one employee type which places
a premium on maintaining a reservoir of good will in their
respective communities.30 Long and perhaps bitter strikes
could easily jeOpardize this good will. A dummy variable
set equal to 1 if the strike was by firefighters, (FIR),

should thus have a negative sign.

Police (POL)

Police also provide a service which is considered
essential. When police strike, the safety and civil order
of a community is jeOpardized. This should exert pressure

on employers to make necessary concessions and on the

49

police to return to work earlier than otherwise might have

been the case.

On the other hand, because of the increased
threat to the safety of the community, employers may be
more willing to search for and to acquire the temporary
services of a trained civil defense unit (public or private).
Interest in maintaining order in the community, regardless

of the added cost involved, would be heightened. Once

replacements are obtained, the pressure on employers to

obtain an early settlement is mitigated. In consideration

of the countervailing effect of these factors, the Sign on

POL, a dummy variable set equal to 1 if the strike was by

poli c e , i s indeterminant .

Sani tation Strikes (SAN)

Sanitation strikes (SAN) could also be inversely

related to duration, but for a different reason. Inherent

in the type of service delivered by sanitation workers is

a Continuous impact on an entire community. When solid

WaSte is not picked up as scheduled, it begins to collect

On Street curbs. Garbage disposal units begin to overflow,

resulting in a threat to the health of citizens. When waste

water treatment plant workers strike, sewerage can be held
in holding tanks for only a few days before it begins to

back flow into businesses and homes. Pressure from citizens,

directed at both the employer and the sanitation worker,

would be expected to mount rapidly. The net result may be

50

a threat to subcontract services with a private sanitation

firm, especially if the employer thinks the union's demands

are unreasonable. This would suggest that the sign on

SAN should be negative.

On the other hand, empirical evidence exists

which Shows a positive Sign on SAN.31 This may be because

San i tat ion workers ,

for the reasons discussed above, have

enhanced bargaining power. Further, since there is a high

prOportion of sanitation workers in private employment, the

counterpart public employment worker may be more prone to

emulate the type of militant behavior found in private

employment. On balance, this reasoning, coupled with the

finding of a positive relationship in prior studies, suggests

an exPectation of a positive Sign on SAN.

Health Care Workers (HOS)

In a study comparing the attitudinal militancy

of nurses and teachers, Alutto and Belasco found that nurses,

as a group, are more disposed toward militancy than teachers.
This finding is not altered when they control for sex

across the two groups.32 Therefore, the duration of

strikes by nurses and other health care workers (HOS) is
expected to be longer than strikes initiated by teachers.

Controls for three other employee groups will

also be considered. No empirical or theoretical basis

exists, however, for making predictions. These are strikes

by street and highway workers (STR) , transportation

51
anployees (TRN) and welfare service workers (WEL). Post
hoc interpretations over the meaning of the signs found on
the coefficients for these dummy variables are also subject
to considerable qualification due to the small number of

strike occurrences for these groups.

Other Employee Types (0TH)

Other bargaining units which initiated a strike
comprise a variety of employee type combinations. In one
case, firefighters and police may have been combined into
the same unit. In another, sanitation workers may be
combined with street and highways workers. The OTH
category thus represents for many cases, the existence of
a mixed occupational bargaining unit. Its correlation
with CM should be strong. As was the case with CM,
the Sign on OTH, a dummy variable set equal to l for
strikes by all other employee groups except education,

should thus be positive.

Degree of Inessentiality (NS)

While it is important to control for unique
factors which characterize the various employee groups, it
is also meaningful to consider the degree of essentiality
commonly associated with the services delivered by the
various employee groups. The greater the degree of
essentiality, the greater should be the pressure for a

timely strike settlement, other things being equal. Previous

52
research on the empirical relationship between strike
duration and the degree of essentiality confirms this
belief, although no attempt was made in these studies to
control for other factors which should also affect duration.33

In this study, public employee services were categorized

into three basic types:

(1) Essential services - police and
firefighters;

(2) Intermediate services - sanitation,
hospitals, transit, water and
sewerage;

(3) Nonessential - streets, parks,
education, housing, welfare and

general administration.

An index of inessentiality (NS) has thus been
constructed which represents varying degrees of essentiality.
The number 1 is assigned if the striking employee group is
either the police or firefighters, the number 2 is assigned
if the employee group falls within the intermediate services
category, and the number 3 is assigned if the strike is
initiated by one of the nonessential employee types. Since
the most essential employee groups are assigned the smallest

number, NS is expected to be positively associated with

duration.

53
The variables 82, FIR, POL, SAN, OTH, HOS, STR,
TRN, WEL, NS and CM constitute the bargaining relationship

vector Bi and thereby represent an importance facet of

t
the overall labor relations environment. A consideration of
the bargaining context, another important aspect of the

labor relations environment, is also important.

The Bargaining Environment
The Pit Vector
Negotiation of the First Agreement (FN)

Once a policy is in effect and a determination of
the bargaining unit has been made, the parties are obligated
to negotiate. Negotiation of first contracts, however, may
be most difficult. Strikes occurring after recognition,
but before a written contract has been obtained, may be
long and bitter. The union tends to get "shunted around
among the conflicting or overlapping authorities of a
department manager, a mayor and a city council."34

Observers of the problems associated with the
negotiation (n5 first contracts have suggested that the
parties, inexperienced often in the give and take of the
collective bargaining process, are less inclined to alter
their original positions, eSpeCially after a strike has
been initiated.35 Further, no contract language exists

over which the parties can haggle. In view of the disagree-

ments which can arise, especially after a bitter recognition

54
campaign, considerable time should be required, even after
the strike has begun, to settle on acceptable contract
language. FN, a dummy variable representing the negotiation

of a first contract, should thus be positively related to

duration.

Union Support for the Work StOppage (US)

Without the support of the labor organization
there is no formal mechanism through which worker grievances
can be heard. Strikes initiated without the support of the
bargaining agent would indeed be expected to be short
lived - The membership resPonsible for an unauthorized
strike would not be eligible for strike benefits, if any
eXiSt-36 Also, depending on the union constitution which
9'0Verns relationships between the international union and
its member locals, unauthorized strike action may be
grounds for establishing trusteeship. Thus, the Sign on
US: a variable representing the presence of union support

for the strike, Should be positive.
The Issues Under DiSpute

Although a majority of strikes in public employ-
mer11: result from disputes over economic issues, this is not
tr‘l‘e of all strikes. One-fourth of the strikes which
OcczlJarred during the negotiation or renegotiation of a

co
I‘1‘t3li‘act during 1977 and 1978 were for non-economic

55
reasons.37 Other concerns important to striking employees
include job security, the administration of work rules,
promotions and demotions, and the administration of the
contract.

No empirical evidence is available on whether
economic strikes in public employment are longer or shorter,
certeris paribus, than the non-economic strike. Recent
evidence, however, suggests that strikes over economic
issues in private sector employment are longer. Economic
issues, in a 196738 study, were defined to include general
wage changes, supplementary benefits (pensions, insurance,
dismissal pay, premium pay) and adjustments to the rate of
pay due to such considerations as job reclassification,
incentive rate adjustments, or changes in policy pertaining
to retroactivity of wage adjustments. One plausible expla-
n"MT-ion is that non-economic issues were less important to
the membership. For this reason, workers were less willing
to Sustain a long strike. This evidence, although not
dj~11‘ectly relevant to the public employment situation,
suggests strikes over economic issues may, as a general rule,
be POSitively associated with duration.

In contrast with the approach taken in the 1967
Strike study, two types of economic issues will be
considered in this analysis--those which resulted primarily
over wages (EC) and those which resulted primarily over a
39

c
orl<=earn with supplemental or fringe benefits (SB).

3 .
upplemental benefits were considered to be: (1) penSlonSr

56
insurance, or other welfare programs, (2) severance or
dismissal pay such as Supplemental Unemployment Benefits
(SUB) or the Guaranteed Annual Wage (GAW), (3) premium
pay (shifts, overtime or weekend pay), (4) travel time,
vacation pay, paid holidays, lunch periods, call in pay,
Sick leave, vacation allowances, funeral leave. Because
only 16 strikes out of the 558 cases occurred because of one
of the issues noted above, any conclusions with respect to
the performance of this variable should be considered very
preliminary. The coefficients on both EC and SB are expected
to be positive.

The existence of a positive relationship may, in
part, depend on the financial ability of the employer to
"give in" to demands. In jurisdictions which are hard
pressed to identify sufficient revenues to meet current
expenditures, job security issues may be more important
bargaining issues than wages or fringe benefits. It is
important to take into account different economic milieu,
especially when economic issues are the principal issue.

The interaction terms EC*WB and EC*SL will thus also be
considered as a relevant factor.

The four variables (FN, US, EC and SB) represent
the bargaining environment of unit i during time period t.
These variables constitute vector Pit'

Other factors merit consideration which cannot
be directly associated with any of the vectors discussed

above. These are the season and location of the strike.

57
fiflnese two general factors have been identified in prior

rwesearch as relevant considerations and are therefore

irucluded in the control vector Kit’

Additional Considerations: Where and When

The K. Vector
st
Location

There is a considerable body of research by
suaciologists which has examined the impact on strikes of

. . . . 40 .
\nar1ous commun1ty character1st1cs. In a recent analys1s

(IE private sector strikes, this trend was expanded to
ichlude a prediction of strike occurrences across Standard

DuetrOpolitan Statistical Areas (SMSA's). This research was

iJlspired by Kerr and Siegel's suggestion in the 1950's that
strikes are associated with the extent to which a community

is tightly or loosely interconnected.4 In the urban, more

congested, less personalized environments, longer strikes
are expected. As suggested forty years ago:
"A society where farmers predominate
is likely to require industrial employees
and employers to co-Operate."42

The Sign on RU, a dummy variable coded as a 1 if the strike

occurred outside of an SMSA, is therefore expected to be

negative. This relationship should be strengthened by the

fact the RU is also likely to be an inverse proxy for union

58

penetration.

Region of the strike should also be relevant.
Strikes which occur in the northern states, where public
employee collective bargaining has stronger political
support, would be expected to be longer than strikes
occurring in the western or southern regions of the country.
The coefficient on NO should thus be positive.

Strikes in southern states, on the other hand,
Should be shorter relative to the length of strikes in the
west or the north. Collective bargaining is not well
established, nor has it been enthusiastically accepted in

southern states.

Season

Skeels hypothesized that strikes during the
winter months should be shorter, other considerations held
constant, than strikes occurring during the summer months.43
Using a dummy variable to represent three of the four
seasons of the year, he finds that, controlling for economic
and political factors, strikes occurring in the winter
lasted significantly fewer days than strikes occurring in
the fall. With winter strikes, there is less Opportunity
to picket or otherwise solicit public support. Since this
reasoning would not be expected to apply to strikes

occurring in southern or western states, the interaction

term NO*WR (where WR represents a strike occurring during

 

 

f‘

(3.

:1

('14

K);

f1

Cc

59
the winter months (December, January and February) is

expected to be inversely related to duration.

The Public Policy Vector
(Lit)

The final vector, is comprised of two

Lit'
dispute resolution procedures which are used to resolve
public employee diSputes (mediation and the court
injunction), and one public policy found in state laws
covering local government employees, the permissive
strike law. Considered first is the relationship of
the two impasse resolution mechanisms on strike duration.
Over the past decade, penalities have been imposed
on workers and on labor organizations who struck, cease and
desist injunctions have been issued by the courts, and
mediators have participated in the resolution of many
disputes. These procedures have been supported and guided
by a variety of state and local laws.44 In some states,
disputes that remain unsettled after the termination of the
contract must be mediated by third parties who are employed
by the state's public sector labor relations commission.
Other states encourage the mediation of disputes by neutrals
from the private sector or from the Federal Mediation and
Conciliation Service (FMCS). As an alternative remedy, the

judiciary can be requested by an employer to issue an

60
injunction when a strike threatens the health and safety of
a community. The penalty, a more punitive measure than a
cease and desist order, is also imposed occasionally on

striking employees or on their respective labor organizations.

The Court Injunction (IJ)

Historically, most public policies have supported
the involvement of the courts in public employee disputes.45
These policies evolved because the threat of an injunction
was thought to be instrumental in preventing strike
occurrences.46 To be a credible threat, however, the
injunction must also be effective in convincing workers who
illegally strike to return to work.

The injunction could, however, aggravate militancy.
One observer has suggested that "a ban on strike is, in
itself, an inducement to strike."47 When employees refuse
to comply with an injunction, the petitioner is placed in
the tenuous position of choosing among several unattractive
courses of action, including further litigation. In most
states, courts have the latitude of imposing penalties on
the members or the union that struck when an injunction is
deliberately disobeyed. Whether threats, coercion, dismissal
or some combination thereof constitute the nature of the
action taken, employees may actually become more uncompro-
mising. Compromise, after all, is critical to obtaining a

settlement. The net effect of IJ, a dummy variable repre-

senting the issuance of a court injunction, is thus

61

indeterminant.

Mediation (ME)
Unlike the court injunction, mediation is an

impasse procedure which is widely practiced.49

It is popular
because of the anticipation that a better mutual understanding
will result over the substantive issues in dispute. The
suggestion of compromises by a mediator also introduces a
new dimension to the bargaining process. The parties are
no longer merely reacting to the same prOposalS and counter-
prOposalS. With an outside prOposal on the table, the
parties, it is assumed, are more likely to settle. Mediators
thus introduce a sense of law and order to the bargaining
process. Skilled third parties can also help inexperienced
negotiators avoid dysfunctional bargaining strategies which
can contribute to a lengthening of a dispute. Earlier
settlements should be facilitated. The sign on the
coefficient of ME, a dummy variable which represents the
presence or absence of mediation, should be negative.

The prediction of a negative relationship for ME
could, on the other hand, be counterbalanced by the assign-
ment of mediators to strikes which are unusually difficult

to settle. The extent to which this effect may (or may not)

be present is the subject of Chapter 5.

Permissive Strike Law (PL)

Nine states have enacted permissive public sector

62
strike laws. Seven of these states experienced at least one
strike during the 1977-1978 period of this analysis. In
Montana, only strikes by firefighters are prohibited. In
Oregon, all public employees except police, firefighters
and guards at prisons and hospitals have a right to strike
as long as all pre-strike procedures prescribed by statute
have been met. Pennsylvania, with a law similar to Oregon's,
also prohibits guards and court employees from striking. In
Vermont, non-teaching municipal employees may strike if 30
days have passed since the issuance of a fact finder's
report and the dispute has not been submitted to an
arbitrator. Minnesota porvides a limited right to strike to
all non-essential employee groups. In Alaska, only "semi
essential" personnel (utility, school, snow removal and
sanitation workers) may strike. In no state is the strike
privilege extended to all public employee groups and in no
state is this right totally unqualified.

Without mandatory legal sanctions against strikes
the courts can not be employed as a mechanism for forcing
workers to return to work. The presence of these laws also
represents public support of and tolerance for strikes by
public employees. The dummy variable PL measures the
existence of a permissive strike law in a state's collective
bargaining policy. It should be positively associated with

strike duration.

63

The Expanded Estimating Model

The expanded form of the estimating model to be

used in this study is as follows:

(1) S. = f(E.

it

Where: S.
rt

it

it

it

it

it

B L.)

C it’ it

P K

tt' it’ it’ it’
represents the number of working

days lost as a result of a strike

in bargaining unit i during time
period t;

is a vector of variables

representing the expectations of bar-
gaining unit i employees during

time period t;

is a vector of variables which
measure: the- economic and political
constraints on the bargaining unit i
employer during time period t;

is a vector of variables

representing the structure of
bargaining unit i during time

period t;

denotes a vector of variables which
describe the bargaining environment
of unit i during time period t:
denotes a vector of control variables

which are expected to affect the

duration of the strike by bargaining

64
unit i employees during time
period t;
Lit is a vector of policy variables
expected to affect the strike
initiated by bargaining unit i

during time period t.

Having discussed and defined the variables to be
considered in this study, the results from entering these
variables in a multiple regression are presented in Chapter
4. As a prelimary to this discussion, the method and

unit of analysis will be reviewed.

65

Footnotes (Chapter 3)

Robert A. McLean, "Coalition Bargaining and Strike Acitivty
in the Electrical Equipment Industry, 1950-1974,"
Industrial and Labor Relations Review, Vol. 30, No. 3
(April, 1977), p. 360.

Orley Ashenfelter & George Johnson, "Bargaining Theory
Trade Unions, and Industrial Strike Activity," Egg
American Economic Review, Vol. 59, No. 1 (March, 1969),
pp. 35-49.

The prOposition that strikes are not economically profitable
to the union membership is not undisputed, however. In

an analysis of 26 strikes in the Vancouver area of Canada,
strikes were found to be profitable to the union membership,
even after conservative assumptions about the probability
of continued employment and the availability of alternative
job Opportunities during the strike were made. See B.
Curtis Eaton, "The Worker and the Profitability of the
Strike," Industrial & Labor Relations Review, Vol. 26,

No. 1, (October, 1972), PP. 670—679. For a critical
discussion of the Ashenfelter and Johnson analysis, see

P. K. Edwards, "Time Series Regression Models of Strike
Activity: A Reconsideration with American Data," British
Journal of Industrial Relations V01. 16, No. 3 (November,
1978), p. 320.

AS further noted by one critic, this model "fails to take
into account that the executive ... must have some degree
of confidence in order to negotiate with employers at all."
Otherwise, the union leadership will "lose all credibility
as a negotiating instrument with management." There is no
strong evidence, however, that there has been a dissipation
in the credibility of the union leadership at the
negotiating table. Malcolm Fisher, op. cit., p. 80.

 

John C. Anderson & Thomas A. Kochan, "Impasse Procedures

in the Canadian Federal Service: Effects on the Bargaining
Process," Industrial & Labor Relations Review, Vol. 30,

No. 3 (April, 1977), p. 288.

 

6

This was initially demonstrated by the insignificance of a
variable used to measure industry profits in the
Ashenfelter and Johnson study, Op. cit.

 

66

This framework was originally used in an analysis of
private sector strike activity, with no assertion that it
was applicable, necessarily, to analyses of public
employee strike activity. In a review of over 50
newspaper accounts of strikes in the public sector which
occurred during 1975, it was found that in more than one-
fourth of the cases, settlements were reached sometime
after the workers returned to the jobs. For a further
discussion of this issue, see Thomas A. Kochan, "Dynamics
of Dispute Resolution in the Public Sector," in Benjamin
Aaron, Joseph Grodin and James Stern, Public Sector

Bar ainin (Washington, D. C.: Bureau of National
Affairs, I979), p. 168.

 

8

This terminology is borrowed from Neil Chamberlain & James
W. Kuhn, Collective Bargaining (New York: McGraw Hill,
1965). pp. 170-173.

 

9

For a summary of recent studies of strike duration which
build on the work by Ashenfelter and Johnson, see Robert
Stern, Op. cit., p. 34.

 

10
This has been described as a form of aspirational
deprivation, in contrast with decremental or progressive
deprivation. Korpi warns that increasing aspirational
deprivation should be positively, but not necessarily
Significantly related to the probability of conflict.
Walter Korpi, "Conflict, Power and Relative Deprivation,"
American Political Science Review, Vol. 68 (1974),
pp. 1575-1576.

 

11
T. R. Gurr, "A Comparative Study of Civil Strife," in
H. D. Graham & T. R. Gurr (eds.), Violence in America
(New York: New American Library), p. 590.

 

12
This idea has been identified in the literature as the
orbits of coercive comparison. See Neil W. Chamberlain
& James W. Kuhn, Collective Bargaining (New York: McGraw-
Hill, 1965). pp. 2034209. These standards are Similar to
the criteria applied by arbitrators. See Frank Elkouri
& Edna Elkouri, How Arbitration Works (Washington, D. C.:
BNA, 1973): pp. 745-796; Howard S. BIOck, "Criteria in
Public Sector Interest Disputes," Proceedings of the 24th
Meeting of the National Academy of Arbitrators (Washington,

-—

D. C.: BNA, 1971)

 

 

 

67

13
Charles Perry, "Teacher Bargaining: The EXperience in
Nine Systems," Industrial and Labor Relations Review,
V01. 33, No. 1 (October, 1979), pp. 10-11.

 

14
Robert N. Stern, "Methodological Issues in Quantitative
Strike Analysis," Industrial Relations, Vol. 17, No. 1
February, 1978), p. 38.

 

15
This was also the conclusion of a summary of recent work
in the area of public policy and labor relations research.
Ralf R. Jones, Public Sector Labor Relations: An
Evaluation of Poligy Related Research (Contract Research
Corporation, 19751, p. 53.

 

 

16
For the construction of different wage change measures,
see Jack W. Skeels, "Measures of U.S. Strike Activity,"
Industrial and Labor Relations Review, Vol. 24, No.4
(July, 1971), pp. 515-525; David Snyder, "Early North
American Strikes: A Reinterpretation," Industrial and
Labor Relations Review, Vol. 30, No. 3 (April, 1977).

 

 

 

17
These dates are an artifact of the fact that the Annual
Survey of Governments, from which these measures are
drawn, is conducted on October 15th of each year. For
a further eXplanation of the sources for the variables to
be used in this analysis, see Appendix A.

18
This same variable has, however, been Operationalized in
an entirely different manner. David Snyder, Op. cit.,
used a moving average of the rate of change in wages
over a Six year period of time to represent the divergence
(or convergence) of expectations. The possibility of
constructing a wage comparison measure such as CE has
not been possible in previous studies because the
bargaining unit initiating the strike was not the primary
unit of analysis. The conception of an expectations
variable is not a new idea. Operationalization of this
variable in this study is, however, a refinement on
earlier strike research.

19
This concept was first introduced by Alfred Marshall in
Principles of Economics, 8th edition (New York: Macmillan
Company, 1949), pp. 383-387; see also Albert Rees, The
Economics of Trade Unions (Chicago: The UniversityTOf
ChICago Press, I960), pp. 70-73.

 

 

68

20
Malcolm Fisher, Op. cit., p. 81.

 

21
Samuel P. Hayes, "Psychology of Conciliation and
Arbitration," in George W. Hartmann & Theodore Newcomb

(eds.), Industrial Conflict (New York: The Gordon Company,
1939) p. 417.

22
A measure of labor costs to total costs has traditionally
been viewed as the prOportion of all labor costs of the
firm to total costs. The WB measure thus differs from
this construction. It has been modified primarily
because public, rather than private sector strikes are
being studied.

23
For a discussion of the implications to a public employer
of granting particularly generous concessions to one
employee group, see Roben W. Fleming, "Public Employee
Bargaining--Problems and PrOSpects," Industrial Relations
Research Association (1978), pp. 14-23.

 

 

24
As public employee bargaining has matured over the past
decade, the parties have undoubtedly become more
SOphisticated in their negotiation strategies. Some
public employee labor organizations have develOped
elaborate auditing procedures for identifying the extent
to which a public employer has the ability to grant the
concessions demanded. The Michigan Education Association,
for example, has develOped a computer program which
analyzes public expenditures in comparison to revenues.
Nevertheless, the identifying of financial reserves to
pay concessions, eSpecially in the larger jurisdictions,
is a practical impossibility. The availability of
uncommited, surplus reserves can be skillfully hidden by
public employers through the establishment of special
accounts for utilities, the commitment of funds to capital
expansion prOgrams which have not been approved by the
governing body, or the refusal to report surplus funds
carried over from one fiscal year to the next. The
variable SL may better represent, then, the extent to
which the union membership believes there "is enough in
the kitty" to grant the concessions sought, than the
extent to which funds may, in fact, be available for this
purpose. As such, it could be viewed as a candidate for
the Eit vector.

69

25
For one of the earliest studies, see Sherill Cleland,
Influences of Plant Size on Industrial Relations
(Princeton, New Jersey: Princeton University Press, 1955).

26
John Shorey, "The Size of the Work Unit and Strike
Incidence," Journal of Industrial Economy (1977), Vol. 23,
No. 2, p. 177.

27
Burton and Krider, op. cit., pp. 156-157.

28
Robert N. Stern, "Toward and Empirical Merger:
Sociological Conceptions of Strike Activity," 28th
Annual Meeting of the Industrial Relations Research
Association Proceeding§7(l975), pp. 59—61.

29
Malcolm Fisher, Op. cit., p. 28.

30
James A. Craft, "Fire Fighter Strategy in Wage Negotiations,"
Quaterly Review of Economics and Business, Vol. 11,

31
Burton and Krider, Op. cit., p. 171.

32
Joseph A. Alutto & James A. Belasco, "Determinants of
Attitudinal Militancy Among Nurses and Teachers,"
Industrial and Labor Relations Review, Vol. 27, No. 2
(January, 1974), pp. 220-222.

33
John F. Burton and Charles Krider, "The Role and
Consequences of Strikes by Public Employees, in J. Joseph
Loewenberg and Michael H. Moskow (eds.), Collective
lgargaining in Government (Englewood Cliffs, New Jersey:
l?rentice Hall, 1972), pp. 278-279.

34
'Neil W. Chamberlain, "Public vs. Private Sector Bargaining,"
in Loewenberg and Moskow, Op. cit., p. 16.

35
Albert A. Blum and Gadi Harel, "The Generic Reasons for
Strikes: An Interpretative Analysis of the Israeli
Case," unpublished manuscript, 1978.

70

36
Sheldon M. Kline, "Strike Benefits of National Unions,"
Monthly Labor Review, Vol. 98, No. 3 (March, 1975),
pp. T7-23.

 

37
This was drawn from preliminary analyses of the machine
readable data file obtained from the BLS which covered
the years 1975-1978.

38

David B. Lipsky and Henry S. Farber, "The Composition of
Strike Activity in the Construction Industry," Industrial
and Labor Relations Review, Vol. 29, No. 3 (April,l976),
pp. 388-404.

 

39
Stern has recently addressed the importance of considering
models which take into consideration different issues
over which strikes occur. See Robert N. Stern,
"Quantitative Strike Analysis, op. cit., p. 41.

 

40
James R. Lincoln, "Community Structure and Industrial
Conflict: An Analysis of Strike Activity in SMSA'S,"
American Sociological Review, Vol. 43 (April, 1978),
pp. 199-220.

41
Clark Kerr and Abraham Siegel, "The Interindustry
PrOpensity to Strike," in Arthur Kornhauser, Robert Dubin,
and A. Ross (eds.), Industrial Conflict (New York, 1954).

 

42
Samuel Hayes, "Psychology of Conciliation and Arbitration,"
in George Hartman and Theodore Newcomb (eds.), Industrial
Conflict (New York: Gordon Company, 1939), p. 410.

 

 

43

Jack Skeels, "Measures of U.S. Strike Activity," op. cit.
For a much earlier discussion, see Dale Yoker No. 4
(July, 1971), pp. 515-525; see also Dale Yoker,
"Seasonality in Strikes," Journal of the American
Statistical Association, V61. 33, No. 204 (December,
1938). pp. 687-693.

 

 

71

44

The most recent summary of these policies can be found in
U.S. Department of Labor, Labor Management Services
Administration, Summary of Public Sector Labor Relations
Policies: Statutes, Attorney General's Opinions and
Selected Court DeciSions (Washington, D. C.: GPO, 1979).

45

A recent summary of penalty laws found across the 50 states
is available in the Hawaii Labor Relations Group, Guide

to Statutory_Provisions in Public Sector Colipctive
Bargaining, "Strike Rights and Prohibitions,"’(Industrial
Relations Center, University of Hawaii, 1978). According
to John Bonner, Director of the Office of Public Labor
Relations, Department of Labor, there are considerable
differences across states in the extent to which

penalties are legally used to shorten the duration of
strikes.

46
An excellent summary of the pros and cons of strike
penalties can be found in Eva Robbins, "Penalties in
Strikes Against a Public Employer, the New York
'COnference Series, Vol. 22, pp. 315-335.

47

William J. Kilberg, "A LImited Right to Strike for Public
Employees," in Loewenberg and Moskow, Op. cit., p. 306;

An alternative view has been offered by Paul Gerhart.
Finding that strike penalties, with explicit prescriptions
for the type of penalty to be imposed, were inversely

and Significantly related to outcomes, Gerhart suggests
that it is not whether strikes are prohibited, but
"whether penalties for engaging in strikes are explicit."
Paul F. Gerhart, "Determinants of Bargaining Outcomes in
Local Government Labor Negotiations," Industrial and

Labor Relations Review, Vol. 29, No. 3 (April, 1976),
p. 344.

 

48

The Department of Labor has recently funded a study
examining the impact of penalties on strike activity,
with particular emphasis on whether they reduce the
occurrence of duration of the strike.

49
For a discussion of the function of impasse procedures,
see Anthony V. SinicrOpi, Dispute Settlement in the

gpblic Sector: The State of the Art (Washington, D.C.:
GPO, 1972)

 

“1.5
. o .

r0
H
(D

rn

rn

Chapter 4

The Analysis

This chapter is composed of two sections. The first
discusses the method of analysis and the data. Results are

presented and analyzed in the second section.
The Method of Analysis and the Data

The unit of analysis to be used is the local
government bargaining unit (or units) which was involved in
a work stOppage while negotiating or renegotiating a collec-
tive bargaining contract.1 Strikes by state government
employees were excluded because of the difficulty in
ascertaining the composition and SCOpe of the bargaining
unit at the state level.2 The analysis will include strikes
which occurred in the United States between January 1, 1977
and October 15, 1978. During this period, 558 stOppages,
resulting in 5,108 days of idleness, satisfy this
description.3

Data on the characteristics of stOppages are
available from the U.S. Department of Labor, Bureau of Labor
Statistics (BLS). Strikes are identified through newspaper
accounts which are monitored on a daily basis by the BLS
district offices, and the Federal Mediation and Conciliation
Service. Because short strikes may not be as newsworthy, the

data represent a biased representation of the larger strikes.

72

'LJ

73
Requests for information are then sent to the public employer
and to the employee association (or union) involved. In
approximately 33% of all instances, responses are received
from both parties. The bulk of the other responses come
from public management. BLS officials estimate that between
95% and 98% of all public employee strikes are included in
this file.

The information requested on the BLS Work StOppage
Report includes a report on when and how long the strike
lasted,4 the type of employee group involved, the union,
and the types of dispute resolution procedures which were
used, if any, to resolve the impasse. Data is also requested
on whether the strike occurred during the first or a
subsequent negotiation period.

These responses, in conjunction with what is
known about the strike from press reports, are coded onto
machine readable data files. This data was acquired with
the support of a National Association of Schools of Public
Affairs and Administration/National Association of Counties
dissertation fellowship.

While this data file identifies the union involved,
the jurisdiction involved remains anonymous. Fortunately,
the Bureau of the Census uses BLS strike reports to identify
the jurisdiction affected. Considerable effort is taken to
insure that the correct jurisdiction is identified. If it
iS not clear on the BLS Work StOppage Report which juris-

diction took the strike, a phone call (or calls) are made to

13‘

jl

t}

t

at

74
insure that an accurate determination is made.

The case number assigned by the BLS, unfortunately,
has no correspondence with the jurisdiction ID number
assigned by the Census. Thus, the extensive data resources
collected by these two agencies could only be merged by
obtaining a lexicon which matched the BLS case number with
the Census jurisdiction ID number. It was learned that the
jurisdiction identification number was written on a COpy of
the BLS strike questionnaire which was used by the Census to
develOp their internal strike files. The Census provided a
listing of the BLS case number with their matched juris-
diction ID numbers for all 1978 strikes. A Similar listing
of 1977 strikes was gathered by the investigator during a
field trip to Washington, D.C. in October, 1979. By merging
the BLS strike files with the Census files on earnings and
collective bargaining coverage, the union which initiated
the strike could be associated with the jurisdiction which
took the strike.

Having identified the jurisdiction involved,
Census data found on machine readable data files were
matched with the data coded onto the BLS strike file. These
data were drawn from the annual employment survey which was
sent to all jurisdictions in the country during 1977, and to
a sample of 15,000 jurisdictions during 1975 and 1976.5 This
survey requests participating governments to report infor-
mation on the monthly earnings by employee group as well as

data on collective bargaining coverage. Finance data were

(V)

75
also drawn from the 1975 Census finance machine readable
data file.

For the variables drawn from the Census data, RE,
EV, WB, NU, SL, and TB (see Appendix A), the number of
strikes in jurisdictions sampled by the Census equaled 412.
For this reason, regression results reported in this chapter
utilize samples of varying sizes, depending on the variables
which are included in the regressions.

Measures to be examined in this study are expected
to be interrelated. It is reasonable to expect that the
same factors which theoretically explain strike duration
will also be related to each other. Introducing modified
versions of the basic model permits the inclusion of highly
correlated variables in separate equations. Their independ-
ent effects on duration can then be evaluated. Reformulation
of the basic model also permits deletion of variables found
to be insignificant. The performance of significant
predictors can then be reexamined.

Moreover, an important purpose served by the

consideration of multiple variables in Ei the expectations

t!

vector, and in Ci the constraints vector, is to enhance

t’
the probability of identifying at least one (or more)
variables for each general factor which contributes to an
explanation of strike duration.

The number of valid cases will vary depending

on the variable in the regression having the fewest number

of nonmissing values. Tests of significance are based on

Ch
th
Th
Ca
Th

de

8:1
an

la

fr.

CO]

76
the number of observations for which all variables were
found to have non-missing values. This is the sample size,

or "n", which will be reported in the tables which follow.

The Results

The regressions reported in Tables 1 and 2 include
as independent variables all of the factors identified in

Chapter 3. The dependent variable, S is strike duration.

it’
Two equations are reported in Table l. The
first includes all "employee types" which were identified

as components of vector Bi (FIR, POL, SAN, OTH, HOS, STR,

t
TRN, WEL). The omitted employee type reference category is
education. As a variable which captures a common
characteristic across the various employee groups, NS,

the inessentiality index, is not entered in this equation.
The essentiality of an employee group is sufficiently
captured by the eight "employee type" dummy variables.

The second equation considers the impact of NS, while
deleting FIR, POL, SAN, OTH, HOS, STR, TRN, and WEL.

CM, the combined bargaining unit variable, is not
entered in the Table 1 equations because the construction
and subsequent inclusion of WB, the labor cost to total
labor cost variable, necessitated the identification of
Specific employee groups. This, by definition, excluded
from consideration employees who were organized into

combined, or mixed bargaining units. Although constructed

...... ....3 Affirm.» :2..=..:.. 13:; .2 ac :t-fifiltt (J... . ......l

NN.

«o.l

ma.

no.

#0.:

ma.1

no.1

Ho.

no.1

Ha.

«H.

«o.1

 

noon

 

non
NA.N
ca
en.
es.
Aeo.mc
recon an
Aen.c
oao.r eo.r
Ash.~c
one. «N.
as
Aa~.c
sea. Ho.r
AeH.V
mmo.r No.r
“so.Hc
~oe.r eH.r
a
Aea.o
nou.r ao.r
Amo.o
owe. so.
Asa.c
one.. no.-
Aen.~o
NAN. mo.
Aoo.~c
emu. SH.
is
Aen.v
a-.r mo.r

 

uaoaowuuooo moon

acuuon no ~o>oa mo. ago on usaoauacqu needs «a Amo.av p sumac
unuuop no Ho>oa oH. any as ucuoauacwua nudes a «H. coo. oxwuuu encode:
nuaoauuuuooo any no road nova: A o manorucouua :« vouuoaou own nouoooru A~o.~v s odquuo Azxav
couscous on eflsou unease uaauuueeunu nose on ooo.~ an eoeu>ue no: on. "mseoz me. .«oae. noeuuuuoeaeauu
noN a Ao~.~c a «sane. seams
ca. «no. accrue: on. uoouun
~n.~ a e
«a Aen.~c + Amoco
AN. we. Neo.n oxauuu Aqueous:
N as
«n. Ado.Hc + Assoc
Hm. «Hun. «sauna woruo
ch.~v + Adds Ana.fiv + andmv
cease. nag ubuaauluoa we. «com. oxuuua cowuauuuoa
Amo.c e Anus Aem.Hc a seems
nNH.r coau0e5nnw me. «one. oxuuua nudged
Aeo.nc r Auto Aeo.fic . «xenon» “memo
. . . no a
canon caduceus: no; «a ammo a u: an new
“he.Hc + noose Away
Aqo.o r Amzeozc nH.r mn~.i huaaaauaouaoua
oHo.r nous“: e sauce
Aon.c Aoa.c + Ammo
Ao~.o r Aomv no. -o. #0. moo. and. as
HoH.r ruse. .n
Aco.~V + Aozv Ame.c Aoo.c a “mac
aamn.r runes so. «no. we. «NH. nouns» uau
Am~.Hc 1 Anne Ae~.fic Ase.c a Anny
oo~.r anusu NH. one. no. con. luau.
Aon.v r Ammo A-.v Ae~.c + Aszc
«ma. Hausa: «a ~o.r noo.r «0.: noo.r canes
.M
An~.uc Amo.~c a . Anne
Amo.Ho + Ammo mu. moo. we. «so. once nos-H
osc.r apnea unmoaon «e «a «mo
AoH.Ho + Aqua Aoq.~v Ann.av + A>uv
hog. ozone one: OH. «ac. dc. “Ho. nouauuoaoo Hanuouaw
Aeo.~c + Amos soo.c “Ne.c + Ammo
..nmo. oneness goes: so. nae. no. sea. Ho>ou «as:
“-.v + Azmc Ano.Hc Aon.Hc + Amos
csa.i coauawuouoc uauwm he no. nnH.H cu. mac.a unease aunt «a
.m .u
uaonuuuuooo omen o~pueu~> sauna» noon uaouonuuooo noon usuauquuooo (deem cap-«nus canoes
oeuooaxu tenuous”

mecca oupawuo> one moousoa uOu n xuucooa< mum

 

 

 

 

 

 

Aoaeaum «He uses
«a nounoauo> :oasu cascades: onerous
.mv coauausv «sebum uo Souuowooua och

van rue: A .H uqm<9

 

 

'7'7

78
differently from CM, OTH, the other, or catchall occupa-
tional category, is a measure which also represents, to a
large extent, the existence of combined units.

Coefficients of multiple determination (R2=.27
and R2=.24) were generated for the two equations,
respectively. This compares favorably with the explanatory
power of the four year pooled regression for 1968-1971
obtained in the Burton-Krider study.6 In that study, an
R2 of .10 was reported. The regressions also had F scores
which were also significant at the 99% level of confidence.
Note, however, that due to missing values for some of the
variables, tests of significance were based on 205 and not
412 cases.

While the expected signs on some of the
coefficients were not predicted by the model, a number of
variables were significant at the .05 level or better in
at least one equation. These included coefficients on
labor costs (WB), hospital strikes (HOS), transportation
strikes (TRN), firefighter strikes (FIR), police strikes
(POL), sanitation strikes (SAN), street and highway strikes
(STR), other employee strikes (0TH), union support (US),
northern strikes (NO) and mediation (ME). The signs found
on FIR, NO and ME were not, however, predicted by the
theoretical model develOped in Chapter 3.

Perhaps most surprising was the finding of a
coefficient on mediation (ME) which was positive and

significant at the .01 level or better. There is no reason

79

to believe, based on this relationship alone, that
mediation contributes to a lengthening of public employee
strikes. In many cases, mediators may only become involved
in the most militant strikes. In light of their importance
from a public policy point of View, the performance of ME

and the other Li variables will be analyzed in Chapter 5.

t
As observed, prior work on strikes has suggested
that noneconomic considerations may be more important
determinants of strike duration than economic factors. The
three significant variables with unexpected signs--
firefighter strikes (FIR), north (NO), and mediation (ME)--
were all associated, however, with noneconomic vectors.
Further, among the 22 variables considered in the bargain-
ing vector (Bith the labor relations environment vector
(Pit)' the legal vector (Lit) and the control vector (Cit)'
less than half were significant in one or both of the
regressions. Given the contradiction between the positive
results for the variables in these vectors which were
anticipated, and the discovery of many insignificant
coefficients, it is apprOpriate to discuss first the
P

performance of the variables in vectors Bi and Ki

t' it' t'

The Bi Variables

t
In view of the relatively small number of strike
occurrences for many of the employee groups (see Appendix

B), the performance of the eight "employee type" variables

0‘

If

V‘
.
.

me

9:

St

80

needstx>be considered when using the 558 sample of cases.
Note that because all variables (with the exception of
CM) were included in Table 1, tests of Significance were
based on a sample of 412 cases. The sample size was
reduced as a result of merging the economic data drawn
from the census data files.

Two additional equations are thus reported in
Table 2. Variables entered in this equation were drawn
primarily from the BLS data source. The first considers

the impact on duration, S of the "employee type"

it'
variables. This is illustrative, since the predictors which
have strong correlations with the "employee type" variables

were associated with the omitted Ci the political and

tr
economic constraints vector. The second regression considers
the impact of NS, the inessentiality index, and CM. Tests
of significance for the Table 2 regressions are based on
514 and 444 observations, respectively. Relative to the
results for the 205 strike samples reported in Table 1,
this narrows the interval required for predictors to be
significant.

The "employee type" control group is represented
by instructional and noninstructional employees in education.
A positive Sign on any of the eight "employee type" variables
means that holding strikes by all other groups constant, this
group experienced longer strikes relative to the length of

strikes in education.

Generally speaking, the evidence in Tables 1 and 2

81
suggests that strikes in education are not longer than
strikes initiated by most other employee groups. For this
to be true, the signs on the "employee type" variables
should be negative. Instead, most signs are positive. The
expectation that strikes in education Should be longer than
strikes initiated by other, more essential "employee groups"
such as firefighters or the police appears to merit
reevaluation.

Significant, positive coefficients were found in
the equation seen in Table l for variables POL (police
strikes), STR (street and highway strikes), HOS (hospital
strikes), SAN (sanitation strikes), TRN (transportation
strikes), FIR (firefighter strikes) and 0TH (strikes by
other employee groups). Prior findings which supported the
prediction of a positive Sign on HOS and SAN appear to be
confirmed. According to these results, strikes by hospital
workers are somewhat longer relative to strikes which occur
in education. The Sign and significance for SAN is
comparable in direction and significance to Burton and
Krider's results.7

Conclusions based on these results alone must be
considered tentative. Only 23 strikes among sanitation
workers occurred during the period of this analysis, and a
scant 8 strikes were found among hospital workers. Without
the support of positive relationships in prior studies, one
would be prone to conclude that these findings are, in and

of themselves, inconclusive.

 

 

oo.
ea
SN.
oo.
“-.no
AN. «no.
cc
“so.o
mo.r ooo.-
Aoo.oo
on. ore.
cc
Aeo.o
No. omo.
AoH.o
oo.r omo.r
Aoo.uo
mo.r oen.-
at
Ann.oo
oo.- Hoo.r
Ao~.o
No. goo.
coco neoouuooooo

«as

o

no.1

AN.

Ho.r

nH.r

 

 

H dandy oom "mmhoz
can
aH.o
as
NN.
no.
Anm.no + Aqmv
coo. 3oH o>woowauoa
cc
Aoo.o a Aouo
owo.r aouuuaoncw
Aoo.oo r Auto
«coca cowuudvea «m
A
Aoo.o r Amneozo
0No.r Sauna: a sane:
Ann.o r Aooo
ono.i Susan
ANo.NV + Aozo
canon i sauce
AnH.Ho r Aomo
HAH.r Anus»
Ann.o r Amso
coo wound: no
.M
ucoaoauuuoo swam omboaun> neuuo>
vouuoaxm

 

coupon «one wooden venues or» long Spout nuapuauo>
.mo cowuauav oxauuo up coauoavouo one

wcqoanuxo car: A

Awo.Ho

mo.r oo~.r
1-.~o
on. now.
«a
Aoo.~o
N”. Hon.
Cd.
Aeo.o
no.1 noH.r
Aoo.No
oo. ooo.
Ga
AaH.Ho
oo.r soo.r
“No.o
no. ooo.
Aon.~o
no. onH.H

 

 

 

Ao~.Ho
no.1 soo.r
xo~.~o
oH. oNN.
cc.
Aoo.no
no. oon.
cc
Aflh.o
no.r moH.r
Aoo.o
so. ooo.
AH~.o
Ho.r nno.r
Aoo.~o
on. ono.
c c
AoH.o
Ho.r Soo.r
Aoo.Ho
so. oon.
Aoo.o
No. ”so.
AoN.o
Ho. ooo.
Aon.o
Ho.- hko.r
Aeo.o
no. «No.
.on.oo
oo. oso.

noon osooouooooo coco neouuoooooo

Aueoaao ono .oso

.N NAQ<H

 

muum

Aomo
wanna uuuoaoa

Aomo
ennui owe:

Amos
appease Sofie:

Azmo
noun-«acme: uuuau

Anus
one: ooaeoaou

Aouoo
oxauuu neunuauouau

Assoc
oxauua ouauHoa

oxen». Anzac
coauauuoaaaauu

queue. Assoc
heinous an. ocean-

as:
orange Haywood:

€98
clan». nozuo

Az<mv
exquua nouuauaeaa

Aexo
Omaha. deduce

noose Aozo
huuudqunoaaona

Aumo
one.

Auoo
ensure an.)

 

own

«on

 

o~n~«ua> nouoo>

€322

83

The only coefficient among the eight "employee
type" variables which was Significant in both Table l and
Table 2 was TRN (transportation strike). This was
surprising, since an a priori prediction was not possible
for this variable. With only 14 strikes in the sample, this
finding could be a function of sampling error. It is
interesting, however, that TRN is significant at the 99%
level of confidence in both regressions. Bus drivers and
other transit workers may be more inclined to wait longer
for a settlement which compares favorably with the earnings
obtained by their private sector counterparts. This finding
may also be influenced by the impact of the New York City
transit strikes during this period.

It is interesting that the positive and Significant
coefficients on FIR (firefighter strike), SAN (sanitation
strike), POL (police strike), STR (street and highway strike)
and OTH (other strike) became insignificant in Table 2.
Signs on STR and POL were even negative. One possible
explanation lies in the relationship seen between these
variables and two of the variables associated with the Cit
vector, SL (fiscal slack) and TB (the local tax burden
index). The correlations between FIR, SAN, POL, OTH, STR
and SL are positive and significant (r=.l7, r=.20, r=.ll,
r=.40, and r=.l9 respectively). On the other hand,
correlations between these variables and TB are negative
(r=-.23, r=-.08, r=-.07, r=-.1l, and r=.02) and, in many

cases, significant.

In

D—l

84

Since many of the "employee type" variables are
inversely correlated with TB, the omission of TB from the
Table 2 regressions means that these variables are probably
picking up some of its variation. Thus, the positive
association between "employee type" variables and strike
duration is probably being reduced when TB is excluded.

Note that the strong positive correlations between the
"employee type" variables and Slack (SL) also infers that
there is a strong negative correlation between the omitted
"employee type" variable (education strikes) and SL. This
correlation was r=-.7l. One inference that can be drawn
from this is that noneducation public employees may have'

an incentive to target strike actions against employers
which have a greater capacity to make concessions. If the
strike decision is a rationally determined process, then one
would expect to see local labor organizations strike
jurisdictions which extract a relatively low level of

local taxes versus total revenues, and who have a relatively
high degree of "slack." As discussed in Chapter 3, the SL
variable (the prOportion of personnel expenditures as a
component of total revenues received) is the variable in this
study which most closely proxies the employer's ability to
pay.

The relatively strong association between the slack
measure and the noneducation employee type variables may be
partly explained by the fact that noneducation union locals

take relatively independent action with respect to the

(A;

(I)

‘11

III

In

85
decision to strike. The National Education Association
(NBA), the sponsor of the largest number of public employee
strikes during this period (256 out of 558), on the other
hand, has conducted a more centrally coordinated bargaining
strategy. In states such as Michigan, the NEA establishes
a contract pattern that is used as a standard against which
to evaluate the progress of negotiations with all school
districts. Districts with a greater capacity to make
concessions, with a higher degree of Slack (SL) and a lower
tax burden (TB), are more likely to settle before strike
action is necessary. Districts with relatively less capacity
to make concessions (or having less slack and a higher tax
burden) may thus be more likely to have strikes.

An alternative explanation for the strong relation-
ship between the "employee type" variables and both the
slack and the tax burden variables is as follows: the
prOportion of nonlabor costs to total revenues across all
school districts which both take and avoid strikes is
lower than for jurisdictions employing noneducation
personnel. Further analysis, outside the scope of this
study, would be necessary to verify this alternative
explanation.

The coefficient on WEL in Table l is positive, but
not significant at the .10 level or better. Its sign in
Table 2 is reversed. It also has a low beta weight (-.01).
This seems to confirm the expectation of an ambiguous Sign

on this variable. With only 13 welfare strikes (see

86

Appendix B), this finding is contaminated with considerable
sampling error.

In contrast to confirmatory findings for Signs
on SAN, OTH, HOS and WEL, the negative sign on NS, the
inessentiality index, was unexpected. Recall that less
essential employee groups were assigned the higher number.

A positive Sign had therefore been predicted for the NS
variable. In the Table 1 results, NS also has a relatively
large absolute beta (-.13) compared to the betas seen on
other variables in this same equation. Findings for the
"employee type" variables in the first regression are also
consistent with this result, since strikes in education

were categorized in Chapter 3 as the least essential employee
group, NS=3.

The construction of the dependent strike duration
measure (Sit) was weighted in favor of the more essential
employee groups (firefighters, police, and hospital workers)
Since these groups provided services to the community on a
continuous basis. The duration of the strike for these
groups was defined as the calender duration of the strike.

To a degree, this would have biased the finding of an inverse
relationship between NS and Sit'

The provision in many state laws for compulsory
arbitration of firefighter and police strikes has apparently
reduced their frequency. The negative Sign on NS suggests,
however, that once firefighter and police strikes are

initiated, there is no reason to believe that their duration

C)

U)

r)

:4

'Y1

(h

In

In

K)

87

will necessarily be Shorter than strikes by teachers or
other, less essential employee types.

It is not unreasonable necessarily to find longer
strikes among more essential employee groups. Employers
would have a greater incentive to identify a readily
available means of providing substitute services if an
essential employee group has struck. Strike replacements,
or even the enhanced credibility of the threat of hiring
replacements, could lengthen a strike as labor and manage-
ment become less amenable to compromise in the process.

The performance of the Size variable (82) was
positive, as predicted, but not significant. The failure
of Size to influence strike duration significantly has
also been found in four previous studies. Two of these
studies constituted an examination of strikes in the private
sector. The other two studies used non-education public
employees as the unit of analysis.

Britt and Galle found that the zero order correla-
tion between strike extensivity (or duration) and average
union size was not significant across eighteen broad
industrial categories.8 This analysis included strikes in
both manufacturing and nonmanufacturing industries. In an
extension of this work, Stern found negative and insignifi-
cant Signs on plant Size and average union size variables
which were considered as predictors of strike duration.9

In their study of public employee strikes discussed

in Chapter 2, Burton and Krider found that a proxy for work

wk
“w.

B

E
nu;

uri
9*

'r

C.

I

NIL

Cs

F\

at

88

group Size, the percentage of employment in Cities over
50,000, was both negative and insignificant in a four year
pooled regression for 1968-1971.10 In a replication of the
Burton and Krider study that considered a more recent
period (1974-1975), a negative and also insignificant sign
was found on a size variable.ll
IS Size really unrelated to strike duration?
Since our hypothesis suggests there should be a relation-
ship, it is possible that 82 is not linearly, but
curvilinearly related. Because of their enhanced bargaining
power, very large units may actually have shorter strikes.
Certainly strike replacements would be difficult to
recruit. One would also expect to find greater SOphisti-
cation in the bargaining skills of both labor and manage-
ment. A respecification of the basic model thus considers

2

SZ , in addition to 82, as predictors of Si These

t'
results are presented in two regressions found in Table 3.
The first regression uses the same model specification as
seen in Table l (the 412 sample). The second regression
is Similar to the model specification used in the Table 2
regressions (the 558 sample).

Since NS represents a common characteristic
of all employee groups, it is entered in the Table 3
regressions in lieu of the eight "employee type" variables.
This eliminates much of the multicolinearity among the

predictors which was evident in the regressions reported

in Table 1 and as evident in the correlation matrix reported

 

as
ANH.oo
oN. ass.
is
AoN.o
no.r noo.r
ASN.oo
me. one.
a.
Aoo.o
Ho. oHo.
AmN.o
Ho.r moo.r
ANN.No
n~.r omn.r
ca
AoN.o
no.r noo.r
AoN.o
No. ooo.
noon uSOHuHuwooo

oHanm can ask

 

«N.

«0.1

0H.

no.

No.1

nH.I

no.1

No.

 

doom

nON
«cmo.N
0N.

Hm.

AeN.no +
«coco

Ann.v A

noH.t

Aoo.No

Ame.
«e

AHN.V r
00H.

AoN.o r
HHH.I

Aoo.Hv +
oom.l
i

ASN.V r
nNH.I

Aoo.o r
ooo.

 

uCOHuHuuooo cmHm

oHanm NHe 0:8

.H 0Hao9 00m

Aaoo
:iH o>HunH5hoa

Aooo
SOHuuaanaH

Auto
coHuoHvoa
Anacozo

woucHa use rune:

Aooo
canon

nozo
capo:

onv
Hons»

szo
unucua

"mmhoz

a
9.x

 

OHAoHua>

Heuoo>

 

Aom.v

«0.1 omN.I
Aoc.Nv
HH. HON.
Gt
Aon.Nv
HH. MHn.
i¢
3.3
no.1 ooo.i
as:
no.1 oNH.I
5.3
cc. nHN.
Cd
Ann.nv
on.l Noooo.l
O.
5.2
on. NHn.
G‘
as:
no. NOH.H

noon uCOHuHuuuoo
OHnflflm Dnn 0:h

on«. you noHuaouuaooaa unoaHHH>uau a

yupHmcou :OHsa «mm new ouHsoou coHnnouwox

 

.n mqm<h

Amo.o
oo.r ooo.r
aNo.Ho
NH. ANN.
C
Ano.oo
nu. ooo.
c
Aon.o
oo.r NN~.r
Aoo.Ho
NH.r oNN.r
AoH.No
or..- 826.:
cc
Aoo.No
No. ooN.
cc
Aon.o
no. ono.
Ano.~o
so. NNo.
Aon.o
oo.r ooo.r
Ano.Ho
NN. ooo.
C
ANN.Ho
oo. Noo.
Ao~.o
so. neo.
.mo.o
No. Noo.~

 

«poo econooooooo
«fiasco Noe use

 

Aomo
oaoaH uHuocoa

Aomo
ooaaH one:

Aooo
uuoansa cows:

Azho
:oHuoHuomoa uauHu

Ronda Amzo
huHHqHucoouoeH

Azuv
ouHas vocHnloo

A nmv
N Nona.

ANoo
ouHa

Anyv
oovusp no»

Aamv
xuaH-

Anzo
uuHa:

Anus
uoou nonaH

.v n>Nv

cooHuadaoo HaaHOuaH

:me

Ammo
HoooH one:

Auoo
encore one:

am

 

OHnoHua>

nauoo>

89

nr

In

In

90

in Appendix C.
The evidence reported in Table 3 lends support
to this respecification. Each of the regressions explain

approximately one-fourth of the variation in Si 82 is a

t'
significant, positive predictor of strike duration when

one controls for the possibility that larger units, because
of their enhanced bargaining power, have shorter strikes.
This is seen by the negative and significant coefficients
on SZ2 in both regressions. The coefficients on SZ

and 3Z2 are both significant at the .01 level. Their beta
weights also are comparable in magnitude. Previous studies
which demonstrated the absence of a significant relation-
ship between Size and strike duration may have failed to
identify a relationship because of the assumption that the
relationship was linear.12

The magnitude of the beta coefficient seen on WB
(labor cost) in Table l is reduced considerably in the
Stable 3 regression (from .46 to .21). Some of the variation
eaxplained by WB variable in Table l is apparently now being
czaptured by the size variables (82 and 822). WB, however,
remains Significant .

Turning to the final variable in the Bit vector,
significant, positive coefficients can be seen on combined
units variable (CM) in Table 2 and in the Table 3
regression. As expected, it appears to take longer to
13

settle strikes initiated by more heterogeneous units.

There are more issues to be settled. Compromises made at

91
the bargaining table may also take longer to sell to the
union membership. In addition, note that controlling for
SZ and 8Z2, the strength of the positive relationship
between combined units (CM) and strike duration does not

diminish. It has a beta in both equations of .09.

The Pit Variables

As expected, US, union support for the work
stOppage, has a positive Sign in the six regressions
reported in Tables 1 through 3. The consistency of this
performance is strengthened by a significant coefficient
at the .05 level or better in five of the six regressions
reported. The absence of international or national union
support for a work stoppage does apparently mitigate
determination among a local union membership to obtain the
desired settlement. The employer may also perceive the
absence of union support for the strike as sufficient
justification for "holding fast." Little or no movement
by an employer during negotiations, dissipating interest
among the local union membership, and the ineligibility to
:receive strike benefits may all be factors which contribute
to Shorter strikes, other considerations held constant.

The wage issue variable (EC) also exhibited a
positive performance, as expected. This variable is more
Significant, however, in regressions which omit Cit Variables.

EC represents the fact that workers are striking primarily

92

over wage issues. The reason that economic considerations
became an issue may be caused by the difficulty in
identifying the revenues needed to fund the concessions
which were demanded. A more salient, stable predictor of
duration may be the financial condition of the employer
taking the strike. The combined effect between EC and two
of the variables in the Cit vector (Slack and tax burden)
was insignificant.l4

The performance of SB, the supplementary fringe
benefits variable, was also consistent across model
specifications, but its sign was unexpected. AS a variable
which represents a Specific type of economic concern, SB
was expected to be positive. These results suggest that
a union membership may be willing to strike longer when the
issue relates to the magnitude of the weekly (or monthly)
paycheck than when it relates to other, supplemental.
benefits. Employers may also be more willing to compromise
when diSputeS concern fringe benefits.

This is because cost savings accrue when group
benefit programs for health, dental or life insurance are
established. As a general rule, the larger the employee
group serviced with a group fringe benefit, the greater is
the economies of scale. The strength of the relationship
is relatively weak, however, and the number of observations
(see Appendix B) makes interpretation for this variable
also subject to considerable qualification.

The last variable in the Pi vector, FN, the

t

93

"first negotiation" variable, also exhibited a consistent
performance across the six equations reported in Tables 1
through 3. Its negative signs, however, were insignificant.
Further, a positive sign had been predicted. This suggests
that during the recognition campaign, many of the more
difficult issues may have been resolved. As a general rule,
both employer and union appear to be no more (or less)
willing than better established units to execute a first
agreement and proceed with administering the new contract.

Given the insignificance of FN, there is no reason
to think that strikes occurring during first negotiation
rounds will necessarily be Shorter. This evidence, however,
does question the general prOposition that strikes which
occur during the negotiation of first agreements Should be
longer relative to the duration of strikes occurring during

subsequent negotiation rounds.

The Kit Variables

Among the five variables included in this vector,
only one, North (NO), was significant and its Sign not
expected. Holding strikes in southern states constant, the
evidence reported here suggests that strikes in northern
states are Shorter than strikes occurring in other regions
of the country. Contrary to expectations, this is consistent
with the prOposition that bargaining units with a longer

(and therefore more mature bargaining history) may be able

 

rim

94
to resolve their differences more expeditiously. This
qualifies the conclusion based on the results seen on FN
that the length of the bargaining relationship is not a
relevant factor.

WR, the variable which represents the occurrence
of a strike during the winter quarter, was positive in the
six equations presented, but it never attained significance.
In preliminary analyses of the data, season appeared to be
a more relevant predictor of the type of employee most
likely to initiate a strike. For instance, a positive
correlation was found between street and highway strikes
(STR) and the winter (WR) (r=.l7) as well as between
education strikes and fall (r=.38). Organized workers
would be expected to prefer strikes during the season that
they think would cause the most inconvenient disruption in
service delivery. A more exhaustive examination of this
possibility is outside the SCOpe of this study. It is
significant to suggest here that no significant or
predictable relationship was found between WR and duration.

The sign on SO (South) was also consistent. It
was negative in the Six regressions reported, but never
attained significance. This is consistent with a priori
expectations, but its contribution to the explanatory
power of the model is miniscule. Not a large number of
strikes actually occurred in south during 1977-1978, so
little weight can be given to this result.

Consistent with the insignificant performance of

rf

tc

Ce

95

WR, the interaction term NO*WR (strikes occurring during
winter in the north), is negative twice and positive four
times in the regressions reported. This variable is also
never Significant. Apparently there are other, more salient
factors which affect the duration of public employee strikes
than season.

Finally, the signs of rural (RU) were negative,
as predicted, but never became Significant. AS one would
expect, strikes in rural areas were slightly shorter than
strikes in more pOpulated areas, but there is no evidence
that this effect is significant. Organized workers in rural
areas who have Opted for strike action may simply be more
militant than has been acknowledged in previous studies.
While this variable is a weak predictor of duration, RU
could be a more salient determinant of strike frequency.

It is also interesting that the magnitudes of the
beta weights on RU drop slightly when it is entered in
regressions having a curvilinear Specification for the
size (822). In the Table 3 regressions, it has a Slightly
lower significance level than in Table l or Table 2.
Controlling for size, it is not surprising that RU becomes
less significant. One justification for the prediction of
its negative Sign was the fact that rural communities tend
to be smaller and more highly interconnected. This is being

captured, to a degree, by 82 and 8Z2.

96

The C. Variables
It

The Insignificance of the Number of Units Variable

Perhaps the most unexpected result among the four

variables associated with the Ci vector was the negative

t
Sign on NU, the number of bargaining units. Recall that a
positive sign had been predicted. This expectation was
based on the assumption that the employer should be the
primary actor affected by the number of units that are
recognized to bargain collectively.

A greater number of units may, on the other hand,
be also a reflection of the enhanced bargaining power of
organized labor in the jurisdiction taking the strike. NU
has a strong negative correlation with RE (r=-.25), the
ratio of comparable employee earnings with the earnings of
the striking bargaining unit. Earnings levels in jurisdic-
tions with a relatively large number of bargaining units
appear to be higher relative to the earnings of their
comparable counterparts. With a greater comparative
earnings advantage, there may be less incentive for labor
to conduct long strikes which may ultimately yield only
modest concessions. The fact that NU also represents the
degree of bargaining leverage held by labor has apparently
moderated the positive Sign expected on NU.

This analysis has thus far considered strikes

initiated in all types of jurisdictions. Many jurisdictions,

97
however, have few bargaining units. This is especially
true of school districts and Special districts.

A second possible explanation for the insignifi-
cance of NU may be that no relationship was found because
of a failure to control for differences in the type of
governments taking a strike. The effect that the number of
units has on the duration of strikes in school districts
may be quite different from the effect found in counties
and cities. Accordingly, separate regressions were run
which considered the duration of strikes occurring in

I I I 15
C1t1es, town and count1es.

In this analysis, the mean
value of NU is increased.

This analysis is important for comparative
purposes. As discussed in Chapter 2, much of the previous
work on the duration of public employee strikes has selected
non-education employees as the primary unit of analysis. A
consideration of the performance of the variables used in

this study can then be compared with the results obtained

in previous studies.

The Duration of Strikes in Towns, Counties and Cities

The results of this analysis are reported in
Table 4. In the left column regression, all variables which
were found in previous analyses to be insignificant were
omitted. In the right column regression, all predictors

which were subsequently found to be insignificant were

We
Ca
Wh.
re]
Co;
dis

Var

98

omitted. "Employee type" variables, however, were also
included in the second regression.

These results Show no appreciable change in the
significance of the number of units (NU). Its sign became
positive, but its beta weight was .00. This evidence,
combined with the results for NU seen in Tables 1 and 3,
suggests that the number of bargaining units may not have
a direct effect on the duration of public employee strikes.
There probably are, however, important indirect effects.

Although NU contributes little to the explanatory
power of the model, these regressions report higher
coefficients of multiple determination than were seen in
Tables 1, 2 or 3. The first regression reports an R2 of
.32, while the second has an R2 of .31.

In comparison to the explanatory power of the
Burton-Krider regression results (r2=.10), a study which
also considered non-education strikes, the explanatory power
of the variables in this regression, and the relative
consistency of their performance, were better.16 The
coefficients of multiple determination reported in Table 4
were, however, comparable to a study which recently repli-
cated the Burton-Krider model by analyzing 1974-1975 strikes.
When using days idle as the dependent variable, this

replication study reported an R2 of .28.1'7

The most
consistent finding across all three studies was the
discovery of a negative and insignificant sign on the size

variable (SZ). SZ2 was also insignificant. In comparison

 

 

 

 

 

 

 

 

 

AoN.o A Aomso
no. AoH. . oxHuu- ouauHoa
Amn.Nc A Azuav
nN. cacao.H oxHuua noHuauuoa-cauu
Hoo.o A «xenon Heamo
oo. ooH. sasono so. noose.
Hoo.Ho + Hoozo
nH. new. «Juno. Houaano:
AAn.Ho + Amoco
0H. «An. oxuuua noruo
Hoo.Ho + Az<mo
H poaH oom "mohoz mH. Nam. oxHuua uOHuuuHuoa
Aoo.Hv A AHomv
NH. oAn. oxwuua ouHHoa
oHH oHH e
Ano.o r Annoy
o¢.n oe.n m ao. oNe. oxuuu- unurmHuouHu
Hm. Nm. Nm AoH.Ho + novaH Amzv
NH.i ooH.r AuHHaHuSUaaoaH
mm. Am. a
AHH.o . ANNoo
no.r ooo.r Nona-
HoH.No Hoo.no + Anew AoH.o + Aumo
. . . . 3a o> on o .i .r
AN eeooA Hm canon H H Hey we moo ouHa «on
ANN.HV N HoHo
HH.: can.i SOHuueoneH AoH.HV A Anny
NH.i OON.r nevus; «cu
Aoo.Ho + Auto
oH. ANN. :OHuaHvoa «a AoA.No ANH.NV A AHmo
H oN acmMA N HN ceNAH N soaHa
Ano.Ho + Aozo Ano.o + Apzv
nH.r HAn.r :uuoa no co. coo. nude:
.u
AAo.oo AAn.nV A Anny
Ao<.o + Aomv no. nNo. An. oNo. uooo nonuH N
co. Hoo. oaouH own: as ea «.0
Aqo.Nv AAm.No + one Amos Aoo.o AnN.Hv + Away
0 o O 0 g o O I I
AH eecNA NN aeHno on an SOH «mm no noA HH was H unease one: «mu
anon mnoHOHuuooo soon ueoHOHuuoou smHm oHoaHua> nouoo> noon uSOHuHuuooo noon ucoHuHuuooo mem oHnaHuo> neuoo>
oouuoaxm peacock»

uoHunsoo van n:30u .ooHuHu cH wcHuuaouo
noxHuua up SOHuouzv or» up coHuquoun 05H .q man<h

99

100

to the mean Size of the 558 bargaining units which initiated
a strike, the mean size of bargaining units which struck
city, town, and county jurisdictions was reduced considerably
(from 406 employees to 254 employees). By deleting from
consideration teacher strikes, previous studies may also

have eliminated from consideration the effect of the larger
units. Having contrasted the results obtained when excluding
the duration of strikes in school districts and special
districts, the results obtained on the other variables as

reported in Tables 1, 2, 3 and 4 are presented below.

The Performance of Other Variables in the Cit Vector

In both regressions seen in Table 4, SL, the slack
variable, attained significance at the .05 level or better.
This is contrasted to the insignificant, but positive signs
found in previous regressions reported in Tables 1 and 3.
Although never significant, SL had relatively strong betas
(.08, .12, .14, and .21) in prior equations. Moreover,
this finding is consistent with the argument that while
it may be strategically advantageous for labor organizations
to strike employers who have a greater capacity to award
concessions, this may result in longer strikes. The greater
the component of non-labor costs to total revenues, the
greater is the Size of the economic pie over which labor
and management have to haggle. With the prospect of gaining

(or losing) more as a result of the final contract, there

101
should be a greater incentive to wait the other side out.

TB, the local tax burden measure, was positive,
but not significant in the 3 regressions found in Tables 1
and 3. It was negative in the Table 4 regression. The
negotiation of contracts during a strike may thus be more
directly affected by public pressure when they occur in
cities, towns and counties than in school districts. This
is because school districts serve a dispersed, more
heterogeneous voting public. This offers one explanation
for the finding of a negative Sign on TB in Table 4.

The finding of insignificant coefficients on TB
in all four regressions can be explained by the possibility
that two countervailing effects are present. The strike
may tend to be longer because employers are less willing to
compromise if concessions can be made only at the expense
of increased tax assessments for an already heavily taxed
public. Labor, on the other hand, may recognize public
resistance to accepting an increased tax burden and thereby
become less adamant at the bargaining table. The theoretical
model in Chapter 3 suggested that the Sign on TB was
indeterminant. Empirically, this appears to be verified.

The final variable in this vector, Labor Cost
(WB), exhibited a strong and significant positive performance
across all regressions seen in Tables 1 through 4. It was
significant at the .05 level or better in four out of five
regressions. The WB beta ranged from .21 to a value of

.46. Viewed as a constraint on the employer's ability to

102

make concessions, the prOportion of the labor costs of the
strikers to total labor costs is thus associated with longer
strikes. When labor costs are high, the employer has good
reason to be more hesitant to compromise. Revenues
necessary to support higher wages would certainly be
expected to be more difficult to identify when large

numbers of workers are involved.

There is also an incentive to take a long strike
because of the greater net savings to the employer.
PrOperty tax and intergovernmental transfer revenues are
unaffected by public employee strikes, but expenditures are
temporarily interrupted. The resultant savings for an
employer when large units strike can be formidable.

Strikers who comprise a relatively small
component of labor costs may also be in a more advantageous
bargaining position, since the employer can more readily
make concessions without also accepting a difficult
financial burden. According to these results, the
magnitude of these labor costs appear to be a more

important factor than the type of employee group who struck.

The E. Variables
1t

All three variables associated with the Eit
vector, CE (the change in mean earnings of other, comparable
employee groups), RE (the ratio of the earning of comparable

employee groups to the earnings of the strikers), and EV

103

(the change in the mean earnings of all other employee
groups working for the same employer relative to the mean
change in earnings of the striking bargaining unit), had
positive, but insignificant signs in the regressions
presented in Tables 1 through 4. In the initial construction
of these measures, they were all expected to represent the
level of expectations held by the bargaining unit which
elected to strike. Thus, the fact that none of these
coefficients were Significant may be due to the fact that
when entered in regressions together, they are competing
with one another to explain the same variation in Sit' the
dependent variable. Accordingly, each of these three
variables are entered separately in a regression with
other predictors found to be Significant in the previous
analysis. These results are presented in Tables 5 and 6.

Table 5 presents regression results when CE and
EV are introduced separately into equations with the other
significant predictors. Table 6 presents the regression
coefficients when introducing RE (and not CE and EV) and
when omitting all three Eit variables.

In none of these regressions do the Significance
levels, the magnitude of the coefficients, or the Size of
the beta weights on the other predictors change as a
result of the introduction of CE, RE or EV. In addition,
the eXplanatory power of the regression which omits the
three Ei variables in Table 6 has an R2 which is comparable

t
in magnitude to the coefficients of determination found in

104
the other regressions.

The EV coefficient, while still insignificant at
the .10 level, exhibits the strongest performance of all
three variables. Employees would be expected to be well
acquainted with earnings concessions awarded to other
workers employed by the jurisdiction. This variable
unexpectedly exhibited considerable variation. This
suggests that a bargaining unit which has elected to strike
may have received no pay adjustments for a considerable
period before the actual strike. Unable to settle on a
contract, the employer may be hesitant to award even
minimal increases because of the concern that it might
diffuse his bargaining leverage.

Thus, rather than capturing merely the extent
to which an employee group may have experienced a loss or
gain in earnings relative to other employee groups, this
variable also picks up the fact that earnings concessions
may not have been awarded to the employees who struck for
a number of months before the strike. With a settlement
held in abeyance as negotiators become more and more
frustrated, the level of discontent among the bargaining
unit membership would be expected to become intensified.

Public employee strikes do not necessarily
occur shortly after the termination of an existing contract
(if one exists). Employees may be willing to linger on
for months working under the conditions set forth in an

exPired contract. The delay between the eXpiration of the

Vector

TABLE 5.

The prediction of strike duration when considering

the independent effects of CE and EV

Variable

 

it

it

it

it

it

NOTES:

wage change
(CE)

wage level
(EV)

labor cost
(NB)

slack

(SL)

size

(32)

size2 2
(82 )

inessentiality
(NS)

union support
(US)

wage issue
(EC)

mediation

(ME)

permissive law
(PL)

See Table 1.

 

 

105

 

Expected
Sign Coefficient Beta Coefficient Bets
.925 .06
+ (1.12)
, .017 .09
+ (1.64)
** at
.005 .19 .006 .21
? (2.31) (2.70)
a as
1.013 .14 1.191 .17
? (1.91) (2.39)
at at
.311 .39 .323 .40
+ (2.79) (3.10)
at **
-.00002 -.37 -.00002 -.40
- (2.81) (3.09)
at a
-.250 -.14 -.213 -.12
+ (2.10) (1.85)
** at
.539 .12 .537 .12
+ (2.11) (2.23)
at an
.299 .13 .284 .12
+ (2.39) (2.42)
** *t
.414 .18 .424 .18
- (3.13) (3.39)
at **
.627 .25 .597 .24
+ (4.30) (4.27)
.48 .47
.23 .23
8.11 8.93
283 318

TABLE 6. The independent effect of RE

 

 

 

 

Expected
Vector Variable Sign Coefficient Beta Coefficient Beta
Eit
internal .039 .01
(RE) comparison + (.14)
C.
tt ** **
labor cost .006 .20 .006 .20
(NB) ? (2.62) (2.63)
** **
slack 1.147 .16 1.143 .16
(SL) ? (2.30) (2.30)
B.
gt ** **
size .327 .41 .324 .40
(SZ) + (3.06) (3.10)
2 ** **
size 2 -.00002 -.40 -.00002 -.40
(SZ ) - (3.08) (3.09)
** **
inessentiality -.243 -.14 —.242 -.14
(NS) + (2.16) (2.16)
P.
1t ** **
union support .541 .12 .538 .12
(US) + (2.23) (2.23)
** **
wage issue .282 .12 .281 .12
(EC) + (2.38) (2.38)
L.
7ft ** **
mediation .422 .18 .421 .18
(ME) - (3.36) (3.37)
** **
permissive law .619 .25 .623 .25
(PL) + (4.43) (4.52)
R 47 47
R2 .22 .22
F 8.77 9.77
n 318 318

NOTES: See Table l

106

107

contract and the initiation of the strike is not uncommon
among teachers who have been known to continue working
over one year after the expiration of a contract before

striking. The EV variable may thus have been more

significant because of the greater discontent that would

have been eXpected to build up among teachers as well

as other employee groups who worked without a contract

for a number of months. In a separate analysis of teacher

strikes not reported in this study, EV was both positive

and significant.
While RE has consistently been positive in all

regressions reported in this chapter, it has also always

Ibeen insignificant. Based on these findings, employers

Inay be less concerned about the prOSpective length of a
.strike if the earnings previously paid to the strikers

liave fallen considerably behind the earnings awarded to

<>ther, comparable employees. Similarly, there is no

.zreason, based on these results, for employers paying
Ik>etter than average wages to be Optimistic that the duration
(Duf a strike, if one occurs, will necessarily be shorter.
(Either factors apparently are more important once the strike

has begun .
The performance of these variables is also

affected by the fact that only the most militant employee

groups have been selected for examination. This reduced

the variation in the expectation variables (Bit) relative

t1<>> the variation for the same variables when considering

108
all bargaining units having recognition. As a direct
result of this selection bias, fewer causal relationships
are apparent from an examination of the correlation
matrix found in Appendix C.

Moreover, these expectation measures are
probably better predictors of strike frequency than of
duration. It is interesting that a similar conclusion
was reached in a replication of the Burton-Krider strike
model18 for the years 1974-1975. In this studylg, an
alternative specification for an expectation measure, the
ratio of change in private sector earnings to the change
in public sector earnings, was found to be a more signifi-

cant predictor of strike frequency and strike breadth

than of strike duration.

Some Concluding Considerations

The summary results seen in the Table 6
regressions suggest there is still much to be learned
about the determinants of strike duration. Note that
among the nine variables considered in the second
regression, two (WB and SL) had signs which a priori were
indeterminant. One variable (822) was not included in
the develOpment of the theoretical model. Two of the
variables found to be significant (ME and NS) had signs

which were predicted to lie in the Opposite direction.

109

The vector, K. was introduced so that factors

zt'
not captured by the other vectors could be controlled.
None Of these variables were found to be significant enough
to be included in the Table 5, 6 or 7 regressions. Some
expected findings, but many unexpected results, were
generated. This is not a negative signal, since it may
be just as important in this stage Of the early develOpment
of public employee strike theory to identify irrelevant
factors as it is to identify significant ones. After
many Of the less important predictors Of strike duration
were omitted in Tables 5, 6 and 7, it is interesting that
the explanatory power Of these regressions was not that
different from the coefficients Of determination found
in Table 1.

On the positive side, the significance Of all
of the regressions was high, as seen by the F scores.
Many Of the predictors behaved consistently, regardless
Of the model Specification. The explanatory power Of the
regressions ranged from one—fourth to one-third Of the
variation in the dependent variable.

From a public policy point Of view, the factors
B

identified by the variables in the Ci and Pi

t’ it' t
vectors can not be affected by public policy once a strike

has begun. The variables in the Li vector, however, do

t
represent Options that are available to employers in the
event a strike occurs. The purpose Of Chapter 5 is to

examine further the implications Of the results reported

110

in Tables 1 through 4 on these public policy variables.

111

Footnotes (Chapter 4)

Recognition strikes, strikes occurring during the term

of the agreement, and wildcat strikes accounted for only
about 15% Of all reported strike occurrences in the public
sector. These strikes are excluded from the analysis since
they are much more likely to be symbolic protests to the
public rather than a result of an interest in Obtaining
better concessions. See, for instance, Ronald W. Glass,
"Work StOppages and Teachers: History and Prospect,"
Monthly Labor Review, Vol. 90, NO. 8 (August, 1967),

pp. 43446.

 

2

Census Officials have reported considerable difficulty in
identifying whether bargaining units at the state level Of
government are organized across occupational lines, or
cut across departmental structures.

3

Eighteen Of the 558 strikes were not continuous events.
In some instances workers struck for a day or more,
decided to return to their jobs, and then soon thereafter
initiated subsequent strike actions. This was defined as
a single strike event. All days Of idleness for the same
bargaining unit were summed together to reflect the total
work days lost. Strikes occurring during the final two
and one-half months Of 1978 were deleted because matched
jurisdiction identification Census numbers and BLS case
numbers were not available.

Respondents are asked to indicate when the stOppage began
and when the employees "returned to work." In calculating
the number of work days idled, it is impossible to
determine how many days falling within the strike period
were holidays, or simply "non-work" days. In determining
the number Of work days idled, the BLS thus assumes that
local jurisdictions Observe the same holidays as federal
government employees. These are New Year's Day (January
lst), Washington's Birthday (3rd Monday in February),
Memorial Day (4th Monday of May), July 4th, Labor Day

(lst Monday of September), Columbus Day (2nd Monday of
October), Veteran's Day (November 11th), Thanksgiving (4th
Thursday Of November) and December 25th.

This sample of jurisdictions includes all cities which
have a pOpulation Of over 25,000, and all counties with a
pOpulation over 50,000. Cities and counties with smaller
pOpulations were selected on a random basis stratified by
geographical area.

112

6

Robert C. Rodgers, "A Replication of the Burton-Krider
Model Of Public Employee Strike Activity," Industrial
Relations Research Association Series (1981), pp. 241-251.

 

7
For a discussion Of these studies, see Chapter 3.

8
David Britt and Omer Galle, "Industrial Conflict and

Unionization," American Sociological Review, Vol. 37
(February, 1972), pp. 53-55.

9
Robert N. Stern, "Toward an Empirical Merger:

Sociological and Economic Conceptions Of Strike Activity,"
28th Annual IRRA Meeting (1975), PP. 58-61.

10
John F. Burton and Charles E. Krider, "The Incidence Of
Strikes in Public Employment," in Daniel S. Hamermesh,
Labor in the Public and Nonprofit Sectors (Princeton,
N.J.: Princeton University Press, 1975).

11
Robert C. Rodgers, Op. cit.

 

12

Depending on the strike deflator when states were used
as the primary unit of analysis, a day's idle dependent
strike measure could also have been adjusted
disprOportionately when the day's idle variable was
divided by the size deflator selected.

13
Although it is impossible to identify from the information
obtained on the work stOppage questionnaire whether
units were combinations Of occupational groups, the BLS
codes onto machine readable data files information
pertaining to the occupational categories Of the strikers.
This is drawn from a review Of the press reports on
the strike by the BLS coding clerks.

14
The interaction terms EC*SL and EC*WB were positive, as
would be expected, but their coefficients were not
significant at the 99% level Of significance. Their
beta weights were .03 and .08 respectively.

113

15

An alternative analytic technique was also considered.
Dummy variables were constructed tO represent each Of

the four major jurisdiction types. With school districts
as the comparison group, the results were consistent with
the regression reported in Table 4. Consistent also with
the findings on "employee type" variables was an inverse
relationship between the duration of strikes in school
districts relative to the duration Of strikes in other
governmental units.

John Burton and Charles Krider, Op. cit.

Robert C. Rodgers, Op. Cit.

 

John Burton and Charles Krider, Op. cit.

 

Robert C. Rodgers, Op. cit.

 

Chapter 5

Analysis Of Public Policy Variables

Two Of the variables in the public policy vector,

L had signs which were consistent with the theoretical

it’
model. These were the injunction (IJ) and permissive law
(PL). The third variable, mediation (ME), was significant
but its positive sign was not expected. This chapter

further analyzes these results by considering first the

performance Of the IJ and PL variables.

DO Court Injunctions Enhance or

Deter the Duration of Strikes?

The injunction variable IJ (representing the
issuance Of an injunction sometime during the course of the
strike) had negative coefficients in the Chapter 4
regressions, but it never attained significance. Injunctions
sought by employers, presumably, do force strikers back to
their jobs earlier than otherwise might have been the case.
Initial discussion for this variable suggested that there
are reasons for believing that injunctions might both
reduce and enhance strike duration. This is why the sign
on IJ was predicted to be indeterminant.

NO information is available on the duration (if
any) of the strike before the injunction was issued. An

heuristic assumption for purposes of this analysis was that

114

115
employers usually seek injunctive relief shortly after a
strike begins. It is important to note, however, that this
assumption may or may not be valid. Employers who face
unusually long strikes may choose to pursue a court remedy
only after all possibilities for a settlement have been
explored. Conceivably, this could take two or more weeks.
This delay would moderate the magnitude and significance
Of the sign on IJ.

Unfortunately, data limitations preclude the
possibility of disentangling these effects. With this
caveat in mind, however, it is worthwhile to explore
three additional questions. The effect of the court
injunction may vary depending on the presence Of compulsory
penalty laws, the size of the striking unit, or the type Of
employee involved. These additional considerations are

discussed below.

DO State Compulsory Penalty Laws Make a Difference?

Some states have compulsory penalties for employees
who strike illegally. The employer, however, must first
request from a judge a cease and desist injunction. Then
the court has an Obligation under the law to fine the
strikers or, depending on the law, their respective
bargaining agent. The existence of a state compulsory
penalty law, PE (coded equal to 1 for states having a

mandatory penalty law and 0 for states without this

116
provision in their collective bargaining statute), may thus
be a significant predictor Of strike duration. This
possibility was examined in a regression analysis which
included PE as well as all Of the significant predictors
of strike duration identified in Chapter 4.

With an expectation that the sign on the compulsory
penalty law variable (PE) should be negative and its
coefficient significant, the regression results confirmed
a negative sign. Its significance was, however,
inconsequential. The coefficient for PE was -.132, its
t-score was .59, and its beta was -.03.

Injunctions were considerably less common in
states with compulsory penalty provisions in their public
employee collective bargaining law than in states which did
not have this provision. As shown in Table 7, among states
with mandatory penalty laws, ten (10) injunctions were
issued during 1977-1978. This constitutes 17% of the total
number Of strike occurrences in these states. This is
actually slightly lower than the prOportion of injunctions
to the total number Of strikes in states which had no
compulsory penalty law. This proportion was 20% (95
injunctions issued out Of 476 strikes initiated in these
states).

The compulsory penalties may have little
relationship to strike duration because in practice,
employers are no more (or less) likely to seek injunctive

relief in states having these laws once a strike has

mo

(“Ne-{Hm

m

 

cosmmH
OOOHuocsnOH
mo Honesz

one

am

am
mm

He
m
cm

mmxauum

mo Honesz

mH<HOH

souwaHnmmz
mommmscme
vsmHmH spasm
OHOm>HhmOOOm
commuo

OHSO

mwmumh 3oz
OHHSmQamm Bmz
HusommHz
camaroaz
mchz
hxosuamm
mmmsmM

maoaHHHH

uOOHuOOOOOU

«HanomaHmu

mmumum

msmH huHmOOm zuOmHOOBOU unoaqu mmumum

.OHHm
sump mwmaaoum xuoz mHm mumuuomom 3mH
noan .Omaom mOHummHo muumaaoo “AmmmH

 

 

..u.o .couwaHnmmSV mmHuHHom mOOHumHOm

 

noan uOuowm OHHnsm mo mumaEOm
.OOHumuuchfiac4 OOOH>uOm unmammmsmz
noan .uOAMH mo unmauumnmn .m.D

 

.B. Isl

N n

m wH

o H

o 0H

0 H

N mH

H N

cmnmmH IMMMNMMWI

mOOHuOssnaH mo umnasz
mo ambasz

"moousom wumn

mH<HOH

sHmsoumHz
xuow smz
mxmmubmz
muommaaHz
wameumz

msmesH

meauoam

mumum

msmH muHmcmm huOmHsmaoo :uHB mmumOm

Amanlmanv
.vmsmmH msOHanOflOH mo

woman: map was moocmuusooo mxHuum mo Hanson Ono an

.Ammv mBmH muHmsmd %uOmHOOEOO usonuH3 cam nuHa mmumum

.5 magma

117

118
actually begun than in states without compulsory penalty
laws. This reaffirms the expectation of loose associations
between state laws and strike duration. This relationship
between PE and strike duration may be insignificant because
the law is not being obeyed or enforced.

What accounts for the fact that 50 public employers
during 1977-1978 violated state law by neglecting to seek
injunctive relief? Strikes may be a preferable Option in
comparison to alternative forms of worker protest. When
workers strike, the conflict is overt. If strikers are
forced to return to their jobs, the basic reasons for their
original discontent will be unchanged and a new source of
resentment will be enflamed. Alternative methods of protest
will then emerge, including work slow downs, blue flu
epidemics, or even "working to rule." These forms Of
protest are not explicitly punishable by law.

By avoiding court involvement, employers may be
attempting to cultivate or to maintain a more viable
relationship with labor. A decision to seek court relief,
especially in light Of the reluctance of other employers
to also see injunctive relief, may strain further what may

be an already hardened labor management relationship.

The Relationship Between the Injunction

and the Number of Strikers

It has been suggested that the effectiveness of

119
the court injunction may vary depending on the size Of the
bargaining unit which has struck. Tom Kochan has suggested
that:
"In general, court injunctions appear
to be more effective in ending smaller
strikes where the union is weak; they
have been less successful in ending
larger strikes by more powerful unions."l
This possibility was investigated by considering
the combined effect Of IJ with SZ as well as with 822 on
strike duration. These results, controlling for other
predictors found to be significantly related to duration,
are as follows:
Sit = .282 - .254 IJ + .000254 sz - .0000000152 522
(-.09) (.32) (-.31)
+.000456 IJ*SZ - .ooooooo271 IJ*822
(.34) (-.33)

(R2 = .24)

The impact on strike duration of the injunction
does appear to covary with size, although the interaction
terms IJ*SZ and IJ*SZ2 are not significant at the .10
level. Beta weights, reported in parenthesis below each of
the coefficients, were relatively large, however.
Interpretation of the net effect Of the injunction is
difficult to ascertain from these coefficients alone.

Table 8 presents a prediction Of the net effect

Table 8. The Effect of the Court Injunction on the Duration

of the Strike
(1977-1978)

Proportion Of Strikes

 

 

Number of Strikers having a fewer number
Involved of strikers
100 .20
406 .50
1334 .67
2668 .95

120

Increase (or decrease)
in duration as a result
of an injunction

(in days)

 

-.75

-.40

+4.0

+35.0

121

of the injunction on strike duration derived from these
coefficients. For purposes of comparison, four different
"number of strikers involved" were considered. These
included the mean number Of strikers in this study (SZ=406).
For strikes involving a fewer number of workers, the net
effect of the injunction was to slightly reduce the strike's
length. For example, if an injunction was issued in a
strike involving 100 workers, its duration, according the
results reported in Table 8, would be on the average .75
days shorter.

This finding is reversed when a larger number Of
strikers were considered. Strikes were four (4) days
longer when 1334 workers were involved. Little weight
should be given to the prediction of a strike which is
predicted to last 35 days longer when 2668 workers were
involved due to the few number Of strike occurrences of this
size which were considered in this study.

Although not significant, these results lend

some support to Kochan's original hypothesis.

The Impact of the Injunction Depending

on the Type Of Employee

The results shown, seen in Table 4 (the
prediction of the duration Of strikes in cities, towns,
and counties), indicated a more significant relationship

between the injunction (IJ) and strike duration than had

122
been seen in other regressions. The mean beta for IJ in
the Tables 1, 2, and 3 regressions was -.035. The beta
weight on IJ in Table 4 was -.11. The primary jurisdiction
type omitted in the Table 4 analysis was the school district.

This suggests the possibility that the injunction
may be associated with longer strikes for teachers relative
to the other employee groups. TO test this possibility,
the interaction term IJ*EDU was constructed, where EDU
is a dummy variable representing a teacher strike.3 This
regression is presented in Table 9.

The variable of interest in this table, IJ*EDU,
is positive and also significant at the .10 level Of
confidence. The negative and significant sign on education
strike (EDU) is consistent with the results reported in
Chapter 4 with respect to the relationship between duration
and the "employee type" variables. Noting that each of the
three variables--IJ, IJ*EDU, and EDU--are dummy variables,
their regression coefficients can be added to generate
prediction of the net effect Of the teacher strike
injunction duration. The injunction, as far as teachers
are concerned, contributes .065 per cent to the duration
of the strike, according to these results.

TO summarize, with a beta weight for the
compulsory penalty law variable (PL) Of -.03, there is
little support for the prOpOsition that making penalties
compulsory has a moderating effect on strike duration. As

seen in Table 7, these laws also do not appear to have

Vector

it

it

it

it

it

Table 9.

9'11””

The prediction of strike duration when
considering the impact of the injunction (IJ)

on teacher strikes (EDU)

Variable

injunction
(IJ)

injunction * education
(IJ*EDU) strike

mediation
(ME)

permissive law
(PL)

internal comparison
(EV)

labor cost
(WB)

slack
(SL)

size

(SZ)

size2 2
(52 )

education strike
(EDU)

wage issue

(EC)
union support
(US)
.48
.23
6.78
284

123

 

Coefficient Beta
-.252 -.09
(1.38)

*
.583 .13
(1.83)
**
.464 .20
(3.48)
**
.571 .23
(3.87)
.016 .08
(1.54)
.003 .11
.885 .12
(1.57)
**
.328 .41
(2.94)
**
-.00002 -.40
(2.90)
-.266 -.12
(1.80)
**
.287 .12
(2.30)
*
.464 .10
(1.83)
Notes: See Table l

124

influenced employer's decisions to seek injunctive relief.
In the smaller bargaining units, injunctions slightly
reduce the length Of the strike. In the larger units, this
effect is reversed. This finding is not statistically
significant, however, and could be biased by a tendency Of
employers from smaller jurisdictions to seek an injunction
earlier than might be true for employers in the larger
jurisdictions. On the other hand, larger units, regardless
of when injunctive relief is requested, are probably more
likely to disobey cease and desist orders than smaller
units. Data limitations preclude the possibility Of
disentangling these effects. Finally, as shown in Table 9:
injunctions issued against teacher strikes tend to be

longer.

The Permissive Strike Law

As predicted, the permissive strike law (PL)
was positive in the regressions shown in Tables 1 through
6 and in Table 9. PL was consistently significant at
the .05 level or better. This variable, with a mean value
Of .233, is dominated however by the 95 strike occurrences
in Pennsylvania. There may be characteristics other than
the existence of a permissive strike law in Pennsylvania
which are also being captured by the dummy variable PL.
Features which characterize Pennsylvania's labor

relations environment could be different from features which

125

characterize the collective bargaining environments Of
other states. Moreover, it would be interesting to compare
Pennsylvania strikes with strikes in another state having

a similar legal environment. Unfortunately, no other
permissive strike state has a sufficient number Of strike
occurrences to make this comparison meaningful.

Nonetheless, Ohio, with 81 strikes, is an
interesting comparison state. A comparison between
Pennsylvania and Ohio strikes will enable a contrast
between determinants of strike duration in a permissive
strike state with a strike prohibition state.

Table 10 presents a comparison Of mean values for
significant predictor variables when considering Pennsylvania
strikes to the mean value of significant predictor variables
when considering Ohio strikes. Note that all variables
having a positive relationship with duration have higher
values for Pennsylvania strikes than for Ohio strikes.

These variables include ME, EC, SZ, and WB. This suggests
the possibility that the coefficient on the permissive
strike law (PL) may be representing considerations other

than the fact Pennsylvania is a permissive strike state.

The Relationship Between

Mediation and Strike Duration

One of the most surprising discoveries was a

positive and significant coefficient on mediation (ME) in

Table 10.

it

it

it

it

A comparison of mean values for significant predictors

when considering Pennsylvania and Ohio strikes.

labor cost
(WB)

slack
(SL)

size
(SZ)

union support
(US)
wage issue

(EC)

mediation
(ME)

Total Number of Strikes

Pennsylvania Strikes

Ohio Strikes

 

 

79.7

.49

622

.98

.75

.88

99

126

66.3

.49

445

.94

.69

.69

84

127

most regressions. With the data available, it is not
possible to determine whether mediation causes longer
strikes, since there is no way of determining (as was the
case with IJ) when the mediator actually became involved

in the dispute (or with what intensity). The variable ME
represents the fact that the employer (or union) respondent
to the BLS questionnaire indicated that mediation was

one, or one among other, methods used to resolve the dispute.

Mediators may have had no involvement at the
outset, but provided assistance sometime after the strike
began. As Observed in Chapter 3, there is reason to
believe that mediators, by virtue Of their role, would be
expected to be involved in the resolution of the more
militant and already longer strikes.

There is evidence which indirectly supports this
possibility. BLS data files indicate that only 65 strikes
out of 101 were mediated in Michigan, a compulsory mediation
state. The mean duration Of strikes which were not
mediated was only 2.8 days. Upon further discussion with
Public Employment Relations Officials in Detroit, it was
confirmed that about one-third of the strikes in Michigan
receive no mediation services, primarily because no
mediators are available when the strike is called. During
September, many education strikes can occur and there are
Often not a sufficient number Of mediators to provide
assistance to all of these strikes. Thus, many short

strikes never receive the services of a state mediator.

128

Further support for this is found when the combined effect
Of ME and EDU is considered (where EDU equals 1 if the
strike is called by educational employees). This interaction
term (ME*EDU) is positive and significant at the .10 level
when entered in a regression with other significant
predictors. Thus, one explanation for this finding is that
since education strikes tend to Occur during a relatively
short period of time in September, mediators tend to
become involvedvonly'in the longer strikes.

Mediation may thus not cause longer strikes,
but rather longer strikes attract mediation. Empirically,
the question Of interest is to determine which Of these
causal relationships is stronger. Without information
on the actual date of involvement, this question is outside
the sc0pe of this study.4

One indirect way Of addressing, but not resolving,
this dilemma is to hypothesize that states which require
mandatory mediation of disputes are more likely to have
mediator involvement from the first day of the strike.
According to most compulsory mediation provisions,
involvement of a mediator is required soon after the
expiration Of the prior contract, and only after a period
Of time has passed (30 days is common) since the issuance
Of a fact finder's report. Under the further assumption
that the defiance of pre-strike procedures is the exception
rather than the rule,5 it is possible to examine the

difference between the length Of strikes in compulsory

129

mediation states and the length Of strikes in states
without this provision in their state statute.

A dummy variable is thus introduced (MUST) which
represents the existence Of a compulsory mediation statute
in a state collective bargaining law. Assuming that
mediators are involved earlier in resolving strikes in
these states, the sign on MUST should be negative.

This analysis is presented in Table 11. The
coefficient on compulsory mediation (MUST) is negative, as
expected, and significant at the .05 level. This lends
support to the prOposition that early intervention in
conflict situations should generate earlier strike
resolutions. Moreover, with early involvement, there
should be less entrenchment by the parties, and a greater
likelihood of ongoing discussion. Mediators may thus be
more successful when their interventions occur at the
early stage Of bluff and counterbluff, when positions are
still being clarified.6 In addition, early involvement of
mediators may also reduce the number of issues to be
settled after a strike has been called. With fewer issues
in dispute, less time may be required to settle them.7

The coefficient on mediation (ME) remains positive
and significant in this regression. Suffice to say, without
information on the time and intensity Of mediator
involvement, it is impossible to make a definitive statement
with respect to the relationship between the duration Of

the strike and mediation.

Table 11. The prediction of strike duration when
considering the impact of compulsory
mediation provisions in state collective
bargaining statutes.

 

 

Expected
Vector Variable Sign Coefficient Beta
L' **
it mediation - .448 .19
(ME) (3. 43)
**
compulsory mediation - -.300 -.14
(MUST) law (2.34)
**
permissive law + .733 .29
(PL) (4.70)
Bit . ‘ *
internal wage + .019 .10
(EV) comparison (1.87)
Cit
labor cost ? .003 .11
(WE) (1.50)
*
slack ? .887 .12
(SL) (1.65)
B. *
it size + .283 * .35
(SZ) (2.53)
2 **
size 2 - -.00002 -.35
(SZ ) (2.55)
Pit **
wage issue + .294 .13
(EC) (2.36)
**
union support + .500 .11
(US) (1.99)
R .48
R2 .23
F 7.51
n 284

Notes: See Table l.

130

131

Nonetheless, there are a number Of justifications
for why the relationship between mediation and strike
duration could be positive. One role Of the mediator is to
develOp "lines Of communication with peOple other than
those representing the parties at the bargaining table."8
The greater the number Of persons who are involved, the
longer, one would expect, it would take to resolve the
strike. Further, public employee strikes may be longer
because the mediator must spend considerable time and effort
in helping the parties appreciate the intricacies of the
bargaining process.9 This adds to the time required to
settle the strike.

It is also possible that some interventions may
actually contribute to a lengthening Of the strike. Many
mediators emphasize the recognition Of the middle ground--
the "zone" where each party needs only to yield a little
gound to meet the other. The problem with this intervention
strategy may be that some issues are simply not amenable to
compromise. If the less skilled mediators force compromise
on just such an issue, more intranscience on behalf Of the
parties may result.10 Conceivably, longer strikes may be
the outcome.

Having discussed the performance of the public
policy variables IJ, PE, IJ*EDU, PL, ME, and MUST, we now

turn to a concluding discussion of the results in Chapter 6.

132

Footnotes (Chapter 5)

1

Thomas A. Kochan, "Dynamic Of Dispute Resolution in the
Public Sector," in Benjamin Aaron, Joseph Grodin, and
James Stern, Public Sector Bargaining (Washington, D.C.:
Bureau of Natibnal Affairs, I979), p. 168.

2

Results on the variables which were omitted from the
analysis are available from the author upon request. The
magnitude and signs Of the coefficients on the omitted
predictors, however, were not appreciably affected by the
introduction of the interaction terms IJ*SZ and IJ*SZSQ.

The variable EDU represents striking teachers who are
organized in instructional bargaining units only. An
alternative specification for EDU which included both

instructional and noninstructional employees generated
similar results.

Contact was made with the public employment relations
Offices in the states of Michigan, New Jersey and
Pennsylvania in an attempt to acquire the time and date
mediators first became involved in strike situations. In
every state, this information was unaccessible. In some
states the information was not available in a useable
form. In Michigan, public law stipulates that any
transactions between a mediator and the parties must
remain confidential. This makes all internal records
unavailable for research purposes.

There are limitations to this assumption, since it is

known that in some states (such as Pennsylvania) pre-strike
procedures are not followed. ,Phone conversation with
John Bomer, P.O.L., Labor-Management Services
Administration, January, 1979..

Linda Dickens, "Conciliation, Mediation and Arbitration in
British Industrial Relations," in Geoffrey M. Stephenson
and ChristOpher J. Brotherton, Industrial Relations: A

Social Psychological Approach (New York: John Wiley and
sons, 9 ’ pp. , o - O o

 

133

7
This argument was made by Ed McMann of the Federal

Mediation and Conciliation Service in a phone conversation,
October 19, 1979.

8
Thomas C. Schelling, The Strategy of Conflict (Cambridge,
Mass.: Harvard University Press, 1960), p. 44.

 

9
Thomas P. Gilroy and Anthony V. SinicrOpi, ”Impasse
Resolution in Public Employment: A Current Assessment,"

Industrial and Labor Relations Review, Vol. 25, NO. 4
(July, 1972) p. 499.

10
Linda Dickens, Op. cit., p. 295

Chapter 6

Summary of Results and Conclusions

The performance Of the variables found to be
significant is discussed in the first section Of this
chapter. In the second section, insignificant findings are
summarized and discussed. In the final section, the
contributions as well as the limitations Of the study are

presented.

Summary of Significant Findings

The average (or mean) beta weights on the
variables which were significant are shown in Table 11.
None of the variables associated with the expectations
vector (Bit) were significant at the .10 level or better.
The expected positive direction Of their signs was confirmed,
however. Two Of the variables associated with the "employer
constraints" vector (Cit) were significant--the prOportion
Of labor costs Of the striking bargaining unit to total
labor costs (WB) and the component Of revenues were not
used by the jurisdiction to compensate personnel, slack or
SL. For all strike occurrences and for strike occurrences
in cities, towns and counties, the beta weights on WB and
SL were relatively strong. In general, the larger the
beta weight, the longer the strike is predicted to last as

a result Of that factor's influence.

134

Table 12.

Vector

Expectations

(Ei

Constraints

(Ci

Bargaining Relationship

(Bi

t

t

t

)

)

)

Variable

no variables

were significant

Labor Cost
(WB)

Slack
(SL)

Size (SZ)
Size2 (822)

Inessentiality
(NS)

Combined Units
(OTH)'

Labor Relations Environment

(Pi

Legal Environment

(Li

t

t

)

)

Union Support
(US)

Wage Issue
(EC)

Mediation (ME)

Compulsory
Mediation
(MUST)

Permissive Law
(PL)

Injunction
Issued Against
Teachers
(EDU*IJ)

135

All Strikes

A comparison of the mean beta weights on variables which
were significant in Tables 1 through 6 for strikes
occurring in all jurisdiction types and strikes occurring
in cities, towns, and counties.

Strikes Occurring
in Cities, Towns,
and Counties

 

 

.26

.14

.39

-.38

-.12

.21

.13

.11

.19

-.14

.24

.13

.40

.24

-.O6

-.03

-.12

.16

.22

.04

.10

.31

136

Note that the sign on the labor cost variable,
WB, was positive in the analyses of all strikes and in
the analysis which excluded strike occurrences in special
districts and in school districts. When the labor costs
Of the striking unit are a small component Of total labor
costs (when WB is small), labor may be able to Obtain
quicker settlements. The total cost Of a concession to
the employer is small, relatively speaking. Labor thus
attains more bargaining leverage with employers because
labor costs constitute a small component Of total labor
costs. This implies that strikes for comparatively small
units are shorter in part because the employer is more able,
and therefore more willing, to make concessions.

An employer may also be more convincing in their
threat to use strike replacements. It is also more
feasible to recruit replacements. An implication Of this
alternative explanation is that strikes may be shorter
because an employer is less likely, not more likely as in
the previous argument, to "give in" to labor's demands.
Here labor may be more receptive to a downward adjustment
of their bargaining demands. With an incentive for each
side to moderate their respective positions, one would
expect to find evidence Of shorter strikes when labor
costs are small.

The positive sign on W3 is consistent also with
an expectation that labor costs of the strikers constitute

a constraint on the employer's ability to adjust their

137
final pre-strike Offers upward. The greater this constraint--
the greater the component Of labor costs Of the strikers
to total labor costs--the longer the strike should last,
according to these results.

When labor costs of the strikers are relatively
high (when the value Of WB is large), there is also an
incentive for the employer to "hold out" longer because
each strike day constitutes an additional cost savings.
Revenues are unaffected when public employees strike.
Payroll expenditures, however, are temporarily interrupted.
Strike occurrences during the 1977-1978 period may thus
have been one, among other, mechanism public employers
used to bring expenditures "in line" with projected
revenues. Long strikes may have been the preferred
alternative for some employers to other Options such as
the imposition of an additional tax levy on an increasingly
unreceptive public.

It is not possible to determine from these results
alone why strikes are longer when labor costs of the
striking unit are high. Is it because of the increased
financial benefits which accrue to the employer as a result
of the interruption in expenditures, or because of the
greater difficulty an employer has in identifying the
additional resources required to fund the concessions
demanded? Both considerations may be equally important
factors.

The slack variable (SL) captures the share Of

138
the economic pie which is available to support labor costs
in lieu of supporting nonlabor costs. To the extent that
a larger share of existing revenues are available to support
nonpersonnel costs, labor should have a greater incentive
to strike as long as necessary to Obtain the concessions
they want. Slack was initially considered as a member Of

the constraints vector (Ci ). When considered as an

t
indicator of the various financial constraints faced by
management, a negative relationship was expected.
These results indicate that the positive sign
on this variable is theoretically consistent when viewed
as an indicator of employee expectations. As such, the
greater the component of revenues which do not support
labor costs, the greater the enhancement in expectations
among the union membership. The probability Of greater
tension in the labor management relationship may also be
enhanced. As Observed, enhanced expectations should result
in longer strikes, other considerations held constant.
Specifically, the positive sign on the beta for
WB (the labor cost variable) means that strikes in education
are longer than strikes by other public employees as a
result Of the higher relative labor costs in education.
Labor costs consumed by non-education public employees are
low relative to the labor costs incurred by school districts.
On the other hand, the negative sign on the
inessentiality index (NS) means that the length of teacher

strikes, controlling for the fact that labor costs in

139
education are higher, are shorter than strikes by other
public employees. Two explanations may account for the
finding that strikes by teachers, as a group, are shorter
than strikes by all other public employees.

Users of public educational services find it
virtually impossible to acquire substitute services when
teachers strike. If nurses strike, patients can be
transferred to Operational medical centers located in a
contiguous jurisdiction. When mass transit workers strike,
commuters have a number of alternative Options available--
they can walk, use car pools, or even ride motorbikes.
School age children, on the other hand, must generally
wait for a strike settlement to be reached before returning
to the classroom. Elementary and secondary public education
is thus not a readily substitutable service when strikes
occur. The net result may be heightened public pressure
exerted by the parents Of school age children to settle
strikes by teachers.

A majority Of education strikes during the 1977-
1978 period were initiated by one labor organization, the
National Education Association (NBA). An inverse relation-
ship between the inessentiality variable (NS) and strike
duration may also be explained by NEA's emphasis on
coordinating the bargaining strategies of local negotiating
teams. It may be that school districts which took strikes
found it extremely difficult to meet NEA's standard contract

demand. Many districts experienced rapidly declining

140
student enrollments during the late 1970's. This was an
additional economic consideration which was not confronted
by other public employers. School systems which
took strikes could thus have been facing more formidable
economic constraints than noneducational jurisdictions
taking strikes. TO the extent that greater economic
constraints result in longer strikes, One would expect to
find evidence Of longer strikes by teachers. This is an
especially interesting finding, since teacher strikes
were initially expected to be longer.

Aside from the significant performance of the
inessentiality index (NS), there were a number of other
significant noneconomic determinants of strike duration.
These factors included size (SZ), the existence Of combined
Occupational units (0TH), union support for the strike
(US) and whether the issue concerned wages (BC). One of
these variables, size, and the square of size (822), were
significant in the analyses Of all strikes. Beta weights
on these variables were .39 and -.38 respectively.
Insignificant beta weights, however, were Observed in these
variables in the analysis of strikes occurring in cities,
town and counties (see Table 11). Consider first the
implications Of these results for establishing legislation
which pertains to all public employee groups.

For the very large units, shorter strikes would
be anticipated. Longer strikes, on the other hand, are

expected from bargaining units which are moderately large

141
in comparison to small units. This means that the
encouragement of broad based units may lead tO longer
strikes in small jurisdictions, but to shorter strikes in
the very large jurisdictions.

How does public policy affect the size Of the
bargaining unit? Criteria for determinating the composition
and size Of bargaining units is established by statute in
some states and by public employment labor relations
boards in others. Representative criteria include such
factors as the degree of (1) common supervision, (2) the
integration and interchange Of the work function, (3) the
desires of the employees, (4) the similarity Of working
conditions, (5) Operational efficiency, (6) the
geographical dispersion Of employees, and (7) the inclusion
(or exclusion) of supervisors from bargaining units.

The effect of the application Of such criteria
is tO encourage either broad-based, large units or small,
narrowly based units. Understandably, both labor and the
employer want units that they believe will be conducive
to Obtaining favorable election outcomes. Employers
generally favor large, broad-based units beCause Of the
belief that they are easier to defeat in representation
elections. The size Of these units, however, may also
affect the subsequent duration Of strikes if and when they
occur.

When considering legislation which applies to

all local government employees or which applies only to

142
teachers, public policy makers should be advised that, at
least based on these results, bargaining unit size is
related to strike duration. On the other hand, when
debating state legislation excludes special district
and school district personnel, there was no evidence Of a
significant relationship between unit size and duration.

In contrast to the difference found in the
magnitude Of the beta weights in Table 11 for size (SZ) across
the two analyses, the beta on the combined unit variable
(OTH) was positive and significant in both analyses
(+.21 and +.16). Regardless Of the employee type being
considered, more heterogeneous units (which also tend to
be larger) were positively associated with the duration
Of public employee strikes.

Two other significant predictors of strike
duration are now considered--the support given by the
international or national labor union (US) and the disputed
issue (EC). When the union fails to support the strike,
its duration is shorter. Negotiations must be carried
on by local negotiators who may be unexperienced. Strike
benefits are not paid. If there is lack Of union support,
doubt may be seeded among the local union membership
concerning the worthiness of the strike. Interest in
continuing the strike should thus dissipate. As also
expected, workers are also willing to strike longer when
the issue concerns wages than when the dispute concerns

other terms and conditions Of employment.

143

Finally, the negative sign on North (NO) means
that strikes occurring in northern states are shorter than
strikes occurring in other regions of the country. This
unexpected result may be due to the possibility that
bargaining units with a longer and conceivably more
mature labor management relationship are able to resolve
their diSputes more expeditiously. Alternatively,
jurisdictions in northern jurisdictions may also have
fewer resources to haggle over, relative to jurisdictions
taking strikes in other regions of the country. This
should dampen labor's interest in sponsoring long strikes.

Four public policy variables were found to be
significant. These included mediation (ME), the compulsory
mediation law (MUST), cease and desist injunctions issued
against teachers (EDU*IJ), and the permissive strike law
(PL). Considered first is the unexpected performance of
the mediation variable. Recall that an inverse sign was
expected, but this variable was consistently positive.

Considerable public support has been committed
to the employment and training of neutral third parties.
The Federal Mediation and COnciliation Service uses an
argument before Congress that mediation contributes to a
reduction in the number of days of idleness as one
justification for continued support. The limited empirical
evidence to date from this and previous studies suggests
that the involvement of outside mediators may not reduce

strike duration, at least if timing of intervention is not

144
identified.

This conclusion merits qualification. The.
existence of a state compulsory mediation law (MUST) was
inversely related to strike duration. This supports the
importance Of insisting on pre-strike mediator involvement.
An important conclusion Of this study is thus that the
timing of mediator involvement is a critical factor in
making a definitive determination of the impact of mediation
on strike duration. A follow-up study is being undertaken
in the summer Of 1981 which will enable the identification
of when mediators became involved in 1979 public employee
strikes.

The injunction, a more punitive policy Option
than mediation, was not significantly related to strike
duration. However, when strikes were initiated by
teacher groups, they were significantly longer. The
beta on EDU*IJ was +.13. Perhaps teachers resent court
injunctions because of the implication that they are being
"unprofessional," or because of the implication that they
have little regard for the best interests of students. If
this reaction is an explanation for the finding of slightly
longer strikes when injunctions are issued against teachers,
what accounts for the greater level of animosity among
teachers relative to the reactions other employee groups
would be expected to have?

Perhaps when teacher injunctions are issued, a

variety of political, legal and other professional leaders

145
become involved. This may be less likely to happen when
small bargaining units strike. This would introduce a
new dimension to the two party bargaining process. As
the arena of negotiations is broadened, it is also
reasonable to expect that it would take longer to discuss,
and subsequently to resolve the disputed issues. The IJ
variable may thus also be picking up considerations other
than the issuance of an injunction.

The performance of the final significant public
policy variable--the permissive strike law (PL)--was
positive, as predicted. According to these results,
permiSsive strike laws contribute only 1.2 days to the
duration Of strikes, all other considerations held
constant. These results suggest that legislatures need
not be overly concerned about encourageing long and bitter
strikes by public employees if they elect to enact

permissive strike legislation.

Summary and Discussion of Insignificant Results

The role of the strikers has been taken into
consideration by the variables associated with the

expectations vector (Bi ). These results suggest that

t
expectations do have a positive effect on strike duration,
but this effect is not as important as other factors. The
employer may be influenced by the earnings of other,

comparable employees, and therefore adjust wage Offers

146
accordingly. This would tend to moderate the impact of
expectations on the duration Of the strike. Expectations
may thus be more salient factors in determining whether
to strike or accept the employer's last Offer.

The role of the public received the least
attention. Only one variable, the local tax burden (TB),
represented the pressure citizens might be expected to
exert on the parties. While the local tax burden, as
viewed in this study, had a slight moderating effect on
strike duration, this effect also was not significant.
The influence of citizens may thus be a less important
factor than other considerations in determining the length
Of the strike.

There are indications, however, that the public
may increasingly exert a greater influence in the collec—
tive bargaining process in the future.

"Employers and the public have become
more willing to resist strike pressures
recently as economic conditions of
cities and states have deteriorated
and the political pendulum has shifted
against public employees."1

The insignificant performance of the other
factors is summarized below.

Egggl (RU): As expected, strikes in rural areas were

slightly shorter than strikes in the more pOpulated areas,

but this effect was not significant. Rural workers who

147
have Opted to organize and who have elected to strike may
simply be more militant than has been acknowledged in
previous studies.

First negotiation (FN): Although strikes occurring during

 

first negotiation rounds were expected to be longer, these
were found to be slightly shorter. Perhaps during the
reCOgnition campaign, many of the more difficult issues
were resolved. These results suggest that, in general,
both employer and union on strike appear to be no more

or less willing than better established units to settle

a first agreement. Alternatively, first negotiation
round strikes in states with no collective bargaining
statute (such as Ohio) may be longer. A reverse effect
may be found for first negotiation rounds in states with
comprehensive laws. This remains to be determined.
Winter (WR): With an expectation of a negative sign on
this variable's coefficient, its sign was positive.
Indirect evidence suggests that season may affect the
decision by certain types Of employees to strike, but not
the subsequent duration Of the strike itself. There is,
for example, a positive and significant relationship
between the winter season (WR) and street and highway
strikes (STR) as reported in the correlation matrix (see
Appendix C).

Supplemental benefit issue (SB): As a variable which

 

represented a unique type of economic issue, SB was

expected to be positive. Instead, its sign was negative.

148
With relatively few strikes over benefit issues, these
results must be considered tentative. However, this
result emphasizes the importance of taking into consider-
ation the various types of economic issues that workers
may be willing to strike over.
Compulsory penalty laws (PE): There is little evidence
from these results that making penalties compulsory has
an effect on strike duration. Further, these laws are
not being obeyed or enforced. Employers are no more
likely to seek court relief in states with compulsory
penalty laws than in states without these laws. While
compulsory penalty laws may affect the prOpensity to take
or initiate a strike, these results give no indication
that they have any significant effect whatsoever on the
duration of strikes that do occur.

Injunctions (IJ): Although insifnificant, injunctions

 

slightly reduce the length of the strike in smaller
bargaining units. In larger units, this effect is
reversed.

Having presented a summary and discussion of
the performance of both the significant and the insignifi-
cant variables, we turn to a concluding discussion of

both the contributions and limitations of the study.

The Contributions and Limitations of the Study

The two years Of strike activity considered

149
(1977-1978) were in a period of economic expansion.
Different relationships may have been found had a period
Of economic contraction been considered. Second, existing
theory and empirical work on strike duration is lacking.
The discovery that many variables were insignificant is a
testament to the lack of a comprehensive theory. Third,
data limitations precluded a consideration of many factors
which may also be relevant considerations. These include,
among Others, the skills and personalities of the
negotiators and mediators, the effect of prior strike
activity by the same bargaining unit, nonwage bargaining
history, and the payment of strike benefits.

On the positive side, the performance Of the
predictors was relatively consistent across model
Specifications. With an explanation for one-fourth to
one-third of the variation in strike duration, the.
"disappointing" results found in the Burton-Krider study
were not replicated in this study.

As expected, many of the more important predictors
of duration were associated with the local bargaining
environment. Region, season, and pOpulation density
contributed less to the explanatory power of the model
than factors such as the issue under dispute, union
support, the number of strikers, the "employee type"
involved, or the proportion of labor costs to total labor
costs. The identification of factors which describe the

local bargaining context is apparently an important

150
consideration in the design of strike duration research
and bolsters an important assumption of this study--that
the bargaining unit as a unit Of analysis is most
apprOpriate.

With a primary focus on the collective
bargaining process rather than its outcomes, these results
indicate that more needs to be known about the processes
associated with resolving disputes. It is necessary, but
not sufficient, to know that a mediator was involved or
a court injunction was issued during the course Of a
strike. It is also important to identify when mediators
began providing conciliation services and with what
degree of intensity. Injunctions may have been sought
by employers and subsequently ordered by the courts,
but it is also important to know whether an injunction
was coupled with a penalty. Conceivably, the impact on
strike duration of the injunction may vary depending on
its timing. This remains an unsettled question.

These results do indicate nonetheless that
strike duration is a potentially fertile and important
analysis to pursue.

In conclusion, it is important to recognize
that this study has directed an emphasis away from
predicting strike occurrences and toward considering a
prediction of their duration. From the perspective of the
parties who are confronted with a strike, resolution of

the dispute is what matters. Since there is no reason to

151
believe that strike prohibition laws will be any more
effective in the future decade than was evident in the
1970's, it is time that industrial relations scholars
began to consider the dilemmas faced by employers, unions
and the public once strikes occur. It is in this Spirit

that this study has been undertaken.

152

Footnotes (Chapter 6)

1
Thomas A. Kochan, "Dynamics of Dispute Resolution in
the Public Sector," Benjamin Arron,- Joseph Grodin and

James Stern, Public Sector Bargaining (Washington, D.C.:
1979), p. 168.

 

APPENDICES

Appendix A. Variable Descriptions and Predictions

 

 

 

Vector Variable Description Prediction
Bit
Wage Change Over the two year period prior to the +
(CE) strike, the change in the mean

it

earnings of the comparable employee
group in the same state

Wage Level Ratio of the mean monthly earnings of +
(RE) the comparable employee group in the
state to the pre-strike, mean monthly
earnings of the striking bargaining

unit
Internal Over the one year period prior to the +
Comparison strike, the ratio of the change in the
(EV) mean, monthly earnings Of all employee

groups working for the same jurisdiction
to the change in the mean monthly
earnings of the striking bargaining

unit
*
Labor Cost Ratio of the wage bill of the striking ?
(WB) unit to total expenditures for
personnel
Units The number of bargaining units recognized +
(NU) to bargain collectively in the

jurisdiction taking the strike

Slack Total revenues minus total expenditures ?
(SL) for personnel as a proportion of
total revenues

Tax Burden An index composed Of the addition of ?
(TB) two variables: the proportion of
prOperty taxes to total expenditures
for personnel and the proportion of
property taxes to total revenues

this variable can only be considered in instances where a

a specific employee group (or groups) can be identified.

By necessity, any strikes which constitute a combination

of employee types were deleted from analyses which considered
this variable

153

Appendix A (continued).

Vector Variable

 

B

it

82

FIR
POL

SAN

HOS

STR

TRN

Inessentiality
Index (NS)

Combined Units
(CM)

Variable Descriptions and Predictions

Description

 

Number of workers involved in the
strike

Firefighter strike (l=yes; O=no)

Police strike (l=yes; O=no)

Sanitation Strike (l=yes; O=no)

A strike by all other employee
groups except education (l=yes;
O=no)

Hospital and health service strike
(l=yes; O=no)

Street and Highway strike
(l=yes; O=no)

Transportation strike
(l=yes; O=no)

Welfare services strike
(l=yes; O=no)

— police or firefighter strikes
— sanitation, hospital, transit, or

water and sewerage strikes

street, parks, education, housing,
welfare or general administration
strikes

Combined occupational composition Of
the bargaining unit (l=yes; O=no)

154

Prediction

 

Appendix A (continued). Variable Descriptions and Predictions

Vector Variable Description Predictions

 

 

 

P.
at . . .
FN Negotiation of the first agreement +
(l=yes; O=no)

US Union support for the work stoppage +
(l=yes; O=no)

EC Primary issue under dispute is wages +
(l=yes; O=no)

SB Primary issue under dispute is the +
payment of fringe and other
supplemental benefits (l=yes; O=no)

K.
1t . .
WR Occurrence of a strike during the -
winter quarter (l=yes; O=no)
RU Occurrence of a strike in a rural, -
non-SMSA area (l=yes; O=no)
NO Occurrence of a strike in a northern +
state (l=yes; O=no)
SO Occurrence of a strike in a southern -
state (l=yes; O=no)
NO*WR Interaction term constructed by the -
multiplication of WR*NO
Lit

ME Federal, state, local or private -
mediator involvement (l=yes; O=no)

IJ Issuance of a court injunction (l=yes; ?
O=no)
PL Permissive strike legislation +

(l=state has a permissive strike law;
O=State has no permissive strike law)

PE Compulsory penalty statute in state -
legislation covering local government
employees

MUST Compulsory mediation statute in state -

legislation covering local goverment
employees

155

Appendix B.

Variable

The Dependent Variable

Sit

The Independent Variables

CE

EV

NU

SL

TB

82

FIR

POL

SAN

HOS

STR

TRN

WEL

NS

Mean

1.647

.111

1.014
.577

65.31

4.22

.478

1.104

406.470

.0305
.0394
.0412
.2007
.0143
.0412
.0251
.0233

2.736

Standard Deviation

 

1.07

.068

.201

5.704

35.81

.525

1334.040

.172

.195

.199

.401

.119

.199

.156

.151

.603

156

Means, standard deviations and data sources
(N = 558)

Data Source

 

Work StOppage Data
File, BLS

Annual Employment
Data File (1975-
1977), Census

Annual Employment
Data File (1977).
Census

Annual Finance Data
File (1975), Census

Work StOppage Data
File, BLS

Appendix B (continued). Means, standard deviations and data sources
(N=558)

Variable Mean Standard Deviation Data Source

 

 

The Indedpendent Variables (continued)

CM .328 .470 Work StOppage Data
File, BLS
EN .1057 .308 "
US .9433 .232 "
EC .6989 .459 "
SB .0287 .167 "
WR .1631 .370 "
RU .2867 .453 "
NO .7885 .409 "
SO .0448 .207 "
NO*WR .1362 .343 "
ME .7025 .458 "
IJ .1846 .388 "

PL .2330 .423 U.S. Department of

Labor, Summary_of
Labor Relations
Policies (Washington:
G.P.O., 1979)

 

157

on

on

ha

wn

mm

vn

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

xHuuuz coHunHouuou one

.o xHocoaac

00 an 30 on 03 03 B3 03 m3 «3 n3 «3 33 03 a 0 s o n v n u 3
00.3 30.- 30. 30. no. 00.- 00.- 00. 33.- mo.- 00. 00. 30.- 00. qo.- 00.- 00.- mo.- 03.- 03.- 03. 00. 00.- 30.- 30.- 00. no. 30.- v3.- 03. to
00.3 00. 00.- 03. ~.- 03. 00.- ~3.- 00.- 00.- 03. no. 03.- o3.- m~.- 03.- m~.- 0N.- 3NH- om.- 00. 0a.- 00. o3.- 30. N0. 03. 00.- 30.- zoo
00.3 n3.- mm. ~3.- 03. so. 00. 00.- 00. 00. n0.- 00. 00. 30.- mo.- 00. 33.- 03.- 00.- so. 00. 00.- so.- A3. 30. 03. 03. 30. 30
00.3 43. 30. 00. 03.- mo. 00.- 00. 30.- 00. 00.- 00. 30.- ~0.- mo.- 30.- 03. mo.- 00. 00.- ~0.- 00.- 30. 30. 30.- m0.- ~0.- 00
00.3 m0.- W3. 00,- 00.- 03.- 00. 03. 00.- 00.- mo. 03.- mo.- 00.- 30.- 00. 03.- 30. v3.- ~0.- 00.- 03. 00. 40.- 00.- mm. m:
00.3 30.- 00.- 03.- 00.- mo. 03.- ~0.- 30.- 00. 40. 03. 00. 00. 33. 33. 30. 03. -.- 00.- 03.- 00. 00.- 00. 30. 00
00.3 00.- no. 30. mo. 03. 03.- 30.- 30.- no. 00.- 00.- mo.- 33.- 03.- no. 03.- 03. 00.- 03. 00. 03. oo. 00.- 02
00.3 30.- 30.- 00.- 00.- 00. 30. 03.- 00. 00. 00.- 30. ~0.-.30. 03.- 30. ~3.- m~.- so. 30. 00. 00. mo.- as
00.3 30. 00. 00.- 00. 00.- 30.- 00. 30.- 30. 00. ~0.- 00. 00.- 33. 00.- 30.- o3.- 30.- 00. 00.- No. a:
00.3 0~.- 30.- ~0.- 00.- 00. 00. 00.- 00. mo.- 00. 30.- m0.- 33. m0.- 00.- 30.- 00. N0. 40. 03.- mm
00.3 30. 03.- 00.- 30. mo. 30. 00. 30.- 33.- 03. 30. 00. 00.- 00. 30. 00.- 00.- 40.- 03. on
00.3 00. s0.- 33.- m0. 00. so.- m3.- 00.- 00.- 00. 33.- 00. 33. 03. 30. 00.- 30.- 03. mo
00.3 43. 00.- 00.- 30. 00.- 30.- 00. 00.- m0.- 30.- 00. 00.- 00. 33.- N3. 30.- s0.- 2t
00.3 ~0.- 30.- ~0.- 00.- n0.- 30.- 00.- 00.- 0m. 03. m3. m~.- 00.- ~0.- so.- 00.» 303
00.3 30.- 00.- 00.- 30.- 00.- 00.- 00. 33. m0.- 00.- 30. 30. 00.- 53. v3. zap
00.3 00.- 00.- 00.- 00.- 03.- 00.- 03. no. ~0.- 00.- 00. 00. 30.- 00.- mom
00.3 00.- 30.- ~0.- 00.- mo.- 00. 00.- am. s3.- 00. 00.- mo.- 00. 00:
00.3 00.- 00.- 03.- 00.- 0m. 00.- 00.- 00.- 00,- 30.- s3. 00. :00
00.3 30.- 03.- m0.- 33. 00.- so. 0m.- 00.- mo. N3. 00.- 300
00.3 00.- 30.- s3. 30.- 30.- 0~.- 00.- 30.- 03. 30.- 030
00.3 No. 00. 33. 00. o~.- 00. m3.- m3.- 00. x00
00.3 03.- mo. 03. 03. 00. a3.- 00.- 00. n
00.3 00. 30. 00.- 00. m0.: 00. «0. 30
00.3 30.- 00. 03.- 00.- 33.- 30.- as
00.3 3~.- 30.- n~.- 00.- 00.- 0:
00.3 no. 00. 03.- m3. 0;
00.3 30. no. 03. >0
00.3 mm. mow- as
00.3 00.. no
_ 00.3 030
Ufiflfldhfl.)

 

o .4 IV to v u: m: h m a- O. H N to v In «3 h» m «m o
r! r: —4 we —c —4 r0 —c an —1 cu r» cu cu cw cw (V «V cu «V In

HNMQMOF‘DO‘

158

n:- 1-“-“ (“r-m"---“1“-