Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
THE EFFECT OF UNIONIZATION ON THE FIRM PROFITABILITY:
THE U.S. MANUFACTURING FIRMS, 1986-1988
by
NEBIYE KARAHASAN, B.A., M.A.
A DISSERTATION
IN
ECONOMICS
Submitted to the Graduate Faculty of Texas Tech University in
Partial Fulfillment of the Requirements for
the Degree of
DOCTOR OF PHILOSOPHY
Approved
Accepted
Dean of the Graduate School
August, 1993
' ^ \ ACKNOWLEDGMENTS ( J ^ W^/^^
I would like to express my most sincere gratitude to my committee
chairman, Dr. James E. Jonish, for his patience, guidance and support during
my studies. I also wish to thank the other member of my committee. Dr. Lewis
E. Hill, Dr. Sujit Roy, Dr. Klaus Becker and Dr. Terry Von Ende for their
discussions and helpful comments.
I also would like to thank to my husband, Rahmi Yamak, for his valuable
discussions and continuous support. Thanks also to the Government of the
Republic of Turkey for its financial support and also to the Department of
Economics at Texas Tech University for its support during my graduate studies.
Finally, special thanks goes to my parents for their love and continuous
support during my graduate studies.
CONTENTS
ACKNOWLEDGMENTS ii
LIST OF TABLES v
CHAPTER
\. INTRODUCTION 1
II. LITERATURE REVIEW 10
2.1 Industry Level Studies 10 2.2 Firm or Business Level Studies 15
2.3 Summary of Literature 33
III. METHODOLOGY AND DATA 36
3.1 Methodology 36
3.2 Data Description 55
IV. EMPIRICAL RESULTS AND ANALYSIS 63
4.1 The OLS Estimation Results of the PCM, ROI and EV Models on Cross-Section Data, 1988 64
4.1.1 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Ratio 68
4.1.2 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Dummies 73
4.1.3 The OLS Estimation Results of the PCM, ROI and EV Models with Industry Union Coverage Ratio and Dummies 79
4.1.4 Summary Assessment for All Cross-Section Estimation Results, 1988 83
4.2 The Estimation Results of the PCM, ROI and EV Models over the 1986-1988 Period 89
4.2.1 The OLS Results of the PCM, ROI and EV Models over the 1986-1988 Period 90
III
4.2.1.1 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Ratio 92
4.2.1.2 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Dummies 97
4.2.1.3 The OLS Estimation Results of the PCM, ROI and EV Models with Industry Union Coverage Dummies 99
4.2.1.4 Test Results for Sources of Union Gains 103
4.2.1.5 Summary Assessment for the OLS Estimation results over the 1986-1988 Period 108
4.2.2 The Results of the Fixed-Effect Regressions over the 1986-1988 Period 111
4.2.2.1 The Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Ratio under the Fixed-Effect Technique 113
4.2.2.2 Summary Assessment for the Estimation Results of All PCM, ROI and EV Models under the Fixed-Effect Technique 118
4.2.3 The Results of the Random-Effect Regressions over the 1986-1988 Period 119
4.2.3.1 The Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Ratio under the Random-Effect Technique 120
4.2.3.2 Summary Assessment for the Random-Effect Regressions 124
V. SUMMARY AND CONCLUSION 126
5.1 Summary of Results 128
5.2 Conclusion 130
REFERENCES 135
APPENDIX 143
IV
LIST OF TABLES
3.1 Variable Definitions and Sources
3.2 Descriptive Statistics of the Variables
4.1 OLS Resu
4.2 OLS Resu
4.3 OLS Resu
4.4 OLS Resu
4.5 OLS Resu
4.6 OLS Resu
4.7 OLS Resu
4.8 OLS Resu
ts of PCM Model with Firm Unionization, 1988
ts of ROI Model with Firm Unionization, 1988
ts of EV Model with Firm Unionization, 1988
ts of PCM Model with Industry Unionization, 1988
ts of ROI Model with Industry Unionization, 1988
4.9 OLS Results of EV Model with Industry Unionization, 1988
4.10 OLS Results of PCM Model with Industry Unionization Dummies, 1988
4.11 OLS Results of ROI Model with Industry Unionization Dummies, 1988
4.12 OLS Results of EV Model with Industry Unionization Dummies, 1988
4.13 OLS Results of PCM Model with Firm Unionization, 1986-88
4.14 OLS Results of ROI Model with Firm Unionization, 1986-88
4.15 OLS Results of EV Model with Firm Unionization, 1986-88
4.16 OLS Results of PCM Model with Firm Unionization Dummies, 1986-88
59
61
69
72
74
ts of PCM Model with Firm Unionization Dummies, 1988 75
ts of ROI Model with Firm Unionization Dummies, 1988 77
ts of EV Model with Firm Unionization Dummies, 1988 78
80
81
82
84
85
86
93
95
96
98
4.17 OLS Results of ROI Model with Firm Unionization Dummies, 1986-88 100
4.18 OLS Results of EV Model with Firm Unionization Dummies, 1986-88 101
4.19 OLS Results of PCM Model with Industry Unionization Dummies, 1986-88 102
4.20 OLS Results of ROI Model with Industry Unionization Dummies, 1986-88 104
4.21 OLS Results of EV Model with Industry Unionization Dummies, 1986-88 105
4.22 Test Results for Source of Union Gains for PCM, ROI and EV Models, 1986-88 106
4.23 Fixed-Effect Results of PCM Model with Firm Unionization, 1986-88 114
4.24 Fixed-Effect Results of ROI Model with Firm Unionization, 1986-88 115
4.25 Fixed-Effect Results of EV Model with Firm Unionization, 1986-88 116
4.26 Random-Effect Results of PCM Model with Firm Unionization, 1986-88 121
4.27 Random-Effect Results of ROI Model with Firm Unionization, 1986-88 122
4.28 Random-Effect Results of EV Model with Firm Unionization, 1986-88 123
A.I Names and Industry Codes of the Firms 143
VI
CHAPTER I
INTRODUCTION
Since 1970s, the beginning of substantial decline of unionization in the
manufacturing sector of the United States, the literature on labor and industrial
economics has devoted much attention to investigating the hypothesized
relationship between the unionization and profitability.
The relationship between unionism and profitability has been a locus
point of the empirical literature since the studies of Freeman (1983) and Clark
(1984), although a particular relationship between unionization and profitability
has been discussed theoretically for a long time. Both studies are influential in
this specific area because the effect of unionism on profitability is being the first
time, empirically investigated by these studies. Until these studies, of course, a
considerable body of literature concerned with the empirical determinants of
profitability exist. Unfortunately, no studies did "explicitly" concentrate on the
relationship between unionism and profitability. For example, no study which
includes unionism, as well as other explanatory variables, on the profitability
equation exists among the forty-six profit studies surveyed by Weiss (1974).
Formally, the first empirical attempt in testing the effect of unionism on
profitability comes from Freeman (1983). By using a sample of industry-level
data for the period 1958-1976, Freeman (1983) scrutinized what the effects of
2
unionism on industry profitability measures, such as price-cost margin and the
ratio of quasi-rents to capital, were. His conclusion was that, regarding both
profitability measures, there was a negative and substantial effect of
unionization on industry profitability. This conclusion was later supported by the
study of Clark (1984), using a sample of North American product-line business
data and rate of returns on sales and capital, separately, as a measure of
profitability. After these pioneering studies, a body of studies exist which deal
with the same hypothesis under different sample of data and different measures
of both unionization and profits. The studies of Hirsch (1991a; 1991b), Becker
and Olson (1992), Bronars and Deere (1990) and Abowd (1989), among others,
are examples of the recent ones.
In general, the effect of unionism on profitability differs, depending on the
sample of data, the measure of profitability, the measure of unionization and the
estimation technique used although there has been an agreement on the
negative union effect on profitability. Therefore, the dispute pertaining to this
literature is ongoing as to the appropriateness of profit measure, union measure
and the estimation technique.
At this point, it is appropriate to categorize the studies into two general
groups in terms of the sample of data: industry level studies and firm level
studies. To categorize studies in this manner is important because the
profitability measure that is chosen largely depends on the availability of
information on the data. The most common measures used in industry level
studies are price-cost margin and rate of return to capital (accounting profit
measures) while in firm level studies, they are Tobin's q and excess value
(market measure) in addition to accounting profit measures. Unfortunately,
none of these measures is theoretically and empirically found to be entirely
unambiguous and satisfactory. Both advantages and disadvantages of each of
these measures have been extensively discussed in the previous studies.
Using a sample of aggregate manufacturing data for 1972, Karier (1985)
estimated that the union captured about 68 percent of the price-cost margin
while, with a sample of firm level data for 1977, Becker and Olson (1992)
concluded that the price cost margin for a fully unionized firm was lower by 30
percent. In comparing the impact of unionization on both accounting profit and
market value measures, Hirsch (1991b, p. 11) made the argument that
...these effects on market value can differ from unionism's impact on current earnings. For example, a union may significantly decrease current earnings but not market value if investors believe the firm can adjust in the future or in some way offset the union's current negative impact. Or, a union may have little immediate impact on earnings but significantly decrease market value if investors expect the union to have a detrimental effect on firm growth and future earnings.
Using firm-level data, in another study, Hirsch (1991a) found that while Tobin's
q (market value measure) for an average unionized firm was lower by an
average 12.4 percent, the rate of return on capital (accounting measure of
earnings) was lower by 9.2 percent, which supported his argument above.
The eariy studies, generally and necessarily, have used industry
unionization as a proxy for firm unionization measure because of the
unavailability of data on firm-level unionization, while most of their data, except
unionization, came from firm level. The studies of Salinger (1984), Connolly et
al. (1986) and Hirsch and Connolly (1987) are examples for those of the group.
The common conclusion of most of those studies was that the effect of
unionism on profitability was negative and statistically significant. As more data
on the firm level became available, recent studies started to devote more
attention to re-investigating the relationship between unionization and
profitability by using more rigorous data on unionization. The studies, which
compared the results, coming from the industry level unionization with those
obtained from the firm-level unionization, argued that to use the firm union
measure as a proxy of unionization in micro level data provided better results
than using the Industry union measure. For example, Hirsch (1991a) and
Becker and Olson (1992) found that the effect of firm union measure on
profitability was always negative and significant while the impact of industry
union measure on profitability was mostly insignificant. Then, in this respect,
Hirsch (1991a) made a conclusion that firm level union measures were
preferable to Industry union measures since industry union measures suffered
from the intra-industry variability.
More Importantly, the conclusion obtained by the previous studies on the
magnitude of the effect of union on profitability appeared to be sensitive, not
5
only to the level of data, but also to alternative measures of the percentage
unionized which enriched the understanding of unionization impact even though
the same spline function was used. In their studies, Hirsch (1991a) and Becker
and Olson (1992) separately used dummy variables to capture the
union-nonunion distinction, rather than basing them solitarily on the variation in
the percentage unionized. By replacing firm unionization with the same spline
function, the findings of Hirsch revealed that a significant portion of union gain
rises across 30 and over percent of unionization for the market value equation,
and over all the range of percentage unionized for the accounting profit
equation. But, later Becker and Olson (1992) found that this occurred over the
first 30 percent of unionization for both profit equations.
The most important one of the limitations in most of the previous
empirical studies on unions and firm profitability has been the difficulty in
obtaining firm-level measures of union coverage. Although recent studies could
obtain firm-level union measure, the data on this variable came from only one
specific year. Therefore, these studies were forced to combine the information
on firm union coverage or union membership which came from only one
specific year with other information related to the firm performance which came
^Some previous studies defended that use of dummy variable which is alternative to percentage unionization has more empirical advantage to alternative one. For example, Clark (1984) made an argument that because of the possibility on the inaccuracy of union data collected use of dummy variable is more realistic to accurate identify whether a business is unionized or nonunionized.
6
from more than one year. Also, this limitation on availability of union coverage
data over time forced the previous studies to use certain kind of estimation
technique. Obviously, since the previous studies could not specify their profit
equation so as to capture the possible effects of omitted or mismeasured firm
specific, industry specific and/or economy wide variables which exhibited
variation across firms and/or through time, and also were correlated with firm
unionization and profitability, the estimated union effects on profitability would
be biased if this was the case. Then, the conclusion on the magnitude of the
union effect on profitability would be misleading. In practice, one feasible
technique to rule out the possibility that the parameter estimates suffer from
unobserved or mismeasured determinants of profit is to use the fixed effect
under panel data. For example, Hirsch in 1991 (1991a, p. 74) argues that
...despite the inclusion of detailed control variables in our regressions, there may be omitted determinants of profitability correlated with the union coverage variable. A potential method to account for omitted variable bias is the use of a fixed-effect or difference model, wherein changes in profitability are estimated as a function of changes in union coverage. But such estimation is not possible here, since union coverage is measured only for 1977.
Also, In another study, Hirsch (1991b, pp. 122-123) in 1991 indicates that
...further study of the relationship of union coverage with economic performance is needed. ...analysis of the performance of union an nonunion companies during a more current period is essential. It is certainly possible that negative union effects
For example, in his recent studies, Hirsch (1991a) combined the sample of one year (1977) on union variable with the sample of more than one year (1972-1980).
on firm performance have been partially mitigated in the 1980s, owing the management and union response both to the forces of domestic and foreign competition and to the poor performance outcomes In the past.
The main objective of this study is to analyze statistically and
comprehensively the effect of unionism on firm profitability and the sensitivity of
such effect to different measures of firm profitability and unionization, and
different estimation techniques.
The specific objectives of the study are:
(1) To analyze the impact of firm unionization on firm profitability
regarding the alternative econometric techniques; that is, ordinary
least squares versus fixed-effect technique.
(2) To examine the impact of firm unionization on firm profitability
regarding the alternative definitions of profit; that is, accounting profit
measures versus market value measures.
(3) To examine the impact of firm unionization on firm profitability
regarding the alternative measures of unions; that is, union coverage
versus union dummies.
(4) To examine the impact of an alternative level of aggregation of union
variable on union-profit effect; that is, firm union variable versus
industry union variable.
The secondary objective of this study is to investigate whether the
returns to research and development investment are still major source of union
8
gains, even though the investment on research and development falls in recent
years.
The basic analytical tools being used to carry out the stated objectives
are ordinary least squares, fixed effect and random effect techniques. In testing
for heteroscedasticity under ordinary least squares technique, both the
Goldfeld-Quandt test (1965) and the White test (1980) are used. In the
presence of heteroscedastic error terms, the White correction procedure (1980)
is employed to correct the residual terms. In addition, the investigation for any
endogeneity between the unionization and profitability is carried out using the
Hausman exogeneity test (1978).
At this point, before starting the subsequent chapter, it is appropriate to
set out the overall plan of this study. The present chapter provides general
information on the statement of the problem, the basic objectives of the study,
and the general analytical tools and methodology used to carry out these
objectives. Chapter II presents a selective review of the literature related to the
effects of unions on profitability. The focus of Chapter II is on the measures of
the profitability and unionism constructed and on the empirical techniques used
in testing the effect of unionism on profitability. Chapter III discusses the
empirical procedures, the analytical tools, the data sets used and the variables
constructed from them. Chapter IV covers the empirical results of this study, an
analysis of the studies, by comparing them with each other and also with
previous studies. Finally, Chapter V provides a summary of the empirical
results, an assertion of the contribution of this study to the existing literature,
the major limitations of this study and the areas suggested for further research.
CHAPTER II
LITERATURE REVIEW
The purpose of Chapter II is to present a selective review of literature
related to the relationship between the profitability and unionization. This
chapter covers the literature from the influential studies of Freeman (1983) and
Clark (1984) to recent studies.•• Since the sample data and the constructed
profitability measures are generally in two forms, the review of literature is
accomplished by dividing the studies into two groups: industry level studies and
firm level studies.
2.1 Industry Level Studies
Since among the industry level studies. Freeman's (1983) was the first to
attempt to investigate the relationship between profitability and unionization
directly, it is appropriate to start the review of literature with his study.
As previously noted, the studies of Freeman (1983) and Clark (1984) are the first to investigate the relationship between the profitability and unionization directly. Undoubtedly, prior to these studies, there is an explosion of research concerning the relationship between the unionization and economic performance such as productivity. But, in these studies, the relationship between the unionization and profitability is examined indirectly. For example, the studies of Brown and Medoff (1978), and Clark (1980) suggest that unionization does not increase production costs and then that unionization does not reduce profits, rather may increase profits.
10
11
By constructing two separate samples of 139 three-digit industries from
Annual Survey of Manufactures, ASM,: 1958-1976 and of 68 three-digit
industries from Internal Revenue Service, IRS,: 1965-1976, Freeman examined
the possible effect of unionism on the industry profitability which was measured
by price-cost margin, PCM, and the ratio of quasi-rents to capital. Under the
ordinary least squares technique, he regressed both profit measures on the
union variable, the ratio of capital to business receipts and other market
structure variables such as industry concentration, advertising intensity,
absolute capital requirements and relative minimum effect scale. According to
his coefficient estimates which were evaluated at the mean of dependent
variable, for the IRS sample, union lowered the rate of return to capital by some
61 percent for a fully unionized manufacturing industry. Regarding the PCM
measure, his results were similar to the rate of return measure; negative and
significant union effect on profitability. Although his findings from the IRS data
set were consistent with those from ASM data set, the size of the effect
obtained from both data sets differed significantly. When union coverage
moved from 0 to 100 percent, both measures for the industry profitability fell
between 13 and 44 percent for the IRS sample while they fell between 13 and
19 percent for the ASM sample. In his same study. Freeman additionally
investigated how the effects of union behaved in concentrated and
In literature, whether advertising is treated as a mean to entry or to create a barrier to entry has been in dispute. For this issue, see Mann (1974), Brozen (1974), Martin (1979), Strickland and Weiss (1976), and Gisser (1991).
12
unconcentrated industries. His findings revealed that the negative effect of
unionism on industry profitability was limited to highly concentrated industries.
In this respect, he concluded that unionism had essentially no impact on
profitability in more competitive industries but a sizeable negative effect in the
concentrated industries.
Later, Karier (1985) extended the study of Freeman to a sample of data
from the 1972 Census of Manufactures and Annual Survey of Manufactures.
For this purpose, he used only the PCM measure for the profitability and
included a set of regional dummies in addition to the explanatory variables of
Freeman into his specification. His ordinary least squares results were in
complete agreement with the findings of Freeman, that is, unions significantly
lowered the PCM. His finding revealed that union captures about 68 percent of
potential monopoly profit. Karier also investigated how this effect behaved in
highly, moderately and lowly concentrated industries. His findings showed that
unions had a negative and significant Impact on PCM in highly and moderately
concentrated industries yet they had little effect on PCM when markets were
more competitive which was also consistent with Freeman's conclusion.
More recently, in another study, Karier (1988) enlarged his previous
study by focusing more on the source of union rent sharing behavior and
therefore on interaction terms of concentration-unionization as well as
concentration-imports. With a sample of 107 three-digit manufacturing
industries over the 1965-1980 period from the same data sources as described
13
above, his generalized least squares results indicated that in addition to unions,
imports also reduced total monopoly profits. His results also revealed that total
monopoly profits were divided among imports (14 percent), unions (47 percent)
and firms (39 percent).
Until 1986, the issue on the endogeneity of unionization with respect to
profitability was ignored in virtually all empirical studies which investigated the
relationship between industry unionization and profitability. Voos and Mishel
(1986a) were the first not only to make the argument that the effect of
unionization on industry profitability was underestimated in the standard single
equation technique if unions were more likely to organize in the most profitable
industries but also to consider the endogeneity of unionization with respect to
profitability in their estimation procedure.
By constructing a sample of 139 three-digit manufacturing industries from
the Annual Survey of Manufactures and Census of Manufactures for the period
of 1968-1972, Voos and Mishel estimated a two-equation model, wherein
unionization was treated as endogenous with respect to profitability and a single
equation model, wherein unionization is assumed to be exogenous. In both the
single equation model and the first equation of a two-equation model, they
regressed the price-cost margin measure on industry union coverage and
market structure characteristics (concentration ratio, measure of capital
intensity, advertising intensity, import penetration, barriers to entry, size of
establishments and sales growth). In the second equation of their two-equation
14
model, they regressed logistic transformation of union coverage on average
characteristics of the workers employed in the Industry (race, sex, education,
experience, types of job), characteristics of industry (concentration ratio,
capital-to-labor ratio, injury rate) that affected the level of unionization and PCM.
According to their OLS estimates which were evaluated at the mean of
dependent variable, a fully unionized industry had a 22.6 percent lower PCM
than a nonunionized industry. On the other hand, their two-stage estimates of
the union impact on profits suggested that union reduced PCM by 35 percent in
the industries that they organized. Since two-stage estimates were found to be
significantly larger than OLS estimates, they concluded that endogeneity was
an important consideration in estimating the union effect on profits.
On the other hand, Domowitz, Hubbard and Petersen (1986) argued that
cross-sectional estimates of the "union-profits" effect were likely to be biased
because of the exclusion of the demand effect in estimating the impact of
unionization on the profitability. Therefore, in addition to industry union
measure and industry market structure variables (concentration rate, the ratio of
capital to output, the ratio of advertising to sales), they also included
economy-wide unemployment rate and some interaction terms between
unemployment rate and rest of the variables to capture the aggregate demand
effect in their OLS estimations. For a sample of 284 four-digit manufacturing
industries from the Census of Manufactures over the 1958-1981 period, their
results indicated that on the average PCM was reduced by about 25 percent if
15
union coverage moved from 0 percent to 100 percent. Moreover, they found
little evidence to substantiate the claim that the negative effect of unionism on
industry profitability existed necessarily in highly concentrated industries.
According to their OLS estimates, both unionized and concentrated industries
had higher PCM than the average PCM during periods of low unemployment
whereas they had lower PCM than the average PCM usually experienced
during periods of high unemployment. Therefore, their conclusion followed that
the effect of unionization on PCM depended on the state of demand.
Recently, in another study, Domowitz et al. (1988) re-examined the
relationship between markup of price over marginal cost and unionization by
using the same sample of data over the 1958-1981 period. They estimated the
markup of price over marginal cost as a function of import-adjusted
concentration ratio and union coverage ratio. Their estimation results provided
evidence that markup was reduced by about 17 percent when union coverage
was equal to 100 percent relative to when it was zero.
2.2 Firm or Business Level Studies
In addition to the industry level studies, there have been a body of
studies concerning the relationship between unionization and accounting profit
measures (and/or market value measures) in the micro level data such as firm
or business level data.
16
Among these studies, Clark's (1984) was the first to examine the
relationship between unionization and performance (profitability, growth of sales
and productivity) in a sample of 900 North American product-line businesses
which were classified as large manufacturing firms and participated in the PIMS
project from 1970 to 1980. In a part of his study related to the relationship
between unionization and profitability, he utilized the rate of returns on sales
and capital, separately, as a function of unionization measures, firm control
variables, industry market structure variables and industry labor market
variables. His OLS results confirmed that the rate of return on sales and on
capital in a fully unionized business was about 18 percent lower than that in
nonunionized business. Moreover, when dummy variables which were
constructed from percentage unionization were used instead of percentage
unionization for a line of businesses, his spline function suggested that a
significant and negative union effect for the businesses wherein unionization
ranged from 0 to 30 percent and no significant union effect for the rest of the
businesses. In order to see whether a firm's market share provided more likely
sources for union rents, Clark re-estimated the same regressions for both
low-market share firms and high-market share firms separately. In contrast to
Freeman (1983) and other previous studies, Clark's results revealed that unions
decreased profits only among the firms which had small market shares (here
firms with less than a 10 percent share). His findings showed that for
low-market share firms, there was about a 40 percent decline in rate of return
17
on both capital and sales due to unionization while for the high-market share
firms, the effect of unionization on profitability measures was statistically
insignificant.
Like Clark (1984), in 1984 Salinger investigated the relationships among
unionization (here, industry union coverage was proxied for firm union
coverage), market structure and profitability measures such as Tobin's q and
rate of return on capital by using large manufacturing firm data from 1979
Compustat tape and PICA data set. First, he argued that Tobin's q was a
better measure of monopoly profits than other single period profit measures
because it measures the long-run monopoly power.^ Then, estimating a
nonlinear least squares profit equation for Tobin's q and rate of return on
capital, he found that the unionization had a negative and significant effect on
Tobin's q while it had insignificant impact on rate of return. His findings showed
that unions capture 11 percent of the monopoly rents (for Tobin's q). In
addition, he found that the negative effects of unionization were limited to highly
^On the other hand. Shepherd (1986) argued that Tobin's q as a measure of firm profitability was more likely to be inferior to other profit measures because: (1) Tobin's q ratio was an event of capital markets, not of the firm itself. (2) Tobin's q ratio was complicated to measure since the market value of the firm, which is in the numerator of the Tobin's q ratio, was a proxy book value, and its accuracy was questionable, and hybrid book-and-replacement value figure, which is in the denominator of Tobin's q ratio, included unknown degrees of error and was likely biased.
18
concentrated industries and that both concentration and entry barriers made it
easier for unions to raise wages."*
Later, Salinger's results for firm level data were also supported by the
study of Voos and Mishel (1986b), by using a sample of data from the U.S.
Supermarket Chains for the 1970-1974 period. The purpose of Voos and
Mishel's study was to analyze whether unionization reduced accounting profits
for large companies in the supermarket industry and whether its impact was
larger when local markets were more concentrated. For this purpose, the ratio
of firm profits to sales was regressed on the market structure and firm
characteristics variables. According to their generalized least squares
estimates, unionization reduced the ratio of firm profits to sales by 76 percent
on average in the supermarket firms in standard-metropolitan area markets but
the reduction in profits for firms in concentrated industries was larger than that
in unconcentrated industries.
In addition, Connolly, Hirsch and Hirschey (1986) scrutinized the
relationship between unionization and research and development expenditure
as well as the association between unionization and market value of firm (here
excess value) by using the same proxy for unionization as in the study of
" For a review of the literature on the relationship between profitability and market concentration rate and/or market share, see Gale (1972), Demsetz (1973, 1976), Weiss (1974), Cowling and Waterson (1976), Phillips (1976), Bond and Greenberg (1976), Carter (1978), Porter (1979), Ravenscraft (1983), Smirlock, Gilligan and Marshall (1984), Schmalensee (1985), Harris (1986), Shepherd (1986), Scott and Pascoe (1986), Uri (1988) and Amato and Wilder (1990).
19
Salinger (1984). For a sample of 367 firms from the Fortune 500 for 1977,
Connolly et al. (1986) specified one equation for each relationship. In the first
equation, excess value, which was a measure of firm profitability, was specified
to be a function of unionization, intangible capital, output market structure and
control variables. On the other hand, in the second equation, research and
development expenditures as a measure of intangible capital were set to be a
function of unionization, excess value and other control variables. Whereas no
evidence on misspecification of excess value equation was found from
estimating it with the ordinary least squares technique, clear evidence on
misspecification of research and development equation was found as a result of
using ordinary least squares technique. Therefore, they used two-stage least
squares technique to capture the effect of endogenous unionization measure
with respect to research and development expenditure. According to ordinary
least squares estimates from the first equation, unions had a negative and
significant effect on firm excess value. From the second equation, both the
ordinary least squares and two stage least squares results revealed that the
impact of unionization on research and development expenditures was also
negative and significant. In contrast to Freeman (1983), Karier (1985) and
Salinger (1984), Connolly et al. (1986) could not find any strong evidence that
union captured rents associated with the output market structure (concentration
ratio, market share). In other words, their estimates rejected the proposition
that market share and concentration were a significant source of union gains.
20
In another study, Hirsch and Connolly (1987) extended the study of
Connolly et al. (1986) to a sample of 367 firms data from Fortune 500 firms in
1977 in order to re-examine the relationship between unionization and firm
profitability. By using Tobin's q and the rate of return on sales as the firm
profitability measure, ordinary and two-stage least squares results provided
evidence that unionization decreased firm profits, but this negative effect was
sensitive to model specification. For a fully unionized firm, union reduced
Tobin's q by 13-20 percent while it reduced the rate of return on sales from 11
to 17 percent. In addition, their estimates indicated that unions captured
monopoly profits associated with firm's expenditure on research and
development, its protection from foreign competition and its market share, that
did not associated with concentration. Obviously, this finding was in sharp
contrast with the conclusion of Freeman (1983), Karier (1985) and Salinger
(1984) whose results indicated that unions captured monopoly profits
associated with industry concentration.
Until Hirsch's study (1990), in virtually all firm level studies, firm
unionization was proxied to industry union coverage because firm level
unionization measure was not explicitly and publicly available. Hirsch was the
first to utilize firm specific union coverage from his own recent survey. By using
his survey data on firm union coverage, market share and concentration, Hirsch
(1990) Investigated the impact of unionization on firm profitability, and tested
whether market power was a primary source of union gains. For this purpose.
21
he matched the 1977 union coverage variable to the other determinants of profit
from different sources and different years over the 1972-80 period. For the
estimation, he used a two-step model because the error term was serially
correlated across the years within firms. First, firm profitability measures such
as the natural log of Tobin's q and the rate of return on capital were regressed
on 246 manufacturing firm dummies and all time varying profit determinants
such as research and development intensity, advertising intensity, firm size,
capital intensity, firm specific sales growth and foreign competition. In the
second step, the coefficients of the firm dummies were first regressed on
time-invariant variables such as firm-specific union coverage, sales weighted
four-firm concentration ratio across industries, firm's sales-weighted market
share across industries and industry union coverage in the firm's principal
two-or three-digit industry. Then, the regression equations were weighted by
the inverse of the standard errors of the firm dummy coefficients. The
estimated union impact on firm profitability in this study was consistent with the
previous estimates. For 247 large manufacturing companies over the period of
1972-1980, he reported that a change from 0 percent to 50 percent unionization
implied a decrease in Tobin's q by 9.0 percent and in the rate of return on
capital by 6.9 percent. However, he found no evidence to support the
hypothesis that industry concentration and firm's market share were major
sources of union gains, but confirmed that the major sources of union gains
were the returns of research and development investment. Therefore, his result
22
was also more likely to support the finding of Clark (1984) which stated that
unions only decrease profits of businesses with small market shares.
In more recent studies, Hirsch (1991a, 1991b) examined the relationship
between firm profitability and unionization by using a firm level coverage
measure for a larger sample of manufacturing firms relative to that in previous
studies. In both studies, he combined various data for a sample of 705
manufacturing firms over 1972-80 period from R&D Master File, Compustat, his
union survey, Word's Business Directory and the Annual Survey of
Manufactures. For the statistical analysis, he used the same estimation method
as in his previous study, a two-step model. By using union coverage at the firm
level, he found a strong negative relationship between unionization and Tobin's
q as well as between the rate of return on capital and unionization, and also
confirmed that union effects on profitability still remained important over the
1972-80 period. His findings revealed that for an average unionized firm,
Tobin's q and rate of return to capital were lower by 12.4 percent and 9.2
percent, respectively. On the other hand, when he used industry union
coverage as a proxy for firm unionization, he found no significant union effect
on firm profitability. Furthermore, by using union coverage dummies
constructed from union coverage ratio, Hirsch found that unionization had
significant impact on the rate of return on capital in all ranges of unionized
percentage, whereas, a significant part of union gains existed across 30 and
over percent of unionization in the Tobin's q equation.
23
In 1992, Becker and Olson (1992) extended the literature on how
unionization affected firm profitability and what the source of union rent sharing
behavior was, by employing a different proxy for firm unionization measure than
used by Hirsch (1991a, 1991b). The firm unionization measure was the ratio of
pension plan enrollment through collective bargaining divided by the total firm
pension plan enrollment, not directly the ratio of unionized employees to the
total firm employment. However, they used not only firm level but also industry
level unionization measures in their estimation. Their main data sources on a
sample of about 300 U.S firms were the 1977 Annual Return-Report of
Employee Benefit Plan (form 5500) and the Compustat data set. As a proxy for
firm profitability, excess value (market value), which was produced by the
difference between the capital market valuation of firm assets, and rate of
return on sales (accounting profit measure) were employed.
In the first part of their study, Becker and Olson searched for the answer
as to how industry and firm unionization measures, market concentration,
intangible assets and other control variables affected the excess value of firm
and the rate of return on sales under the ordinary least squares technique.
According to their parameter estimates, only the firm unionization measure was
associated with a substantially lower rate of return on sales and excess value,
although both firm unionization and industry unionization had a negative effect
on both measures of firm profitability. More specifically, their coefficient
estimates suggested that a fully unionized firm had an 18 percent lower excess
24
value below the mean than that of a nonunionized firm while at the mean of the
dependent variable, a 100 percent unionized firm had a 30 percent lower rate of
return on sales than that of a nonunionized firm. When they investigated the
possible variations in the union effect on firm profitability over the range of
unionization, they found that a negative and significant effect existed for the first
30 percent of unionization. Finally, from the estimations of their profit equations
with the interaction terms between unionization and research and development
expenditure, they concluded that research and development expenditure was
an Important source of union gains in the excess value equation, though
significantly less so in the rate of return equation.
In the literature, there have been some studies which dealt with the
profitability-unionization relationship for the manufacturing firms of countries
other than the U.S. manufacturing firms. Among the recent ones, using the
establishment level data from the 1980 and 1984 Workplace Industrial Relations
Surveys which covers 2040 and 2019 British establishments with 25 or more
employees, respectively, Machin and Steward (1990) estimated the effect of
unionization on financial performance of establishments under statistical method
and financial performance measures different from the previous studies. In
estimating the specified performance equation, they used "the ordered probit
estimate method" since their financial performance measure was qualitative
such as below, above or about industry average of financial performance. They
regressed this qualitative dependent variable on establishment-level variables
25
(unionization, market share, capital intensity and growth in the demand
variables) and on industry level variables (five firm concentration ratios, industry
level price-cost margin variables). Although the profitability measure and the
estimation method were quite different from the standard approach, their
findings supported the previous results of U.S. firm level studies: a negative and
significant union effect on firm performance. More specifically, unionized
establishments were more likely to have about average and especially, below
average levels of industry performance. Their results also indicated that unions
significantly restricted financial performance for more likely establishments
which had a larger market share in terms of employment.
In another study, Machin (1991) used a sample of 290 large British
manufacturing firms from the Data Stream and Exstat data bank. Census of
Production and Survey of Companies for 1984 and 1985 in order to investigate
the relationship between unionization and firm profitability. His study was the
first to employ a sample of British firms based on the actual profit data.
The ratio of accounting profit to sales was used as a proxy for the firm profit
measure while dummy variables were constructed for the firm unionization
measure. In addition to firm and industry unionization measures, he employed
market share (in terms of firm sales), industry concentration ratio, and some of
the firm level variables as independent variables. The coefficient on the
recognition dummy for firm unionization was estimated to be negative and
statistically significant, suggesting that unions reduced the rate of return on
26
sales by some 2.3 percentage points. His OLS results also confirmed that the
ability of unions to capture monopoly rents was not related to concentration, but
it was related to the firm specific market share (firms which have a high market
share) and to the degree of union activity in the firm's operating industry
(industries which are highly unionized).
In the case of Japanese manufacturing firms, more recently, Brunello
(1992) analyzed the relationship between unionization and firm performance
(profit, wage and productivity) for 979 unionized and nonunionized
manufacturing firms which were drawn from the 1987 issue of the Yearbook of
Japanese Unlisted Companies. In his profitability equation, the rate of return on
invested capital and the profits-to-sales ratio were set to be a function of
technological change measures (capital-to-sale ratio), market structure
measures (market share of the firms), labor quality measures (average age of
employees, the ratio of female to male employees), firm union measure which
took the value of 1 if the firm is unionized, 0 if not unionized, and some
interaction terms.
By using both OLS technique and the combination of Heckman's
two-step procedure with instrumental variables for factor inputs, IVM, Brunello
found similar results to most of U.S. firm level studies: the effect of unions on
firm profitability was substantial and negative. However, if the dependent
variable was the rate of return on invested capital, the unionization effect on
profitability was -19.56 percent with OLS estimates and -23.80 percent with
27
IVM. On the other hand, if the dependent variable was the profits-to-sales ratio,
the effect of unions on profitability was -36.5 percent with OLS estimates and
-37.3 percent with IVM estimates. Finally, his results revealed that the union
effect on firm performance was substantially smaller in small-and medium-size
firms than that in large firms.
Although in the literature, the profitability measures, which were
constructed and widely used by most of the previous studies, are the price-cost
margin, the rate of return on capital, and the market value of firm, there have
been some studies which used completely different firm profitability proxies (for
example, equity value of the firm and unexpected changes in shareholder
wealth) as well as different unionization measures (petition date, certification
date and unexpected changes in unionized workers wages).
For example, in 1984, Ruback and Zimmerman (1984) investigated the
relationship between unionization and equity value of the firm (abnormal
monthly common stock returns) in the respect of union representation election.
In other words, they analyzed how petitions and certification of election affected
the stock prices of firms. The expected return of a firm's securities, which was
the difference between the actual market return of securities and the estimated
market return of securities was proxied to the abnormal common stock returns
of firms. With a sample of 253 manufacturing firms in the New York Stock
Exchange over the period 1961-80, they regressed the abnormal returns on a
series of binary variables representing petition date, election date, different
28
unions and different industries. For the estimation of the effect of different
unions and industries on the equity value of firms, they used weighted least
squares in order to correct for heteroscedasticity. The evidence presented in
their study suggested that union representation election tremendously lowered
the equity value of firms; that is, stock prices fell by 1.38 percent when unions
petitioned for representation elections. They also found that the combined
abnormal return on the petition and certification month fell 3.8 percent when the
union won the election and 1.3 percent when the union lost the election.
Recently, by analyzing which firms experienced the largest declines in
firm's equity value because of new union representation elections and which
factors helped to incur these large equity losses resulting from those elections,
Bronars and Deere (1990) extended the study of Ruback and Zimmerman
(1984). Bronars and Deere used the same data source as did Ruback et al.
(1984), but used a different sample (firms in all sectors except service sector)
and event. By using a sample of unionization elections, which were limited to
those occurring in firms in the New York Stock Exchange firms between
February 1962 and September 1980, from National Labor Relations Board
Files, they picked up the sample of 255 elections at 137 different firms. Two
different dependent variables, the percentage change in the firm's equity value
in certification month and the percentage change in the firm's equity value in
petition month are set to be a function of a vector of observable characteristics
of firm and industry. According to their weighted least squares estimates, union
29
election events had a negative and significant impact on the firm profitability
only during the month in which the petition for an election was filed but, the
effects during the certification month were much weaker than the petition month
as Ruback and Zimmerman indicated. In addition, they found that equity losses
were the largest in industries where union wage gains were the highest and
unionization rates were the largest; that is, a 10 percent point increase in an
industry's average unionization rate caused an additional decline in equity value
of 46-87 percent. They also confirmed that labor-intensive firms suffered from
the largest declines in equity value from union organization elections, and the
presence or absence of right-to-work laws had no effect on a firm's expected
losses. Finally, the hypothesis of union spillover effects was supported by their
study.
On the other hand, using a sample of about 1200 large private
manufacturing and nonmanufacturing firms, Becker and Olson (1989) analyzed
the relationship between unionization and firm performance in different aspect
from that of Ruback and Zimmerman (1984) and Bronars and Deere (1990).
Specifically, they examined how union and nonunion firms differed in their
allocation of business risk and financial performance between shareholders and
employees over the period 1970-81. By combining unionization data from the
1977 Annual Return-Report of Employee Benefit Plan (form 5500) with CRSP
(Center for Research on Security Prices) stock price and return data base and
by applying the OLS technique, they estimated the shareholder returns as a
30
function of stockholder systematic risk, unionization and sometimes two-digit
industry dummies. Their results showed that risk-adjusted returns earned by
shareholders in unionized firms were significantly lower than those earned by
shareholders in nonunionized firms.
One of the studies which examined the relationship between unionization
and profitability in a different aspect from the standard approach came from
Abowd (1989). By using a sample of both manufacturing and
nonmanufacturing firms, Abowd Investigated how the unexpected changes in
union labor costs were related to unexpected changes in the market value of
the common stock of the employer firm. More specifically, the purpose of his
study was to find the answer to the following two questions: (1) was there a
negative and a dollar-for-dollar tradeoff between these two variables?, (2) did
collective bargaining maximize the sum of shareholder's wealth (shareholder's
wealth was measured by the value of common stock of the firm) and union
members' wealth (union wealth change was measured using the wage rate
growth method)? In his methodology, first, he converted information on wages
rates, settlement data, unit size, and industry classification from collective
bargaining agreements Into an estimate of the labor cost. Second, he
estimated the present value of the total labor costs that decompose the cost of
realized collective bargaining into expected and unexpected components.
Third, he estimated the unexpected change in shareholders' wealth by using
security price movements around the time of settlement. Finally, by applying
31
the OLS technique, he regressed unexpected changes in union labor cost on
the unexpected changes in common stock value of the firm. His findings
indicated there was a negative and a dollar-for-dollar tradeoff between the value
of common stock and unexpected changes in collective bargaining labor costs.
He also found that the collective bargaining maximize the sum of shareholders'
and union members' wealth.
Unfortunately, there is no study which investigated the impact of
unionization on only regulated manufacturing firms despite the fact that there
has been explosive research on the issue of union-profitability. However, there
are some studies which investigated the unionization-profitability relationship for
regulated trucking and airiine firms. Among these studies, for a small number
of pre-regulated trucking firms. Rose (1985a) partly examined the relationship
between a firm's share price responses to regulated reforms and unionization.
By using 15 firms which were continuously trading general commodity carriers
and 6 Interstate Commerce Commission events which were related in entry and
rate over the 1977-79 period, he estimated the share price responses to
regulatory reforms as a function of unionization (percentage of traffic handled
by union drivers) and other company's operation characteristics. He found that
unions had a negative effect on share price responses to entry and rate events.
His findings supported the hypothesis of union rent sharing, and also confirmed
32
that unions were not expected to reduce their claim proportionally with a loss of
monopoly rents because of the deregulation of regulated trucking firms.^
Similarly, in another study. Rose (1985b) found support for the
conclusion which stated that deregulation caused the reduction of the union's
share of pro-regulation rents by examining the wages of unionized workers In
trucking industries. On the other hand, the study of Hendricks, Feuille and
Szerszen (1980) showed that unionized labor in airline firms captured the part
of regulation rents while deregulation of airiine industry would have little impact
on the power of organized workers. In later study. Rose (1985c) argued that
the previous studies, which examined the relationship between union-nonunion
wage differentials in trucking industries in order to find the share of unions in
regulated industries, might help to quantify the magnitude of union's effect on
wages, but they did not clearly reveal the spline of rents between the union and
trucking companies. He also argued that his previous study (Rose, 1985a)
could provide only a lower bound on rents captured by the union. Therefore,
he tried to complete the shortcomings of his previous study by using a more
appropriate profit measure and equations for the sample of 21 trucking firms.
In addition to Rose's study, some studies, which investigated the relationship between union-nonunion wage differentials, found that some portion of rents which resulted from regulation of some industries were captured by organized labor. For instance, Moore (1978) found that the oven/vhelming portion of the regulation rents in the trucking industry was captured by unions. Another similar example is in the study of Hendricks (1975) which suggested that unions had big power to get part of the rents in regulated electric utilities.
33
In his study, he first used five sample periods: 1973, 1974-75, 1976-77,
1978-79 and 1980-82. Then, he estimated his equation over the regulatory
period, 1973 through 1977 because he expected that the 1978-79 and 1980-82
results would be quite different since in these periods there was deregulation in
the trucking industry. He regressed the ratio of excess market value to sales
on unionization rates, the proportion of total revenues accounted for by motor
carrier revenues, the ratio of predicted loss in firm value to intercity motor
carrier revenues, a dummy variable equal to 1 if the firm is a general freight
carrier and the ratio of firm average number of employees to motor carrier
sales. In general, his estimation results showed that teamsters receive
between 60 and 70 percent of regulated trucking industry rents.
2.3 Summary of Literature
Despite the studies concerning the relationship between unionization and
economic performance that has existed for a long time, the relationship
between profitability and unionization has been examined directly since the last
decade. The studies of Freeman (1983) and Clark (1984) were the first to
investigate the union effect on profitability under industry and line of business
level unionization measures, respectively. The findings of both studies are that
there is a negative and significant union effect on profitability whereas the
magnitude of this effect differs depending on the measures of unionization and
profitability and on empirical methodologies.
34
After these two pioneering studies, explosive research started to extend
the underiying relationship under a different sample of data, different profitability
measures such as accounting profit type and market value type, and different
unionization measures. However, the main findings of Freeman and Clark are
generally supported while the size of the union effect differed in virtually all
studies. Some examples of the firm level analysis include the studies of
Ruback and Zimmerman (1984), Voos and Mishel (1986b), Hirsch and Connolly
(1987), Hirsch (1990, 1991a, 1991b), Bronars and Deere (1990), and Becker
and Olson (1989, 1992).
In general, the existing studies produced satisfactory results for the
negative union effect on profitability. But, the size of the negative effect differed
depending on the sample of data, profit measures, union measures and
empirical methodologies used. However, the findings of virtually all the studies
suffer from the problems caused by the unobserved or omitted determinants of
profitability which are correlated with the union measures because the sample
of data used in these studies was generally not appropriate to allow for the
estimation technique capturing such determinants. In addition, since the
samples of data used significantly differed from study to study, it is quite difficult
to compare the findings of one study with those of another study in terms of the
sensitivity of profit-union relationship to the measures of profitability and
unionization, and to the estimation techniques. Undoubtedly, to overcome this
35
problem, the best strategy is to apply all possible of measures of unionization
and profitability, and the estimation techniques only under a sample of data.
CHAPTER III
METHODOLOGY AND DATA
The purpose of this chapter is to present the empirical procedures and
the analytical tools which will be employed in this study in order to investigate
the relationship between firm profitability and unionization. This chapter
consists of two subsections. In the first subsection, econometric models are
described and discussed to explore the association between firm profitability
and unionization. In the final section, the data sets used and the variables
constructed from them are presented.
3.1 Methodology
This study utilizes the standard profitability equation which is widely used
by Hirsch and Connolly (1987) and by Becker and Olson (1992) among others
in order to examine the relationships among union coverage, firm profitability
and other determinants of firm profitability. To make the findings on the
relationship between union status and profitability more comparable to the
results of previous studies, the specification of the profitability equation used by
previous studies is followed. The standard definition of profit (Karier, 1985) is
presented in the following equation (3.1).
36
37
PR-QP-RK-WL-OE, (3-1)
where 0 is firm output, P is the price of firm output, R is firm depreciation rate,
K is the value of firm capital, W is average wage in firm, L is labor hours in firm
and OE is other expenditures of firm such as material cost. If equation (3.1) is
reordered in terms of profit per sale, the following equation (3.2) results.
PR__QP-RK-WL-OE (3.2) TR TR ^
where total revenue, TR, is equal to PQ.
In this study, three distinct measures are employed for firm profitability
separately; the price-cost margin, the rate of return on invested capital and
excess value.^ As is well known, the first two measures are based on
accounting profits while the last one is based on a combination of accounting
and market-based profits. The rationale for adopting three separate measures
in this study is twofold. The first one is to be able to compare it with the bulk of
prior work. In literature, satisfactory agreement in choosing the appropriate and
unique measure for profitability has not been yet reached by the empirical
In addition to those profit measures, the use of Tobin's q was attempted. However, due to the many missing observations for calculating profit measure it was not used.
38
studies. In practice, the utilized measures have changed from study to study
because of the lack of availability of data on such measures and of the
researcher's preference for the chosen measures. For instance,
accounting-based measures such as price-cost margin and return on
investment have been widely used by Freeman (1983), Clark (1984), Karier
(1985), Voos and Mishel (1986a, 1986b), Hirsch and Connolly (1987), and
Becker and Olson (1992). On the other hand, in their studies Connolly, Hirsch
and Hirschey (1986), and Becker and Olson (1992) have used a hybrid
measure which is based on the combination of accounting-and market-based
profits.
Although several profitability measures have been examined in practice,
none of the measures used in the literature is found to be entirely unambiguous
and satisfactory. The principal advantages and disadvantages of each of the
possible profitability measures have been extensively discussed by the previous
studies. For example, one of the disadvantages of using price-cost margin as
an acceptable profit measure has been pointed out by Weiss (1974). Weiss
argued that the price-cost margin is an acceptable definition or a satisfactory
measure for profitability as long as variations in the value of the capital-sale
ratio are controlled. Similarly, Fisher (1987) argued that even under constant
returns, the profit-sales ratio cannot be equal to the price-cost margin because
of the problems connected with the valuation of capital. Then, he proposed to
use a non4lnear transformation [In(l-profiVsale)] instead of profit-sale ratio as
39
price-cost margin. But, by using Fisher's measure as the price-cost margin,
Machin (1991) found no clear-cut difference between the standard and Fisher's
proposed measures.
On the other hand, one of the principal advantages of using the
price-cost margin measure for profitability is its availability on the firm level. In
order to avoid making a biased choice among the profit measures, all three
measures are used separately.
The firm profitability will be treated as a function of a set of explanatory
variables including firm specific variables, industry specific variables and 16
two-digit industry dummies, which will then be augmented by a union variable
so that the basic estimated equation will be
6 2 1(
/f«i k•^ k-^
M e . iw
VPoyf^Pi/f 'iy^^E P2/*f^E Psy/w / - E P4/w2;/« eyf (3.3)
j=1...,N t=1...,T,
where k and t index the number of periods and the number of variables in the
corresponding groups, respectively, and j indexes the firm. In this study, three
models in terms of firm profit measure, O , are used to estimate the relationship
For discussion on the advantages and disadvantages of price-cost margin measure, see Liebowltz (1982), Collins and Preston (1968), and Cowling and Waterson (1976).
40
between the firm profitability and unionization. The first model employs
the price-cost margin (PCM) as a dependent variable. The second model
utilizes the rate of return on invested capital (ROI) as a measure of profit.
Finally, the third model uses excess value (EV) as a proxy for firm profit.
Under this model formulation, Oj, represents the profitability of firm j in
period t. Poj, Is an intercept. Z j is firm (or industry) union measure such as the
ratio of workers who are covered by union contracts to the total number of
employed workers of the firm (F-UN) (or the proportional of eligible workers who
are covered by union contracts in each firm's principal three-digit industry
(l-UN)) and p j, is its estimated coefficient. Zgjkt includes firm specific variables
such as the natural logarithm of capital-sale ratio (FCAP/S), firm sales growth
(GROWTH), advertising expenditures (ADVES/S), research and development
expenditures (RDES/S), natural logarithm of number of employees (FS) and the
market share of firm (MSF)."* pgjkt 'S the estimated coefficient vector
corresponding to a matrix of firm specific variables. Z3jkt represents industry
specific variables and ^^^ is their estimated coefficient vector. The industry
^As defined in Becker and Olson (1992), excess value measures the difference between the capital market valuation of firm assets reflecting their future profit potential and their cost.
" For the argument that research and development expenditures should be treated as current expenditures or investments in intangible assets, see Mansfield (1968), Bally (1972) and Branch (1974). In addition, see Palda (1964), Peles (1971), Schmalensee (1972), Mann (1974), Brozen (1974) and Comoner and Wilson (1979) for the argument that advertising expenditure should be treated as current expenditures or investments in intangible assets.
41
specific variables included in Z^^^ are the ratio of value of four-digit industry
inventory to value of four-digit industry shipments (IINVA/IS) and annual growth
rate in the four-digit industry sale (IGROWTH). The variables included in Z^-^^^
are the 16 two-digit industry dummies and p>^^ is the corresponding estimated
coefficient vector. Finally, EJ, is an error term with zero mean and constant
variance. Here, all variables except FCAP/S and FS are in linear form.
FCAP/S and FS are in natural logarithm form.^
All three profit models will be run for the same independent variables
except FCAP/S. While the first two models do incorporate FCAP/S ratio as an
explanatory variable. Model III does not include FCAP/S ratio^. So, the simple
specification of each of three models without 16 two-digit industry dummies will
be of the following form:
Model I:
PCM = Bo-B,iUN^^B^i^^)-B,{GROWTH)^B^i^^^yBs{^^)
^Be(FS) •'B.iMSFj -^BeC-^) B^{IGROWTH) +E.
^FCAP/S and FS variables are taken in natural log form since some observations in these variables were unevenly scattered; relative to the mean values of these variables.
^Becker and Olson (1992) found that FCAP/S variable did not have any economic impact on EV. They argued that the economic and statistical significance of the unionization variable did not change when they add FCAP/S variable to their estimation specification.
42
Model II:
ROI = B; B;{UN) ^ B ; ( ^ ) .B;(GRomH) .BK ^^^^^) ^B;{^^^)
If l l j
Model III:
i*v^o^u/T-LJv.T,-/>^0\/E5v ^**,RDES, EV = Bl'^B;\UN)^Bl\GROmH)^B;\^^^^^^^yBl\^^^^^^)^^^^^ s s
IINV, Bi{MSF)^B:r\^yBl\IGROWTH)^E*\
where UN represents firm union coverage (F-UN) or industry union coverage
(l-UN) or dummies which are constructed from firm union coverage and industry
union coverage.
Before presenting the empirical procedures used to estimate the above
models, it is appropriate to review the association between profitability and
explanatory variables. The hypotheses associated with firm profitability and
explanatory variables are as follow;
Unionization measure (UN): the relationship between unit costs (as a
result of increase in union wage) and profit rates depends on the nature of the
bargaining curve settlement (wage-employment pair lies on labor demand or
contract curve), the product demand elasticity and market structure. As
discussed by Clark (1984), under the case that wage-employment outcomes lie
43
on the labor demand curve (in respond to union wage gains, the firm adjusts
upward along its labor demand schedule), the association among all chosen
profitability measures and unionization measures can be negative unless
increase in union-labor productivity is grater than union wage increase. On the
other hand, if unions increase productivity, the relationship between unionization
and profitability can be positive or negative depending on the elasticity of
substitution.
If the wage-employment pair lies on the contract curve and if there is no
offsetting productivity increase, the rate of profit falls regardless of which
profitability measure is used. An efficient bargaining situation on a vertical
contract curve implies that union and firm will maximize the total value of the
enterprise (the sum of firm profits and union rents) and then barging over the
division of the profit. Under this barging regime, the firm makes production
decisions as though it faced the union wage. Thus, level of employment, the
stock of capital, the level of output and price level are unchanged after
unionization. Only the wage varies along the contract curve. Since the level of
employment, output and prices and the stock of capital are not affected by
union, given product demand and the opportunity cost of labor, a union wage
increase, with no offsetting productivity increase, will decrease the profit rate,
whether measured by PCM, ROI and EV (Addison and Hirsch, 1989). The
presence of union productivity effects in the bargaining contest may lead to
ambiguous results. Because the firm makes its decisions as though it faced
44
nonunion wage rates, the union productivity effect has the same impact as
would neutral technological change in the nonunion setting. The firm behaves
as though marginal costs had declined (and output had increased). When the
elasticity of product demand exceeds one, the ratio of capital to labor will be
constant because both inputs increase proportionally. If the productivity effect
is large enough, it may increase total profits sufficiently to leave the firm's PCM
unchanged after division with the union. However, ROI and EV will fall since
the stock of capital will increase.
Measure of capital intensity (FCAP/S): the ratio of firm capital investment
to sales serves as a proxy for the firm capital intensity. There is no certain
hypothesis regarding the coefficient on FCAP/S. The FCAP/S is included in the
first two models (PCM and ROI) to account for differences in firm returns due to
differences in capital intensity. The inclusion of capital intensity measure in EV
Model is more speculative, because capital intensity is measured In the
denominator of EV (Hirsch and Connolly, 1987).
Growth in firm and industry sales (GROWTH and IGROWTH): firm sales
growth and industry sales growth are included in all three models to capture the
changes in firm-level and industry-level demand. If industries and firms
experience large increase in demand, firms can potentially capture short-run
rents until new firms enter the market or existing firms expand their productivity
capacity. High growth rates might also indicate more recent and productivity
capital which may be more profitability than older capital. Under above
45
conditions. It is accepted to have a positive association between all chosen
short-run and long-run profit measures and GROWTH (or IGROWTH).
Advertising expenditures (ADVES/S): advertising expenditures are
defined as the ratio of the stock value of firm's advertising expenditures to firm's
sales. The common argument was that when advertising expenditures create
the product differentiation barriers to entry, there is a positive association
between advertising expenditures and profitability.
On the other hand, the positive correlation between the advertising
expenditures and profitability would be statistically insignificant when the effects
of advertising on profitability are assumed to be very long lived, implying that
when depreciation rate of advertising intensity is very low.
Research and development expenditures (RDES/S): research and
development expenditures are defined as the ratio of the stock value of firm's
research and development expenditure to firm's sales. Economies to scale
barriers have been postulated to arise from research and development
expenditures because investment in it is costly, risky, and subject to economies
of specification in personnel and equipment (Phillips, 1966; Mueller and Tilton,
1969). By taking advantage of high-yield, high risk research and development
opportunity, the firms may increase their profits in two ways. First, profit
relative to firm size may be increased by an above-average return on research
and development projects. Second, research and development created new
products may permit increased growth. Also, to the extent that capital market
46
imperfections exist, the need to undertake large scale research and
development investments can also be a source of absolute cost barriers
(Grabowski and Mueller, 1978).
Firm size (FS): firm size is defined as the number of employees.
Although average firm size may be an important determinant of profit rates, the
direction of the effect is unclear a prior. If we assumed that Baumol's
proposition (1959) which states that large firms have all of the options of small
firms and can invest in lines requiring such scale that small firms are excluded,
is true, the effect of firm size on all three profitability measures should be
positive. On the other hand, if Liebenstein's argument (1966), which states that
X-inefficiency imposed productivity in large firms, is valid, it should be expected
that the association between firm size and profitability measures should be
negative.
Market share of the firm (MSF): market structure in all three models is
characterized by market share of the firm, defined as the firm's sales divided by
four-digit industry sales. If it is assumed that firms with high market share
individually or collectively elevate prices by restricting output, someone could
expect that the association between the profitability and market share is
positive. If the sources of profitability are related more to firm-specific factors
than to the size distribution of competitors, it is possible to expect an
inconclusive association between the market share and the profitability
47
(Hirschey, 1985). Therefore, the expectations on the signs of market share of
the firm for all three models are ambiguous.
Inventory investment (IINV/VIS): inventory investment is defined as the
ratio of the value of industry inventory to the value of industry shipment.
IINV/VIS is used to capture the effect of industry-specific and/or economy-wide
business cycles. It is hypothesized that the relationship between IINV/VIS and
profitability is negative.
In this study, three different estimation techniques are exercised for each
model in order to exploit the relationship between firm unionization and
profitability: ordinary least squares, OLS, technique for cross-section data and
panel data separately, the fixed-effect technique for panel data and the
random-effect technique for panel data.
The first approach is to estimate equation (3.3) using the OLS technique
on cross-section data only for one year. Since the cross-section data of 1988
is the most recent data in our sample, the preference is to regress equation
(3.3) only for this year. Moreover, the main purpose in estimating equation
(3.3) for the cross-section data of 1988 is to compare the results with those of
the previous studies which have used cross-section data. However, because
the number of observations in the cross-sectional sample is changing between
85 and 102, the OLS estimates could suffer from the shortages of degrees of
freedom, implying that the information obtained from this sample is not
48
completely enough to meet the information requirements of the specified
models.^
In practice, the conventional way to avoid the inherent problems
associated with the use of data sets containing a smaller number of
observations, and also to improve the efficiency of the cross-section estimates
as well is generally accomplished by using panel data. Since the sample
consists of both cross-section and time series observations, the second
approach will be to apply the OLS technique to equation (3.3) under the panel
data. However, In addition to the above advantages of using panel data. It
does not necessarily follow that the application of the OLS technique to panel
data can always provide a perfect solution or reduction of all econometric
problems. The key econometric problem would be that the estimates under the
application of the OLS technique to panel data and cross-section data could be
biased when the omitted (or unobserved) determinants of firm profit are
correlated with the included explanatory variables in a profit equation such as
firm unionization. If this is the case, then the biased estimated coefficients
would destroy the conventional b.l.u.e. property of the OLS estimators.® One
way to account for omitted firm specific or time specific effect is the possible
^See Hsiao (1991) for more information on this topic.
®As shown in Johnston (1984, pp. 260-261), the bias in regression coefficients would be dissipated if the covariances between the excluded and included variables are zero, but the estimated disturbance variance would be biased and so that the statistical inferences would still be misleading.
49
inclusion of detailed control variables in all models. When panel data is
available, fixed-effect technique (the observed variables are transformed by
subtracting out the appropriate time-series means or cross-firm means, and
then applying the OLS to transformed data) or least squares dummy variable
technique (dummies are used to capture the time specific effects or firm specific
effects) is another way to eliminate omitted or unobserved firm specific or time
specific effect.® To increase the likelihood of obtaining unbiased estimates,
especially in the fixed-effect technique, the profitability equation is specified
such as that there may be some omitted or unobserved variables whose value
remains constant across individual firms at a given time, but exhibits variation
through time (time effects) or whose value remains constant through time for a
given individual firm but varies across individual firms (individual effects).
Obviously, here the main objective of applying the fixed-effect technique will be
to capture the possible effects of omitted or mismeasured firm specific, industry
specific, or economy-wide variables which are correlated with the firm
®Both approaches in their specifications allow for the intercept term to vary over time and to vary over cross-section units. If the slopes are to vary as well, pooling data would be inappropriate and each separate cross-section regression would involve a distinct model or a single equation model whose all the coefficients are treated as different for different cross-sectional units in different time periods. However, since the number of the parameters which will be estimated will be greater than the number of observations, it is not possible to estimate any meaningful coefficient. But, it is possible to allow for random variation in slope parameters and then to estimate them in an efficient manner by using the random-coefficient model of Swamy (1970). For more details for the random-coefficient model, see Hsiao (1991, pp. 131-153) and Swamy (1970, pp. 311-323).
50
unionization and profitability.^^ Here, the length of the time period (three
years) is not enough to lead the parameters of the model to change
significantly. Therefore, we necessarily assume that the slope coefficients of all
explanatory variables are constant over the period 1986-88. Thus, under this
assumption by employing the OLS technique to panel data, the following
specification is estimated to allow a varying intercept over time or across
individual firms and a common vector of slope coefficients over time and across
individual firms.
e 2
<l>/r-*rBiy,{Zi -Z, Y: B^ Z^ tr' yw) E BsyJ- s/ - /w) H^r^i (3-4) ^-1 k'^
^^he effects of all omitted or unobserved variables are classified by three types of variables. First, the effects of those omitted variables that are specific to individual cross-section units (or firms) but stay constant over time. Examples of such omitted variables are attributes of individual-firm management, ability, gender and socioeconomic background variables. Second, the effects of those omitted variables that are specific to each time period but are the same for all cross-sectional units. Examples of such variables are industry or economy-wide prices, interest rates, and wide-spread optimism or pessimism. Finally, the individual time-varying variables are the variables that vary across cross-sectional units at a given point in time and also exhibit variations through time. Examples for those variables are attributes of firm profits, sales and capital accumulation (Hsiao, 1991).
51
or
6 2
where j=1..,N, t=1986,1987,1988; 0,=(1/N)I%0j„ Op(1/T)Z\^iOj,;
Z,.=(1/N)Z%Z,,, Z,=(1/T)lViZ,,; Z,^=(VN)t'.__,Z^,„ Z^,={m)i:\,Z^^;
Z3^=(i/N)z%Z3j^, z,^=(^^•)t'•^__,z,^^ £ , = ( I / N ) I % £ J , £J=(I/T)ZV,£J,
The difference between two specifications results from the construction
of the means of the variables which represents time-effects or individual-effects.
The computation method of the mean of the variables under both specifications
is performed in the following manner. In the first specification with time-effects,
each variable is averaged across the firms for a given year while in the second
specification with individual-effects, each variable is averaged across three
years for a given firm.
On the other hand, since the fixed-effect technique provides unbiased
and consistent parameter estimates, the central issue associated with pooling is
one of efficiency. So, in the fixed-effect estimation technique, the effects of
omitted firm specific (or time specific) variables are treated as fixed over time.
Therefore, the case may be that the individual-effects or time-effects like ej,
could be random variables. If this is the case, then the fixed-effect estimation
technique may not yield efficient parameter estimates. Therefore, in addition to
the fixed-effect technique, the random-effect technique is applied to the
52
profitability equations." This technique allows for the possibility of comparing
the results of both techniques (fixed-effect and random-effect techniques) to
each other. The estimation of the random-effect technique is a generalization
of the weighted least squares technique since it weights observations in an
inverse relationship to their variances. A two-stage estimation process is used
to obtain the weighing. In the first stage, the OLS will be run on the entire
pooled sample for equation (3.3). Then, the OLS regression residuals are used
to calculate a sample estimation of the variances' components. In the final
stage, the estimated variances obtained from the first step are used to obtain
the generalized least-squares parameter estimations. ^
In addition to the association between firm unionization and profitability,
the source of union gains is an important issue and has been in dispute for a
long time. The previous studies on this topic generally found that negative
union effects were restricted to highly concentrated industries (Freeman, 1983;
Karier, 1985; Salinger, 1984). With subsequent research, the conclusion that
industry concentration provides the major source for union rents appears to be
incorrect and the source of union gains may relate to more firm-specific factors
"Under the random-effect technique, the error term from equation (3.3) is composed of three effects; e,,=|Lij+X,+T|j„ where [i is the individual effect, X is the time effect and TJ is the purely random effect. If there is only individual effect, Ej will be equal to the sum of |Xj and r\.^. On the other hand, if there is only time effect, e, will be equal to the sum of \ and \.
^ More of the details are available in Johnston (1984) and in Rats User's Manual.
53
than to the size distribution of competitors (Connolly, Hirsch and Hirschey,
1986). More recently, Hirsch and Connolly (1987) argued that quasi-rents on
intangible assets are more Important source of union gains than monopoly
profits. This situation could cause unionized firms to have lower research and
development investments than do similar nonunion firms. This argument was
later confirmed by Hirsch (1991b, 1992). Furthermore, while Hirsch and Link
(1987) found product innovation activity to be less important among a sample of
unionized businesses than among similar nonunionized businesses, Acs and
Audretsch (1988) found fewer innovations in highly unionized industries. These
previous findings provide the motivation to investigate whether the research and
development expenditures are still an important source of union rent sharing
behavior. Therefore, this hypothesis will be tested over the period 1986-1988.
However, the validity of this hypothesis is not examined for any individual year
(cross-section) because of many missing observations in each individual year
which may produce less accurate results.
Finally, the possibility of bias arising from the simultaneous determination
of profits and unionization is considered explicitly. The issue of simultaneity
and the concomitant property of simultaneous equations bias in the estimated
unionization-profit relationship are raised by Voos and Mishel (1986a) for
manufacturing industries. Under the ceteris paribus assumption, they argued
that the industries In which unions are likely to organize have higher potential
profits than other industries. Among firm level studies, by employing collective
54
bargaining agreements in the firm's primary three-digit census industry as a
proxy for firm unionization, Hirsch and Connolly (1987) could not reject the null
hypothesis which states that there is no misspecification from using OLS. They
confirmed that two-stage least squares estimates were similar to OLS
estimates. More recently, Hirsch (1991b) found that there was some evidence
for simultaneity between firm union coverage and Tobin's q, but no evidence for
simultaneity between firm union coverage and rate of return on invested capital
by using the firm level unionization measure. Therefore, he suggested that
exogeneity of firm unionization might not be an inappropriate assumption.
Moreover, although in their studies, Clark (1984) and Becker and Olson (1992)
did not test the endogeneity of unionization at the firm level data, they made the
argument that there is no direct simultaneity between the establishment of
bargaining units in the firms which were probably established a long time ago.
They strongly argued that the measure of firm unionization is cleariy
endogenous if firm decisions about union and nonunion operations are a
function of the union effect on profitability if this investment change the relative
share of unionized workers. And they also continued to argue that even though
the endogeneity is assumed to be between the firm profitability and
unionization, it is almost impossible to capture that endogeneity because no
such data is available which captures the dynamics of investment, and makes a
clear distinction between union and nonunion employment in the firm. Besides
the unavailability of such data, the other major difficulty in accounting for union
55
endogeneity is to identify and measure the instruments that influence union
coverage, but not profitability. Even though those problems may exist in the
sample data, we will still test the endogeneity of union with respect to
profitability for all three models by applying the Hausman specification test.
3.2 Data Description
For the period 1986-88, Standard & Poor's Compustat tape. Current
Wage Developments published by Bureau of Labor Statistics, Annual Survey of
Manufactures and Census of Manufactures published by Department of
Commerce were the major sources for the cross-section and time series
database. The sample were necessarily constructed by matching the firms
whose union coverage information exists in Current Wage Development with
those in Compustat data set. Therefore, all firms in the sample are unionized
to some extent. The number of firms obtained by matching the two data sets is
104. However, because of missing values, the sample for regressions that
include particularly advertising, and research and development expenditures are
smaller than the sample for the other regressions. The data related to firm
profitability measures such as PCM, ROI and EV are drawn from the
Compustat tape. The FCAP/S, GROWTH, ADVES/S, RDES/S and FS are also
obtained from the Compustat tape. In order to compute the stock values of
advertising expenditure, and research and development expenditure, each stock
is weighted in an average of past investment outlays, where the weights decline
56
exponentially as one goes backward in time.^^ In computing the stock values
of such variables, a constant depreciation rate which is recommended by
Grabowski and Mueller (1978) is used. In their study, Grabowski and Mueller
(1978) found that the employing of different depreciation rates did not change
the effects of the values of research and development, and advertising
expenditures on profitability significantly.""^
Data used in computing MSF, the ratio of firm sale to the value of
four-digit industry shipment to which the firm belongs, comes from the Annual
Survey of Manufactures and from Standard & Poor's Compustat tape.
In order to get firm level unionization data. Current Wage Development
and the Compustat tape are used. Firm level union coverage data is
constructed in the following manner. First the number of covered workers is
aggregated across the listed contracts of a given firm from Current Wage
Development. Then, the number of those covered workers is divided by the
total number of employed workers in the firm.
^^For the calculation of ADVES and RDES, we used the equations which are used by Grabowski and Mueller (1978). These equations are as follows: ADVES = I~o(1-^R)Vk. RDES=Z~o("l- A)''at.k. K and X^ are depreciation rates for research and development expenditures and advertising expenditures, respectively, r, and a are research and development expenditures and advertising expenditures in period t, respectively.
" See Neriove and Arrow (1962), Mansfield (1968) and Schmalensee (1972) for the argument which was whether or not research and development and advertising capital are depreciated at a constant proportional rate.
57
Industry level data is added to the sample based on a firm's primary SIC
industry code and associated census industry codes. The industry specific
variables, IINV/S and IGROWTH, are drawn from the Annual Survey of
Manufactures. Industry union coverage data at the three-digit level is obtained
from Curme, Hirsch and Macpherson (1990). Curme, Hirsch and Macpherson
estimated the union membership and contract coverage density for industries,
occupations, states, and metropolitan areas."* Although their estimates are
based on calculations from the Current Population Survey tapes over the period
1983-88, annual estimates only for 1986-88 are based on twelve monthly
surveys in each of three years. In their study, they estimated industry union
coverage density, which is the proportion of eligible workers who are covered
by union contracts, in each firm's principal three digit census-coded industry
over the period 1986-88. In this study, industry union coverage density will be
used rather than union membership density in order to make both firm and the
industry union measures more comparable.
Finally, GROWTH and IGROWTH variables in panel data are deflated by
Gross Domestic Product price deflator to obtain real firm and industry growth
^ In order to calculate membership and contract coverage densities, Curme, Hirsch and Macpherson (1990) used the following method which is also used by Freeman and Medoff (1979), and Kokkelenberg and Sockell (1985): U=(Zi8jCOi/IiCOij)*100 where Uj represents the percentage of union members or workers covered by a contract in group j (where i is industry or occupation or state or metropolitan area), Sjj is a binary variable equal to 1 if worker i is a covered (or union) member, cOjj is the sampling weight assigned by the CPS to worker 1.
58
variables. Since all other variables are defined as the ratio to another nominal
variable of the corresponding year, they are implicit in real terms in panel data.
Therefore, other variables are not deflated or inflated again. In addition, to get
real firm and industry growth variables for cross sectional estimates with 1988
data, those variables are not deflated since data for this estimation are
themselves in static form.
The definitions and sources of all variables are presented in Table 3.1
while the descriptive statistics of all variables for 1986, 1987 and 1988 are
reported in Table 3.2.
59
TABLE 3.1 Variable Definitions and Sources
Variable Description
Price-Cost Margin; Gross operating profits before depreciation and taxes PCM divided by firm's sales.
Rate of Return on The ratio of gross operating profits before depreciation Invested Capital; and taxes to total assets of the firm. ROI
Excess Value; EV
Firm Unionization; F-UN
[(Market value of firm's equity+Book value of debt)-Book value of tangible assets]/Firm's sales.
The ratio of workers who are covered by union contracts to the total number of employed workers of firm.
'Capital; FCAP/S The ratio of net value of plant to firm's sales.
Growth in Firm's [(Firm's sale in t-Firm's sale in t-3)/Firm's sale in t-3]. Sales; GROWTH
The Ratio of RDES The stock value of firm's research and development to Firm's Sales; expenditures divided by firm's sales. RDES/S
The Ratio of ADVES The stock value of firm's advertising expenditures to Firm's Sales; divided by firm's sales. ADVES/S
*Firm Size; FS The natural log of number of employees.
The Market Share Firm's sales divided by the value of four-digit industry of the Firm; MSF shipments in which the firms is located.
The Ratio of IINV to VIS;IINVA/IS
The ratio of four-digit industry inventories In which firms are categorized according to the value of four-digit industry shipments in which firm is located.
60
TABLE 3.1 (Continued) Variable Definitions and Sources
Variable Description
Growth in four-digit [(Value of four-digit industry shipmentSj-Value of four-digit Industry Shipments; industry shipments,)/Value of four-digit industry IGROWTH shipments,].
Industry The proportional of eligible workers who are covered by Union Coverage; union contracts in each firm's principal three-digit l-UN census-coded industry.
* Only variables in natural log form.
Sources: PCM and EV: Compustat, Becker and Olson (1992); ROI: Compustat, Brunello (1992); F-UN: Current Wage Developments and Compustat; FCAP/S: Compustat, Brunello (1992); GROWTH: Compustat, Connolly, Hirsch and Hirschey (1986); RDES/S and ADVES/S: Compustat, Salinger (1984); MSF: Compustat and Annual Survey of Manufactures, Hirsch (1991); IINV/VIS: Annual Survey of Manufactures, Census of Manufactures; IGROWTH: Annual Survey of Manufactures, Hirsch (1991); l-UN: Curme, Hirsch and Macpherson (1990).
Note- All variables are calculated over 1986-88. RDES=r,+(1-.10)r,.i+ (1-.10)^2; ADVES=a,+(1-.30)a,.,+(1-.30)%.2; r, and a, are research and development expenditures and advertising expenditures in period t.
TABLE 3.2 Descriptive Statistics of the Variables* (Standard Deviations in Parentheses)
61
Variables
PCM
ROI
EV
F-UN
FCAP/S
GROWTH
RDES/S
ADVES/S
FS
MSF
IINVA/IS
IGROWTH
l-UN
1986 Mean
0.123 (0.058)
0.146 (0.063)
0.191 (0.292)
0.271 (0.257)
-0.503 (0.320)
0.148 (0.341)
0.060 (0.045)
0.026 (0.050)
1.412 (0.509)
0.215 (0.213)
0.158 (0.101)
-0.019 (0.118)
0.434 (0.261)
N
104
104
104
104
104
104
82
93
104
104
104
104
104
1987 Mean
0.135 (0.063)
0.153 (0.058)
0.206 (0.289)
0.230 (0.218)
-0.492 (0.274)
0.243 (0.249)
0.044 (0.045)
0.024 (0.051)
1.423 (0.486)
0.220 (0.219)
0.170 (0.200)
0.102 (0.204)
0.428 (0.245)
N
104
104
95
104
104
104-
104
104
104
104
104
104
104
1988 Mean
0.130 (0.066)
0.144 (0.057)
0.276 (0.358)
0.196 (0.180)
-0.476 (0.273)
0.188 (0.204)
0.051 (0.047)
0.023 (0.050)
1.420 (0.449)
0.216 (0.225)
0.157 (0.112)
0.120 (0.111)
0.414 (0.255)
N
104
104
102
104
104
104
92
104
104
104
104
104
104
62
TABLE 3.2 (Continued) Descriptive Statistics of the Variables* (Standard Deviations in Parentheses)
Note: *N represents the number of observation in the sample. The means of all variables over the period 1986-88 are ; 0.129 for PCM, 0.148 for ROI, 0.224 for EV, 0.232 for F-UN, -0.490 for FCAP/S, 0.193 for GROWTH, 0.052 for RDES/S, 0.024 for ADVES/S, 1.418 for FS, 0.217 for MSF, 0.161 for IINV/VIS, 0.067 for IGROWTH, 0.425 for l-UN.
CHAPTER IV
EMPIRICAL RESULTS AND ANALYSIS
The purpose of this chapter is to present and to examine the empirical
results of the relationship between the firm unionization and profitability, and the
sensitivity of the union-profit relationship to the types of firm profit measures
(PCM, ROI and EV) to the firm unionization measures (F-UN, l-UN, F-DUMMYs
and l-DUMMYs which are constructed from F-UN and l-UN ratios), to the type
of estimation techniques (OLS, fixed-effect and random-effect) and to the
database (cross-section and panel data).
This chapter is divided into two main sections. In the first section, the
OLS results of cross-section data, 1988, for each of three models separately is
presented and analyzed. By applying the OLS technique for 1988, we first
looked at how the firm union coverage ratio, F-UN, and the alternative
measures of firm union coverage ratio, F-DUMMYs, affect PCM, ROI and EV
measures. Then, by using the industry union coverage ratio, l-UN, and
l-DUMMYs, which are constructed from l-UN, as a proxy for the firm union
coverage, the analysis of union-profitability relationship is also performed.
In the second section, we present and investigate three models
throughout the 1986-88 period. This second section is divided into three
subsections. The first subsection starts with the presentation of the OLS results
63
64
for each of three firm unionization measures (F-UN, F-DUMMYs and
l-DUMMYs) and for each of three models separately. Then, the test results for
sources of union gains are presented for each model. This is followed by the
presentation and the evaluation of the results of all three models with firm union
coverage ratio, F-UN, under the fixed-effect technique in the second subsection.
Finally, in the third subsection, the results of all three models with firm union
coverage ratio, F-UN, under the random-effect technique are reported.
4.1 The OLS Estimation Results of the PCM. ROI and EV Models on Cross-Section Data. 1988
Since this study used three distinct firm profitability measures and three
distinct firm unionization measures, a brief outline of the specifications of all
three models is discussed first. In all specifications of each of three models,
the union variable is used as the permanent explanatory variable. Besides this
variable, a number of variables which are important in explaining the firms
profitability are also employed in all three models. In this section, the stock
value of the research and development expenditures, RDES/S, as an
explanatory variable is excluded from the specifications, since many missing
observations in the RDES variable caused the degree of freedom in the sample
to decrease significantly. Moreover, the inclusion of the RDES/S in the
specifications resulted in a statistically insignificant coefficient for this variable in
^While whole sample data covers 102 firm observations, the inclusion of RDES/S on the models caused the sample to be 67.
65
each of three models. Even though the coefficient of RDES/S is found to be
insignificant in the cross-section regressions, the conclusion that the effect of
RDES/S on the firm profitability is insignificant does not necessarily follow.
Because of the data limitations, the regressions could not capture the possible
effect of this variable on firm profit measures. However, it is possible to obtain
a more robust result concerning the effect of this variable on the measure of
firm profitability from the regression equations which will be estimated over the
1986-88 period. In addition, the statistical significance of the unionization
variable is found to be unaffected when the MSF variable is included in all
regressions, implying that the omission of this variable from the regression
specifications has a limited influence on the findings. Therefore, this variable
is excluded from the regression specifications in this section.
In order to observe how unionization and the other explanatory variables
affect PCM, ROI and EV, the following strategy is implemented. In the first
step, each of PCM, ROI and EV models are estimated with each of the union
measures (F-UN, l-UN, F-DUMMYs and l-DUMMYs) separately and with all
other explanatory variables except RDES/S and MSF, for which the results are
shown in the first specification of each table. In the second step, first, the
backward stepwise regression procedure is used in order to search for the best
^However, it is not implied that the market power is irrelevant in explaining firm profitability. This result occurs because the MSF variable is a poor proxy for market power of the firm and the data required to properly calculate an acceptable proxy for market share is inaccessible.
66
subset of explanatory variables to be included in the regressions to obtain
optimal results. In this stage, the backward stepwise regression procedure is
used rather than forward stepwise regression procedure because some
statisticians argue that the backward stepwise search method is preferable to
using the forward stepwise search method for the pool of potential explanatory
variables containing small and moderate numbers of variables.^ In the final
stage of the second step, each model is estimated by including the best subset
of explanatory variables which remains after stepwise procedure into regression
equations, the results of which are reported in the second specification of each
table. In the final step, each model is reestimated by including variables which
are used in the second specification, and the 16 two-digit industry dummies into
regressions, the results of which are found in the third specification of each
table.
To ascertain whether the OLS estimates suffer from heteroscedastic
error terms, two different techniques are applied to all model specifications
separately: Goldfeld-Quandt and White tests.* The calculated F-statistics from
both procedures did not reject the null hypothesis of no heteroscedastic error
terms at the 1 percent level in all specifications of each model. Additionally,
unionization is tested as to whether it is endogenous with respect to each profit
^For a more complete argument about this subject, see Neter, Wasserman and Kutner (1989, p. 458).
"^Goldfeld-Quandt test and White's test are described by Goldfeld, S.M. and Quandt, S.M. (1965) and White, Halbert (1980), respectively.
67
measure. For this purpose, the Hausman specification test procedure is
employed.^ The calculated test statistics for each of three profitability
measures indicate that the null hypothesis which states that unionization is
exogenous with respect to all profit measures is not rejected at any acceptable
level.
The OLS results for PCM, ROI and EV Models with the firm union
coverage variable, F-UN, are reported in Tables 4.1, 4.2 and 4.3, respectively.
In addition to F-UN, each model is reestimated by employing unionization
dummies constructed from F-UN as alternative measure of the firm union
coverage ratio in order to examine the variation in the union effect on firm
profitability over the range of unionization. Therefore, the range of firm
unionization is divided into three categories; F-DUMMY1 for F-UN>0.60,
F-DUMMY2 for 0.60>F-UN>0.30 and F-DUMMY3 for 0.30>F-UN.^ The
regression results using the firm unionization dummies for all three models are
presented in Tables 4.4, 4.5 and 4.6. In Tables 4.7, 4.8 and 4.9, the OLS
estimation results of three models with industry union coverage ratio are
reported. Besides industry union coverage ratio, the same models are also
estimated by employing industry unionization dummies which are constructed
^For the testing procedure, see Hausman (1978) and Rats User's Manual (1992).
®By taking the studies of Hirsch (1991a) and Becker and Olson (1992) as a reference, we divided firm unionization into such ranges.
68
from l-UN. The OLS results with industry union coverage dummies are shown
in Tables 4.10, 4.11 and 4.12.
4.1.1 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Ratio
Table 4.1 shows the OLS results of all specifications for PCM model. As
seen from Table 4.1, union coverage at the firm level on all specifications was
found to be associated with substantially lower short term firm profits, PCM. In
all specifications of Table 4.1, the estimated coefficient of F-UN was found to
be negative and statistically significant at least at the 10 percent level.
Evaluated at the mean, the effect of F-UN on PCM ranged from -6.6 percent to
-10 percent for 100 percent unionized firm.^ The effect of firm unionization on
PCM was found to be not significantly sensitive to the inclusion or exclusion of
the other explanatory variables. Yet, the magnitudes of the estimated
coefficients of F-UN and of the rest of variables became smaller and less
significant when the 16 two-digit industry dummies are included into the
regressions.
^By using the means of variables from Table 3.2, the relationship is evaluated between all profit measures and union variable (and sometimes other explanatory variables) in elasticity form. If both dependent and independent variables (say PCM and F-UN) are in linear form, the average percentage effect of F-UN on PCM is calculated by [Pp.uN(mean of F-UN/mean of PCM)], where PF-UN 'S estimated coefficient of F-UN. On the other hand if dependent variable (say PCM) is in linear form and independent variable (say FCAP/S) is in log form, the average percentage effect of FCAP/S on PCM is calculated by [pFCAP/s(1/mean of PCM)], where PFCAP/S "S estimated coefficient of FCAP/S. For more information on this subject, see Gujarati (1988 pp. 154-155).
TABLE 4.1 OLS Results of PCM Model with Firm
Unionization, 1988 (t-Statistics in Parentheses)
69
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R2
N
Industry Dummies
(1)
0.142* (6.62)
-0.067'' (2.56)
0.049* (4.68)
0.069' (2.43)
0.038 (0.37)
0.034* (2.95)
-0.068 (1.30)
0.078 (1.45)
11.54
0.51
85
no
(2)
0.147* (7.73)
-0.064* (2.86)
0.054* (6.55)
0.076* (2.88)
—
0.030* (3.00)
-0.039 (0.87)
0.082" (1.67)
16.65
0.51
102
no
(3)
0.125* (4.20)
-0.044' (1.75)
0.045* (3.78)
0.054' (1.84)
—
0.030' (2.52)
-0.015 (0.26)
0.068 (0.83)
5.64
0.59
102
yes
* a Significant at the 0.01 level; ''Significant at the 0.05 level; 'Significant at the 0.10 level.
70
From the same table, the control variables, FCAP/S and GROWTH, were
found to have a positive and statistically significant effect on PCM. The
estimated coefficient of FCAP/S in the first specification was found to be 0.049
and implies that a 10 percent increase in the ratio of FCAP/S would cause PCM
to increase by 3.76 percent. Similariy, the estimated coefficient of GROWTH in
the same specification is 0.069 and it is suggested that a 10 percent increase in
GROWTH would cause PCM to increase by 0.99 percent. The findings for
those control variables are consistent with those of most of the previous
studies.
On the other hand, the coefficient estimate of ADVES/S, one measure
of the intangible assets, is found to statistically insignificant in all regression
specifications. In addition to the stock value of advertising, its current
expenditures, and its lag values as a measure of the intangible assets are used
separately. However, the findings did not change significantly, and the
estimated coefficients of advertising became more insignificant. The results of
the effect of ADVES/S on PCM are not consistent with those of Hirsch and
Connolly (1987) and Salinger (1984).
In addition, as pursued in all specifications in Table 4.1, the firm size,
FS, which is measured as the number of employees, is found to have a positive
and statistically significant effect on PCM. From the first specification, the
estimated coefficient of FS is 0.034, implying that PCM is increased by 2.61
percent as a result of a 10 percent increase in FS. The effect of the ratio of the
71
four-digit industry inventory value to the value of industry shipments, IINV/VIS,
which is purposed to capture the effect of the industry or economy wide
business cycles on the short run profit measure, PCM is also examined. It is
expected that the effect of IINV/VIS ratio on PCM is negative. The estimated
coefficient of IINV/VIS ratio is negative as expected, but is not statistically
different from zero at any acceptable level. The coefficient estimate of the
IGROWTH variable is positive and statistically significant at least at the 10
percent level in specification (2), but insignificant in specifications (1) and (3).
When the 16 two-digit industry dummies are included in specification (2), it is
observed that both magnitudes and statistical significance level of the effect of
unionization and the remainder of explanatory variables on PCM became
smaller.
The estimation results of the ROI model which are shown in Table 4.2
reveal that F-UN had a negative and statistically significant impact on ROI. The
coefficient estimate in all three specifications implies that ROI decreases as the
unionization increases, that is, ROIs of the firms that are 100 percent unionized
are changing between -12.3 percent and -13.7 percent.
The estimated coefficient of FCAP/S (=0.021), one of the control
variables, was found to be positive and statistically significant only in
specification (1) at the 10 percent level, implying that there exists a positive
association between FCAP/S and ROI. The coefficient estimate of FCAP/S
from specification (1) demonstrated that a 10 percent increase in this variable
TABLE 4.2 OLS Results of ROI Model with Firm
Unionization, 1988 (t-Statistics in parentheses)
72
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R2
N
Industry Dummies
(1)
0.162* (7.05)
-0.091* (3.25)
0.021' (1.89)
0.021 (0.69)
0.179 (1.64)
0.015 (1.27)
-0.076 (1.36)
0.068 (1.19)
6.03
0.35
85
no
(2)
0.199* (13.3)
-0.101* (4.24)
0.017 (1.56)
—
0.161 (1.46)
-0.089 (1.65)
8.79
0.30
85
no
(3)
0.154* (6.10)
-0.098* (3.42)
0.001 (0.03)
—
0.203° (1.70)
-0.033 (0.56)
3.01
0.46
85
yes
* a Significant at the 0.01 at the 0.10 level;
level; ''Significant at the 0.05 level; 'Significant
73
would cause ROI to increase by 1.45 percent. In contrast to the PCM model,
GROWTH, FS and IGROWTH have insignificant effect on ROI. While both
ADVES/S and IINV/VIS have a nearly significant impact on ROI, the effect was
positive for ADVES/S and negative for IINV/VIS.
Finally, Table 4.3 presents the OLS results for the EV model. In all three
specifications, F-UN was detected to be significantly and negatively related to
EV. The coefficient estimates of F-UN in all specifications suggested that firms
that are 100 percent unionized have a lower EV between 16.7 percent and 26.7
percent as compared to nonunionized firms.
We also found that ADVES/S had a statistically significant and positive
effect on EV in all specifications. The estimated coefficient of ADVES/S was
4.20 in specification (1), 4.14 in specification (2) and 4.01 in specification (3).
Finally, for the rest of explanatory variables (GROWTH, FS, IINV/VIS and
IGROWTH) in specification (1), any significant effect on EV was not detectable.
4.1.2 The OLS Estimation Results of the PCM, ROI and EV Models with Firm Union Coverage Dummies
As seen from the first two specifications of Table 4.4, when all possible
explanatory variables are included into PCM model, the unionization effect on
PCM was negative and statistically significant at the 5 percent level, especially
for the firms whose union coverage ratios are between 0.30 and 0.60. The
estimated coefficient of F-DUMMY2 in specification (1) is -0.035 which indicates
74
TABLE 4.3 OLS Results of EV Model with Firm
Unionization, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-UN
GROWTH
ADVES/S
FS
IINV/VIS
IGROWTH
F-Statistics
R2
N
Industry Dummies
(1)
0.099 (0.82)
-0.285' (1.91)
0.138 (0.85)
4.206* (7.25)
0.064 (0.97)
0.120 (0.43)
-0.058 (0.19)
13.37
0.50
85
no
(2)
0.256* (5.36)
-0.377* (3.01)
—
4.149' (7.40)
39.57
0.49
85
no
(3)
0.133" (2.04)
-0.236* (1.74)
—
4.013* (6.95)
—
—
—
7.46
0.65
85
yes
* ai Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
75
TABLE 4.4 OLS Results of PCM Model with Firm
Unionization Dummies, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R2
N
Industry Dummies
(1)
0.134* (6.47)
-0.032 (1.50)
-0.035" (2.47)
0.048* (4.47)
0.080* (2.85)
0.023 (0.22)
0.035* (2.97)
-0.072 (1.37)
0.073 (1.37)
9.98
0.51
85
no
(2)
0.141* (7.57)
-0.036" (1.99)
-0.029" (2.25)
0.053* (6.32)
0.088* (3.37)
—
0.029* (2.75)
-0.041 (0.90)
0.074 (1.51)
13.79
0.50
102
no
(3)
0.116* (4.10)
-0.024 (1.26)
-0.018 (1.26)
0.043* (3.63)
0.065" (2.18)
—
0.029" (2.34)
-0.014 (0.24)
0.070 (0.84)
5.24
0.59
102
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY3 is 0.30>F-UN.
76
that the overwhelming portion of the union gain occurs between 30 and 60
percent of unionization.
From all specifications of Table 4.5, the estimated coefficients of the
union coverage dummies also suggest that the effect of union on ROI varies
with the extent of firm union coverage. The coefficients of both F-DUMMY1
and F-DUMMY2 in all specifications are statistically significant at least at the 5
percent level. The estimated coefficients in the first specification are found to
be -0.049 for F-DUMMY1 and -0.042 for F-DUMMY2.
In EV model as in PCM and ROI, the estimated coefficient of
F-DUMMY2 in all specifications is negative and statistically significant, and is
between -0.172 and -0.207. These estimates indicate that union effect on
profitability for the firms that are moderately unionized is greater than that for
the firms that are highly and less unionized.
As seen from Tables 4.4-4.6, besides union coverage dummies, the
estimated coefficients of all other explanatory variables followed the same
pattern as they did in Tables 4.1-4.3. So, moderate unionization which is
represented by F-DUMMY2 (0.30<F-UN<0.60) yields more meaningful result
than high unionization, F-DUMMY1 (0.60<F-UN) and low unionization,
F-DUMMY3 (F-UN<0.30).
11
TABLE 4.5 OLS Results of ROI Model with Firm
Unionization Dummies, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
FCAP/S
GROWTH
ADVES/S
FS
IINV/VIS
IGROWTH
F-Statistics
R 2
N
Industry Dummies
(1)
0.151* (6.77)
-0.049" (2.13)
-0.042* (2.75)
0.019' (1.68)
0.034 (1.14)..
0.162 (1.45)
0.016 (1.31)
-0.083 (1.47)
0.059 (1.03)
4.99
0.34
85
no
(2)
0.159* (7.31)
-0.048" (2.14)
-0.036" (2.45)
0.017 (1.51)
—
0.140 (1.24)
0.022' (1.74)
-0.110" (1.98)
—
5.91
0.31
85
no
(3)
0.096* (2.99)
-0.045' (1.90)
-0.027' (1.62)
0.007 (0.45)
—
0.178 (1.46)
0.027' (1.97)
-0.052 (0.85)
2.77
0.48
85
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY3 is 0.30>F-UN.
78
TABLE 4.6 OLS Results of EV Model with Firm
Unionization Dummies, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
GROWTH
ADVES/S
FS
IINV/VIS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.085 (0.74)
-0.148 (1.25)
-0.207" (2.59)
0.189 (1.20)
4.081* (7.07)
0.059 (0.89)
0.125 (0.45)
-0.024 (0.08)
12.22
0.52
85
no
(2)
0.240* (5.74)
-0.223" (2.14)
-0.224* (3.04)
—
4.016' (7.14)
27.61
0.50
85
no
(3)
0.130" (2.21)
-0.167 (1.60)
-0.172" (2.25)
—
3.895' (6.77)
7.38
0.66
85
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level. F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY3 is 0.30>F-UN.
79
4.1.3 The OLS Estimation Results of the PCM, ROI and EV Models with Industry Union Coverage Ratio and Dummies
While Tables 4.7-4.9 present the OLS results of PCM, ROI and EV
models with industry union coverage ratio, respectively. Tables 4.10-4.12 report
the OLS results of the same models with industry union coverage dummies.
From Tables 4.7-4.9, the coefficient estimates of industry union coverage
ratio are found to be statistically insignificant in all firm profit measures. The
comparison of the estimated coefficients of l-UN in Tables 4.7-4.9 with the
estimated coefficients of F-UN in Tables 4.1-4.3 provides clear evidence that
firm union coverage has strong advantage over the industry union coverage in
terms of supporting the hypothesis of a negative union-profit association.
Obviously, these findings confirm the results of Hirsch (1991a) and Becker and
Olson (1992). By using different time periods, different profit measures and
different estimation methodologies from each other, both studies concluded that
firm union coverage measure had a negative and statistically significant impact
on firm profitability. On the other hand, when industry union coverage ratio is
used as a proxy for firm unionization, they found that unionization had a weak
and statistically insignificant impact on firm profitability.
In all regression specifications of Tables 4.7-4.9, all explanatory variables
except FS are found to be in the same pattern in terms of the magnitudes and
statistical significance levels of the coefficients as those in the previous
regressions with F-UN. Therefore, those variables are not focused upon here.
In comparing the results of the PCM, ROI and EV models in Tables 4.1-4.3 with
TABLE 4.7 OLS Results of PCM Model with Industry
Unionization, 1988 (t-Statistics in Parentheses)
80
Independent Variables
CONSTANT
l-UN
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.103* (4.52)
0.026 (1.15)
0.052* (4.80)
0.087* (3.06)
0.090 (0.87)
0.046* (4.20)
-0.066 (1.21)
0.020 (0.39)
10.11
0.47
85
no
(2)
0.110* (5.66)
0.025 (1.30)
0.057* (7.07)
0.102* (4.22)
—
0.040' (4.13)
21.50
0.47
102
no
(3)
0.100* (4.16)
-0.001 (0.45)
0.042* (3.76)
0.075* (2.69)
—
0.037' (3.45)
5.91
0.57
102
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
TABLE 4.8 OLS Results of ROI Model with Industry
Unionization, 1988 (t-Statistics in Parentheses)
81
Independent Variables
CONSTANT
l-UN
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.120* (4.91)
0.006 (0.25)
0.027" (2.31)
0.046 (1.49)
0.246" (2.15)
0.031* (2.61)
-0.079 (1.33)
-0.001 (0.01)
3.99
0.26
85
no
(2)
0.124* (5.13)
0.006 (0.26)
0.027" (2.32)
0.046 (1.59)
0.245" (2.19)
0.031* (2.63)
-0.079 (1.37)
—
4.72
0.26
85
no
(3)
0.075" (2.47)
-0.038 (1.28)
-0.010 (0.63)
0.027 (0.83)
0.245" (2.03)
0.036* (2.68)
-0.017 (0.28)
—
2.60
0.46
85
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
TABLE 4.9 OLS Results of EV Model with Industry
Unionization, 1988 (t-Statistics in Parentheses)
82
Independent Variables
CONSTANT
l-UN
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R 2
N
Industry Dummies
(1)
-0.083 (0.69)
0.130 (1.03)
0.208 (1.31)
4.417* (7.58)
0.115' (1.84)
0.112 (0.39)
-0.307 (1.03)
12.52
0.49
85
no
(2)
-0.091 (0.87)
0.095 (0.78)
0.146 (0.98)
4.430* (7.89)
0.125" (2.06)
—
—
18.53
0.48
85
no
(3)
-0.162 (1.41)
0.173 (1.19)
0.002 (0.01)
4.042* (7.07)
0.122' (1.91)
—
—
6.83
0.66
85
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
83
those in Tables 4.7-4.9, It is observed that the estimated coefficient of FS is
positive and statistically significant at the 1 percent level only for the PCM
model with F-UN, while it is positive and statistically significant at the 10 percent
level for all three models with l-UN.
Tables 4.10-4.12 explore the impact of unionization on profitability,
utilizing three industry union coverage dummies such as I-DUMMY1,
I-DUMMY2 and I-DUMMY3 instead of industry union coverage ratio. Among
dummies which are constructed from industry union coverage ratio, particularly
I-DUMMY1 and I-DUMMY2 (I-DUMMY3 in intercept) are employed as is
employed for F-DUMMYs. When different combinations of industry dummies
are included in the specifications, the results yield negative and statistically
significant coefficient estimates for I-DUMMY2 and, but not for I-DUMMY1.
For both PCM and EV, from the first and second specifications of Tables 4.10
and 4.12, the estimated coefficients of I-DUMMY2 are found to be negative and
statistically significant at the 10 percent level. So, the industry unionization
spline for all three models except ROI reflected the similar pattern to the results
of firm unionization spline. Here, the significant portion of union gains again
occurs between 30 and 60 percent of unionization.
4.1.4 Summary Assessment for All Cross-Section Estimation Results, 1988
In general, the cross-section results for 1988 provide significant support
for the argument that the level of union coverage at the firm level has a
84
TABLE 4.10 OLS Results of PCM Model with Industry
Unionization Dummies, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R2
N
Industry Dummies
(1)
0.121* (5.98)
0.041* (1.03)
-0.015' (1.77)
0.053* (5.07)
0.086* (3.14)
0.074 (0.73)
0.042* (3.97)
-0.059 (1.13)
-0.055 (1.07)
10.27
0.51
85
no
(2)
0.131* (7.30)
0.033* (1.49)
-0.012' (1.89)
0.059* (7.50)
0.106* (4.50)
—
0.037* (3.93)
—
—
19.19
0.49
102
no
(3)
0.116* (4.49)
0.007 (0.34)
-0.017 (1.14)
0.046* (4.01)
0.076* (2.77)
—
0.036* (3.32)
—
—
5.79
0.58
102
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
TABLE 4.11 OLS Results of ROI Model with Industry
Unionization Dummies, 1988 (t-Statistics in Parentheses)
85
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
FCAP/S
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.137* (6.01)
0.007 (0.30)
-0.020 (1.54)
0.029" (2.48)
0.048 (1.55)
0.243" (2.15)
0.029" (2.46)
0.073 (1.24)
0.026 (0.46)
3.96
0.29
85
no
(2)
0.136* (6.29)
0.004 (0.37)
-0.018 (1.44)
0.035* (3.26)
0.058" (2.01)
0.271" (2.50)
0.026" (2.22)
—
—
4.91
0.27
85
no
(3)
0.101* (3.30)
-0.037 (1.41)
-0.049' (1.98)
-0.001 (0.90)
0.030 (0.97)
0.325* (2.74)
0.033* (2.64)
—
—
3.10
0.50
85
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
86
TABLE 4.12 OLS Results of EV Model with Industry
Unionization Dummies, 1988 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
GROWTH
ADVES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.019 (0.18)
0.241* (1.44)
-0.113' (1.77)
0.204 (1.35)
4.319* (7.79)
0.092 (1.54)
0.015 (0.56)
-0.087 (0.30)
13.15
0.54
85
no
(2)
0.030 (0.32)
0.232 (1.48)
-0.120' (1.95)
0.181 (1.26)
4.270* (8.02)
0.101' (1.75)
—
—
18.66
0.54
85
no
(3)
0.014 (0.12)
0.169 (1.33)
-0.168' (1.97)
-0.002 (0.01)
4.260* (7.70)
0.108' (1.77)
—
—
7.70
0.70
85
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
87
negative impact on profitability regardless of different measures of firm
profitability and firm unionization. The findings, which are based on the most
recent sample data on firm level, are generally consistent with most of the
previous studies. Upon examination of the first specification of all three models
with F-UN in which all explanatory variables are included, it is noted that PCM
is reduced by 10 percent, ROI by 12.3 percent and EV by 20.2 percent for 100
percent unionized firms. These results are consistent with the findings of
Hirsch and Connolly (1987). By constructing the sample from the 1977
Compustat tape, Hirsch and Connolly (1987) found that unionism reduced
Tobin's q by 13 percent-20 percent, rate of return on sale by 11 percent-17
percent.
Furthermore, estimation of all three models suggests that as the
profitability measure changes, the effect of unionization on the corresponding
profitability measure varies to some extent. Obviously, the size of the effect is
found to be larger on EV than on both PCM and ROI. It is also found that it is
less on PCM than on ROI.
Now, question is why the sizes of the effect on profit vary with profit
measures. Actually, if the magnitude of the unionization effect with its
corresponding profitability measure is examined, it can be argued that this
result is related to the time characteristics of chosen profitability measures. As
previously discussed, the first two measures are accounting type profitability
measures, while last one is combination of both market and account type.
88
Since the last profitability measure consists of the components related to past,
present and future, it is generally long-run type profit measure. On the other
hand, the first two are short run type measures because they consist of the
components related to past and present. Therefore, it is possible and natural to
find a higher unionization effect on long-run type of profit measures than that on
short run type of measures.
This study employed four different firm unionization measures, that is,
firm and industry union coverage ratios (F-UN and l-UN) and dummies for F-UN
and l-UN, it is appropriate to compare the results on unionization effects on
profitability with each other. First of all, while F-UN ratio suggests a negative
union effect on profitability, l-UN ratio does not.
When F-UN enters into all regression equations, the estimated
coefficients are detected to be negative and statistically significant. In general,
the estimated coefficients of F-DUMMYs in all regression specifications suggest
that union has significantly negative effect on profitability if the firm union
coverage ratio is especially between 30 and 60 percent (F-DUMMY2).
Yet, when l-UN enters into all regression equations instead of F-UN, the
estimated coefficients of l-UNs for all three models are generally discovered to
be statistically insignificant. On the contrary, a different conclusion results when
l-DUMMYs constructed from l-UN are used as a measure of the firm
unionization. Only the estimated coefficient of I-DUMMY2 is negative and
89
statistically significant at the 10 percent level in PCM and EV models but
Insignificant in ROI.
Consequently, the overall results of this section indicate that the union
effects on profitability differ depending on both measure of profitability and
unionization. When the firm union coverage ratio, F-UN, is used in all
specifications, negative and significant union effects on profitability are found
regardless of the profitability measures, which supports the hypothesis of a
negative association existing between unionization and profitability. On the
other hand, industry union coverage ratio was proxied to firm unionization in all
three models and the findings do not support the hypothesis that there is a
negative relationship between unionization and profitability. So, if the results
from both ratios, F-UN and l-UN, are compared, it can be concluded that F-UN
seems to be a more reliable measure than l-UN in terms of supporting the
hypothesis of negative union-profitability association. On the other hand, a
comparison of the results with both firm and industry union coverage dummies
reveals that both measures provide similar information on the relationship of
profitability-unionization.
4.2 The Estimation Results of the PCM. ROI and EV Models over the 1986-1988 Period
To overcome the idea that cross-section estimates of the union-profit
relationship are likely to be biased, three models are reestimated over the
1986-88 period. As we previously noted, the nature of the panel data allows for
90
control of unobservable individual firm and industry effects in the empirical
analysis. Therefore, it is expected that the results, especially on the association
between unionization and profitability under panel data are more robust than
the previous results. Once more, when including market share of firm (MSF) as
an explanatory variable into the regression specifications, negative but
statistically insignificant coefficient estimates are most often found. This finding
again implies that the market share measure is a poor proxy for the market
structure. Therefore, this measure is excluded from the rest of the empirical
estimations. In addition, in this section the estimation results of the regression
specifications with industry union coverage ratio, l-UN, are unreported because
there were not any statistically significant coefficient estimates of this variable
for all three models in the prior section or in this section. On the other hand, in
the previous section a very significant effect of union on the profitability is found
when l-DUMMYs are entered into the regression specifications as a proxy of
the firm unionization. As a result, the regression equations with l-DUMMYs are
used for all three models in this section as well.
4.2.1 The OLS Results of the PCM, ROI and EV Models over the 1986-88 Period
In this first subsection, applying the OLS technique to panel data, all
specifications of three models described in the first part of this chapter are
estimated. Here, the potential problem is that the disturbance variances of the
regressions may not be constant across firms. Using the Goldfeld-Quandt test
91
procedure, all specifications of three models are tested to confirm whether the
OLS regressions suffer from heteroscedasticity. Not surprisingly, the results
confirmed that all specifications of the three models suffered from this
problem.® Therefore, the White correction technique to the standard errors is
applied for all reported specifications of three models.®
The OLS estimations of all three models with F-UN, F-DUMMYs and
l-DUMMYs are displayed in Tables 4.13-4.21. The first specification of each
table presents the estimations of models with all explanatory variables except
16 two-digit industry dummies. Then, the explanatory variables that are
insignificant at the 10 percent level were eliminated by a stepwise procedure.
The second specification in each table is, therefore, designed to provide the
OLS results of models with some explanatory variables which remain after the
stepwise procedure. Finally, the last specification provides the OLS results of
models with explanatory variables which are used in the second specification
and with the 16 two-digit industry dummies. In addition. Table 4.22 presents
the test results of all three models for the source of union gains.
®ln spite of the fact that the choice of how many central observations (firms) to drop is largely subjective. Following the evidence suggested by Harvey and Phillips (1974), only 1/3 of the observations (firms) are not used.
^White's method is to obtain unbiased point estimates of estimated coefficients for all variables using OLS and then to estimate omega, Q., matrix as a diagonal matrix with the jth squared OLS residuals as the (j,j) th element in omega matrix. If the estimated equation is 0=zp+e in matrix notation. So the formula Var(PoLs)=(Z'Z)-"'z'aZ(Z'Z)"'' is used as a consistent estimator of the variance-covariance matrix of the OLS estimator-regardless of the precise form of the heteroscedasticity (White, 1980).
92
4.2.1.1 The OLS Estimation Results of the PCM. ROI and EV Models with Firm Union Coveraoe Ratio
Tables 4.13-4.15 summarize the results of least-square estimations of all
three models with F-UN variable after correcting the variance-covariance matrix
of the OLS estimators for heteroscedasticity.
From Table 4.13, the effect of F-UN on firm profitability was found to be
negative and statistically significant which is consistent with the results of our
cross-section study. The coefficient estimate of F-UN reported in the first
specification implies that an increase of F-UN from 0 to 100 percent is
associated with a decrease of 7.1 percent in PCM. As seen from the first two
specifications, FCAP/S, GROWTH, RDES/S, FS and IGROWTH have positive
and statistically significant effects on PCM, while ADVES/S and IINV/VIS have
small and statistically insignificant effects on PCM. The coefficient in
specification (1) is estimated to be 0.060 for FCAP/S, 0.062 for GROWTH,
0.020 for FS and 0.080 for IGROWTH. In contrast to the results of the
cross-section estimation, RDES/S is an important determinant of PCM. The
estimated coefficient of RDES/S is positive and statistically significant at least
at the 5 percent level. The coefficient estimates of RDES/S were 0.139 and
0.153 in the first two specifications, respectively. Finally, the inclusion of the 16
two-digit industry dummies causes the magnitude and significance level of the
estimated coefficients of most explanatory variables to be reduced significantly.
For example, the addition of these dummies to the second specification reduces
TABLE 4.13 OLS Results of PCM Model with Firm
Unionization, 1986-88 (t-Statistics in Parentheses)
93
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R
N
Industry Dummies
(1)
0.161* (10.4)
-0.040* (3.46)
0.060* (10.9)
0.062* (5.62)
0.032 (0.80)
0.139" (1.89)
0.020" (2.52)
-0.019 (0.79)
0.080* (3.57)
0.49
204
no
(2)
0.164* (11.9)
-0.042* (4.01)
0.063* (11.7)
0.068* (6.25)
—
0.153" (2.28)
0.020* (2.81)
—
0.070' (1.96)
0.46
243
no
(3)
0.126* (6.64)
-0.024' (1.76)
0.047* (6.35)
0.050* (5.16)
—
0.104 (1.25)
0.029* (3.56)
—
0.058 (1.42)
0.53
243
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
94
the absolute value of F-UN coefficient from -0.042 to -0.024 and the
significance level of this coefficient from 1 percent level to 10 percent level.
Table 4.14 reports the OLS results of ROI model with F-UN. The
coefficient estimate of F-UN in the first specification indicates that a fully
unionized firm had a 7.2 percent lower ROI than a nonunionized firm. As found
in cross-section estimations, FCAP/S has a positive and significant impact on
ROI and its estimated coefficient in the first specification is 0.034. Both
GROWTH and IGROWTH which are designed to capture the effects of changes
in the firm and industry level demand have positive and statistically
significant effects on ROI. From the first specification, the statistically
significant coefficients of GROWTH and IGROWTH are 0.061 and 0.077,
respectively. In addition to F-UN, FCAP/S, GROWTH and IGROWTH, the
effect of ADVES/S on ROI is detected to be statistically significant and positive
in all specifications except the last specification. The estimated coefficients of
ADVES/S reported in specifications (1) and (2) are 0.186 and 0.209,
respectively. On the other hand, in all specifications of ROI model, RDES/S,
FS and IINV/VIS appear to have an insignificant impact on ROI. The results
obtained by including industry dummies to the last specification are found to be
similar to previous estimation results.
Table 4.15 reports OLS estimates of EV model with F-UN variable. The
coefficient estimates of F-UN in all specifications are negative and statistically
significant at the 1 percent level. An increase In the firm unionization rate from
TABLE 4.14 OLS Results of ROI Model with Firm
Unionization, 1986-88 (t-Statistics in Parentheses)
95
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
R2
N
Industry Dummies
(1)
0.160* (9.72)
-0.046* (3.09)
0.032* (4.99)
0.052* (3.17)
0.186* (3.83)
0.087 (0.88)
0.011 (1.29)
-0.027 (1.06)
0.071* (2.48)
0.29
204
no
(2)
0.170* (18.0)
-0.069* (4.32)
0.021* (3.52)
0.081* (4.77)
0.209* (4.40)
—
—
—
0.010* (5.86)
0.27
258
no
(3)
0.111* (9.44)
-0.052* (2.63)
0.019 (1.33)
0.070* (4.25)
-0.046 (0.70)
—
—
—
0.003' (1.92)
0.46
258
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
96
TABLE 4.15 OLS Results of EV Model with Firm
Unionization, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-UN
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R
N
Industry Dummies
(1)
0.139" (1.95)
-0.243* (4.14)
0.094 (1.64)
3.030* (9.99)
0.410 (0.59)
0.041 (0.83)
-0.007 (0.08)
-0.064 (0.52)
0.44
173
no
(2)
0.236* (9.90)
-0.338* (6.15)
—
3.634' (9.61)
0.47
228
no
(3)
0.118* (3.16)
-0.222* (3.51)
—
3.173* (5.32)
—
—
—
—
0.58
228
yes
* *Signlfleant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
97
0 percent to 100 percent implies 25.1 percent decline in EV in specification (1),
34.9 percent decline in EV in specification (2) and 22.9 percent decline in EV in
specification (3). In addition to the F-UN variable, among all other explanatory
variables, only ADVES/S, one measures of intangible assets, is found to have
a positive and statistically significant effect on EV. Even though the magnitude
of estimated coefficients of ADVES/S in this subsection are smaller than those
of ADVES/S in our previous cross-section results, these coefficients are still
large in terms of the magnitude. The estimated coefficients of ADVES/S in
three specifications are 3.030, 3.634, and 3.173, respectively.
4.2.1.2 The OLS Estimation Results of the PCM. ROI and EV Models with Firm Union Coverage Dummies
Tables 4.16-4.18 demonstrate the OLS estimates of all three models with
F-DUMMYs, the alternative measures of F-UN. The OLS results of the PCM
model are reported in Table 4.16. From the specifications (1) and (2), the
estimated coefficients of F-DUMMY2 are negative and statistically significant at
the 1 percent level. The estimated coefficient of F-DUMMY2 is -0.024 in
specification (1). Again, it is detected that significant union gains arise
especially over the range of between 30 and 60 percent of unionization.
Besides F-DUMMYs, in both of the first two specifications all other explanatory
variables except ADVES/S and IINV/VIS have positive and statistically
significant impacts on PCM. The coefficients of these explanatory variables
98
TABLE 4.16 OLS Results of PCM Model with Firm
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R
N
Industry Dummies
(1)
0.158* (10.7)
-0.017 (1.58)
-0.024* (2.87)
0.062* (11.0)
0.068* (5.92)
0.024 (0.60)
0.174" (2.35)
0.019" (2.42)
-0.019 (0.77)
0.085* (3.82)
0.49
204
no
(2)
0.161* (12.0)
-0.019" (2.23)
-0.022* (2.72)
0.063* (11.6)
0.074* (6.38)
—
0.182* (2.66)
0.018* (2.70)
—
0.002 (1.25)
0.46
243
no
(3)
0.121* (6.66)
-0.011 (1.08)
-0.010 (1.12)
0.047* (6.29)
0.065* (5.27)
—
0.119 (1.27)
0.028* (3.54)
—
0.001 (1.16)
0.53
243
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY3 is 0.30>F-UN.
99
which are reported are found to be similar in terms of size and sign to those In
Table 4.13.
The results of ROI and EV models with firm union coverage dummies
are summarized in Tables 4.17 and 4.18, respectively. Once more, as seen
from all of the specifications of both tables, the estimated coefficients of
F-DUMMY2 are negative and statistically significant at least at the 1 percent
level. AFrom specification (1) of each table, it was seen that the estimated
coefficient of medium union coverage, F-DUMMY2, was -0.033 for ROI and
-0.168 for EV model. As shown in Table 4.17, in addition to union coverage
dummies, all other independent variables except RDES/S, FS and IINV/VIS
have positive and statistically significant impacts on ROI. Unfortunately, from
Table 4.18, only ADVES/S among all other explanatory variables in the EV
model has a positive and statistically significant coefficient at least at the 10
percent level. Once more, the magnitudes and signs of the estimated
coefficients of statistically significant explanatory variables which are presented
in Tables 4.17 and 4.18 are similar to those in Tables 4.14 and 4.15.
4.2.1.3 The OLS Estimation Results of the PCM. ROI and EV Models with Industry Union Coverage Dummies
The OLS results of PCM, ROI and EV models with industry union
coverage dummies are presented in Tables 4.19, 4.20 and 4.21, respectively.
For the PCM model, the estimated coefficients of medium union coverage,
I-DUMMY2 is negative and statistically significant. While the spline function
TABLE 4.17 OLS Results of ROI Model with Firm
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
100
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R2
N
Industry Dummies
(1)
0.157* (9.63)
-0.010 (0.81)
-0.033* (3.36)
0.034* (5.41)
0.061* (3.58)
0.175* (3.52)
0.137 (1.38)
0.010 (1.16)
-0.026 (1.00)
0.077* (2.83)
0.31
204
no
(2)
0.164* (18.4)
-0.036* (2.69)
-0.040* (4.22)
0.020* (3.29)
0.084* (4.71)
0.189* (3.92)
—
—
—
0.010* (5.39)
0.28
258
no
(3)
0.105* (9.63)
-0.035" (2.39)
-0.026* (2.78)
0.020 (1.33)
0.073* (4.37)
0.057 (0.86)
—
—
—
0.004" (1.99)
0.46
258
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY3 is 0.30>F-UN.
TABLE 4.18 OLS Results of EV Model with Firm
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
101
Independent Variables
CONSTANT
F-DUMMY1
F-DUMMY2
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R
N
Industry Dummies
(1)
0.143* (2.17)
-0.150* (3.49)
-0.168* (4.59)
0.126* (2.22)
2.896* (9.57)
0.574 (0.82)
0.026 (0.54)
-0.018 (0.19)
-0.013 (0.11)
0.46
176
no
(2)
0.220* (10.0)
-0.218* (5.80)
-0.196* (6.13)
—
3.499* (9.06)
—
—
—
—
0.43
228
no
(3)
0.089" (2.49)
-0.148* (2.83)
-0.135* (3.34)
0.113" (2.03)
3.040* (5.24)
—
—
—
—
0.59
228
yes
* a, Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* F-DUMMY1 is F-UN>0.60; F-DUMMY2 is 0.60>F-UN>0.30; F-DUMMY2 is 0.30>F-UN.
102
TABLE 4.19 OLS Results of PCM Model with Industry
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R 2
N
Industry Dummies
(1)
0.140* (10.42)
-0.023 (1.46)
-0.019* (3.42)
0.058* (9.86)
0.064* (5.77)
0.064 (1.46)
0.054 (0.71)
0.031* (4.23)
-0.011 (0.58)
0.089* (4.12)
0.52
204
no
(2)
0.130* (7.06)
-0.029 (1.53)
-0.020* (3.00)
0.048* (7.30)
0.001 (1.26)
—
—
0.035* (5.06)
—
0.004" (2.24)
0.33
305
no
(3)
0.098* (5.93)
0.004 (0.41)
-0.005 (0.72)
0.031* (4.30)
0.001 (1.29)
—
—
0.036* (5.39)
—
0.002 (1.40)
0.48
305
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
103
shows a statistically significant negative effect of only medium unionization on
PCM, it is observed that the effect of medium unionization is smaller than that
of high unionization. From the first specification, the estimated coefficient is
found to be -0.023 for I-DUMMY1, -0.019 for I-DUMMY2. All other explanatory
variables except intangible asset measures (ADVES/S and RDES/S) and
IINV/VIS have a positive and statistically significant impact on PCM.
Similarly from Tables 4.20 and 4.21 for both ROI and EV models, the
spline function indicates a strong negative union effect for firms in where the
industry union coverage ratio is between 30 and 60 percent, and then indicates
a weak significant slope thereafter. The estimated coefficient of I-DUMMY2 in
the specification (1) is -0.015 and -0.141 for ROI and EV models, respectively.
These results imply that unions in the firms in where the industry union
coverage ratio is between 30 percent and 60 percent capture the major portion
of ROI and EV. The impact of all other explanatory variables on ROI and EV
is found to be qualitatively and quantitatively similar to those in Tables 4.11
and 4.12.
4.2.1.4 Test Results for Sources of Union Gains
Evidence that research and development intensity still provides an important
source for union gains remains strong in all three models.''^ From Table 22,
^^0 test whether research and development intensity is an important source of union gains, three profit measures which are PCM, ROI and EV are separately estimated as a function of only F-UN, RDES/S and F-UN*RDES/S.
104
TABLE 4.20 OLS Results of ROI Model with industry
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
R2
N
Industry Dummies
(1)
0.138* (8.90)
0.013 (0.99)
-0.015" (2.17)
0.031* (5.00)
0.056* (3.29)
0.226* (4.37)
0.042 (0.42)
0.021" (2.39)
-0.020 (0.91)
0.074" (2.46)
0.29
204
no
(2)
0.125* (7.98)
0.022 (1.17)
-0.013" (2.33)
0.022* (3.53)
0.082* (4.37)
0.259* (5.35)
—
0.018" (2.34)
0.011* (6.03)
0.24
258
no
(3)
0.071* (4.51)
-0.016 (0.19)
-0.012 (1.31)
-0.020 (1.13)
0.071* (3.70)
0.150 (1.47)
—
0.021* (2.86)
—
0.004" (2.10)
0.46
258
yes
* *Significabt at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
* I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
105
TABLE 4.21 OLS Results of EV Model with Industry
Unionization Dummies, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
I-DUMMY1
I-DUMMY2
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
R
N
Industry Dummies
(1)
0.049 (0.90)
0.103* (1.50)
-0.141* (4.61)
0.129" (2.23)
3.271* (10.3)
-0.052 (0.07)
0.099* (2.45)
0.035 (0.49)
0.027 (0.22)
0.49
176
no
(2)
0.008 (0.14)
0.091* (1.46)
-0.095 (2.61)
0.159" (2.33)
3.832* (10.3)
—
0.106* (3.09)
—
—
0.49
228
no
(3)
-0.024 (0.35)
0.059 (0.79)
-0.090' (1.85)
0.128" (2.02)
3.139* (5.26)
—
0.084' (2.71)
0.60
228
yes
• a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level. I-DUMMY1 is l-UN>0.60; I-DUMMY2 is 0.60>I-UN>0.30; I-DUMMY3 is 0.30>I-UN.
106
TABLE 4.22 Test Results for Source of Union Gains for PCM, ROI and EV Models, 1986-88
(t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-UN
RDES/S
F-UN*RDES/S
R2
N
PCM
0.144* (21.7)
-0.025 (1.39)
0.149' (1.69)
-1.404* (3.29)
0.13
243
ROI
0.153* (24.1)
-0.016 (0.82)
0.231" (2.24)
-1.624* (3.52)
0.15
243
EV
0.232* (5.11)
-0.019 (0.16)
2.415* (2.77)
-11.06* (3.88)
0.20
212
•k a Significant at the 0.01 level; "Significant at the 0.05 level; Significant at the 0.10 level.
107
the coefficients on the interaction term of F-UN*RDES/S are negative and
statistically significant at least at the 1 percent level. No surprisingly, the
estimated coefficients of F-UN for all three profit measures become statistically
insignificant when the interaction of F-UN*RDES/S is added to specifications
because the significant part of union coverage effect on profitability is captured
by this interaction term. The results are consistent with those of Hirsch (1990),
Becker and Olson (1992) and others."* Hirsch (1990) provided evidence that
sources of union gains did not result from output market structure (market
concentration rate, market share of the firm). In his study, he found that the
interaction term of the research and development intensity and firm unionization
measure had a negative and statistically significant impact on both Tobin's q
and ROI. In the same study, he also reported that the interaction term between
the firm unionization measure and firm market share had a positive end
significant effect on both profit measures. Hirsch's later finding supported the
finding of Clark (1984) which states that unions reduced profits only among
those businesses with small market shares even though Clark used completely
different data source from Hirsch's. Surprisingly, the finding of a positive and
significant estimated coefficient on the interaction term between unionization
and market share was criticized by Hirsch and Connolly (1987), Freeman
^ For the extensive survey on sources of union gains, see Hirsch and Addision (1986, p. 214) and Addision and Hirsch (1989, pp. 92-95).
108
(1983).'^ At that time, Hirsch and Connolly (1987) and Freeman (1983)
argued that Clark's argument and finding on this issue was specific to the PIMS
data.
4.2.1.5 Summary Assessment for the OLS Estimation Results over the 1986-88 Period
in this subsection, a framework is developed for analyzing the effect of
unionization on PCM, ROI and EV over the 1986-88 period by using OLS after
applying the White correction to error terms of all specifications. As mentioned
eariier, the analysis again provides clear evidence that unionized firms earn
substantially lower return than nonunionized firms. From specification (1)
reported in Tables 4.13-4.15, for fully unionized firms relative to lowly unionized
firms, PCM is lower by 7.1 percent, ROI by 7.2 percent and EV by 24.3
percent. The size of this negative effect is comparable to the size found by a
number of previous studies which are also used the sample data to cover more
than one year. By using aggregate data for manufacturing industries and
evaluating his findings at the mean of a dependent variable, not at the mean of
both dependent and independent variables. Freeman (1983) found that unions
reduce PCM by about from 39 percent to 44 percent for the Internal Revenue
^ Even though the measure on market share of firm is weak measure, when the models are estimated with interaction term between F-UN and MSF, significant and positive coefficients for this interaction term for all three models are found, which confirms the finding of Clark (these results are not shown on the table).
109
Service sample, by about from 13 percent to 19 percent for the Annual Survey
Manufactures sample. On the other hand, using Japanese manufacturing firms
as a sample and applying OLS technique, Brunello (1992) found that PCM was
lower by 36.5 percent and also ROI by 19.5 percent for fully unionized firms.
Hirsch (1991a) used firm level data and estimated that the ROI was 9.2 percent
lower in unionized firms than in nonunionized firms (at the mean of both
dependent and independent variables). Using micro data Becker and Olson
(1992) also found that EV and PCM were lower by 18 percent and 30 percent
(at the mean of the dependent variable), respectively.
On the other hand, only RDES/S among the explanatory variables in this
subsection experiences the mixed figure as in the first section of this chapter.
As in the first two specifications of Table 4.13, the estimated coefficients of
RDES/S are positive and statistically significant at least at the 5 percent level,
suggesting the significant effect of RDES/S on PCM measure. However, it is
found that the effect of RDES/S on both ROI and EV measures is insignificant,
which is also consistent with the results of the cross-section estimation.
As seen from Tables 4.16-4.21, when F-DUMMY and l-DUMMY which
are entered to the specifications separately, it is evident that the basic results
from PCM, ROI and EV are little affected by the new form of the firm
unionization measures. However, some interesting patterns emerged In the
comparisons of the coefficient estimates of each measure. According to the
estimated coefficients of both F-DUMMYs for all three models, the important
110
share of the union gain is distributed over the range of between 30 percent and
60 percent of the unionization. This result is partly consistent with that of
Hirsch (1991a) but inconsistent with the result of Becker and Olson (1992).
Hirsch (1991a) found that important portion of union gains occurs across 30
percent and over percent of unionization in Tobin's q equation, whereas
unionization had statistically significant impact on the rate of return on capital in
all ranges of unionized percentage. On the other hand, Becker and Olson
found that the union gains occur over the first 30 percent of unionization, not
over the first 60 percent.
In addition, the estimated coefficients of industry union coverage
dummies, l-DUMMYs, in all regression specifications also suggest that the
notable portion of union gains occurs over the range of between 30 and 60
percent of unionization. The findings from all specifications with industry
unionization dummies are in conflict with the findings of Becker and Olson
(1992). In their study, they found statistically insignificant coefficient estimates
for all of F-DUMMYs. The difference between these results and Becker and
Olson's, who obtained the industry union measure used from Kokkelenberg and
Sockell (1985), could be attributed to the source of data for the industry union
variable. On the other hand, even though Hirsch (1991a) used the same data
source for l-UN as does this study, comparable results for I-DUMMY1,
I-DUMMY2 and I-DUMMY3 do not occur, since in his study, he only considered
the effects of industry union coverage ratio, l-UN, but did not consider the
I l l
effects of the industry union coverage dummies (I-DUMMY1, I-DUMMY2 and
I-DUMMY3).
In general, all findings on the union-profit relation from all three models
under the cross-sectional data are qualitatively and quantitatively repeated in
this subsection. Again, all three models suggest that the negative and
significant union effect on profitability occurs for the firms that are moderately
unionized. Similarly, when l-DUMMYs are included in the specifications of all
three models, the above conclusion was repeated: a negative and significant
union effect on profitability for the firms whose Industry union coverage ratio are
between 30 and 60 percent. In addition, the comparison of cross-section
estimates of PCM, ROI and EV models to the estimates of all three models
over the 1986-88 period provides evidence that the number of explanatory
variables which has significant impact on PCM and ROI during 1986-88 is
greater than that during 1988.
4.2.2 The Results of the Fixed-Effect Regressions over the 1986-1988 Period
In this second subsection, the results of PCM, ROI and EV models under
fixed-effect technique over the 1986-88 period are discussed. Unfortunately,
the estimated coefficients of all three models under the fixed-effect technique
with individual effects are statistically insignificant. The insignificant coefficients
of all are found for all explanatory variables because large degrees of freedom
112
are lost when the firm means are removed. ^ On the other hand, when
regression equations are estimated with time-effects, the coefficients of most of
the explanatory variables are statistically significant at least at the 10 percent
level. '* Therefore, only the estimations of all three models under the
fixed-effect technique with time effects are presented. Here, to correct
heteroscedastic error terms no method is applied because one way to deal with
this problem is to use the fixed-effect technique itself.""
Additionally in this subsection three models are estimated with only firm
union coverage ratio."" The estimation results of all PCM, ROI and EV
models under the fixed-effect technique are shown in Tables 4.23-4.25,
respectively. Specification (1) of each table presents the estimation results of
each of three models with F-UN and the rest of the explanatory variables
whereas specification (2) reports the estimation results of the same model with
F-UN and some of explanatory variables which remain after stepwise
^ Under the fixed-effect technique with individual firm effects, the degrees of freedom of the sample is equal to the number of observations minus number of explanatory variables minus number of firms.
" Under the fixed-effect technique with time effects, the degrees of the freedom of the sample is equal to the number of observations minus number of explanatory variables minus number of periods (t=3).
^ For more information on this topic see Hsiao (1991).
^ Under the fixed-effect technique, dummies which are alternative measures of firm union coverage ratio cannot be used because the changes in firm profitability measures, PCM, ROI and EV are estimated as a function of changes in union coverage measure (F-UN) and in the rest of explanatory variables.
113
procedure. Also, in this section the estimation results of the PCM, ROI and EV
models with industry union coverage ratio are unreported since there were not
any statistically significant coefficient estimates of this variable for all three
models.
4.2.2.1 The Estimation Results of the PCM. ROI and EV Models with Firm Union Coverage Ratio under the Fixed-Effect Technigue
According to the coefficient estimates of F-UN, all three models strongly
support the hypothesis of negative and strong union effect on profitability. The
estimated coefficients of F-UN in both specifications of all three models are
found to be negative and statistically significant at the 1 percent level. More
specifically the coefficient estimate of F-UN is -0.048 for PCM model, -0.054 for
ROI model and -0.29 for EV model. Evaluated at the mean level, these
coefficients indicate that PCM, ROI and EV for a fully unionized firm relative to
a nonunionized, are reduced by 8.6, 8.4 and 30 percent, respectively.
Furthermore, all other explanatory variables except GROWTH and
IINV/VIS are found to have positive and statistically significant effect on PCM.
From specification (1), the estimated coefficient of FCAP/S is 0.036. Moreover
the estimated coefficients of FS and IGROWTH in the first specification are
0.031 and 0.104, respectively.
A comparison of the OLS and fixed-effect regression results for the PCM
model indicates that both ADVES/S and RDES/S become statistically more
114
TABLE 4.23 Fixed-Effect Results of PCM Model with
Firm Unionization, 1986-88 (t-Statistics in Parentheses)
Independent Variables
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
R
N
(1)
-0.048* (4.12)
0.036* (3.52)
-0.001 (1.19)
0.066' (1.69)
0.148" (1.90)
0.031* (3.87)
-0.035 (1.21)
0.104* (3.59)
0.36
201
(2)
-0.048* (4.06)
0.037* (3.59)
—
0.070' (1.76)
0.155" (1.96)
0.032* (3.94)
-0.037 (1.21)
0.057* (2.82)
0.35
201
* ai Significant at the 0.01 level. "Significant at the 0.05 level. 'Significant at the 0.10 level.
115
TABLE 4.24 Fixed-Effect Results of ROI Model with
Firm Unionization, 1986-88 (t-Statistics in Parentheses)
Independent Variables
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
R 2
N
(1)
-0.054* (3.47)
0.005 (0.69)
-0.001 (1.45)
0.196* (3.53)
0.152" (1.84) .
0.016' (1.81)
-0.055' (1.73)
0.106* (3.20)
0.24
201
(2)
-0.054* (3.47)
—
—
0.191* (3.50)
0.163" (1.98)
0.016' (1.75)
-0.065' (1.85)
0.056" (2.28)
0.23
201
* a Significant at the 0.01 level. "Significant at the 0.05 level. 'Significant at the 0.10 level.
116
TABLE 4.25 Fixed-Effect Results of EV Model with
Firm Unionization, 1986-88 (t-Statistics in parentheses)
Independent Variables
F-UN
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
R
N
(1)
-0.298* (4.50)
0.001 (0.01)
3.570* (9.46)
0.032 (0.09)
0.068 (1.41)
0.013 (0.14)
-0.092 (0.59)
0.49
183
(2)
-0.290* (4.79)
—
3.604* (10.0)
—
0.060 (1.49)
0.47
228
* a Significant at the 0.01 level. "Significant at the 0.05 level. 'Significant at the 0.10 level.
117
significant and more Important determinants of the firm profitability under this
technique. From specification (1), the estimated coefficient is 0.066 for
ADVES/S and 0.148 for RDES/S.
For the ROI model, besides F-UN, the estimated coefficients of all other
explanatory variables except FCAP/S and GROWTH are found to be
significantly different from zero. Once more, both intangible capital measures,
ADVES/S and RDES/S become statistically more significant and more notable
determinants of ROI under the fixed-effect technique and estimated coefficients
in specification (1) are found to be 0.196 for ADVES/S and 0.152 for RDES/S.
While the estimated coefficient of FS in specification (1) is 0.016, the estimated
coefficient of IGROWTH is 0.106. On the other hand, the value of industry
inventory, IINV/VIS, is found to have a negative and statistically significant
impact on ROI, and the estimated coefficient of this variable in specification (1)
is -0.065.
Finally, in both specifications of EV model, in addition to the firm union
coverage variable, only ADVES/S among all explanatory variables has a
positive and statistically significant estimated coefficient and the coefficient of
ADVES/S in the first specification is estimated to be 3.570.
118
4.2.2.2 Summary Assessment for the Estimation Results of all PCM. ROI and EV Models under the Fixed-Effect Technigue
Using panel data, comparison of the OLS and fixed-effect estimates
reveals some changes in the results for all three models. The estimated
coefficients of F-UN for all three models under both the OLS and fixed-effect
techniques are similar in terms of sign, but differ in terms of magnitude and
significance level. The estimated coefficients of F-UN, ADVES/S, FS and
IINV/VIS for all three models under the fixed-effect technique are larger and
more significant than those under the OLS technique. Also, the coefficient
estimates of RDES/S for the both PCM and ROI models under the fixed-effect
technique are larger than those under the OLS technique.
Perhaps the most interesting additional information included in the fixed-
effect results for PCM and ROI models is the negative, insignificant and small
coefficients of the GROWTH variable and unstable coefficients of IGROWTH
from one specification to another. Actually, this unexpected and unstable
behavior of the coefficient estimates signifies that both GROWTH and
IGROWTH measures are not appropriate to explain the profitability under the
fixed-effect technique. First of all, both measures are constructed by taking
percentage changes of both variables, that is, they are already once
differenced. Since the fixed-effect technique differences all variables once
more, then the GROWTH and IGROWTH variables actually represent second
difference of sales and value of Industry shipments, respectively. Undoubtedly,
119
the second difference caused both variables to lose significant information on
the relationship between profitability and GROWTH (or IGROWTH). Therefore,
the result on the relationship of GROWTH (or IGROWTH) and profitability under
the fixed-effect technique cannot be reliable. From all findings in this section, it
can be concluded that the results of all models under the fixed-effects
technique seem to be more robust than OLS results because the fixed-effect
technique provides the possibility to capture the effects of omitted and
unobservable variables which are related to both firm profits (especially PCM
and ROI) and other explanatory variables (especially F-UN, ADVES/S, RDES/S,
FS and IINVA/IS).
4.2.3 The Results of the Random-Effect Regressions
over the 1986-88 Period
The results of PCM, ROI and EV models under the random-effect
technique are pursued in Tables 4.26-4.28, respectively, in this third subsection.
In this technique the same specifications are also used for all three model as in
the previous part of this chapter. While the estimates of all three models with
all explanatory variables are shown in specification (1), the estimates of the
same models with some of independent variables which remain after the
stepwise estimation procedure are presented in specification (2). The final
specifications in those tables explore the estimates of the 16 two-digit industry
dummies and explanatory variables which are already used in the second
specification. Here again, we did not apply any method to correct
120
heteroscedastic error terms since the random-effect technique itself corrects for
this problem.
4.2.3.1 The Estimation Results of the PCM. ROI and EV Models with Firm Union Coverage Ratio under the Random-Effect Technigue
Once more, in all specifications of all three models under the
random-effect technique, the estimated coefficients on the F-UN variable for all
PCM, ROI and EV models still remain negative and statistically significant at the
1 percent level. From the first specification of Tables 4.26-4.28, PCM is
reduced by 8.0 percent, ROI by 7.3 percent and EV by 29 percent when firm
union coverage, F-UN, is increased by 100 percent."*^
As seen from the first specification for the PCM model, while the
coefficient estimates of FCAP/S, GROWTH, FS and IGROWTH are all positive
and significant at least at the 5 percent level, the coefficients of ADVES/S,
RDES/S and IINV/VIS are not significantly different from zero at least at the 10
percent level. From the first specification of Table 4.26, the coefficient is
estimated to be 0.040 for FCAP/S, 0.029 for GROWTH, 0.028 for FS and 0.127
for IGROWTH. The comparison of the estimation results of PCM model under
the fixed-effect technique with those under the random-effect technique reveals
^^When F-DUMMYs and l-DUMMYs are employed rather than union coverage ratio under the random-effect technique, the similar result are found as under the OLS; the significant portion of union gains occurred over 30-60 percent of unionization.
121
TABLE 4.26 Random-Effect Results of PCM Model with
Firm Unionization, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R 2
N
Industry Dummies
(1)
0.128* (9.04)
-0.045* (3.38)
0.040* (5.79)
0.029* (3.52)
0.076 (1.27)
0.101 (1.38)
0.028* (3.87)
-0.024 (1.06)
0.127* (4.28)
16.59
0.40
201
no
(2)
0.145* (10.7)
-0.047* (4.02)
0.047* (9.05)
0.030" (2.58)
—
0.134' (1.93)
0.028* (4.00)
—
0.109* (4.26)
24.79
0.39
236
no
(3)
0.156* (7.97)
-0.039 (2.63)
0.048* (6.93)
0.015' (1.73)
—
0.045 (0.70)
0.029* (4.27)
—
0.065* (2.57)
14.16
0.58
236
yes
* *Significant at the 0.01 level; "Significant at the 0.05 level;'Significant at the 0.10 level.
122
TABLE 4.27 Random-Effects Results of ROI Model with
Firm Unionization, 1986-88 (t-Statistics in Parentheses)
Independent Variables
CONSTANT
F-UN
FCAP/S
GROWTH
ADVES/S
RDES/S
FS
IINVA/IS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.147* (9.15)
-0.047* (2.87)
0.013' (1.69)
0.058* (4.52)
0.210* (3.09)
0.091 (1.10)
0.015' (1.79)
-0.051" (2.04)
0.024 (0.71)
11.03
0.31
201
no
(2)
0.172* (12.1)
-0.052* (4.19)
0.014" (2.15)
0.051* (4.58)
0.167" (2.47)
—
0.014'' (2.12)
-0.065' (2.67)
14.30
0.25
258
no
(3)
0.097* (5.27)
-0.043" (2.38)
-0.024 (1.33)
0.041* (3.81)
0.149" (2.11)
—
0.019" (2.50)
-0.037 (1.55)
—
8.21
0.42
258
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
TABLE 4.28 Random-Effect Results of EV Model with
Firm Unionization, 1986-88 (t-Statistics in Parentheses)
123
Independent Variables
CONSTANT
F-UN
GROWTH
ADVES/S
RDES/S
FS
IINV/VIS
IGROWTH
F-Statistics
R
N
Industry Dummies
(1)
0.148' (1.93)
-0.281* (3.41)
0.120' (1.90)
3.533* (10.2)
-0.127 (0.30)
0.068 (1.58)
-0.016 (0.14)
-0.147 (0.82)
25.4
0.50
183
no
(2)
0.229* (8.17)
-0.289* (4.65)
0.096' (1.87)
3.641* (11.7)
—
—
—
—
68.80
0.47
228
no
(3)
0.100* (3.00)
-0.174" (2.45)
0.073 (1.58)
3.339* (11.3)
—
—
—
—
22.75
0.66
228
yes
* a Significant at the 0.01 level; "Significant at the 0.05 level; 'Significant at the 0.10 level.
124
that the intangible capital measures, ADVES/S and RDES/S, become important
determinants of PCM under the fixed-effect technique, but not under the
random-effect technique.
From the first two specifications of Table 4.27, besides F-UN, FCAP/S,
GROWTH, ADVES/S, FS and IINV/VIS among all explanatory variables are
found to have a statistically significant impact on ROI. As pursued in the first
specification, the estimated coefficient was found to be 0.013 for FCAP/S, 0.058
for GROWTH, 0.210 for ADVES/S, 0.015 for FS and -0.051 for IINV/VIS.
Again, no significant effect of RDES/S on ROI is found.
On the other hand, from Table 4.28, besides unionization only
GROWTH and ADVES/S have a positive and statistically significant impact on
EV. The estimated coefficients of GROWTH and ADVES/S in specification (1)
are 0.120 and 3.533, respectively.
4.2.3.2 Summary Assessment for the Random-Effect Regressions
Once more, all three models under the random-effect technique
supported the proposition of a negative and significant union effect on
profitability. A comparison of the random-effect and fixed-effect estimates
reveals no significant qualitative and quantitative change in supporting the
hypothesis of negative union effect on firm profitability. However, when both
techniques are compared in terms of the magnitudes of the estimated
coefficients from one specification to another, the coefficient estimates under
125
the fixed-effect technique are more stable than those under the random-effect
technique. On the other hand, the fixed-effect technique suggests a positive
and significant effect of stock value of research and development on both PCM
and ROI measures while the random-effect technique indicates that the effect of
that variable on both profitability measures is not statistically different from zero.
CHAPTER V
SUMMARY AND CONCLUSION
Since the 1970s, unionization in the manufacturing sector of the United
States has begun to decline substantially. With this substantial decline, the
literature on labor and industrial economics has shifted its focus on the
relationship between unionization and profitability. Although in the literature, the
profitability in terms of its determinants has been theoretically investigated for a
long time, the formal empirical investigation on the effects of unions on
profitability starts only at the beginning of 1980.
The pioneering studies of Freeman (1983) and Clark (1984) in this
specific literature were the first to attempt to investigate the effects of unions on
profitability. By using a sample of industry level data for the period 1958-1976,
Freeman (1983) concluded that there was a negative and significant union
effect on both price-cost margin and the ratio of quasi-rents to capital during the
underlying period. Just afterward, the findings of Clark (1984), which come
from the use of a sample of North American product-line business data, and the
use of rate of returns on sales and capital as a measure of profitability,
supported Freeman's conclusion.
After these two influential empirical studies, an explosive research on the
union-profitability effect has developed. In general, there has been agreement
126
127
on the negative union effect on profitability despite the sample of data, the
measure of profitability, and the measure of unionization which differ from
studies to studies. But, the dispute pertaining to the union-profit effect is
ongoing as to the size of the negative union effect on the profitability,
depending on the measure of profitability, the estimation technique, the
measure of unionization and alternative measures of unionization since the
effect of unionism on profitability differs depending on the sample of data, the
measure of profitability, the measure of unionization and the estimation
technique used.
In addition, the estimated union effect on profitability in the previous
studies would be biased since these studies could not specify their profitability
equations so as to capture the possible effects of omitted or mismeasured firm
specific, industry specific and/or economy wide variables which exhibit variation
across firms and/or through time, and also correlated with firm unionization and
profitability.
In the light of the shortcomings of the existing literature on the
union-profit effect, the main objective of this study was to analyze statistically
and comprehensively the effect of unionism on firm profitability and the
sensitivity of such effect to different estimation techniques which were the OLS,
fixed-effect and random-effect techniques; to different profitability measures
which were PCM, ROI and EV; to different unionization measures which were
firm and industry union coverage ratios; and to alternative measures of
128
unionization which were union dummies and coverage ratio. In addition to the
main objective, the secondary objective of the study was to examine whether
the returns of research and development investment were still a major source of
union gains.
The fundamental analytical tools being used to carry out the stated
objectives were ordinary least squares, fixed-effect and random-effect
techniques. In the presence of heteroscedastic error terms under OLS
technique, the White correction approach was employed to correct the residual
terms. Finally, the investigation for any endogeneity between unionization and
profitability was carried out by using the Hausman exogeneity test procedure.
5.1 Summary of Results
In this study the effect of unionization on firm profitability was
investigated in some detail. By any of three profit measures, union measures
and estimation techniques, the effect of unionization on each was found to be
negative and significant, even though unionization declined significantly on
average in the firms which were in the private sector for the last ten years.
Regarding the union effect on profit measures, comparing three
estimation techniques in terms of the effect of unionization on profitability
measures, the effect which was estimated under the fixed-effect technique
differed from that under both the OLS and random-effect techniques, implying
129
that some unobservable or omitted variables which were related to both
profitability and unionization measures existed.
Also, the size of the effect of union coverage on profitability differed
depending on profit measures. Comparing the effect of firm unionization on
both PCM and ROI with that on EV, such effect on the long-run profitability
measure was greater than that on short-run measures. For example, when
excess value is proxied to firm profits, the findings from Tables 4.23-4.25
indicated that profitability for a fully unionized relative to a nonunionized firm
was reduced by 30 percent. In the case of price-cost margin, the results
suggested that profitability for a fully unionized firm relative to a nonunionized
firm was reduced by 8.6 percent. On the other hand, a comparison of the
union effect on ROI with that on PCM indicated no significant difference
between two short-run measures regarding the unionization effect.
From the findings of this study, it was also observed that the effect of
unionization on each profit measure differed depending on the union measures.
When the firm union coverage ratio was proxied for firm unionization, all three
models supported the proposition of negative union effect on firm profitability
regardless of which measures were used as a proxy for profitability. On the
contrary, all models provided no evidence of a negative union effect on
profitability when the industry union coverage ratio was proxied for firm
unionization. On the other hand, when both the firm and the industry union
coverage dummies were proxied for firm unionization, the findings from both
130
alternative measures showed no significant change. For example, the
estimation results from both measures provided strong evidence on the
negative union effect on profitability. Also, according to the estimation of both
measures, the spline function of both firm union coverage and industry union
coverage provided evidence that a significant portion of union gains existed
especially over the range of between 30 and 60 percent of unionization.
Finally, the estimations throughout the 1986-88 period showed that
research and development investment was still a major source of union gains in
spite of the fact that the investment on research and development has declined
in recent years.
5.2 Conclusion
Findings of this research on a sample of large U.S. manufacturing firms
for the 1986-1988 period provide strong support for the hypothesis of negative
union effect on profitability. Despite differences among analysis in
methodology, measures of profitability and unionization, time period, data
sources and unit of observations, the findings in this study are consistent with
most of previous studies in terms of supporting the hypothesis of a negative
union effect on profitability. The main conclusion of this study is that the effect
of unionization on firm profitability is negative and substantial regardless of
different estimation techniques, different profit measures and different
unionization measures. This study also concludes that the size of the negative
131
union effect differs depending on the estimation techniques, measures of
profitability and unionization measures as well as the alternative measures of
unionization.
Most importantly, the union effect on profitability differs depending on the
estimation technique, although the negative effect is found in all three
estimation techniques; the effect estimated by the fixed-effect technique is
found to be larger than that estimated by both the OLS and the random-effect
techniques. Theoretically, the existence of the difference between the size of
the effects estimated by the OLS and the fixed-effect techniques is an indication
of the existence of the omitted and/or unobserved variables such as worker
quality which are correlated with included explanatory variables such as union
status in profitability equations. Even though random-effect technique captures
individual and/or time effects in error terms, there is no justification for treating
individual effects and/or time effects as uncorrelated with other included
explanatory variables, as is assumed in the random-effect technique. The
random-effect approach may, therefore, suffer from the inconsistency due to the
omitted variables. In the light of the above explanation, the fixed-effect
technique is preferable to any alternative techniques used in this study.
The magnitude of the negative union effect on profit also differs
depending on the profitability measures. In this study, the size of the effect
under long-run profitability measure such as EV is found to be significantly
larger than that under the short-run measure such as PCM or ROI. This
132
difference indicates that investors expect the union to have a detrimental effect
on firm growth and future earnings. Theoretically, use of long-run profit
measures is preferable to short-run profitability measures because of the
dynamic characteristics of long-run profit measures. On the other hand, since
excess value incorporates businesses' expectations about unionism's effect on
future profits , it has an unknown degree of error. Then, empirically, the union
effect on profitability can be overestimated or underestimated. By taking into
account both theoretical advantage and empirical disadvantage of EV, and
empirical advantage and theoretical disadvantage of short-run measures such
as PCM or ROI, this study is indecisive on the preference of profitability
measures.
Finally, the union effect on profitability differs depending on unionization
measures; that is, the effect under the firm union coverage ratio is negative and
significant while it is insignificant under industry union coverage ratio. On the
other hand, the effect under the dummy variables, regardless of which union
measures are used to construct them, is negative and significant only for the
firms whose the underlying union coverage ratio is between 0.30 and 0.60. As
is understood from these findings, when the firm union coverage ratio is
compared to the industry union coverage ratio in terms of performance to test
the underlying hypothesis, this study shows that the firm union coverage ratio is
more favorable than the Industry union coverage. But, for the cases of the
dummy variables, this study is not conclusive on the choice between firm and
133
industry union coverage. Finally, our study suggests that the choice between
the firm union coverage ratio and the dummies constructed from such coverage
will depend on the purpose of the researcher; that is. If someone prefers to look
at the union effect on profitability only for those firms which have unionization
ratios between certain ranges, the dummy approach would be preferable to the
alternative one, but if the purpose is to find the effect of unionization on
profitability for all firms in the sample, the use of the coverage ratio will be
certainly favorable to the dummy approach.
In the light of the findings of this study, it is argued that in recent years,
the poor economic performance of the U.S. manufacturing firms is likely and
partially caused and mitigated by enhancing union organizing and bargaining
power of union, i In the absence of appropriate productivity increase, union
wage increases will likely continue reducing domestic and foreign
competitiveness of the U.S. manufacturing firms. Therefore, some
modifications in the U.S. labor law may be needed for reallocating bargaining
power between firm and union managements.
Even though this study is conclusive for all measures of profitability and
unionization except the alternative measures of unionization, and for all
estimation techniques, there are some limitations to this research. The first
one is the difficulty in obtaining the measure of firm union coverage ratio. Since
the firm union measure is not publicly accessible, this study was forced to
construct the firm union coverage ratio by aggregating the number of covered
134
workers across all of a firm's listed contracts from the publications of the
Bureau of Labor Statistics and by dividing that number by the total workers in a
firm from the Compustat tape. Matching two data sources in this way forced us
to choose large firms nonrandomly. So, the results cannot be easily
generalized to the whole U.S. manufacturing sector. Second, the endogeneity
of unionization with respect to profitability is examined subsequently in a less
satisfactory manner; that is, we are forced to use predicted value of
unionization as a proxy for the firm unionization because it is extremely difficult
to find the proxies which influence the union coverage ratio but not profitability.
For these reasons, it is recommended that further research on the
union-profit effect be done constructing firm union measures "over time" and
applying the techniques which could capture the possible effect of the
omitted/unobserved determinants of profitability which are related firm
unionization.
In addition to the estimation of the union effect on profitability under
fixed-effect technique, the major contribution of this study to the literature is the
fact that the findings with the alternative measures of unionization, profitability
and the alternative estimation techniques provide strong support for the
hypothesis of a negative union effect on profitability and also provides evidence
on how the size of the negative effect of unionism differs depending on the
measures of unionization, profitability and estimation techniques.
REFERENCES
Abowd, John M., "The Effect of Wage Bargains on the Stock Market Value of the Firm," The American Economic Review. Vol. 79, (1989): 774-799.
Acs, Zoltan J. and David B. Audretsch, "Innovation in Large and Small Firms: An Empirical Analysis," The American Economic Review, Vol. 78, (1988): 678-690.
Addison, John. T and Barry T. Hirsch, "Union Effects on Productivity, Profits, and Growth: Has the Long Run Arrived?," Journal of Labor Economics, Vol.7, (1989): 72-105.
Amato, Louis and Ronald Wilder, "Firm and Industry Effects in Industrial Economics." Southern Economic Journal, Vol. 57, (1990): 93-105.
Bally, Martin N., "Research and Development Costs and Returns: The U.S. Pharmaceutical Industry." Journal of Political Economy. Vol. 80, (1972): 70-85.
Baumol, W. J., Business Behavior. Value, and Growth. MacMlllan Book Company, New York, (1959).
Becker, Brian E. and Craig A. Olson, "Unionization and Shareholder interests," Industrial and Labor Relations Review. Vol. 42, (1989): 246-261.
, "Unions and Firms Profits." Industrial Relations. Vol. 31 (1992): 395-415.
Bond, Ronald S. and Warren Greenberg, "Industry Structure, Market Rivalry, and Public Policy: A Comment," Journal of Law and Economics, Vol. 19, (1976): 201-204.
Branch, Ben, "Research and Development Activity and Profitability: A Distributed Lag Analysis," Journal of Political Economy. Vol. 82, (1974): 999-1011.
Bronars, Stephen G. and Donald R. Deere, "Union Representation Elections and Firms Prnfitflhility." Industrial Relations, Vol. 29, (1990): 15-37.
Brown, Charfes and James Medoff, "Trade Unions in the Production Process," Journal of Political Economy, Vol. 86, (1978): 355-378.
135
136
Brozen, Yale, "Entry Barriers: Advertising and Product Differentiation," in Harvey J. Goldschmid, H. Michael Mann and J. Fred Weston (eds). Industrial Concentration: The New Lftaming, Little, Brown and Company, Boston, (1974): 115-133.
Brunello, Giorgia, "The Effect of Unions on Firm Performance in Japanese Manufacturing," Industrial and Labor Relations Review. Vol. 45, (1992): 471-486.
Carter, John R., "Collusion, Efficiency, and Antitrust." Journal of Law and Economics. Vol. 21, (1978): 435-444.
Clark, Kim B., "The Impact of Unionization on Productivity: A Case Study." Industrial and Labor Relations Review. Vol. 33 (1980): 451-469.
, "Unionization and Firm Performance: The Impact on Profits, Growth, and Productivity." The American Economic Review. Vol. 74, (1984): 893-919.
Collins, Norman R. and Lee E. Preston, Concentration and Price-Cost Margins in Manufacturing Industries. University of California Press, (1968), Berkeley.
Comanor, William S. and Thomas A. Wilson, "The Effect of Advertising on Competition: A Survey," Journal of Economic Literature. Vol. 17, (1979): 453-457.
Connolly, Robert A., Barry T. Hirsch and Mark Hirschey, "Union Rent Seeking, Intangible Capital, and Market Value of The Firm," The Review of Economics and Statistics. Vol. 68, (1986): 567-577.
Cowling, Keith and Michael Waterson, "Price-Cost Margins and Market Structure." Economica. Vol. 43, (1976): 267-274.
Curme, Michael A., Barry T. Hirsch, and David A. Macpherson, "Union Membership and Contract Coverage in the United States, 1983-1988," Industrial and Labor Relations Review. Vol. 44, (1990): 5-11.
Demsetz, Harold, "Industry Structure, Market Rivalry, and Public Policy," The Journal of Law & Economics, Vol. 16, (1973): 1-9.
, "More on Collusion and Advertising: A Reply," Journal of Law and Economics. Vol. 19, (1976): 204-08.
137
Domowitz, Ian, R. Glenn Hubbard and Bruce C. Petersen, "The Intertemporal Stability of The Concentration-Margins Relationship," The Journal of Industrial Economics. Vol. 35, (1986): 13-34.
"Market Structure and Cyclical Fluctuations in U.S. Manufacturing," The Review of Economics and Statistics. Vol. 70, (1988): 55-66.
Fisher, Franklin M., "On the Misuse of the Profits-Sales Ratio to Infer Monopoly Power." The Rand Journal of Economics. Vol.18, (1987): 384-396.
Freeman, R.B., "Unionism, Price-Cost Margins, and the Return to Capital," NBER Working Paper. No. 1164, (1983).
and James L. Medoff, "New Estimates of Private Sector Unionism in the United States," Industrial and Labor Relation Review. Vol. 32, (1979): 143-174.
Gale, Bradley T., "Market Share and Rate of Return," The Review of Economics and Statistics. Vol. 54, (1972): 412-423.
Gisser, Michal, "Advertising, Concentration and Profitability in Manufacturing," Economic Inguiry. Vol. 29, (1991): 148-165.
Goldfeld, S. M., and R. E. Quandt, "Some Tests for Homoscedasticity," Journal of the American Statistical Association, Vol. 60, (1965): 539-547.
Grabowski, Henry G and Dennis C. Mueller, "Industrial Research and Development Intangible Capital Stocks, and Firm Profit Rates," Bell Journal of Economics. Vol. 9, (1978): 328-343.
Gujarati, Damodar N., Basic Econometrics, McGraw-Hill Book Company, New York, (1988).
Harris, Frederick H. Deb., "Market Structure and Price-Cost Performance Under Endogenous Profit Risk," The Journal of Industrial Economics. Vol. 35, (1986): 35-59.
Harvey, A. C , and G. D. A. Phillips, "A Comparison of the Power of Some Tests for Heteroscedasticity in the General Linear Model," Journal of Econometrics. Vol. 2 (1974): 307-316.
Hausman, J. A., "Specification Test in Econometrics," Econometrica. Vol. 46, (1978): 1251-1270.
138
Hendricks, W., "The Effect of Regulation on Collective Bargaining in Electric Utilities," The Bell Journal of Economics. Vol.16, (1975): 451-465.
, Feuille, P. and C. Szerszen, "Regulation, Deregulation, and Collective Bargaining in Airlines," Industrial and Labor Relations Review, Vol.34, (1980): 67-81.
Hirsch, Barry T., "Market Structure, Union Rent Seeking, and Firm Profitability," Economic Letters, vol. 32, (1990): 75-79.
, "Union Coverage and Profitability Among U.S. Firms," The Review of Economics and Statistics, Vol. 73, (1991a): 69-77.
, Labor Unions and The Economic Performance of Firms, W. E. Upjohn Institute for Employment Research, (1991b), Michigan.
, "Firm Investment Behavior and Collective Bargaining Strategy," Industrial Relations. Vol. 31, (1992): 95-121.
and John T. Addison, 'The Economic of Analysis of Unions." Allen & Unwin Inc., (1986), Winchester.
and Robert A. Connolly, "Do Unions Capture Monopoly Profits?," Industrial and Labor Relations Review. Vol. 41, (1987): 118-136.
and Albert N. Link, "Labor Union Effects On Innovative Activity," Journal of Labor Research. Vol. 8, (1987): 323-332.
Hirschey, Mark, "Market Structure and Market Value," Journal of Business. Vol. 58, (1985): 89-98.
Hsiao, Cheng, Analysis of Panel Data. Econometric Society Monographs, Cambridge University Press, Cambridge, (1991).
Johnston, J., Econometric Methods. McGraw-Hill Book Company, New York, (1984).
Karier, Thomas, "Unions and Monopoly Profits," The Review of Economics and Statistics. Vol. 67, (1985): 34-42.
, "New Evidence on the Effect of Unions and Imports on Monopoly Power." Journal of Post Kevnesian Economics. Vol. 10, (1988): 414-427.
Kokkelenberg, Edward C. and Donna R. Sockell, "Union Membership in the y r i t ,o ' '"^"^^'''31 and Labor Relations Review. Vol. 38, (1985): 497-532.
Leibenstein, Harvey, "Allocative Efficiency Versus X-Efficiency," American Economic Reviftw Vol. 56, (1966): 392-415.
Liebowltz, S. J., "What do Census Price-Cost Margins Measure?," Journal of Law & Economir.c;, Vol. 25, (1982): 231-246.
Machin, S.J., "Unions and the Capture of Economic Rents: An Investigation Using British Firm Level Data," International Journal of Industrial Organization. Vol. 9, (1991): 261-274.
139
and M. B. Steward, "Unions and The Financial performance of British Private Sector Establishments," Journal of Applied Econometrics. Vol. 5, (1990): 327-350.
Mann, H. Michael, "Advertising, Concentration, and Profitability: The State of Knowledge and Directions for Public Policy," in Harvey J. Goldscmid, H. Michael Mann and J. Fred Weston (eds). Industrial Concentration: The New Learning. Little, Brown and Company, Boston, (1974): 137-155.
Mansfield, Edwin, "Industrial Research and Technological Innovation." W. W. North & Company, Inc., New York, (1968).
Martin, Stephen,"Advertising, Concentration, and Profitability: the Simultaneity Problem.", The Bell Journal of Economics. Vol. 10, (1979): 639-647.
Moore, Thomas G., "The Beneficiaries of Trucking Regulation," The Journal of Law and Economics. Vol. 21, (1978): 327-343.
Mueller, D. C. and J. E. Tilton, "Research and Development Costs as a Barrier to Entry." Canadian Journal of Economics. Vol. 2, (1969): 570-579.
Neriove, Marc and Kenneth J. Arrow, "Optimal Advertising Policy Under Dynamic Conditions." Economica. Vol. 29, (1962): 129-142.
Neter, John, William Wasserman and Michael H. Kutner, Applied Linear Regression Models." Richard D. InA in, Inc., Boston, (1989).
Palda, Kristian S., The Measurement of Cumulative Advertising Effects. Prentice-Hall, Englewood Cliffs, N.J, (1964).
140
Peles, Yoram, "Rates of Amortization of Advertising Expenditures," Journal of Political Economy. Vol. 79, (1971): 1032-1057.
Phillips, Almarin, "A Critique of Empirical Studies of Relations Between Market Structure and Profitability," The Journal of Industrial Economics. Vol. 24, (1976): 241-249.
"Patents, Potential Competition, and Technical Progress," The American Economic Review. Vol. 56, (1966): 301-310.
Porter, Michael E., "The Structure within Industries and Companies Performance." The Review of Economics and Statistics. Vol. 61, (1979): 214-227.
Ravenscraft, David J., "Structure-Profit Relationships at The Line of Business and Industry Level," The Review of Economics and Statistics. Vol. 65, (1983): 22-31.
Rose, Nancy L., "The Incidence of Regulatory Rents in the Motor Carrier Industry." Rand Journal of Economics. Vol. 16, (1985a): 299-318.
, "Union Wage Gains Under Regulation: Evidence from the Trucking Industry." MIT Sloan School of Management Working Paper. No. 1683-1685, (1985b).
, "Unionization and Regulation: The Division of Rents in the Trucking Industry," MIT Sloan School of Management Working Paper: No. 1684-85, (1985c).
Ruback, Richard S. and Martin B. Zimmerman, "Unionization and Profitability: Evidence from the Capital Market," Journal of Political Economy. Vol. 92, (1984): 1134-1157.
Salinger, Michael A., "Tobin's q. Unionization, and the Concentration-Profits Relationship." Rand Journal of Economics. Vol. 15, (1984): 159-170.
Schmalensee, Richard, "Do Markets Differ Much?," The American Economic Review. Vol. 75, (1985): 341-351.
, The Economics of Advertising. North-Holland Publishing Company, Amsterdam, 1972.
141
Scott, John T. and George Pascoe, "Beyond Firm and Industry Effects on Profitability in Imperfect Markets," The Review of Economics and Statistics. Vol. 68:2, (1986): 284-292.
Shepherd, William G., "Tobin's q and the Structure-Performance Relationship: Comment," The American Economic Review. Vol. 76, (1986): 1205-1210.
Smirlock, Michael, Thomas Gilligan and William Marshall, " Tobin's q and the Structure-Performance Relationship," The American Economic Review. Vol. 74, (1984): 1051-1059.
Strickland, Allyn D. and Leonard W. Weiss, "Advertising, Concentration, and Price-Cost Margins.", Journal of Political Economy. Vol. 84, (1976): 1109-1121.
Swamy, P., Efficient Inference in a Random Coefficients regression Model," Econometrica. Vol. 38, (1970): 311-323.
Un, Noel D., "A Re-Examination of the Relationship between Industry Structure and Economic Performance." Applied Economics. Vol. 20, (1988): 1383-1400.
U.S. Department of Commerce, Bureau of the Census, Census of Manufactures. Washington, (1987).
, Bureau of the Census, Annual Survey of Manufactures. Washington, annually, (1986-1988).
U.S. Department of Labor, "Statistics on Compensation Changes," Current Wage Developments. Washington, monthly, (1986-1988).
Voos, Paula B. and Lawrence R. Mishel, "The Union Impact on Profits: Evidence from Industry Price-Cost Margin Data," Journal of Labor Economics. Vol.4, (1986a) 105-133.
^ "The Union Impact on Profits in the Supermarket Industry," The Review of Economics and Statistics. Vol. 68, (1986b): 513-517.
Weiss, Leonard W., "The Concentration-Profits Relationship and Antitrust," in Harvey J. Goldschmid, H. Michael Mann and J. Fred Weston (eds). Industrial Concentration: The New Learning, Little, Brown and Company, Boston, (1974):184-233.
142
White, Halbert, "A Heteroscedasticity-Consistent Covariance Matrix Estimator and A Direct Test for Heteroscedasticity," Econometrica, Vol. 48, (1980): 817-829.
APPENDIX
TABLE A.I Names and Industry Codes of the Firms
Name of the Firms 4-Digit Industry Codes
Philip Morris Cos. Inc. 2111 Quaker Oats Co. 2000 IBP Inc. 2011 Heinz (H.J) Co. 2030 Hershey Foods Corp. 2060 Anheuser-Busch Cos. Inc. 2082 Fieldcrest Cannon 2211 Georgia-Pacific Corp. 2430 Mead Corp. 2600 Weyerhaeuser Co. 2421 Boise Cascade Corp. 2621 Champion International Corp. 2621 Intl. Paper Co. 2631 Kimberly-Clark Corp. 2621 Potlatch Corp. 2621 Scott Paper Co. 2621 Union Camp. Corp. 2621 Westvaro Corp. 2621 Federal Paper Board Co. 2631 Longview Fibre Co. 2670 Minnesota Mining & MFG Co. 2670 New York Times Co. 2711 American Cyanamid Co. 2800 Dow Chemical 2800 FMC Corp. 2800 Olin Corp 2800 Goodrich (B.F) Co. 282 Hercules Inc. "^JT Union Carbide Corp. 2821 Johnson & Johnson 2834 Colgate-Palmolive Co. 2840 Procter & Gamble Co. 2840 PPG Industries Inc. 2851 Ashland Oil Inc. 291^ Atlantic Richfield Co. 291 ^ Chevron Corp. 291
143
144
TABLE A.I (Continued) Names and Industry Codes of the Firms
Name of the Firms 4-Digit Industry Codes
Exxon Corp. 2911 Mobil Corp. 2911 Phillips Petroleum Co. 2911 Sun Co. Inc. 2911 Texaco Inc. 2911 USX Corp. 2911 Goodyear Tire & Rubber Co. 3011 Armstrong World Inds. Inc. 3290 Lone Star Industries 3241 National-Standard Co. 3310 Worthington Industries 3310 Armco Inc. 3312 Bethlehem Steel Corp. 3312 Cyclops Industries Inc. 3310 Inland Steel Industries Inc. 3312 LTV Corp. 3312 Asarco Inc. 3330 Aluminum Co. of America 3334 Reynolds Metals Co. 3334 Stanley Works 3420 Briggs & Stratton 3510 Brunswick Corp. 3510 Cummins Engine 3510 Outboard Marine Corp. 3510 Deere & Co. ^^23 Ingersoll-Rand Co. 3560 Dresser Industries Inc. 3530 Timken Co. ^^^2 Interlake Corp. 3569 Tecumseh Products Co. 3585 Westinghouse Electric Corp. 3812 General Electric Co. 3600 Maytag Corp. 3630 Whirlpool Corp. ;^^;J^ Chrysler Corp. ^2? \ General Motors Corp. "^^l^ General Motors-Class H 3812 Navistar International 3711
145
TABLE A.I (Continued) Names and Industry Codes of the Firms
Name of the Firms 4-Digit Industry Codes
Dana Corp. 3714 Smith (A.O) Corp. 3714 TRW Inc. 3760 Textron Inc. 3720 Boeing Co. 3721 General Dynamics Corp. 3721 McDonnell Dougles Corp. 3721 Teledyne Inc. 3724 United Technologies Corp. 3724 Rohr Industries 3728 Todd Shipyards Corp. 3730 Huffy Corp. 3751 Lockheed Corp. 3760 Martin Marietta Corp. 3760 Rockwell Intl. Corp. 3760 Litton Industries Inc. 3812 Loral Corp. 3812 Honeywell Inc. 3822 Fischer & Porter Co. 3823 Xerox Corp. 3861 Willamette Industries 2621 Acme Steel Co-Del 3312 Laclede Steel Co. 3312 Wyman-Gordon Co. 3460 Trico Products Corp. 3714 Emhart Corp. 3452 Cameron Iron Works 3533 Champion Spark Plug 3690 Owens Corning Fibrglas 3290 Unisys Corp. 3570
Note: This four-digit Standard Industry Classification (SIC) Codes designates the product or service contributing the largest percentage of company sales (Compustat, Section 10, pp. 3).