PLACE IN RETURN Box to remove this checkout from your record.
To AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c'JCIRCIDaIeDInpss-p. 1 5

 

THE EFFECTS O?

in par

 

SChoc

 

THE EFFECTS OF TRADE UNIONS ON TOTAL COMPENSATION
IN KOREA FOR 1987-1999

By

JEONGHYUN LEE

A DISSERTATION
Submitted to
Michigan State University

in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
School of Labor and Industrial Relations

2003

THE EFFECTS

71““ 1‘) ‘thv\ ‘1"
.uicm’p‘i \‘ M ‘
4' 1‘.-

L'iz;.‘;::; “3:25 and . t
. . I v ‘. .".,i
71222132 F‘IIILJII L...,

i 5

~-.. ‘ '. ‘ , o I;
in»: AM! Links.
“.‘tv - ,vov I ‘r \
in." LOJECC...1M U ..

32:13: \lhcthcr Kt -'

' ";~‘\ , ‘ ' a
fail; arm mu~l 5t

33:: II: mien CECE:
56:33: KIWI.

r'?‘1’"~ flu ,Osc. this \
in Sam ill) \\ I;

 

ABSTRACT

THE EFFECTS OF TRADE UNIONS ON TOTAL COMPENSATION
IN KOREA FOR 1987-1999

By

JEONGHYUN LEE

The purpose of this study is to examine the effects of trade unions on total compensation
including wages and benefits in Korea for the period 1987-1999. In 1987, free collective
bargaining between employers and unions began, following political democratization and
massive labor unrest. The experiences of Western countries indicate that trade unions are
likely to undertake converging roles regarding total compensation. Therefore, in order to
determine whether Korean trade unions undertook similar roles for the period, four
specific areas must be considered: the union wage effect, the union wage-equalizing
effect, the union effect on human capital and other wage determinants, and the union
benefit effect.

For the purpose, this study is based on three kinds of data, individual-level data from the
Basic Survey on Wage Structure for the union wage effect and union effect on human
capital and other wage determinants, establishment-level data, constructed from the
BSWS, for the union wage-equalizing effect, and establishment-level data from
Enterprise Labor Cost Survey in Korea. The Ordinary Least Square, Quantile Regression,
and Two-Stage Least Squares estimation method are used to analyze the data.

Regarding the union wage effect, results show that positive wage effects are found in

Korea for 1987-1999 and the size of the union wage effect is around 5%, implying

. a ‘
, r I . ‘
'J ”we‘l‘ ’OLK to A” ——
h...“ .-

.'. I” , 1'; oh.\\g
.4: 5."ka frlu .l': “IL

. ' ~r ‘ ,. n
.- .. “‘ .y- is
(Wilma WWW"

.l. A

“I; ‘Lr..:.": Iliumix“

h
1 _ ,2.
mi mixers 1r.

Ip'w- ! m
.

‘ v- v
,.\r.t. if .X‘ ' ‘
Ar‘AH. ugh“ .\ \kiln\\4 \K'

. ‘ :1: Jill“. “KC-C;
5 .

-_...; M s; , t
«. .rst‘m;.:\ “:1.“ "

I

1.4;: 0i ll‘ii’iil‘ll‘. . ‘7

gm HI “Jim I

"7"" are l ‘
uni“ “HPJHC‘Y‘ trlx‘.‘" ‘
I . - ”I .

l
I

11$." .’ ,- . .
"I l\ I) )h\\'¥\r‘ 2L ‘0 a;
‘ bot“. .
2'1 I
‘ru

; T" V...
H»; m
i q“\ 1,) .1 ~
L\C [II l‘""‘ ,5‘
w; .:

 

Ii,
:7‘"!
“"3. Chin
dr‘dl. u ‘— I
‘ Al‘n m
39»
a Fan jII,
- .‘u

 

relatively low to 10-25% in the United States. While the union wage effect is larger for
the lower end in the wage distribution for 1987-1995, in addition, the union wage effect
for less-paid workers became less, compared to better-paid workers, from 1996-1999.
The change indicates that Korean trade unions move from the organizations for less-
skilled workers in the late-19805 and early-19905 to the organizations representing
mainly more-skilled workers, who receive higher wages, in the late-19903. With respect
to the union wage-equalizing effect, the results indicate that unions reduced the degree of
wage inequality within establishments by 7%-25% in the 19903, when measured by
variance of logarithm of wages and by 7%-14%, when measured by standard deviation of
logarithm of wages. The result implies that Korean unions have succeeded in preventing
their employers from introducing pay-for-performance wage systems in the 19903.
Also, it is shown that the union reduces the return for education and tenure in all years
and gender wage differentials in most years studied. Regarding union benefit effect,
results imply that unions increase the level of voluntary benefits, compared to non-
organized establishments. In addition, the union organizing rate within establishments
and the individual union’s affiliation to the national union confederation are shown to

affect the level of benefits and nature of benefits, provided by employers.

(Key Words: Total compensation, union wage effect, wage distribution, wage inequality,
human capital, union benefit effect, Korean industrial relations, quantile regression, two-

stage least squares, and ordinary least squares)

Copyright by
JEONGHYUNLEE

2003

To H, mm pm“ raw...

 

c - Cl 'i‘ .

1m .1: lo .lunl R-
It} committee. for .*
4:210“ 10 llnle‘

u

3:: committee muff .
p L. t .

1:11 1. Basis mr if".
a; mi 9‘ '

«1“..th driclwn.

'\
IA

1
“l.

w: .11: 10 than. a"
0* l

w

“Marti \lllcrlcl er I
gilt?“

will.

 

Acknowledgements

I would like to thank Richard N. Block, my advisor for the past six years and chairperson
in my committee, for his endless patience and encouragement. Without him, I could not
imagine how to finish my study and this dissertation. Also I am deeply grateful to all of
other committee members: to Karen Roberts, to Peter Berg, to Richard C. Hill, and to
Robert F. Banks for the careful reading and invaluable critiques. All errors and rooms
remained to be developed should be mine.

I would like to thank all in the bibliography. In addition, my deep thanks go to all I could

remember whenever I think of industrial relations in Korea and while writing this
dissertation.

vi

 

[its oITabIes .........
Liv of figures .........
Keys to .lbhrei iations
Thapter One: Introdu

Chapter Two: LIIt'rilIl

ll. \Jttrcwzdt
I. l'ntunllagcl
1.3. l'nionlt‘tctt
3.1. l'nionltltttt
.5. l'nionllcncf'

. J

ChapterThrre: Data

-1. Inditide-l c
M -
ESldl‘llxlimt‘i‘.

«a

.4. Establtxhrticr‘.

Tthter Four. l'ninn
.l. Intrtdttttiun.
42 liodclandlit.

........

4.1. ~

fltrpterFiie:
5.1
‘3. llodclandl
Data ......
:-- Results .....
Concl

I‘nion I
- nerdttt‘lton

U$ltin _

 

llzlpter Sewn: l'nin
1i. lntrOdUL‘tiitn'l
I: Itklt‘lankll

J

 

 

Table of Contents

Lists of Tables ...................................................................................... ix
List of Figures ...................................................................................... xi
Keys to Abbreviations ........................................................................... xii
Chapter One: Introduction ....................................................................... 1
Chapter Two: Literature Review
1.1. Natures and Goals of Trade Unions ................................................... 5
1.2. Union Wage Premium in the Wage Distribution .................................... 8
1.3. Union Effect on Wage Dispersion .................................................... 15
2.4. Union Effect on Wage Determinants ................................................. 25
2.5. Union Benefit Effect ................................................................... 29
Chapter Three: Data
2.1. Individual-Level Data from BSWS .................................................. 52
2.2. Establishment-Level Data from BSWS .............................................. 54
2.3. Establishment-Level Data from ELCS .............................................. 55
Chapter Four: Union Wage Effect in the Wage Distribution
4.1 . Introduction .............................................................................. 5 7
4.2. Model and Data ......................................................................... 57
4.3. Results .................... 66
4.4. Conclusion ............................................................................... 99
Chapter Five: Union Effect on Wage Dispersion
5.1. Introduction ............................................................................ 103
5.2. Model and Estimation Method ...................................................... 104
5.3. Data ..................................................................................... 106
5.4. Results .................................................................................. 112
5.5. Conclusion.................... ......................................................... 131
Chapter Seven: Union Effect on Four Wage Determinants
1.1 . Introduction ............................................................................ 1 35
1.2. Model and Data ........................................................................ 136
l .3. Results .................................................................................. 137
6.4. Conclusion ............................................................................. 1 76
Chapter Eight: Union Benefit Effect
7.1. Introduction ............................................................................ 1 77
7.2. Model and Data ........................................................................ 178

vii

7.3. Results ........
7.4. Conclusion _.

ThapterNine: Conelu~
8.1. Summsr} oil

81. Prospects for ‘

Appendix: Telephone I
Regarding l

Bibliography ...........

 

 

 

7.3. Results .................................................................................. 190

7.4. Conclusion ............................................................................. 206
Chapter Nine: Conclusion
8.1. Summary of Four Empirical Studies ............................................... 210
8.2. Prospects for Near Future Korean Industrial Relations .......................... 219
Appendix: Telephone Interview Protocol
Regarding the Union Benefit Effect ........................................... 229
Bibliography ..................................................................................... 23 1

viii

 

T3331. Later L051: .

1351512: 51.1183 OT X."
1180. 1 WV

7313‘ ‘ Tile 1‘1" '
1.1.1533. 1‘ “\“wL‘l

pt‘T CITY?10};‘1

TableLl. Dexnptltex
unionized s:

Table 13. Size till 111
for a 8.1m; .

231313. Size 011 ‘n:
for the ma;

Abel-4. Size 01‘ l 'n:
for the ten: I

.13. Ways to 1):;

across firm:

 

   
 
 
   
  

liltlai - .
“val Dt‘k‘fll‘llu‘
:md 19W In.

Femle \l'n:

Tn“: .
emf .
'~‘- Trends in ::

1“Cell 1].“

inlet _ .
- Ebllmc-llL'b I
“39% in
Tn

List of Tables

Table 2-1. Labor Costs in All Industries, 1986-1999 .......................................... 33

Table 2-2. Shares of Sub-categories in Statutory and Voluntary Benefits,
1986-1999 .............................................................................. 36

Table 2-3. The shares of statutory and voluntary benefits in total labor costs
per employee in many countries .................................................... 44

Table 4-1. Descriptive statistics for all variables in all, non-unionized, and
unionized sectors for 1987-1999 (unweighted) .................................. 64

Table 4-2. Size of Union Coefficient in OLS and Quantile Regressions
for a Samples for 1987-1999 ......................................................... 67

Table 4-3. Size of Union Coefficient in OLS and Quantile Regressions
for the male sample for 1987-1999 ................................................. 79

Table 4-4. Size of Union Coefficient in OLS and Quantile Regressions
for the female sample for 1987-1999 ............................................... 79

Table 4-5. Ways to Determine Wage Increases by Unions
across firm size in 1990 .............................................................. 93

Table 5-1. Descriptive statistics of data for 1987, 1990, 1994, 1997,
and 1999 between unionized and non-unionized sectors ....................... 109

Table 5-2. Standard Deviation of Log Wage for All, Male, and
Female Workers in All Industries for 1987-1999 ................................ 113

Table 5-3. Trends in the Three Wage Inequality Measures
between Union and Nonunion sectors, 1987-1999 .............................. 1 14

Table 5-4. Estimates of the Union Effects on the Variance of Log Hourly
Wages in 1987, 1990, 1994, and 1999 by OLS and ZSLS ..................... 117

Table 5-5. Results from first-stage in ZSLS in equation with the variance
of logarithm of wages as the dependent variable in 1987, 1990,
1994, 1997, and 1999 ............................................................... 123

ix

Tell: it. Estimztes e3~
Log lit-Lil}
1M. 1W".

Ti‘é: 5-7. The resets 1'
Sllrztldfcl elk":
in 1‘18“. 19'

731311. Deseriptixe:
81131111101115:

T132: 7-2. Trim 1’ :Tee‘
1995-] Wu 1..

1b" ”.1 ' . -~
1.: . .. “age-renter

tveightedi .

11115-1. Logistic m;.'
the prm lxlu
lei-1:

.8-1. Progressex i
lnt‘iustml l.

 

 

Table 5-6. Estimates of Union Effects on the Standard Deviation of
Log Hourly Wages Within Establishments in 1987, 1990,
1994, 1997, and 1999, estimated by OLS and 2SLS ........................... 126

Table 5-7. The results from first-stage in ZSLS in equation with

standard deviation of logarithm of wages as dependent variable
in 1987, 1990, 1994, 1997, and 1999 ............................................ 128

Table 7-1. Descriptive statistics for all variables in all, non-unionized,
and unionized sector for 1995-1999 ............................................... 189

Table 7-2. Union Effects on the Level of Voluntary Benefits for
1995-1999 for All Samples (weighted) ............................................ 192

Table 7-3. Wage-related equations for unionized samples for 1995-1999
(weighted) ............................................................................ 195

Table 7-4. Logistic multiple-regression for union effects on
the provision of various voluntary benefits for 1995-1999 .................... 201

Table 8-1. Progresses in Changing from Enterprise Unions to
Industrial Unions within the F KTU and KCTU ................................. 223

TL-Ilfi-l. l'nion \\ 4;;
Tiger: ‘ 3. l'niun 11;;
T1113 l'nion 11):;

lignel-I Compinwi:
including a

\‘EL-"idl‘lm li‘

:i-guie til. The coef‘tie
graduates. '
and union“;

E" g) i I‘

1.51. ta. Return rate-
estimutei E
and union.

tart-rt: R

~ ~ ‘ -‘ Llum TLITC
regression

10:19:11?

”Eleni/ed I

 

 

 

List of Figures

Figure 4-1. Union Wage Premiums for All Sample in 1987 and 1999 ...................... 69
Figure 4-2. Union Wage Premiums for All sample from 1988 to 1998 ..................... 71
Figure 4-3 Union Wage Premiums for men and women for 1987-1999 ................... 81

Figure 4-4 Comparison of union coefficients in between the equation
including occupational variables and the equation without the
variables for all in 1987, 1993 and 1999 .......................................... 90

Figure 6-1. The coefficients of university graduates and high school
graduates, estimated by the OLS, between the non-unionized
and unionized sectors for 1987-1999 ............................................. 138

Figure 6-2. Return rates for high-school graduates and university graduates,
estimated by quantile regressions, between the non-unionized
and unionized sectors for 1987-1999 ............................................. 143

Figure 6-3. Return rates for a year of experience, estimated by quantile
regressions, between the non-unionized and unionized sectors
for 1987-1999 ....................................................................... 151

Figure 6-4. Return rates for a year of tenure, estimated by quantile
regressions, between the non-unionized and unionized sectors
for 1987-1999 ...................................................................... 158

Figure 6-5. The coefficients for sex (female workers), estimated by
quantile regressions, between the non-unionized and
unionized sectors for 1987-1999 ................................................ 166

xi

ZSLS
AMG
BLS
BSWS
CDMA
CLMC
COLA
CPS
DLP
DSP
ECI
ELCS
EOI
EPI

F DI
FKI
FKTU
GM
GNP
GSM
HCI
ILO
IMF
IRRC
ISI
KCTU
KCTU
KEF
KLI
KSLI
KTUT
LAV
MAD
MOL
MWC
NCES
NCTU
NSO

Key to Abbreviations

2-Stage Least Square

American Military Government

Bureau of Labor Statistics

Basic Survey on Wage Structure

Code Division Multiple Access

Central Labor-Management Council
Cost-living Adjustment

Current Population Survey

Democratic Labor Party

Democratic Society Party

Employer Cost Index

Enterprise Labor Cost Survey
Export-Oriented Industrialization
Economic Planning Board

Foreign Direct Investment

Federation of Korean Industries
Federation of Korean Trade Unions
General Motors

Gross National Product

Global System for Mobile communication
Heavy and Chemical Industrialization
International Labor Organization
lntemational Monetary Fund

Industrial Relations Reform Committee
Import-Substitution Industrialization
Korean Congress of Trade Unions (1945-1947)
Korean Confederation of Trade Unions (1995 ~)
Korea Employers Federation

Korea Labor Institute

Korean Labor and Society Institute
Korean Trade Union for Teachers
Least-Absolute Value

Minimum Absolute Deviation

Ministry of Labor

Minimum Wage Committee

National Council for Economy and Society
National Congress of Trade Unions
National Statistical Office

xii

0:10
0:5
3381‘ 0
pi:
iii-LC D

1’1"
L71

r-\-
l
l»:

 

 

OECD Organization for Economic Co-operation and Development

OLS Ordinary Least Square

POSCO Pohang Steel and Iron Company

QR Quantile Regression

TFT-LCD Thin Film Transistor-Liquid Crystal Display
UAW United Auto Workers

UN United Nations

xiii

.5. . ' .
T...::n tea. in Iv
.. x .
LN; CUCTTSW Omar.

nicer. labor for l”?

I
n u»,

in of trade

31‘. 198‘ in Korea

it ’31-: Kn'

.2311 CCU?

l'
’1": .
sn‘.l :1; l“ 1.

a netted ,5 a

$333: 1. .
n1:10 athtex e r
L.

ll“: “ v
‘L3héh0Ul {he 1‘)“

 

 

CHAPTER ONE

INTRODUCTION

Fifteen years in Korea have passed since the political democratization and massive
labor offensive occurred in 1987, which can be compared to the wage offensive by
American labor for 1933-1937 and the British General Strike in 1926. In spite of the no-
short history of trade unions and free collective bargaining guaranteed by labor laws in
1953, 1987 in Korea was the first year of free union establishment and free collective
bargaining.

As the Korean economy had developed rapidly for the three decades prior to 1987,
Korea was viewed as a developing country with labor acquiescence (Koo, 2001). Thus, it
is the lack of organized labor and labor acquiescence that has enabled the Korean
economy to achieve rapid industrialization during the past a few decades.

Throughout the 19905, Korea, along with probably South Africa and Poland, remained
one of the most dynamic industrial relations systems in the world, showing frequent
conflicts between employers, trade unions including workers, and government especially
during the Korean financial crisis of the period of 1998-1999 (Koo, 1996; Chang, 1999).
Korea in the 19908 was described as ‘a nation of endemic strikes’ by the New York Times
(January 17, 1997).

Most likely, the emergence of a new landscape in Korean industrial relations was
affected by the political situation surrounding democratization in the late 19803. Some

workers built trade unions with political orientations,.but it was also possible to join trade

. - “Cl
Mao-t: TOT n00 Ck
”Tahiti b\ the ”L .
Clam-5‘ '

1 ‘fi
we: the “7?" “
M.

p.131: increase 01 “ “‘
‘ra'ecede’fiit‘tl 53.". CS '
‘+ 5

:23: unions 51::

e 1‘“

r)

r to union men:

The: a century ago

116E111”. 140115 355003110 :1

11211110113 01 their en:

fiance wages and em

h Luring

and Workers

 

 

unions for non-economic reasons. Furthermore, employers and government were
challenged by the new environment in industrial relations during that time period.

Despite the importance of political, cultural, ideological or personal approaches, the
economic approach is the most important factor in explaining how Korean trade unions
have survived to date. For example, 50.3% of all demands raised by workers were related
to the increase of wages and various allowances in 1987, when massive strikes and an
unprecedented series of union establishments occurred (Chang, 1999). Have the Korean
trade unions since 1987 performed the economic roles as expected and as found in other
developed countries? Moreover, are the economic roles of the unions enough to attract
workers to union membership?

Over a century ago, Sidney and Beatrice Webb (1894) defined a trade union as “a
cOntinuous association of wage-earners for the purpose of maintaining or improving the
Conditions of their employment (p.1).” The purpose of trade unions is to protect and
adVamce wages and employment for their union membership in general, in case, a threat
to unions and workers comes from market competition (Commons, 1909). However,
fOllowing these goals of trade unions, have Korean trade unions satisfied their union
membership by increasing wages and improving working conditions? If not, we can
expect that the lack of enhancing economic utilities of union membership will result in a
decrease of union membership in the near future and the dynamics of Korean trade
mlions cannot be repeated in the 21St century any longer.

This study focuses on the importance of wages and benefits. The unions’ goals are to
Pufsue such objections as wages, benefits, safer workplaces, and more freedom in union

actiVities, sympathetic and swift grievance handling for workers, community activities,

ndpalitzcalaetnities. ‘
in he mend-tile u‘
ensued reguding \\.;
conned to other tirt‘

. .2.
etzntges are den: n...

”W 11th an :nterv.
teatime “ages. The
inner. for then r: err,”
113338 implied 11‘. the

lT'l“ ‘ .‘ \ \P\
Dian.“ LhL pin“,

3

"‘9 'w A
365.1314!“

#830101er

lags tor Korean \\ o:-

m he :10) :‘mment Lit:

ill “3338 Thus the

r
1‘“ 5

TI; §
tamed workers. Ir.

1“
ugly} N

embers exnec‘.

Emmi- The up.

. :01] the “11101)

5,, .
microns“
, .
‘10” S‘vste
:2; ,
Vi 9p

 

and political activities. Among them, the issue of wages is the most visible and calculable
for the rank-and-file union members. They can feel immediately what their union has
achieved regarding wages during the current year, compared to the previous year or
compared to other firms. While other goals such as grievance handling and political
activities are demanded by the limited scope of union members, in addition, all union
members have an interest in wages and are involved in determining what the unions want
regarding wages. The wage issue is an appropriate measure to reflect how unions
fianction for their membership, what union members demand from their unions, and how
they are involved in the decision-making process of unions.
During the period of 1987-1999 was the first time that free collective bargaining
Was guaranteed to workers via the government and labor laws. Prior to 1987, the level of
Wages for Korean workers had been low because workers were not allowed to organize
and the government adhering to export-driven economic development restricted based on
low wages. Thus the opening of the freedom to establish unions and free collective
bargaining in 1987 meant the chance for unions to increase their wages for most
OI'ganized workers. In this historical setting, we may assume that many rank-and-file
union members expected the unions to demand an increase in workers’ wages as their
firSt priority. The expectation held by most union members may function as considerable
Pressure on the union itself and its leaders. In the context of Korea for 1987-1999, most
mlions focused on collective bargaining at the level of the firm, due to the mandated
enterprise union system. Even though over time political activities beyond the boundary
of enterprise have been recognized as important as economic activities within firms

thrOugh collective bargaining procedures by many observers and union leaders, political

:1" 1121: not bee
inns.

in though this 5'.
in: economic aspects 1
L15 cud admits that 1
geals regrding “353.
We: econnnne .

.a cemett pure ec

extent: deCtsmn-nt.»

‘ I

w. in: different

‘51

e1

e1 antlers. \\ ithir‘

The first emptncal :

tr: Q
1“ I11”.:

011 wage pTL‘T

1

e. (“‘5‘
. 1
W51 .

Litton of New

1‘»: ' -
“-‘Lfiil' .
- «in Is CORSltlL’T.
3:55
a.“

“La - .
. .1. tour impor‘..;“

 

activities have not been in the direct interest of union members and the goal of most
unions.

Even though this study focuses mainly on the aspects of wages, it does not assume
that economic aspects or goals are more important goals than other union goals. Rather,
this study admits that the other goals of the unions may be more relevant than economic
goals regarding wages. With respect to the Dunlop-Ross debate, this study does not
support the economic aspects of unions argued by Dunlop. Rather this study will argue
that seemingly pure economic issues like wages within unions may be not related to pure
economic decision-making within unions but may be related to political compromise
between the different worker groups such as the union leaders, the less-skilled, and more-
skilled workers, within the unions.

The first empirical part of this study examines the existence of union wage effects and

how union wage premiums are distributed to the various groups of workers in the
diStribution of wages. In the second empirical part, how trade unions affect wage
dispersion is considered. The third empirical part examines how trade unions in Korea
aff‘ect the four important wage determinants, education, tenure, experience, and gender in
the process of wage determination. The final empirical question in this study concerns
whether trade unions in Korea have affected the level of voluntary benefits, an
increasingly important part of total compensation in Korea. Also the roles of additional
union-related variables, union organizing rate and the nature of national union
Confederations, which each individual union is affiliated to, are considered, regarding the

1eye] and contents of voluntary benefits.

Waxes and Goals o‘

Tn: stain: non '. .
anions in histof

1.1: o. ‘nde um!»

 

3.025 85 OTL'lfllltlln
‘

I
1

3:11:33 inclu

r;-

lili \\

“-3qu
e"- ~l 1

I ~ ‘ I
. s 131: 111110' n

I
‘i".
Allaul \\

CHAPTER TWO

LITERATURE REVIEW

1. Natures and Goals of Trade Unions
Classical Explanations on the Nature and OrLgin of Unions

The starting point of this study is to consider the nature and origin of trade unions as
organizations in history. From the 19th century, many have explained the origin and
nature of trade unions. For example, Webb and Webb (1894 & 1987) defined trade
unions as organizations for wage earners to maintain and even enhance working
conditions including wages. In order to achieve the goal, trade unions are required to
control the labor market and to set standard working conditions.

American observers seem to go along with the thoughts of Webb and Webb with
regard to the goals of unions. For example, Perlman (1928) discems the consciousness of
job scarcity as the main force to create trade unions, while Commons points out the
expansion of the market as an origin of trade unions. Regarding the role of unions, the
thoughts of Perlman and Commons are not too different from Webb and Webb in that
they also emphasize the maintenance of job and enhancement of working conditions.

The Marxist perspective (Hyman, 1975; Goldfield, 1987) towards unions also admits
that trade unions, on the one hand, are, by nature, the economic organization for workers
to enhance their wages and working conditions. In other words, trade unions are viewed
by workers’ organization as improving contracts between employers and workers.

Mentioning that the origin of trade unions is capitalism in itself, however, Marxism

ms at the sane 1-

‘5an.
.

magnet: 01 tapas..-
.‘OIx-Itni Lula Oh_-.
L111 unions “AAAAL Laik‘
SCSI 3912312711 110 \x.
:tcttcnets tn the 10'
at : .. ,
owersersthln. ,n.
‘l T‘ g is. ,' u
Eli. ell; llii.m.llc 110‘!

animation in this st ..

71...‘
A." l‘i‘.l I. -- ‘ i
“I R035 Dirndl: ‘

1:11: mos: such L-

 

Lama; 3115118 . on i

:.« ~.
.‘
“in. '5'

....1c institutions

'13", r»
‘1‘ KIIOLml 0f “ 211'"

stohe close to 1»

.
r-'
Q4'

. ‘ ‘1 ,l
Tet actor tn-i-

.

n»

53’ thin
as “m0“s I}:

E3‘é‘na1~ ‘
lei-1

.3 1, and
JTl:

.iqp T

 

argues at the same time that the origin of trade unions should contribute to the
termination of capitalism which is opposite to the long-run interests of workers, and not
limit unions while they enhance working conditions for workers. Marxists believe that
under capitalism no way exists to achieve better and fairer contract between employers
and workers in the long run, except the termination of the relations between employers
and workers (Marx, 1947). However, the potential gap between the origin of trade unions
and the ultimate goal of unions, inherent in Marxism, exceeds the scope of empirical

examination in this study.

Dunlop-Ross Debate over the Nature of Union

While those such as Webb and Webb, Marx, Commons, Hoxie, and Perlman provide
classical answers on how trade unions are created and what the purposes of the unions
are, the debate between John T. Dunlop and Arthur M. Ross narrows down the specific
purposes of trade unions and the natures of unions. Dunlop (1944) describes unions as
economic institutions to maximize a wage bill for workers. The “wage bill” means the
total amount of wages combined with size of employment. The concept of the wage bill
seems to be close to the concept of profit to employers. Dunlop assumes trade unions are
as a single actor trying to achieve an economic goal. The contents of an economic goal
for unions narrows the broad working conditions including wages, benefits, employment,
industrial security, and so on, which are emphasized by Webb and Webb (1984 & 1987),
Perlman (1928), and Marxism noted above, to the wage bill, when it comes to Dunlop

(1944).

Opposzng the 3:53.“
11:15.11 regards trade ;.
.. goals Valued .t:
pineal goal of trade

is focuses on the \

titer. Lhe unions and '.

tie consent of union .-.

\..t
.....op muses. One

at tent-s L'TOtlpS in “

§

"7";‘(3 .4 r ‘
1....it ufDLlltlL‘Tl'. 00 .

1‘3“»; . I
...-.0llll15 Stub
Someone might tit:

he 01 trade unto

if“:
“>53 0n the union

 

 

Opposing the argument concerning the goal of unions presented by Dunlop, Ross
(1950) regards trade unions as the political organizations created to compromise the
various goals valued and pursued by various groups within unions. Ross does suggest the
political goal of trade unions’ alternative to the economic goal by Dunlop.l However,
Ross focuses on the various and even conflicting goals held by many internal groups
within the unions and the political process of decision-making within unions. Ross rejects
the concept of union as a single actor and the ultimate goal to maximize the wage bill,
Dunlop espouses. One of the strengths of Ross (1950)’s argument is to find the existence
of various groups in the unions and to assume that the goal of unions is not fixed, but

variable dependent on compromise among the interest groups.

Stance of This Study

Someone might think that this study is based on Dunlop’s thoughts on the economic
purpose of trade unions maximizing the wage bill for membership in that this study
focuses on the union effects on wages. However, this study does not argue that either
Dunlop or Ross is better regarding the issue of the purpose and nature of trade unions.

Even though the focus of union activities is in maximizing economic utilities for
their memberships, a union’s target is to move from individual bargaining with their
employers to government, when social security issues involving government emerge
more importantly than individual bargaining with employers. For example, trade unions

may place more emphasis on changing laws requiring increased premiums paid by the

 

' At this juncture, the meaning of ‘political’ is very different from usual usage such as ‘political unionism,’
which means unions’ activities including enactment of labor-related laws favorable to labor, political
protests, or any political activities beyond collective bargaining with employers. However, the meaning of
the political nature of trade union, argued by Ross, does not exclude political activities by unions towards
government and other social institutions.

n:1oters. Nth recite

t

he

“U

m of trade 1
M the role ot’c
11202361111165.

The role of labor is.

t: fztorahie to take: I".

1e: proto're'd hi the '.

{21111163 aimed to rti.

mid b} the trai- -

\

51’3“.“de in than-gt?

1’31.le _ L
: “--15“0Tl{‘e"§ C1"
‘~

1. examining the '

 

 

)7
a:
(1'3
(t!
O.
._. .
. 5
r .fi
'—

 

employers, with regard to workers’ compensation or unemployment benefits. The larger
the priority of trade unions over those legislations are, the greater the possibility to
misjudge the role of collective bargaining is, when evaluating the overall outcomes of
union activities.

The role of labor laws has been great in Korea. Union activities to amend labor laws to
be favorable to labor have been an annual agenda. Also the general strikes in early 1997
were provoked by the unilateral amendment of labor laws by the ruling New Korea Party
at that time. Despite the importance of labor laws, it is not relevant to focus on the union
activities aimed to make labor laws favorable to labor in order to evaluate outcomes
achieved by the trade unions of 1987-1999, because no case is presented that the unions
succeeded in changing labor laws including increased premiums paid by the employers
regarding workers’ compensation, unemployment benefits, health insurance, or pensions.

In examining the effects of trade unions on economic outcomes from the period of
1987-1999, this study focuses on the four major issues: the relative distribution of union
wage effects among worker groups, the union effects on wage dispersion, the union

effects on how to determine wages for workers, and the union effects on benefits.

2. The Union Wage Premium in the Wage Distribution

Union Wage Differential
Union wage differentials have been and continue to be the topic of much concern by
many including employers, unions, workers, and even policymakers, and researchers, in

that a large union wage differential may distort the labor and product markets. In

 

axon. the Size 01 u ..
lain-or tester For ll‘.‘.S

:e.'.:;‘~.tert the most bj.

 

 

I“. _ ‘ .

page: are the no
who dfr§4\' ‘n-~.v
tLLu y .-.s th>\c\
fig": ‘ . .
.L...-...'.e the LT“..\‘!‘-..\

hitfl.lghhl.()n fl

condensttt and co"..-

 

Tfif when “use ;

y\l.
“we.

“T“ We tine:

xfiun: .
Inga DECI‘OY k."‘
Ragst
t» le\el “hen nU

YT713:- 'a _
: ntfl ”llhe 1.

Sing - at
JET t0 . ‘
1 afhj Tit)?

I
v-I -,
\L f

"433:1
...on sector

3:. t“ .
*csnti the CM:

 

addition, the size of wage differentials is an indicator, which reveals union power in the
labor market. For this reason, the area dealing with union wage differentials has been
highlighted the most by labor economists.

Wages are the most important outcome of collective bargaining. On the one hand,
union density causes wages and an excessively large union wage premium may
undermine the unions’ bargaining power by diminishing employment (Linneman and
Wachter, 1986). On the other hand, large union wage premiums are seen as boosting
union density and collective bargaining.

The union wage differential2 is defined as “the percentage of wage advantage of
.worker represented by a union over a worker with comparable skills and nonpecuniary
job attributes but not represented by a union” (Johnson, 1984; 3). Even though the
definition appears easy to understand, it contains two difficult issues. First, to measure
union wage differentials do not exclude the direct and indirect impact of unions on the
nonunion sector. Union wage differentials are correct only if we know the predominant
wage level when no union is present in the labor market. Because we do not know the
wage level in the labor market when no union exists, the wage difference between the
union sector and nonunion sector does not exclude the indirect union effect on wages in
the nonunion sector (Pencavel, 1991)

Second, the expression of “comparable skills and nonpecuniary job attributes” in the
definition cannot hold up in empirical study. Thus this difficulty in the definition
produces two different approaches regarding selectivity bias as follows:

The traditional method used to find the union effect on wages is to place a union

dummy variable in the equation for wages and observe the size and significance of the

 

2 The concept of union wage differential and union wage premium will be interchangeably in this study.

 

CPTCTEUI for union L

53"

. ‘ L .-
.,.3::on abs-Limes L.

, ..
tfi;i.l'.1Ctll.lt’I". "re. 51.1.

' _ U"
assnptzon means
so: disarm tame.
:snhers tPencatel. I‘

W‘K

excludes a poss.
in union wage dztte.’

tsst important st ‘ "

m4) .
teed on the tradition

teens; is that the Ui‘.‘

liit

...-1. 1r. the L'.S.. or

‘
affix-9 "“

”eteanemeub.

l- ‘9' ..
\ . s
‘~ u R J.

~ F“\¥I

4.1., , _
““1 "‘10“ that union

Rf’e’TCh 0f the u:

71* ~
\ ‘9
A uh. :‘Ol’a

~tnenme

 

coefficient for union dummy variable. Using a union dummy variable in the wage
equation assumes that other independent variables such as individual characteristics
(education, tenure, skill, etc.) are independent of worker’s union status (Lee, 1978). This
assumption means that the sizes of coefficients for independent variables except the
union dummy variable in the wage equation are identical for union and nonunion
members (Pencavel, 1991). Thus, the traditional method for the finding union effect on
wages excludes a possibility of interaction between wage differentials and union status in
that union wage differentials affect union status. Lewis’ (1963) study is regarded as the
most important study, which has motivated other following studies so far. His study is
based on the traditional approach using the union dummy variable. The most important
finding is that the union wage differential has varied considerably over time (Johnson,
1984). In the U.S., union wage differentials reached a peak in the periods, 1930-1934 but
almost disappeared between 1945-49. In addition, the size of the coefficient for the union
is greater between 1925-29 but smaller between 1940-45. Freeman and Medoff (1984)
also show that union wage differentials were above 20% in the US. until 19705.

Research of the union wage premium has faced the problem of selectivity bias.
Therefore, the effects of individual and organizational characteristics of union/nonunion
workers are not the same and explanatory variables impact the workers’ decision to be
union 3 member or nonunion worker and further wages differently. As a result, the union
and nonunion (and/or unionized and nonunionized firms) members are not the same in
terms of the characteristics of their demographics and organizations. The fact that union
status selection process is not random results in selectivity bias which means the ordinary

least square should not be applied.

10

 

In addztton. {he 5":
him: from each Gil-”4
35.1an “35'“ mid-ll».
aeticients of indn 1c-
Ltdlettttfi. “hen cow:
:5: studies on union
5.16: as education and
:ezhton members (l
fate: of the indzx idu.
edged. empltw) ers in
fitted size of the m:
“age equations imi
53mm and Leah

.5 .l‘

A number of me
331113095 for the selet
mm or Eliminatec
:3me

S‘
t n ' ,
pcrhlolr am:

If" . -
.‘4; d a"
‘Q‘r 0f eQUallt‘ti

31ft“:
‘ ' me Unlon SI

 

 

 

In addition, the selectivity bias may make the unionized and nonunionized sample
different from each other for two reasons. First, the union’s standard-rate wage policy by
reducing wage inequality among workers within and across firms reduces the size of the
coefficients of individual characteristics associated with workers’ skill such as education
and tenure, when compared to size of the coefficients in nonunionized sample.3 In fact,
most studies on union wage differentials report the fact that the size of the coefficients,
such as education and tenure, in wage equations for union members is smaller than for
nonunion members (Freeman and Medoff, 1984; Pencavel, 1991). Second, once the
impact of the individual characteristics in terms of skill for workers in unionized firms is
reduced, employers in unionized firms are likely to hire more skilled workers because the
reduced size of the impact of skill, expressed in the coefficients of education and tenure,
in wage equations indicates marginal costs for hiring more skilled workers are cheaper
(Duncan and Leigh, 1980).

A number of methods to reduce the selectivity bias are present. Two kinds of
solutions for the selectivity bias are present (Lewis, 1986). First, selectivity bias can be
reduced or eliminated by using panel data. Second, simultaneous equation estimates are
the most popular approach to the bias. Simultaneous equations consist of wage equation
(or a pair of equations by union status) and an equation determining union status. In
practice, the union status equation must first be fitted to data and the predicted value of
the union status variable is attained for each worker. After that, the value is entered in

place of (or in addition to) the observed union status on the wage equation. Lee (1978)

k

3 Chapter 5 and 7 is related to this issue.

11

sees this method to

,1“ v‘\ .p

. 1' . 1
titted 2111a”...

{'33: “'32:“ PTCT‘TZS
B‘" '

1 approaches.
:etzec 31mm: to 11‘;
532-5 '11 that both 5e:
2...; .
an 51.5111 on the as:
J‘tt‘itin'

......11 across the

ultltr. no RX!»

:6 l8 bt’lt

 

 

uses this method to calculate union wage differentials based on a two-stage estimation

procedure (Heckman, 1979) using probit and ordinary least square estimation.4

Qnion Wage Premium in the Distribution of Wage

Both approaches, traditional method using union-dummy variable and the recent
method aiming to deal with the selectivity bias, regarding union wage premium are the
same in that both seek to find the accurate size of the union premium at mean values.
Both stand on the assumption that the regression coefficients regarding union status are
constant across the entire conditional wage distribution (Chamberlain, 1994).

However, no reason is available to assume the uniformity of the union wage effect.
Traditional wisdom in industrial relations states that a trade union is likely to benefit the
' workers at the bottom of wage distribution by boosting their wages more than
comparable workers in the non-unionized sector (Lewis, 1963; Freeman and Medoff,
1984). The median voter model has been used to support the idea that, in determining
Wage demands, the median voter will be favorable to workers at the bottom of wage
distribution rather than the workers at the upper part of wage distribution because the
median wage is below the mean wage (Freeman and Medoff, 1984; Booth, 1984). A
recent finding (Budd and McCall, 2001) appears to support the prediction of the median

Voter model that unionization changes affected only above 25"1 percentile in the wage

¥

‘ However, all researchers do not agree with a need for solutions to correct selectivity bias such as the
Heckman and Lee technique. For example, some influential researchers in union wage effect, such as
Lewis ( 1986) and Freeman and Medoff (1984), do not think that the Heckman and Lee technique is
Superior to the traditional ordinary least square (OLS) estimation. It is very difficult to identity union status
equation, which is required in the Heckman and Lee technique, with variables that do not affect wages.
Such a technique yields divergent results depending on the structure chosen (Freeman, 1994; 294).” For
instance, sometimes the technique by using selectivity variable produces implausible results as the
technique produces 107% of union wage differentials in a study (Renaud, 1998). For this reason,

Ichniowski and her colleagues (1989) did not attempt to estimate simultaneous equations in their study on
union compensation differentials for policemen.

12

o l ‘ 1 '
teth1sme1h0u to 111.

1 . v ‘, . "-
10:65:38 thttlman. 1

33th approaches. 1?.
23111111131111; to deal '-
in: 111 hat both see'
11111111111111 Lose"
mean: across the entzr

hotteter. no reasot:

Taitronal Wisdom tn '.

11,1115 at the hotttv

E'7‘i7-3s’1‘1hleworkers t:‘

“9111 The median 111‘

isidcmzmds. the m;

asiactton rather llldl
1:111 “is: is beret.
711: 11111.1 18111111 1'

.1.1 T1C'(l€l lhat Uni-0'

 

 

 

uses this method to calculate union wage differentials based on a two-stage estimation

Procedure (Heckman, 1979) using probit and ordinary least square estimation.4

Union Wage Premium in the Distribution of Wage

Both approaches, traditional method using union-dummy variable and the recent
method aiming to deal with the selectivity bias, regarding union wage premium are the
same in that both seek to find the accurate size of the union premium at mean values.
Both stand on the assumption that the regression coefficients regarding union status are
constant across the entire conditional wage distribution (Chamberlain, 1994).

However, no reason is available to assume the uniformity of the union wage effect.
Traditional wisdom in industrial relations states that a trade union is likely to benefit the
‘ Workers at the bottom of wage distribution by boosting their wages more than
comparable workers in the non-unionized sector (Lewis, 1963; Freeman and Medoff,
1984). The median voter model has been used to support the idea that, in determining
Wage demands, the median voter will be favorable to workers at the bottom of wage
distribution rather than the workers at the upper part of wage distribution because the
median wage is below the mean wage (Freeman and Medoff, 1984; Booth, 1984). A
recent finding (Budd and McCall, 2001) appears to support the prediction of the median

Voter model that unionization changes affected only above 25th percentile in the wage

k

4 However, all researchers do not agree with a need for solutions to correct selectivity bias such as the
Heckman and Lee technique. For example, some influential researchers in union wage effect, such as
Lewis (1986) and Freeman and Medoff ( 1984), do not think that the Heckman and Lee technique is
Superior to the traditional ordinary least square (OLS) estimation. It is very difficult to identity union status
equation, which is required in the Heckman and Lee technique, with variables that do not affect wages.
Such a technique yields divergent results depending on the structure chosen (Freeman, 1994; 294).” For
inStance, sometimes the technique by using selectivity variable produces implausible results as the
technique produces 107% of union wage differentials in a study (Renaud, 1998). For this reason,

IChniowski and her colleagues (1989) did not attempt to estimate simultaneous equations in their study on
union compensation differentials for policemen.

12

153111111011 from 1084'
capers that 11 age 11:513.
that the 11 age d1:

.1 number 01’ recent
trig-111310;: dttTercht c
1.37.5111th Chamber;

1871.515 tstth less 11111

Ha.

108‘ in US. is

:9 fl‘ ‘. Pr -
‘l‘J‘-oj. .30. ~‘ :1" I
. _ ‘ 1.\.

acres lOMftl q ‘

rational tttsdom. $1

Wists at the b11111"

5'“ 1 .
112177111011

Etehthoueh the t"
35.31 .1 a
. when the 11111

«times I i
are 5111] 1mm»

, 1L

Warned oh

5.

“-1.

iiiion—i
dl “1)dom Si”.

1? ." I
““I{$‘I
. be p0

ssrh'

.~
”111.1713
~ :1

11.

a
4‘5

'1 .
“35 (ltsttihutto'

 

 

distribution from 1984 to 1994 in the US. grocery store industry. Renaud (1998) also
supports that wage differential is larger for unskilled workers than for skilled workers,
finding that the wage differential declines as skill level increases.

A number of recent studies adopting quantile regression indicate that the OLS model
may ignore different effects of unions upon wages, depending on conditional wage
distribution. Chamberlain (1994) found that the union wage effect for both inexperienced
workers with less than 9 years and more experienced workers with more than 20 years
during 1987 in US. is higher at quantile q=.10 in wage distribution than other quantiles
at q=.25, .50, .75, and .90. and shows a monotonic decrease after q=.10 as a quantile
increases toward q=.90. An extreme case in South Afiica appears to support the
trztditional wisdom. Schultz and Mwabu (1998) report that the union wage premium for
workers at the bottom is 145%, while it is 19% for workers at the top of wage
'- distribution.

Even though the two studies (Chamberlain, 1994; Schultz and Mwabu, 1998) indicate
they support the traditional wisdom and appear to reject the OLS assumption, the two
Studies are still incomplete. The quantiles examined are few and arbitrary and thus, we
Cannot draw any conclusions from the results based on such few quantiles.5 Also the two
Studies examined only one-year data. Thus, we can not say if any changes occurred in
tIaditional wisdom supporting unions that benefit less-paid workers in years.

It might be possible for unions to boost mainly the well-paid workers at upper end of
the wage distribution scale rather than the lesser-paid workers. For example, the craft

unions are expected to boost wages for skilled workers more because those unions open

¥

5 Chamberlain (1994) uses five quantiles (q=.10, .25, .50, .75, and .90) and Schultz and Mwabu(1998)
examine results at q=.10, 50, and 90.

13

he:memhersh1p to :
nee-11.111 of the ssh-g:
31:05 and the darn:
when as their ment’:
setpected to he s:
ieeTc-pzng or underd;
fit-paid and skllcc‘l
hitter. economic ta
‘2‘“ LB yet 10 ht“ '1'
5:50 lilf “3ch ll

Eggs-113d b." the em

 

 

 

their membership to a limited number of skilled workers, who are likely to be at the
upper half of the wage distribution scale (Freeman, 1980). However, the decline of craft
unions and the dominance of industrial unions accepting unskilled or semi-skilled
Workers as their members in the late 20th century imply that the influence of craft unions
is expected to be small in an economy. Under authoritarian corporatism in the
developing or underdeveloped countries such as Brazil, it is likely that unions benefit
better-paid and skilled workers the most as a result of political co-optation (Song, 1991).
However, economic favor for better-paid and skilled workers as a result of political co-
optation has yet to be tried by the Korean government.

Also the wages for the workers at the bottom part of the distribution may be
aggravated by the emergence of two-tiered wage agreements by unions, which mean
‘Vvage settlements that decrease the pay rates of future hires while they maintain or
increase the pay rates of existing employees (Katz and Kochan, 1992; 513).” However,
the twootiered wage agreements are not available in Korea, and thus, little possibility for
it exists in Korea.

The issue of how the union premium is distributed in the wage distribution is also
important because it may reveal which group of union workers governs unions. This
paper assumes that it is the most important goal of unions to increase the wages for their
rank-and-file members. Also this study excludes the possibility that other issues such as
employment, working conditions including health and safety, autonomy in the workplace,
etc may be the more important goals of unions to be achieved. We cannot tell which
group of workers in the wage distribution governs trade unions and sets bargaining

demands of unions to employers. However, we can tell at least which group of workers

14

hes in most from

race dtstrthtttton 15 a tct
r’rtoshtch group of \s
17312111111 Therefore. 111
not premtum can be 1

11.7215 he won effecti

The chapter 4 exam

‘1 I
1

50111011 of “aces m
:111. examine umon pr

:1“ I- *7 ‘
1111:1111) chcmees o

l‘

.. ears trom the Grea‘
in; 3‘1
1111.1“). ] am open
, [ROI] premtur‘.‘

lil‘ie. ‘
Ageing “thh SU"*'
cis'

rhttton and smart.

n

.

tpii‘ a .
1.1.11 11111111 or

3%m

ESOllthL'n

 

 

 

benefits the most from unions. Assuming that a distribution of union premiums in the
Wage distribution is a result of governance structure within unions, we may know ex post
facto which group of workers governs trade unions, through the size of the union wage
premium. Therefore, when examining union premium in wage distribution, the issue of
union premium can be closely connected with the governance structure within unions as
well as the union effects on labor markets in general.

The chapter 4 examines if the union wage premium changes across quantiles in the
distribution of wages in Korea. To overcome the weakness of using only a few quantiles,
I will examine union premiums at every quantile. In addition, I will attempt to determine
if any yearly changes occur in the union premium at every quantile by using data with the
1 3 years from the Great Labor Offensive in 1987.

Initially, I am open to both possibilities including the OLS assumption, implying an
equal union premium regardless of quantiles in wage distribution, and traditional
Wisdom, which suggests larger union premium for workers at the bottom of wage

distribution and smaller for the workers at the upper of wage distribution, even though it

is unlikely that the OLS assumption will be found valid.
3. Studies on the Union Effect on Wage Dispersion

Meanings and Importance of Wage Dispersion
While the union wage effect focuses on the level of wages, the union effects on wage
dispersion place more attention on the relative aspect of wages between and within

employees (and organizations). Both the compensation theory (Milkovich & Newman,

15

reEesant others no ls‘r‘
fighters. Through C
see hots ernplosers ‘
lichei. 20(0). The tt'
Itch isthe kes to or
1re {Lazear gt 8
testifies is to secure

‘Jisll'd. 201111

 

 

 

 

 

1999) and the equity theory (Adams, 1965) emphasize the importance of the relative
aspect of wages in that employees care greatly about their wages compared to those of
relevant others no less than about the absolute amount of wages they earn from their
employers. Through comparing their wages to those of the relevant others, employees
care how employers value their work and how employers treat themselves (Bloom &
Michel, 2002). The tournament theory asserts that it is the concept of wage dispersion,
which is the key to understanding the relationship between compensation and employee
outcomes (Lazear & Rosen, 1981). In this sense, one of the important functions of union
activities is to secure fair treatment in the employment relations for their membership
(Godard, 2000).
Wage dispersion here is defined as “the amount of difference (inequality) in wage

created by a firm’s wage structure” (Milkovich & Newman, 1999). A more dispersed
Wage structure means relatively rich rewards for fewer employees at the top of the wage
Structure and relatively poor rewards for the rest. A more compressed wage structure is a
more egalitarian wage structure, where a difference between wages for the fewer
employees at the top of the wage structure and wages for the rest exists. While a more
dispersed wage structure is aimed to foster competition among employees in pursuit of
rich rewards for fewer employees at the top of the pay structure, in other words, less
Wage dispersion is aimed to facilitate cooperation among employees rather than
competition. As a result, wage dispersion may affect product quality (Cowherd & Levine,
1992), turnover of college and university personnel (Bloom & Michel, 2002; Pfeffer &
Davis-Blake, 1992), and employee tenure. Dispersed wage structures increase retention

of a firm’s most talented employees at the expense of lowering retention and increasing

l6

>-at - 1' ”1,3
enter among
2:11 , -.

f'iill S '3 7‘05:

”I ~1WJ :" ‘:"'
ok‘lel'" ‘1; IS; I ‘

Or. the letel ot‘ 1}:
9312s, the) rarse 11:
his s2:es are red;
:Tplsjted 1n la'oor r1
li'itl; Part. 198417

..cene tnequaltts in t

l 1th regard to th:

{mi-,1, .
wae“ag€1ncqu“l
‘u‘lt

”Tim

I

REV

131:1:
311d H 6}“ (10d

111:5 .
hue i0 lost

Incl

‘deI dek Cl 5

 

 

turnover among the employees who remain (Lazear & Rosen, 1981). Reduced wage
differentials are posited to signal that all employees are important, thereby creating more
employment stability.

On the level of the whole economy, trade unions redistribute income from capital to
labor as they raise union wages. Wage dispersion becomes more important as it reflects
how wages are redistributed within labor. As a result, wage inequality between the
employed in labor market explains much of income inequality in an economy (K00,
1996; Park, 1984).6 For instance, wage inequality in Korea accounts for over 50% of

income inequality in the 19903 (J eong and Choi, 2001).

Wage dispersion and unions

With regard to the effect of unions on wage structure, it is well known that unions
reduce wage inequality by reducing the dispersion of wages within and across firms in
organized sectors as compared to unorganized workers (Freeman and Medoff, 1984;
Belman and Heywood, 1990). One study (Lemieux, 1993) reports that trade unions in the
US. contribute to lowering wage inequality by reducing the overall dispersion of wages
in the labor market by about 6.3 percent. Unions’ standardization policy, such as single
rate or automatic-progression modes of wage payment, in regard to wages is said to
decrease the magnitude of impact of principal variables such as education, tenure, sex,
and firm size, which are expressed in the regression coefficients (Freeman, 1980).

However, the effect of unions reducing wage inequality is not simply unidirectional

as a result of opposing forces. On the one hand, union wage effects increase wage

g

6 Income here includes nonwage earnings from all kinds of propeny such as land and financial interest as
well as earnings from wages.

17

 

heealtts 111 some st.
.2115 1ncre2se 112;;
1:71-12:11: sorters. l, r

O .31. ‘ .
Siren? ETTTWOW‘ALTT

Etrttard pressure 0
Lrhcsort. 19‘51. The
stressed stage meg;
irritantliiol. p. l

:as more unec; cal h

3‘19,
1...!th3.

1
3.3”"

.3575 1n 111101"; ~.'

.3“...
1-1.111 new untot‘s C

511 . .
i..1¢tdCurrne. l‘ih

it); ‘
L. .9
:5 11

equality 1n um

i

. “UPON the re] .~
Three

Lu. SQUI‘CES Of \1

v
u
'2’,

..isen firms. and

 

lets
111, 19931’

 

 

inequality in some ways. For instance, many studies on union wage effects assure that
unions increase wages of blue-collar union workers more than those of blue-collar
nonunion workers. Union success in enhancing the union/nonunion relative wage is likely
to reduce employment in the unionized relative to non-unionized industries, thus placing
downward pressure on nonunion wages and increasing wage inequality (Lewis, 1963;
Johnson, 1975). The so-called ‘crowding effect’ (Kahn and Curme, 1987) indicates
increased wage inequality by unions. Focusing on this function of trade unions, Milton
Friedman (1962, p. 124) accused unions of having “made the incomes of the working
class more unequal by reducing the opportunities available to the most disadvantaged
workers.”

On the other hand, union wage effects decrease wage inequality in other ways. For
employers in unorganized firms, they are willing to introduce union wage policy to
forestall new unions of their employees and compete for skilled workers (Rosen, 1969;
Kahn and Currne, 1987). This ‘threat effect’ (Kahn and Curme, 1987) is likely to reduce
Wage inequality in unorganized sector. Thus, the net effect of unions on wage inequality
depends upon the relative strength of these opposing effects.

Three sources of wage inequality are, wage dispersion within firms, wage dispersion
between firms, and the white-collar/blue-collar wage differential (Freeman, 1980;
Lemieux, 1993).7 Industrial unions lower wage dispersion between firms by realizing
wage standardization within an industry or local product market. Also unions can
contribute to reducing wage dispersion within firms by reducing the number of job

categories and wage differences between job categories within firms. Finally, unions can

g

7 Doeringer (1984) uses different expressions, vertical equity and horizontal equity. While vertical equity
describes how to reduce wage inequality within occupations, firms, or job categories, horizontal equity
means wage equality between occupations, industries, firms, or other categories.

18

 

ferhihe-collar “ '07
The “35’? COT-7
J

httutrons depend
Freeman and Med-1
bier than the as c
13:15 to reduce \s.
:::i2:12c contlzcts
-"n piece or
personal rates or
wean 10.811; 4 1)
one “3) used 11'
Mm“ (or tartaric
ife'ence betttecn t.

1
5.3.5, .
“WMJI

they found

 

affect wage dispersion between blue-collar and white-collar workers by enhancing wages
for blue-collar workers.

The wage compression by unions is realized by fact that unions are political
institutions depending on the preferences of the median voter (Farber, 1978; Booth, 1984;
Freeman and Medoff, 1984). Because the wage level of the median voter in unions is
lower than the average wages in unionized firms or industry, the union’s wage policy
tends to reduce wage gap among members. Also the fact that union leaders tend to
minimize conflicts between members support standard-rate wage policy, defined as a
“uniform piece or time rates among comparable workers across establishments and
irnpersonal rates or ranges of rates in a given occupational class within establishments
(Freeman, 1980; 4).”

One way used to find union effect on wage dispersion is to calculate the standard
deviation (or variance) of log earnings of union and nonunion workers and compare the
difference between the two (Freeman, 1980; Bratsberg & Ragan Jr., 1997). Using the
method, they found the union effects on wage dispersion in the US. Other ways to
measure the degree of wage inequality are to use Gini-coefficient, a summary statistic
deriVed from the Lorenz curve (Hyclak, 1979; Flaherty & Caniglia, 1992; Bloom &
Michel, 2002), quintile share calculating share of the 20% of the relevant population in
the earnings distribution (Flaherty & Caniglia, 1992), coefficient of variation (Rowthom,
l992; Rodriguez- Gutierrez, 2001), or the difference between ninetieth and tenth
Percentiles of the earnings distribution (Bratsberg & Ragan Jr., 1997; Leslie & Pu, 1996).

A few recent studies (DiNardo, Fortin, and Lemieux, 1995; DiNardo and Lemieux, 1997;

19

Budd and McCall. 3i "
\ttouinmualit) imiz.
T3: trend of unions
3M: l'nitt‘d States, I
EDIE on “age dyer»;
imii W 1m cousin

his US, standard

 

333111th members in
buster. the rule t

LfiHEUCEHF. if ill 1311 '

71;.
-.. mien: ‘
-. monupm

\

W101 .-
wl Glmcrsxon \

1‘4

3:}
"£3333 ‘ 1993 ) rcpor
Brim ;

atom for all “1

§avl
tiers

Is: A?

h i‘Cl 0f prObahzl

“CW (304'

)0). b)-
Mil-'33 .

is

 

e bluficoll

d.

 

 

sum
SEC [013
Bins r
an: S}

Budd and McCall, 2001) focus on the entire distribution of wages, but do not examine
Various inequality indices mentioned above.

The trend of unions reducing wage dispersion and wage inequality is not confined only
to the United States. Freeman’s comparative study (1994) indicates that the effects of
unions on wage dispersion also are found in all the other nine countries in his study.
Among the ten countries, the largest union effect on wage dispersion is found in the US.
In the U.S., standard deviation of log earnings of union members is less than that of
nonunion members by 25%.

However, the role of the union in reducing wage dispersion has not been consistent.
Theoretically, if highly paid workers are well organized by labor organizations such as
craft unions, monopolistic aspects of unions protecting well-paid workers will dominate
and wage dispersion will be increased (Freeman, 1980). In effect, however, Flaherty &
Caniglia (1992) report that American unionism appears to increase inequality in earnings

distribution for all women because unions exert a positive impact on full-time female
workers and thus broaden the wage dispersion between full-time and part—time. Adding
the effect of probability of being dismissed into theoretical framework, Donohue and
Heywood (2000), by using the data from the Current Population Survey (CPS) in 1990,
found a similar result that increased unionization increases in fact the dispersion of wages
for female blue-collar workers. Also, Arbache (1999) finds that while unions increase
Wage premium, the larger wage dispersion is found in unionized sectors than in
unorganized sectors in Brazil. Therefore, whether union reduces wage dispersion in

Korea should be determined by empirical studies.

20

lite literature on
he in “age dlsp:
first is supplyszde d;
station. female I;

, v
urn-.1 ‘ 3 1 ‘ '
gunk-1011 C dL" m

l

1"!”

...ujer‘s Slth. The
”<qu of ether slzllJ
mart competition (1
2150:" um marl et tr.
Stator of union lug
man also at‘feetx

4,-zzttontzed sectors t

fill a. ' ‘ 0
an that lntlt‘tbt‘tl \

was of union

tlt‘t‘a

 

 

 

The literature on wage dispersion presents at least three major determinants for the
Change in wage dispersion (Gosling and Machin, 1995; Budd and McCall, 2001). The
first is supply-side determinants of wage dispersion such as the change in the return to
education, female labor force participation, immigration, and so on. The second is
demand-side determinants such as a change in the demand for skilled workers from the
employer’s side. The increased demand for skilled workers by employers is viewed as a
result of either skill-based technical change (Bound and Johnson, 1992) or increased
import competition (Murphy and Welch, 1992). A third explanation is to focus on the
role of labor market institutions like trade unions and minimum wages. Union density, an
indicator of union bargaining power, affects wage dispersion. In addition, the union wage
premium also affects the difference in wage dispersion between unionized and
nonunionized sectors (Budd and McCall, 2001). Many involved in industrial relations
explain that increased wage inequality in the US. in the last two decades is a result of the
decline of union density and eroded minimum wages (Freeman, 1992; DiNardo and

Lemieux, 1993)

Mgflfispersion in Korea

Traditionally, East Asian countries like Korea, Japan, Taiwan, and Singapore have
been relatively egalitarian societies, compared to the ‘winner-take-all’ society such as the
US. and Western European countries, in terms of wage dispersion (Cowherd & Levine,
1992; Koike, 1988). Cultural explanations emphasizing collectivism is said to lead to the
egalitarian societies in the East Asian countries (e.g., Hofstede, 1980; Adler, 1997; Berg

and Chang, 2000). For example, wage inequality in 1970 was reported to be lower in

21

 

l

Km by 350111 3T"
readil) decreased i
terrific/905 in l
-Wton The reas
adecrease in return

iiled “utters u it}:
if: and lapel. 1W5
along with an 1;
stain-e ot‘real m;

if". Thus Korea u

\

Eli; 1993).
SINCE the late-1W

iii“ l"
ac.) tntem'm en \\ '

 

 

 

Korea by about 30-40% than in the US. (Kim and Topel, 1995).8 Wage inequality has
steadily decreased in Korea since the 19705 (Park, 1984). The phenomenon continued to
the mid-19908 in Korea, while the US. experienced a considerable increase of wage
dispersion. The reason for the phenomenon prior to the Great Labor Offensive in 1987 is
a decrease in return rate for education, which resulted from a shortage of supply for less-
skilled workers with low education level as the level of education increases (Park, 1984;
Iiim and Topel, 1995; Nahm, 1997).9 It is a surprise that the reduced wage dispersion has
gone along with an increase in real wages in Korea, compared to the US. which showed
a decline of real wages and a increase in wage dispersion (Kim and Topel, 1995; Nahm,
1997). Thus Korea was praised as an unusual example of ‘growth with equity (World
Bank, 1993).

Since the late-19605, the relative low level of the wage inequality in Korea has been
closely interwoven with the low wages under the export—driven strategy. The government
could appease workers’ discontent with low wages through depending on low wage
inequality before the year of 1987 (Song, 1991; Lee, 1993). The external fairness of
Wages, achieved by the principle of ‘generous to the less-paid and strict to well-paid,’ had
been an important tool for the government to sustain low wage advantages (Bae, 1995).

Also, Korean trade unions also have kept the principle of ‘generous to the less-paid
and strict to well-paid.’ In addition, unions have regarded uniform and automatic
Progression of wage payment, reducing wage gap across occupations, educations, and

53X, as goals to be achieved with employers at the bargaining table (Park, 1995; Koo,

\

8 The measure of wage inequality is the differentials between the ninetieth and the tenth percentiles of the

:Vage distribution (Kim and Topel, 1995).

3 3111-011 of the 20-24 age group enrolled in post-secondary schools and universities increased from 6% to
7 /o m 1988 (Nahm, 1997).

22

H*t In reality ”“7
113ml within occul‘
3:035 occupations ark
rerierot “351‘ P337771
rag: for “triers. The

grtgesston of Mg: ,9
urgession 0t urge r
iterate progression

grips. and of sex.

Vttt' few studies c
19931 malyzed data 1
33‘ It“ Whether trade
“Fillies the tort
73111110115 did not .t
"it teem,” he 1;
tritium gm CIT."
tilt tear 1986 t,-

w . I

 

 

1996).10 In reality, most firms have multiple forms of automatic progression of wage
Payment within occupations and job groups, producing wage inequality and competition
across occupations and job groups (KLSI, 1998). Generally speaking, the higher the
number of wage payments within firms, the greater employers’ discretion in determining
wage for workers. The goal of unions is to change from a multiple number of automatic
progression of wage payment within occupations and job groups to a single automatic
progression of wage payment within occupations and job groups and further to single
automatic progression of wage payment within firms regardless of occupation, job
groups, and of sex.

Very few studies on union effect on wage dispersion in Korea are available. Lee
(1993) analyzed data in 1986 from the Basic Survey on Wage Structure (BSWS) in order
to see whether trade unions reduce wage dispersion. Even though he focused on the
employees who work in the large firms, which hire at least 500 employees, he found that
trade unions did not affect wage dispersion significantly. The lack of effect of unions on
wage dispersion he found made sense because the data used were gathered in 1986 when
anthoritarian government repressed free activities of unionism and collective bargaining
and the year 1986 is just one year prior to the Great Labor Offensive of 1987 in Korea
(Park, 1984).

In the analysis of data in 1988 from the BSWS, which was conducted by the Ministry
of Labour, Song (1994) examined degree of inequality of wages. He found that the
Standard deviation of wages for union members is smaller by 5% than the standard

deViation of wages for non-union members. However, he concluded that the number (5%)
\

to
Also numerous trade unions in Korea mentioned uniform and automatic progression modes of wage
Pamem as one of major accorrrplishments realized throughout bargaining with employers in their web-
ased~home—pages.

23

J": a 3
Cl elilfrfilLC ll". Shift...

ELCil smaller than C\

amnion sector.
from the data of:

smith for pert?

xv. 9‘ '
.,ers.onmthtn n‘t

‘
q
us

:5‘121Ie'b , 3“"
«Maul. A \ “ 1‘
' 5. 0 I

ll

:5:??th ‘
.....sats ot \w

in t We). hon e\ e

1..
R3; ll“. '
.“r .....h that th th

kiwi».
:tb\rlnL‘ {0 b
~ "7' '
L «1.3.1::

2": .
'Hl h." ‘
eQHsldcnn‘O ll
3
L."

0f difference in standard deviation of wages between union and non-union members was
much smaller than expected after considering effects on wage dispersion in the United
States. Using data fiom the Minimum Wage Committee (MWC) in 1988, Kim (1991)
reports that wage dispersion in union sector is 73% for men and 68% for women in
nonunion sector.

From the data of the BSWS in 1993, K00 (1996) examined the union effects on wage
dispersion for permanent workers in the metal industry. The result was that wage
dispersion within, measured by variance of log wages, establishments in the union sector
is smaller by 23% than that in the nonunion sector, after controlling for differences in the
characteristics of workers and their establishments between the two sectors. According to
K00 (1996), however, the wage dispersion between the establishments in union sector is
larger than that in the nonunion sector due to the intrinsic feature of enterprise unionism
neglecting to bargain at the industry level beyond the enterprise level. He concluded that
the wage dispersion in the union sector was less by 11% than in the nonunion sector
When considering the positive effect of unions on wage dispersion within establishments
and the negative effect on wage dispersion between establishments.

Lee and Kwon (1995) also reported that in 1994 unions appeared to decrease wage
diSpersion in the union sector by 4.6% rather than in the nonunion sector in all industries.
However, wage dispersion for manufacturing workers in the union sector was much
Smallet by 24.6% than in the nonunion sector. Therefore, the union effects on wage
dispersion are much stronger in the manufacturing industry than in any other industry,

due to the high union density.

24

Employers tr} 1
or mployes 1n orde
sitters. unions the
shII tr order to mom.

Even though som'

meal the same ltmt'...

 

n: its period of We '
.1, . r! t
.tlt-ttlttal norrers I
ml} the metal 17'
«are dispemon are

are: be genernlt/eti

lltn) questions st
‘\ i. l . '
1 rar. Haxe lxoreon

«LT: to skill and et:

anothose questtor.

:1:
.non Ellect on R.

 

The employers try to set wages comparable to the educational attainment and skill of
the employees in order to motivate the employees. As a result, when bargaining with the
employers, unions make an effort to reduce the return to educational attainment and/or
skill in order to maintain wage equality for labor solidarity.

Even though some searchers have examined union effects on wage dispersion, all
reveal the same limitations that do not explain how unions have affected wage dispersion
for the period of 1987-1999. Also, many studies limit the boundary of interest regarding
blue-collar workers (Kim, 1991) or workers in manufacturing (Lee, 1993) or more
nfilt‘l‘owly the metal industry (Uh and Lee, 1992; K00, 1996), where the union effects on
Wage dispersion are likely to be the largest. Even then, the results from prior studies
Cannot be generalized.

Many questions still remain. For example, how has wage dispersion in Korea changed
so far? Have Korean unions been successful in resisting market pressures to broaden the
return to skill and education attainment and bigger wage inequality? This study aims to

ansWer those questions.

4- Union Effect on Return Rates for Human Capital

wot influence of unions over wage determinants

Basically, the question, whether trade unions create wage premiums for their union
members, is equal to the question, if trade unions can be a separated independent variable
in the wage equation. The question, regarding union wage effect, which will be dealt

InaiIlly in the chapter 4, assumes that the directions and sizes of coefficients representing

25

stator. H0'~"~" '37
nation may 9"

trade unions rea.

mono} chx‘aet;

"etor. Re'r' ' '

' 1
:4. 11‘.
»

 

all independent variables, such as industry or establishment characteristics, personal
characteristics, and human capital, are identical between unionized and non-unionized
sector. However, the sizes of coefficients for many independent variables in the wage
equation may be different between the two sectors, due to the effects of trade unions. If
trade unions really affect determinants of wages, the verification of union effect on wage
determinants can be another topic of research, different from union wage effect.

The question in this section is whether trade unions affect the rates of return for
human capital for workers regarding education, experience, and tenure, and the effects of
personal characteristics such as gender, regarding wages, compared to the unorganized
sector. Regarding the whole period of this study, 1987-1999, whether any significant
changes in the union effect on the return rates for those variables are evident is interest,
also. Also, it is interesting to examine how the union effects on wage determinants have
changed from 1987 to 1999. Furthermore, it might be interesting to see whether trade
unions may affect wage determinants differently, depending on the relative location of

employees ranging lowest, median, and highest part of wage distribution.

Union and various determinants of wages
Previous studies on union effects on the wage structure conclude that trade unions
reduce returns to all individual employee characteristics such gender, marital status, and
human capital including experience, tenure, and education (Bloch and Kuskin, 1978;
Duncan and Leigh, 1980; Johnson and Youmans, 1971; Freeman and Medoff, 1984).
Previous studies have explained how trade unions reduce returns to personal

characteristics such as gender and human capital including educational attainment,

26

experience. and tens
mm: nay b5 of
tag: rates. than 1‘.
scones simplified. ‘

ferment amon 1 Cl’.

 

retirements for the 4

rhwa

-.....nee in \\ aees t
1;. ‘
t... and pay tor p

ya. , a
....orrnante. trade t

:‘ ,{ic‘c‘a '
.....erence m ll‘um

\f“ \ a
9'» s . X
will“: 0. ultre

The four \andh‘t

err ‘
w) are ltheh' it

g

eet the to.

1‘.‘ f
. Eat

Set ~'
laililt’lflSil'V r

0

3071mm” ’
ahOrt
rat

Eff Fa- ’
L. tlgn in
Vii? "|

experience, and tenure (Freeman, 1982; Maranto, 1985). According to them, the most
important way by which trade unions affect wage determinants is establishment of single
wage rates. Under the single wage rates, differentiation of wages among employees
becomes simplified. Within a category of wages, difference in the level of educational
attainment among employees can be minimized or little, if employees meet minimum
requirements for the job. That is, over-qualification in human capital may not result in
difference in wages (Maranto, 1985). While employers seek to wage plan, merit review
plan and pay for performance, reflecting all difference in human capital and work
performance, trade unions seek to establish standard wage policy, which is less sensitive

to difference in human capital and employee’s performance.

Importance of educationirgmience, tenure. and sex

The four variables, education, experience, tenure, and sex, are selected in that trade
unions are likely to affect the effects of the four on wages. Also many other variables that
may affect the level of wages for workers: for example, organizational characteristics
such as industry, region, and establishment size, industry-wide variables such as industry
concentration ratio and the rate of unemployment in the industry, macro-economic
variables such as unemployment rate and the rate of economic development, strategies of
government over wages and collective bargaining, etc. The nature of the enterprise union
system and the lack of resources held by the industry-level or national level union
confederation in Korea compels us to expect that the unions can impact on the four
variables more than any other variables on the organizational-level, industry-level, or

national level.

27

long the four. et.‘
Korea hate pointed I
wage gap hetnoen t
|
neuron is the enter
t robs \‘S. po.
moaned about ti'.
granites more than -_-

Also gender has h.
irf'ernrle oorkers h..
....e there has hel
“Slim m llldn}. llll‘,‘
L’i‘ttrsit}; female at,

a unnersnv in or

3}" "’Tan
(.....tnt from lllt‘

 

LY?"
... .5018 \‘C
, ars. compo

«normed out that l

.1413: ...
.a trade unions

Win: n' '
1th less 1nd

1

We. mal= "
L and ten'

1
'9

1.:

3tell known th'

Among the four, education is the most important variable that prior studies on wages in
Korea have pointed out (Song, 1991; Bae, 1995; Chang, 1999). It is well known that the
wage gap between high school graduates and university graduates has been large and
education is the criterion dividing occupations into white-collar vs. blue-collar and well-
paying jobs vs. poor-paying jobs (Bae, 1995 ; Chang, 1999). Also, workers have
complained about the unfair wage gap between high-school graduates and university
graduates more than in any other aspects regarding wages, except the low level of wages.

Also gender has been one of interest regarding wages (Song, 1991). However, wages
for female workers have been much lower than wages for comparable male workers. For
instance, there has been a separate scheme for female workers from the scheme for male
Workers in many firms in Korea (Jeong, 1993; Park, 1995). In spite of graduating from a
university, female workers usually earn lower wages than the male workers graduated
from a university in many firms. F urtherrnore, the situation in the manufacturing sector is
nOt different from the non-manufacturing sector. To be promoted, female workers wait
for more years, compared to male workers. Many researchers, as well as female workers,
have pointed out that the economic discrimination regarding wages dependant on sex. As

10ng as trade unions are expected to oppose any type of discrimination and to protect
Workers with less individual bargaining power, unions are likely to affect the wage gap
between male and female workers in the organized sector.
It is well known that tenure is one of important determinants of wages in Japan, where
enterprise union systems have developed. The seniority-based wage system is evident in
the United States and Japan showing the importance of tenure in determining wages there

(Maranto, 1985). Korea also has seniority-based wage scheme in most firms. Thus, we

28

:“r expert that tenure
tic-otter. the a\ CT“ .33
error in Korea smee
trimming rages r.‘
seller role of ten..'
education and se\ d'e'.
stirring. such as ter.
diam

Also etperience 15 t
Stu-specific sltrils ;
Static. and rather
31116 1111] bt’ Value

1'," n
“J“ rmg are de'.

>4.

11355333“ Oflcnt ..

Ellie 4

ampomm th‘l;

llni
gluon Benefit E.

 

ll’l .
E 'mP'OttancL

it‘d-l
grd'c

t _
.Qj ,
”Balm. the *

t
1

Roy} .

lrm
11.1:

_. I
a). Anon),r y
5 J

 

can expect that tenure may play a major role in determining wages in Korea, on one hand.
However, the average years of tenure per workers have been low in the manufacturing
sector in Korea since the 19708. Therefore, we can anticipate that the role of tenure in
determining wages might be less in Korea than expected. Song (1991) mentioned the
Smaller role of tenure in determining wages in Korea than in Japan, arguing that
education and sex determines wages in Korea, while the variables that are related to costs
of living, such as tenure, age, and numbers of families, are major determinants of wages
in Japan.

Also experience is one important determinant of wages. While tenure is related to more
firm-specific skills among the four variables, experience is related'to the less firm-
Specific, and rather more general skills usable in the entire labor market. As a result,
tenure will be valued when determining wages within firms if internal labor markets
Within firms are developed. If internal labor markets within firms are not developed and

lesser years of tenure among workers are found, the variable, experience, will emerge as

more important determinant of wages.

5- Union Benefit Effects

The importance of the union effect on fringe benefits, when examining its effect on
tOtal compensation, has increased and is expected to increase further in the near future.
For instance, the percentage of fringe benefits in total compensations for blue-collar
Workers in the US. increased from 17% in 1951 to 30% in 1981 (Freeman and Medoff,

1984). Another source of data reports a rapid rise in employee benefit costs, increasing

29

ion 34.". of pane
13,2995 tllrllot'ieh ..
CS. but is also foes;
hhorea the share .
costs pad h} the er“:
135530111 lid- 1'0 Q:
lnsntte of the it).

v

non etfeet on \\ tic-i

 

Joints prooomO“ "

llthi
:,.L ‘ “

33321157133110“ and 11::

"re
‘~?TOP07110ndl to ei‘

 

from 24.7% of payroll costs in 1959 to 31.1% in 1969 to 38.4% in 1990 and then to 42%
in 1995 (Milkovich and Newman, 1999; 395). The phenomenon is not confined to the
U.S. but is also found in Canada (Renaud; 1998) and even in Korea (Uh and Lee, 1992).
In Korea, the share of benefits including statutory and voluntary benefits in total labor
costs, paid by the employer, has increased from 8.26% in 1987 to 11.82% in 1999, while
the share of cash earnings paid by employers in total labor costs has decreased from
83 -55% in 1987 to 72.97% in 1999 (MOLb, various years).

In spite of the increasing importance of fringe benefits, most studies concerning the
union effect on wages have been based on an assumption that unions would affect fringe
benefits proportionally as it did on wages (Alpert, 1982).H The changes in wages and
filrther a union effect on wages can be a rough proxy showing changes in total
cOmpensation and the union effect on total compensation so long as increases in wages
are proportional to changes in benefits (Smith and Ehrenberg, 1983). If the assumption of
preponional effect of unions on fringe benefits on wages is not appropriate, it means that
Studies on the union wage effect ignoring fringe benefits miscalculate union effects on
tOtal compensation. As a result, many studies declared a potential overestimation or
underestimation of union effects on total compensation (Hyclak, 1979; Lewis, 1986;
Bellante and Long, 1981; Moore and Raisian, 1987). Recent studies have attempted to fill

the gaps in the previous studies regarding fringe benefits (Renaud, 1998; Wunnava and
Ewing, 1999).
A general belief is that trade unions affect fringe benefits through increasing the level

of benefits, the share of benefits in total compensation, and the probability that a specific

\

n . . . . . .
While Alpert (1990) called this assumption as ‘proportronal assumption of benefits,’ Triplett (1983)
hauled it as ‘consistency hypothesis.’

3O

 

PM of benefits st-
truth chanmg 1h.

tttttihutatt' plats tt

stitch cost sharing .

 

 

: is public sector 1
Deer conducted ttt
11;;i'etschmtdt. 1W

531‘). Austrai ta 1 K0

:“v-

2: Benefits ttt _l<

laments. and hen

”IT _ ' .
“A “Item: tror

it Eps 311d do“

fluid 10 83.5;

r- A ~
‘31‘i1“ “

'5' from the

 

types of benefits such as pensions and health insurance are provided by employers,
through changing the amount of cost sharing between employers and employees from
contributory plans towards non-contributory (entirely employer-financed) plans with no
explicit cost sharing (Gentry and Peress, 1994), and through changing the fimding pattern
in the public sector (Mitchell and Smith, 1992). As a result, the union benefit effect has
been conducted in many developed countries including Canada (Swidinsky and
Kupferschrnidt, 1991; Gordon et al., 2003), UK. (Brown, Deakin, Nash and Oxenbridge,

2000), Australia (Komfield, 1993), as well as the United States.

Frirrge Benefits in Korea

Table 2-1 presents the changes in the portions of various categories consisting of total
labor costs in Korea from 1986 to 1999. Cash earnings consist of regular wages, overtime
payments, and bonus payments.12 Cash earnings represent the largest share of total labor
costs, ranging from 67.90% in 1998 to 84.40% in 1989. However, their share indicates
the ups and downs over the period. The share of cash earnings, 78.63%, in 1986
increased to 83.55% in 1987 and to 84.40% in 1988, due to a sharp rise in wages
resulting from the Great Labor Offensive of 1987 (Choi, 1992). Compared to the declines
of the portions of other categories, such as severance payments, statutory benefits, and
voluntary benefits, in total labor costs, the increase in the share of cash earnings in 1987
and 1988 implies that the demands of employees and unions at that time targeted to
increase cash earnings including bonus payments, while paying little attention to non-

cash earnings, especially voluntary benefits. As employees’ desires to achieve higher

___

’2 Paid holidays, vacations, and bonus, which are classified as benefits in U.S., are regarded as cash
earnings in Korea. Thus, the share of benefits in total compensation in Korea would be considerably
mcreased when following the scheme of classification of labor costs in United States.

31

 

l «o -
W353 T3153. 31 it’ll).

defused 111 the 1‘?"

cat. emitters in the '

The share of c.-.

 

mares 01 \t tees .

 

Wits. For instinct
1the the shares 0
eaattt e13 stable tt

Pl'ton Olcash ear

"3333113 150m
E
33‘. ‘
T‘h’i'u‘n: b
c tne
.i’

wage rates, at least in part, was realized, the share of cash earnings in total labor costs
decreased in the 19908, reaching 72.97% in 1999. It means that it took a few years for
both employers and unions including union members to start to increase the share. of non-
cash earnings in the total labor costs.

The share of cash earnings in Korea can be compared to that in Japan because the
structures of wages and total labor costs are very similar to those in Japan. In 1988 the
share of cash earnings are similar in both Korea (84.40%) and Japan (83.68%). However,
the difference in the share of cash earnings between Korea and Japan occurred in the
19905. For instance, the numbers in 1995 are 82.26% in Japan and 74.42% in Korea.
While the shares of cash earnings and other components comprising total labor costs is
relatively stable in Japan during the past two decades (Hart and Kawasaki, 1999), the
portion of cash earnings (and other components) experienced greater changes in Korea.

The reduction of cash earnings in total labor costs in the 19908 implies the greater
importance of various non-wage labor costs (Song, 1995; Park, 2002). Among the
various components of total labor costs, three components, statutory benefits, voluntary
benefits, and severance payment.

With minor variations, the share of statutory benefits over the period is stable, ranging
from about 3 percent to 5 percent. As mentioned previously, the decrease in the share of
statutory benefits for 1987-1991 can be attributed to the absolute increase of total labor
costs due to the wage offensive since 1987. Comparably, the increase of statutory
benefits borne by employers in 1992 and 1993 can be explained by a new category in
statutory benefits, an allotted charge for promoting employment of the handicapped.

Also, an addition of employment insurance and an increase in the insurance premium,

32

I I I II I
......1: .. .. ... . :
. .. - ru.1L—..L..a.d..~
3:2... ._ 9.12;...) >..:.::....J. .1 cc 1.2.3::2: 2... 9.2.29»:

 

u.....-.

-...CUDI

:1... »

 

0.233... 1.39:: 225...... 91.22.437.91 a a<

 

«.33 . -CI? ~ .7.-4...:r.:?.: :ax Ix 7.11.... v £3214 .~ 1” .¢~.~.-

...otaE 05 8 8528...: 8: we: 888:8 .8» 2: E muqm on. 0883 nose—6.6 8c. 88:. .<.Z 68 Z

.o\ooo_ 888 8: on .38 mon 5.. .8 09.2.3.8 .8 EB of. 63332. m. 2.88.. 2: 0883 cog—88 fl 8:35-me .8 ES. 2:. 882
380 .83 .850 + $80 “:35?on + $80 wage“. +

$80 8:88 §==_o> + .80 amazon boaafim + 3:258 3:883. + 8:38.-»8 + $5.58 :80 u 380 893 “802

.80» some .30.): .33 no 8&5: .3 $048 actsm .80 .83 3.8.85 05 688m

 

 

 

cod co; :6 $6 3.3 a _ .om .<.Z 3&5 coa—
vod co; Nm.m cud mm. _ N coda .<.Z 85¢ 82
o_.o 3: mm... oz. 3.2 328 .<.Z no.2. 82
m _ .o mod 0:. mm... 33. mafia .<.Z 3.2. com—
omd mag 36 and Q..— . no.3 .<.Z NYE. 32
o _ .o m: 3.» mod cod Eém .<.Z 3.2. $2
_N.o NM: mmd 86 :d m _ .mm .<.Z 8.3. m3—
emd 3..— Nms med end 3.? .<.Z 31K ~03
and _n._ 8.5 mad Nb.» .<.Z .<.Z ands 32
Ed .1.— ood 84m :8 .<.Z .<.Z 86.. com:
26 2.6 «6.0 mod cod 3.2” 2.: 2.. _ w awo—
Ed 8.— vd _o.m m~.n 2.6m and. 3.3 awo—
m_.o 3.. $6 Sam Sun 32 5.: 3.3 So—
n _ .o 36 3.6 and 3.5 no.3 SN— medn 32
380 380 $52.80 $5.50

Ea... wmwwawh ammwummwfi? 05:03 Wowwfimm :80 go 5.8 .8 o\.. mwflmwam

£38m . . boHBSm o\.. 8 888m 8 288.5
.280 8%— .SS mo 082588 a m<

 

 

amaTcwE .8589: =< 5 $80 .893 .TN 035.

33

 

 

«huh is a hard.”
I
accounts for 015..”
One of 316 mm".
proportion 0119;?
countfics such 3..
(mm: for \\ c11-
Thc Korean 7
structure of xx 3.:
of Korean find:
total labor cw.
or even a “21;:-
CM} CMing\_
5W5 induflfi.
dismisgcd. A.
H3000 in 1.
fftumed 10 Q
uin’CL‘Cdcm;
training and
Slam“. .
fiOm 1987‘
mmance

}.n{.
kl L «.-
lkdppr

 

 

 

 

 

which is a burden born by employers due to a sharp rise in the unemployment rate,
accounts for other increases of share of statutory benefits in 1996 and 1999, respectively.
One of the most conspicuous characteristics of Korean labor cost statistics is that the
proportion of statutory benefits within the total cost is much lower than those of other
countries such as Japan and the U.S., let alone continental European countries which are
famous for well-designed welfare systems (Song, 1995).

The Korean financial crisis, which started in November 1997, enormously affected the
structure of wages and total labor costs for 1998 and 1999. Table 2-1 confirms the effects
of Korean financial crisis on the total labor costs. While the importance of cash in the
total labor costs dropped in 1998 and returned to over 70% in 1999 due to a wage freeze
or even a wage-cut, the importance of bonus payments in 1999 increased to 30.18% of
cash earnings, the highest percentage ever. Massive layoffs and dismissals in almost
every industry made employers pay severance payments to their employees who were
dismissed. As a result, the share of severance payments reached 21.55% in 1998 and
14.30% in 1999. In addition, voluntary benefits after the crisis were cut massively and
returned to the level in 1989 and 1990 (Bang, 2002). A drought of cash flow and the
unprecedented unemployment rate forced firms to reduce their expenditures on employee
training and recruitment by nearly half in 1998 relative to the previous year.

Statutory benefits consist of medical insurance, workers’ compensation, pensions
from 1987, fimds for pneumoconiosis in the mining industry for 1986-89, employment
insurance from 1995, and an allotted charge for promoting employment of the
handicapped from 1991. Table 2-2 shows sub-categories under statutory benefits and

voluntary benefits and the variations in their shares for 1986-1999. Until 1993, it was

34

 

wists £1151an"
sire ot'xtorlcrs c
bentiLs as medic-
1987.

Pensions acct-
their sum and sh
minim} benefits ‘
percent in 1999.
lPSdHnCt‘. which
Remit dCX t‘lx‘;‘
Writs of smut
“$1.13.

The [\‘nSion j
mllolcrs. mp,-
W and 1998 3
5% W 69.0 to
implement ( um.
the [on] P311011.

.1130 Table 3-;I
perm; housin:

mpl‘l‘t'c

- -Cuh._

 

Md] 8 life in“

workers compensation that accounted for the largest share of statutory benefits. Then the
share of workers compensation began to decrease and fell below 20 per cent of statutory
benefits as medical insurance did. The national pension plan in Korea was launched in
1987.

Pensions accounted for a considerable share (26.3 per cent) of statutory benefits from
their start and showed a continuous increase. Since pensions became the largest item of
statutory benefits in 1994, the share of pensions in the statutory benefits increased to 47.9
percent in 1999. Another important sub-category of statutory benefits is employment
insurance, which was launched in 1995. Although it started quite late, considering recent
economic development of Korea, employment insurance has become one of the important
sources of statutory benefits like medical insurance and workers compensation in the late
19805.

The pension premium currently is 9% of the total payroll, 4.5% for employees and
employers, respectively. The steep increase of share of pension of statutory benefits in
1994 and 1998 in Table 2-2 resulted from an increase of pension premiums from 3% to
6% and 6% to 9% (Kim, 2002). While employees pay 0.5% of their payroll, the
employment (unemployment) insurance premium paid by employers is 0.9~1.6 percent of
the total payroll, depending on the establishment size.

Also Table 2-2 presents twelve sub-categories listed under voluntary benefits over the
period: housing-related benefits, medical- or hygiene-related benefits, meals for
employees, culture, sports, and entertainment, supports for other insurance for employees

such as life insurance, expenditures for congratulations and condolences, supports for

35

 

 

 

. \ ll..J—.i\s\\:\\l\.ln,i\.flll|l
\.\.\»\ .r.\.\»\ r \.3\ .\\.3\ V\.\.\ ‘\~\.\ ~.\.\.\ Pl\~\»\ \33\ \\\.~\..\ \~.r,\»\ .(.I.\~\ r .r.\»\ (.I.\.\ \

.avavav - I.\x«.- ~ JI~H~U=Un~ >1L-:::~.u\/ uJ-sz \fi,:vu:::.f.. nu.- 33.5:v1.J-~.dlr--.Iv..\gv Zak-:5". ~-|Mrt rvxas=N

 

ooze—9:”;

 

av v.m d.m m.v d.m a.m m.m ad d.v <.Z ow. Wm m6 o6 3:22:80
Eons—3350
S _d_ a.w a.m dd a._ d; o._ a._ .<.Z od ..m dd a._ 8:ng $50
EoEEatoEm
ad v.m in ad v.0 d.v v.v Wm v.m .<.Z v.a ad md_ c.» Q .3.—cam
.2330
:m w.mm o. _ m mdm van o. _ m ddm van d.mv .<.Z oav 9mm mdv van £82
a.m mv mv 9m d.m 9m v.m dd ad .<.Z wd v.m ad ad 2.3m»:
@1860:
ad. ad «.3 2: «:2 oa— mg 0.8 vd. .<.Z v.a_ o.a dd_ Q: wcfisom
sol
332:3;
d.a_ ad. mi of ma 8ng
Eofizoifim
3386.83
v._ dd ad. ad ad ad ad 2 d._ 05 a0
Boga—afim
..o «6 ed v. _ 50280835
a.av adv cam a.am Nov adv d._m add add .<.Z m.mm ddm mad o._ 8280a
ad. 0. _ d mad a.vd m .n d wan adv can v.mv .<.Z 5m mam vdv m.vm coznmcoafioo
£3.83
a.v_ «.2 Wm. 2d _.dd odd 9% mdd add ....<.Z add mad add _.mv 8:83:—
.3802
Sena
Essen

aaad maa~ aaad baaN nan: vaad maad daa~ ~aa~ aaa~ ama~ Quad amad £22

 

 

22.83 .325m bs§o> 2a boaaam s 853333 .8 83% and 2.3

36

 

l l l I:
_ C N C v. C n C . C _ C . 3 ~ 3 ~ C .u.:...<<.:\.»
r. x. ch :8 ca xdn a... ch. :4... 0.9 3:: ....~.:..:...:
—A~.~m\l-.-a~\
NY?— N.3. v. \r. m??— h... NYC. V... 3;. pd; .<.Z .3. 3.»; Cs; Qda r..u.:Z.€.:.:>.\
N‘ssxx.::\.u.’.
C 3 V 3 C C C C C C ... C h C .i a u. a <.Z ..r Cfl n.» .1 u vat: :.../.

 

..o_na=a>< 82. £82: .. ”802
Boa some can“: .8 >532 ,3 304.3 actsm #60 ~33 emceecm 05 60.55

 

3 v6 d.v a.a o._ a.a a6 n._ 3 mascara—08m
ooaoEEm

md ed a.m 0m a.v a.a. ed _.a a.v coz8m>
Eggs—om

_.o dd a.a a.a _.o _.o _.o 3 _.o 28230
Wm— o.a a.a a.a add x: 9m _ a.a _ a.a ES..— fiancom
£5-82:

d.v_ d.a_ «:2 «.3 a._ _ a9 v.2 a.: md_ .<.Z d.v_ o.v_ o.v_ od— mflamvcoaxm
332%

cd ma ed ed a.a ed a.a d._ v._ .<.z dd ed ed fin mw£>am

 

 

37

mm?“ ”W
chiidcarc or nu:

.- J
mtmém W ‘“

make. some 4‘

llllllllm bent.
for emplO}L‘k‘ 3
on inn-firm l
.‘W ex en the!
Indicates 1h.
iii: insurance T

l‘nc upcs :
Hie increase 0
employees for

10 better find

employee savings, expenditures for employee schooling, intra-firm benefit fund,l3
childcare or nursery facilities, expenditures for employee relaxation or vacation except
payment for days not worked, and the employee stock-sharing program. Among the
twelve, some are worthy of attention. In spite of the recent decrease of its importance,
meals still remain the largest item of voluntary benefits. It accounts for about one-third of
voluntary benefits in the 19908. It is shown that housing—related benefits and expenditures
for employee schooling have been important benefits throughout the period. Expenditures
on intra-firm benefit fund became the second largest item of the voluntary benefits in
1999 even though they revealed large variations depending on the year. Finally, Table 2-
2 indicates that the share of employers’ support for other insurance for employees like
life insurance has increased in recent years.

The types and numbers of voluntary benefits have increased over the period as seen in
the increase of subcategories classifying voluntary benefits in Table 2-2. The needs of
employees for voluntary benefits become diverse as time passes. The unions are expected
to better find employees’ needs for voluntary benefits and will channel more effectively
the voice of employees to employers (Freeman and Medoff, 1984). The table implies that
the portion of "basic or peripheral’ benefits such as meals, congratulations and
condolences, and sports and entertainment, is declining in the 19905, while newer types
of benefits such as other insurances, schooling expenditures, and intra-firm benefit fund

are being created or are becoming important currently (Park, 2002).

 

‘3 Intra-firm benefit fund is aiming to increase the level of benefits for workers, by employers’ contribution
after collective bargaining between employers and unions. By law, the maximum limit of the fund is 5% of
profit before tax in the previous year. Because the operation of the fund is independent of employers, it is
expected that the fund is valuable to workers, especially when layoff or a delay in wage payment occurs. In
many cases, the fund is used to support schooling expenditures of workers’ families, expenditures for
congratulations and condolences, and loans to workers purchasing a house or stocks of the firms. The
number of firms having the fund is 862 and the total number of the funds are estimated to be
3,428,400,000,000 Won, 2.85 billion in U.S. dollars, in 2001 (Dong-A 11b0, July 25, 2001)

38

In on

n

locresr

on \015:

In order to examine the effects of trade unions on non-wage benefits, this chapter
focuses on voluntary benefits. Most studies regarding the union effect on benefits focus
on voluntary benefits except for a few studies examining statutory benefits along with
voluntary benefits. Regarding severance payments, Japan is said to have a considerably
larger proportion (4.01 per cent) of severance payment than the United Kingdom,
Germany, France, and the United States, within the total costs in 1992 (Hart and
Kawasaki, 1999). The share of severance payment in Korea is much larger than in Japan,
as well as in other industrialized Western countries. However, severance payments in
nature are determined by the wages and tenure.

Severance payment in Korea (and in Japan) has characteristics of both statutory and
voluntary benefits. On one hand, severance payment may be classified as statutory
benefits in that basic formula determining the amount of severance payment for
employees leaving firms is set by labor laws. However, tenure is a critical factor
determining severance payment because monthly regular wages are multiplied by tenure
(Labor Standard Act, Article 34). On the other hand, severance pay share the
characteristics of voluntary benefits in that many employers pay higher severance
payment beyond the lower bound, set by the Labor Standard Act as a result of collective
bargaining between employers and unions. No matter whether severance payment should
be regarded as statutory benefits or voluntary benefits, a remarkable aspect of severance
payment is that severance payment may work similarly to voluntary benefits because
Korean severance payment is very close to the European retirement-related contributions

by employers, which is classified as non-statutory benefits (Hart and Kawasaki, 1999).

39

 

Tnerefore. it i5

061161.18. 111 K0?

lmmn: l r77:

l'sing unp
benefit diff-cm
Oil‘lt‘fi Writer:
Pffimium is l:

\\

mien streets.
premiums. lift
B) using
ilk difference
ifif‘wr‘cn Ulllt)‘
hamlets are
5) lhe plan. ,1
included in h
the Plan IS on
A study
on bent»

ills ,-

Therefore, it is difficult to say that severance payments are one of pure types of voluntary

benefits in Korea.

Imp_ortant Empirical Studies

Using unpublished 1972 BLS data, Alpert (1982) found small but significant union
benefit differentials for manufacturing production, non-manufacturing production, and
office workers. Freeman and Medoff (1984) suggest that the size of the union benefit
premium is larger (20% ~ 30%) than the union wage premium. In addition, distinct
union effects on fringe benefits in the U.S. are found especially in pensions, holiday
premiums, life insurance, and medical insurance.

By using the April 1993 CPS Employee Benefits Supplement, Budd (1998) shows
the difference in fi'inge benefits especially in employee’s retirement and health insurance
between union members and nonunion members. He states that only 41.8% of nonunion
members are covered by a retirement plan, while 79.9% of union members are covered
by the plan. Also while 85.8% of workers covered by collective bargaining agreement are
included in health insurance, the percentage of nonunion members who are included in
the plan is only 57%.

A study by Renaud (1998) in Canada adds further evidence of the impact of unions
on benefits and total compensation. His results from micro data from the Canadian
General Social Survey (GSS) of 1989 show that the union effect on total compensation
12.4%, while the effect on wages is 10.4%. However, the size of benefits in total

compensation in the sample is only 6%. Surprisingly, the size of the union benefit

40

 

dirlércntials in it
found in puhiit‘ ~
The size 01 ‘.

diScmercd that ‘

 

suhsnnplcs. the
value estimated I
'he size 01 uniur

\lhilc those 5
01 benefits or or
Nance are i

carnpensation

ieui

ile. licndr
ili‘stion of M

monized and

l 1&1an

. is th.

 

differentials in his results turns out to be 45.5%. Union effects on the level of benefits are
found in public sector, also (Mitchell and Smith, 1992; Cohen and Cohen, 1998).

The size of the union effect on benefits is still unclear. Belman and Heywood (1990)
discovered that when using two separate probit equations for the union and non-union
sub-samples, the probability of receiving fringe benefits is approximately a third of the
value estimated from a single equation. With these results, they raise the possibility that
the size of union effect on benefits by previous studies may be considerably exaggerated.

While those studies mentioned above mainly focused on the union effects on the level
of benefits or on the probability that a specific type of benefits such as pension and health
insurance are provided by employers, union effects on the share of benefits in total
compensation is another issue. Several studies (Lester, 1967; Schiller and Weiss, 1980;
Feuille, Hendricks, and Kahn, 1981; Alpert, 1982; Woodbury, 1983) raise an important
question of whether differences in wage/benefit mix exist in total compensation between
unionized and non-unionized sector and, if any, why does the difference appear? Their
finding is that unionism is positively related to the share of benefits taken in total
compensation 1984 (Woodbury, 1983; Fosu, 1984). For example, full collective
bargaining coverage is reported to increase the benefit share of total compensation by 1.6

to 2.2 percentage, compared to nonunion firms (Woodbury, 1983).

Reasons for Lager Effect of Unions on Benefits than on Wagg
What can explain the phenomenon that unionized firms pay more of their
compensation in the form of benefits? One explanation comes from a monopoly model of

unions regarding bargaining power. Unions with strong bargaining power are able to

41

 

induce the cm?
Hendricks. and
. lationship 561‘

A second a;
colicczhc lung;
larger pmpimh'
oilsrs mini 1m
young margins
benefits using ll
firms. the uniun
“Plum in st
enhance the sha
V‘l‘ifir. are inclirl
“LES.“ h SCL‘I
fiii‘trent from 1

Wages and hem

 

mums bcn “
CII'

fe

 

“ SlUdlt‘S ex.

 

induce the employer to pay not only higher wages but also more fringe benefits (Feuille,
Hendricks, and Kahn, 1981; Freeman and Medoff, 1984). This idea assumes a positive
relationship between union effect on wages and on benefits.

A second explanation for the larger effect of unions on benefits than on wages is that
collective bargaining is driven by the preferences of older workers who generally seek a
larger proportion of benefits (Freeman & Medoff, 1984). While non-organized firms
offers minimized benefits and maximized wages that are necessary to attract relatively
young marginal’ workers with little seniority, unionized firms have to offer better
benefits using the influence of unions which are based on the older workers. In unionized
firms, the union’s bargaining goal is heavily influenced by the preferences of the median
employee in seniority distribution. Freeman (1981) offers a reason why union may
enhance the share of benefits. Trade unions, which are likely to be operated by median
voter, are inclined to follow older workers with longer tenure, who prefer benefits to
wages.14 It seems that this explanation, which is based on the median voter model, is
different from the first explanation in that it assumes a negative relationship between
wages and benefits.

A third explanation points out that union leaders may have their own incentives to
increase benefits more than wages (Lester, 1967; Mabry, 1973; Freeman, 1981). While a
few studies examining union wage effect mention the role of union leaders, the literature
on union benefit effect pay more attention to the 'role and separate preferences of union
leaders. Union leaders are said to have more interest in fringe benefits in that the

administration of such benefits may result in a bureaucracy, which tends to helpful for

 

“ In spite of an explanation by Freeman (1981 & 1984), a piece of advice by Lester (1967), Schiller and
Weiss (1980), and Woodbury (1983), arguing that a theory about why trade unions might to alter wage-
benefrt mix is still needed, is still correct.

42

 

them to he nu.
501nm by he
time to hide c\'

A lounh ex,"
olfringc hand
that empltw cr-

Shlllt‘tl \\ 01h L'l

generous hen

micanis l,\

53an

In (“de

(‘3?th app?

mam'llnil‘uf

them to be re-elected. Also it could be possible for union leaders to maintain labor
solidarity by keeping up with the level of wages across firms or industries and at the same
time to hide extra rent as a less visible form of benefits rather than wages (Mabry, 1973).
A fourth explanation offers that employers also may be involved in the recent increase
of fringe benefits in the compensation packages in many countries. Especially, it is likely
that employers offer various and generous benefits for older workers in order to retain
skilled workers and reduce turnover rate where employers invest significantly in the firm—
specific training of their employees (Becker, 1964; Mabry, 1973).15 On the contrary, it
might not be good for employers expecting to attract marginal workers to provide more
generous benefits because fringe benefits are less visible than wages to outsiders like job

applicants (Mabry, 1973).

Conditions to produce union benefit effect in Korea

In order to assure whether the previous studies on the union benefit effect in the U.S.
can be applied to Korea, we must ask whether union effect on fringe benefits in Korea is
meaningful. The importance of benefits, that is, the size of fringe benefits in total
compensation needs to be considered.

In Korea, the share of fringe benefits in total compensation appears lower than in the
U.S. (MOLb, 1999). However, the importance of the union effect on fi'inge benefits
should not be overlooked for two reasons: a poor national welfare system and

government intervention into wage bargaining between employers and unions.

 

'5 It was Henry Ford, Welfare Capitialism, and Human Relations Movement in the 19203 that recognized
the importance of benefits and argued to make an effort to provide benefits for workers in the United States
(Davis, 1986; Katz and Kochan, 1992).

43

1

...—«I
234

in [11$

(14
q)

 

 

Sumter} H 3
Benefit (3.1 "

Vole-7L1!) : (l
Benefit 1 15 ’

 

Not: Values in {up
mueslapannr L
Sine Korea 1 at»

Korea has. :1
la'ole 3-3 or

he '

regud lo the shl

Among the St“.
10W in Kore.
llT‘RIon ol \oln-

combined “ill.

[mm elleel t

gill Cmmem n, l

hlh ‘
elm 1mm 3

iii -
‘srenlnls a-I

 

 

Table 2-3. The shares of statutory and voluntary benefits in total labor costs per employee
in many countries

 

 

U. K. A ustria German Belgium Japan Korea
Statutory 14.3 20.5 20.4 23.4 7.9 4.6
Benefit (84.6) (71 .7) (88.7) (92.8) (73.8) (33.1)
Voluntary 2.6 8.1 2.6 1.8 2.8 9.3
Benefit (15.4) (28.3) (1 1.3) (7.2) (26.2) (66.9)

 

 

Note: Values in parentheses are the share each item in total benefits. The reference year for European
countries, Japan, and Korea is 1984, 1988, and 1993, respectively.
Source: Korea Labor Institute (1995) and Hong (1996).

Korea has a much more limited national welfare system (Song, 1995; Choi, 1992).
Table 2-3 presents a simple comparison between many countries including Korea, with
regard to the shares of statutory and voluntary benefits in total labor costs per employee.
Among the seven countries, the share of statutory benefits in total labor costs is the
lowest in Korea, indicating an extremely weak social welfare system. On the contrary, the
portion of voluntary benefits is quite higher in Korea than in other countries.‘6 That fact,
combined with no labor party or pro-labor government, supports the importance of the
union effect on benefits. The lack of a strong social welfare system supported by the
government makes the importance of welfare system relating to employment relations
within firms greater (Hong, 1996). Also the fact excludes a possibility that union benefit
differentials are a poor indicator to supplement union wage differentials because union

effects on benefits are realized through welfare legislation, not through collective

 

'6 The shares of statutory and voluntary benefits in total labor costs for Korea in 1993 in Table 2-3 are
exactly same as the shares in Table 2-2, earlier. Table 2-2 implies that the share of voluntary benefits in
total labor costs and in total benefits is higher in 1993 than any other year in the period of 1986-1999. In
spite of some exaggeration in Table 2-3, the shares of voluntary benefits in total labor costs in Korea are
still higher. It is because the shares of voluntary benefits in total labor costs are larger than the shares of
statutory benefits in every year between 1986- l 999.

44

 

brgaining. as in 7:
tin: the toiee ch.
in addition. .
enterprise letel "
oilered h} the :
concern about it
unions. and so \
increase ting-cs 4
Various fringe h
regretted as in t“
Giten the h;

031 benefits \\ h,
Effect on “also.

ha“

locus in n:;

C0

moonshine ,

“he“ fOllt .-

ttthzmge fm t

 

bargaining, as in the Scandinavian countries (Freeman, 1994). In this vein, it is expected
that the voice effect of unions in Korea is greater on fringe benefits than on wages.

In addition, since 1987 the government has intervened in wage bargaining at the
enterprise level through forcing firms to follow so-called ‘one-digit wage guideline’
offered by the government (Choi, 1992). The government continued to have a deep
concern about inflation, which may be caused by a steep increase in wages sought by
unions, and so strongly prefers an increase in welfare program or fringe benefits to
increase wages (Park, 2002). Consequently, the unions have sought to increase or create
various fringe benefits, which are free from the government wage guidelines, and not
regulated as in the case of wages (Choi, 1992; Bae, 1995; Park, 2002).

Given the historical conditions mentioned above, I hypothesize that the union effect
on benefits which have existed since the Great Labor Offensive in 1987 is larger than the
effect on wages as in the U.S. Determining the size of the union benefit differentials will

be a focus in my study in the near future.

Compensating wage theory expecting no union effect on benefit

When following the compensating wage theory which implies lower wages in
exchange for more generous benefits (Schiller and Weiss, 1980; Rosen, 1986; Olsen,
1994 & 2002), positive union effects on wages may result in negative union effects on
shares (and/or levels) of benefits in total compensation, holding other factors constant.
Assuming a perfectly competitive labor market with perfect information and no
uncertainty, all workers are to be paid by their marginal product. If wages are the only

form of compensation, the same wages will be given to all workers having the same

45

ability or mg?
I
in the eonrpetit:
I
compensation l}
|
tinns littering h
I
the lions otter:-
costs. Alter all.

the long run. l.‘

oen [hottgh tin.

writers in the
ton‘uol the lex
“agelenelit rr.
Thus. mun)
indispensohle :
etie‘

. \l Oil 10m] \

c0mpenSallt‘tn 1

dining “11h lt'

 

 

 

ability or marginal product. Even though fringe benefits are introduced into some of firms
in the competitive labor market, compensating wage theory predicts that the level of total
compensation for all workers having same marginal product should be the same between
firms offering benefits and other firms offering no benefits. If not, the theory expects that
the firms offering more generous wages due to benefits will face the liability of labor
costs. Afier all, the compensating wage theory says that there should be no difference, in
the long run, in total compensation for all workers having the same marginal product,
even though there might be a different wage/benefit mix across firms or preferences of
workers in the short run. It also implies that union benefit effects do not exist when we
control the level of wages, even though trade union may impact on the mix of
wage/benefit mix.l7

Thus, many studies on the union effect on fringe benefits support the belief that it is
indispensable to consider the union effect on fringe benefits in order to have the union
effect on total compensation even though the share of fringe benefits for workers in total
compensation is much smaller than that of wages. Fortunately, the number of studies
dealing with total compensation, not just wages, has been increasing. The trend of studies

on the union effect on wages should be to include fi'inge benefits and other non-pecuniary

benefits, which have not yet been explained.

Characteristics of Benefits in Korea

 

The most conspicuous characteristics of fringe benefits in Korea is that statutory

benefits cover major benefits but deprive employees and unions of bargaining for the

 

'7 In reality, the compensating wage theory assuming competitive labor market is challenged by existence
of preferential tax treatment by government (Gentry and Peress, 1994), different administration costs for
benefits according to firm size, union bargaining power to boost both wages and benefits, and so on.

46

lerel of those 1
are the most it
benefits in the
prefe ences of
ogenerall} 1»
Clark 198m.
B)‘ exeludin
employers and
1th5 to be u
hoelits in gen
”flit \‘oluntnr
film benefit l‘u
The ml or to]:
Hm mwgh lt'
Dioxide Val'ltlu
Willis that On
[We in Older
filttllhcr flit“
lS Slider“ in l
retired mm“

:1

ll~ ,
Ni 53 H.
maid “12h c-

t‘.

level of those benefits with their employers. For example, pension and health insurance
are the most important two benefits, which are likely to be related to union effects on
benefits in the United States. Also the two benefits are very likely to relate to the
preferences of older and skilled workers. Also eligibility for the magnitude of the benefits
is generally associated with seniority or job tenure in their firms (Mabry, 1973; Allen and
Clark, 1986).

By excluding pension and health insurance from the private bargaining between
employers and unions due to statutory benefits, the role of seniority, age, and skill is
likely to be reduced with regard to benefits. It may undermine the union effects on
benefits in general in Korea because older skilled workers especially do not always favor
major voluntary benefits such as meals, housing, schooling expenditures, and an intra-
firm benefit fund than younger workers do. The situation would be similar to employers.
The major voluntary benefits cannot be regarded as a kind of deferred wages in the long
run enough to retain skilled workers. It may reduce the incentives for employers to
provide various and generous benefits. In fact, a study on 20 types of voluntary benefits
reports that only 19% of voluntary benefits are related to the clause of a minimum year of
tenure in order for the employee to be eligible (Hong, 1996).18

Another factor discouraging unions’ responsiveness to the preferences of older workers
is evident in Korea. In the U.S. some unions such as the United Mine Workers allow
retired workers to participate in internal decision-making processes, resulting in a greater

voice for older workers in determining the contents of benefits (Farber, 1978; Allen and

is The study by Hong (1996) is based on small size of sample (N=240) in 1993. Thus, the results should be
Interpreted with caution.

47

 

...,=
r“.

.a
.v.

Onl-

[Sm

Clark, 1986). Under enterprise unionism, retired workers are not allowed to be members

of unions at all. Also lower union benefit effects in Korea are expected.

Previous studies in Korea

Only a limited number of studies (Choi, 1992; Song, 1994, Park, 1996, Hong, 1996;
Choi, Y.S., 2000) on union effects on benefits are available in Korea. This scarcity of
research on union effects on fringe benefits results from the lack of data. Compared to
data availability for union wage effects, it is difficult to find data available regarding the
union effects on fi'inge benefits (Song, 1994).

Therefore, due to the lack of appropriate data, Song (1994) adopts an indirect way to
test union effects on benefits, by examining the degree of workers’ satisfaction regarding
fringe benefits or the degree of resource investment in enhancing working conditions
including fi'inge benefits by unions in order to estimate union effects on fringe benefits
(Song, 1994). Choi (2000) attempts to find union effects on the level of benefits from
433 firms listed on the Korean Stock Exchange. Both studies found no significant
difference in the fringe benefits between union and non-union members (Song, 1994;
Choi, Y.S., 2000) thus making it impossible to determine the differences in fringe
benefits between union and non-union members.

Using two kinds of data, aggregated industry-level data and establishment-level data,
Choi (1992) found significant positive effects of unions on the level of benefits and the
share of benefits in total compensation in 1988 from establishment-level data, though no

union effect for 1969-1989 from time-serious data based on aggregated industry-level is

48

atuhhh.l
nunherol“
‘henurnher
it} the lotul
related to th
ol'arninimt
WS- Peri.
Exchange. :
sized tinns.
lhe preti
310i lllClUdt‘
If“). No 1
lenelits. o
c\lrnines z
u Lahles sit

Union l‘cdcr

“if ma

[f5 .33
@133 (h

it: of it».
1“; mil Lt
-F Vinol'lel‘i

5': the at 3:

< u

will - t
., 74.1
1m,

t

i

Q7

l

available. Hong (1996)l9 examines variety of benefit types as measured by the total
number of voluntary benefits and the degree of the exclusivity of benefits as measured by
the number of benefit types without the clause of a minimum of a year of tenure, divided
by the total number of benefit types. He found that the existence of unions is positively
related to the variety of types of benefits and the percentage of benefits without the clause
of a minimum year of tenure to be available to employees in the total number of benefit
types. Park (1996) used data from large firms, which are found in the Korean Stock
Exchange, and shows a limitation that her findings cannot be applied to mid- or small-
sized firms.

The previous studies on union benefit effects reveal many limitations. Some studies do
not include the level of wages as independent variable in the benefit equation (Choi,
2000). No study considers the possibility of an endogeneity problem between wages and
benefits, which will be considered by adopting 2SLS in this study. In addition, no study
examines any differences in benefits within the unionized sector by using additional
variables such as union organizing rate and the individual union’s affiliation with national

union federations as seen in this study.

Research Questions

The main research question in this study is whether or not the union effect on fringe
benefits exists in Korea. Many studies on union wage effect and union benefit effect

implies the existence of the union benefit effect. A few even argue a larger union effect

 

'9 One of feature in Hong (1996) is that it focuses on organizational characteristics such as whether a firm
is monopoly in the industry, whether productivity for blue-collars and white-collars can be measured, how
large the added-value per employee is, whether the firm is belong to Chaebol, and whether the firm has
Specialized department regarding personnel management or human resources.

49

on benefits
negatite rel.

Armm
the suit-cote
and intro-tit

lhird. t't
‘ougainin;
Ol'glni/ing
limit err
national u
determine
Compositi

Bargain“
Pittman i

mplt‘tc.

a

L A4
>—_‘
,—*

iOn
he be

Fill

t .
10\\'y

on benefits than wages (Renaud, 1998). The compensating wage theory predicts a
negative relationship between union and benefits.

A second question raised by this study is how unions affect the probability of each of
the sub-categories provided by employers such as meals, housing, schooling expenditure,
and intra-firm benefit fund.

Third, this study also will examine how two more critical variables, related to union
bargaining power, affect benefits as well as the union dummy variable does. The union
organizing rate, defined as the number of union members divided by the total number of
regular employees, in the establishment and enterprise union’s affiliation with either two
national union confederations (F KTU/KCTU) are the two variables, added in order to
determine whether any variations in level of labor costs including total compensation and
composition of labor costs within unionized establishments are present.

Bargaining outcomes vary according to union bargaining power. One of the most
popular indicators showing union power is union density measured by the percent of
employees who are union members in the bargaining units (Feuille et al., 1981). In
addition to union density, many variables measuring union structure and union rivalry
have been used in previous studies (i.e., Feuille et a1, 1981; Fiorito and Hendricks, 1987).
Following a line of previous studies, this study expects that union density may affect
union preferences towards levels and shares of benefits in total compensation (Alpert,
1982; Woodbury, 1983; Mitchell and Smith, 1992; Kahn, 2000).

Two national labor camps in Korea, F KTU and KCTU, are know to be very different
from each other in terms of bargaining strategy and the type of mobilization of union

membership. While the FKTU has depended on more peaceful negotiations without the

50

rank-md-tile
shun to text
Song. 109g;
rrxed and
ditlerenee lg
ciroetensti
W are in

Hill mos
ll‘i’lleg Or 1“
leaders and
dlrrllulgd U
Chamclt‘tisl
might be St

Ollgml L‘ '01
0bent‘lits
Commit

arm1

rank-and-file union members’ visible collective actions such as strikes, the KCTU has
shown to resort to more massive collective actions from the union membership (Kim and
Song, 1998; Kwon, 1997). As a consequence, most of the union leaders who were
arrested and sentenced were affiliated with the KCTU rather than the FKTU. The
difference between the two national confederations continues in terms of the individual
characteristics of the firll-time union leaders working for each. While the union leaders,
who are in their fifiies and are high school graduates with longer tenure, dominate the
FKTU, most of the full-time union leaders in the KCTU are university graduates in their
thirties or forties but have shorter tenure (Lee et al., 2001). Regarding benefits, the union
leaders and rank-and-file union members place more emphasis on benefits in the FKTU-
affiliated unions than in the KCTU-affiliated unions (Kwon, 1997). These different
characteristics of the unions under the two national peak organizations indicate that there
might be some different levels of total compensation and labor costs and the composition
of total compensation achieved between the FKTU and KCTU. With respect to the level
of benefits and total compensation, it is expected that the average level of benefits and
compensation will be higher in the unions affiliated with the KCTU than in the unions
under the FKTU, because the KCTU-affiliated unions are more responsive to the voice of

rank-and-file union members and more militancy in bargaining.

51

ln dealir‘
htrgaining (
inditidunl-lt
used for ur:
tanthles. et
letel data it
and anon.
“ll-1011 beer
lh'ee kinds

pm‘t'l'tled in

l. lndi

CHAPTER THREE

DATA

In dealing with the four empirical parts regarding union effects on economic
bargaining outcomes for 1987-1999, this study is based on the three kinds of data:
individual-level data from the Basic Survey on Wage Structure (BSWS), which will be
used for union wage effects on the distribution of wages and union effects on four
variables, education, experience, tenure, and sex, in determining wages; establishment-
level data from the BSWS, which will be utilized for union effects on wage dispersion;
and establishment-level data from the Enterprise Labor Cost Survey (ELCS) used for
union benefit effects. This chapter explains briefly the nature and characteristics of the
three kinds of data for the four empirical parts. Extra information regarding data will be

provided in the relevant chapters later.

1. Individual-level data from BSWS

The data set used here comes from the Basic Survey on Wage Structure (BSWS),
which has been conducted by the Ministry of Labour (MOL) in Korea every year since
1972. The BSWS consists of individuals who are hired in at approximately 2,700-5,500
establishments, drawn by a stratified random sample method, with at least ten regular

Workers (MOLa, various year). The data source does not include agricultural workers,

52

“0th
WThC

lh
intern
tenure
fans.

l’u’lltf‘fl

workers in foreign organizations, and civil servants including polices, armed forces, and
workers in public educational institutions, school, hospitals, and libraries.

The BSWS data as originally individual-level data contain decidedly in-detail
information on individual workers such as wages, hours worked, sex, martial status, age,
tenure, career, occupation, and level of skill held by workers, and on the characteristics of
firms, which respondents work for, such as region, industry, firm size, and existence of
unions (Lee & Kwon, 1995). In addition, the BSWS data are highly comprehensive in
that the number of observations ranges from 383,000 at the minimum to 585,000 at the
maximum. This study uses a randomly extracted 10% of the original number of
observations from the BSWS. Finally, the annual BSWS is the only data in Korea, which
may satisfy researchers who have interest in a period, not limited to a specific year, with
regard to union wage effects.

For these reasons, most previous studies (Bae, 1990; Lee & Kwon, 1995) on union
wage effects depend on the data set: 1976 (Song, 1989), 1980 (8.1. Park, 1984), 1986 (Y.
B. Park, 1993), 1987 (Jeong, 1991), 1988 (Bae, 1990; Song, 1994), 1989 (Jeong, 1991;
Chae, 1992), 1994 (Lee & Kwon, 1995), and 1987-1999 (Shin, 2001). Thus, it is
possible to utilize most of the previous studies on the same topic because most of them
are based on the data.

This study has selected and uses only regular employees, ages 19-65. Also the data
used are limited to employees who have non-professional and non-managerial
occupations, since the membership of trade unions is unlikely to be opened to

professional and managerial occupations.

53

 

 

or 3“"
chapter Ml
mot“
‘t‘eeLt W
to determlm

rtonilnlon set

The P“
lenieen Unl'

equalizing 9‘

rented tariul

sorting lndl‘

esnhiishmt’”
hung at leit
Also inlonna
a‘mge \‘flfli

when tlilll

latte Within
approxi

1115 desirahle

The annual individual-level data from the BSWS for 1987-1999 will be used in the
chapter dealing with the question of how the size of union wage premium changes in the
distribution of wages. Also the data will be used in the chapter examining how union
affects four values including education, experience, tenure, and sex, which are expected
to determine wages. In the chapter, the annual data will be divided into union and

nonunion sectors.

2. Establishment-level data from BSWS

The part dealing union effects on wage dispersion within establishments aims to
answer the questions, whether there is a significant difference in wage dispersion exists
between unionized and non-unionized sectors and, if any, how big the union wage
equalizing effects are. Those questions need establishment-level data including wage-
related variables. Therefore, I constructed establishment-level data from the BSWS by
sorting individual-level data from the BSWS after considering the locations of
establishments and the establishment identification numbers. After that, establishments
having at least five employees who are found in the BSWS as observations, are selected.
Also information relating wages within are calculated, such as the average of wages, the
average variance of wages, the percentage of females, salesmen, skilled, technician
workers within establishments as well as the variance of education, experience, and
tenure within establishments.

The approximate number of establishments constructed in a year is 2,000. Even though

it is desirable to construct establishment-level data for every year for the period of 1987-

54

1999, r

eoddl

meBS‘

'1)

1999, establishment-level data for only five years, 1987, 1990, 1994, 1997, and 1999
could be constructed because the process of constructing establishment-level data from

the BSWS is considerably time-consuming.

3. Establishment-level data from ELCS

The data used for union effect on benefits are the Enterprise Labor Cost Survey
conducted by the Ministry of Labor.

The survey has been conducted annually since early 19803 and contains items inquiring
about the existence of unions, the amounts of obligatory and non-obligatory benefits paid
by individual firms, etc. The population of the survey is all firms hiring at least 30
permanent employees (10 employees from 1997) and the size of the sample for the
survey is 2,406 in 1997, for example. Data from the Enterprise Labor Cost Survey
(ELCS) have strength, compared to other data set used in other countries. First, the data
are not limited to a specific type of benefit such as pension and health insurance or a part
of benefits, which employers offer to employees, but cover comprehensively all
expenditures regarding benefits including statutory benefits, voluntary benefits, and even
training-related expenditures and recruitment expenditures. It may allow us to observe the
union effect on the entire features of benefits, not on specific benefits, and not to
overlook any misunderstanding resulting from a potential relationship between a specific
benefits examined and other benefits, when focusing on a part of the benefits.

However, two important limitations of the data are present also. Compared to

individual-level data like the CPS data, one of the limitations of ELCS data in this paper

55

 

is that the
amount
human cap?
mitigated h.
threedigit i
tnd inform.
limitation c
hiahlhl‘tmt‘
explained it
Beeause r.
Unfortunate.

establishmo

is that the data in the sample lack information on human capital such as experience,
education, and tenure. It is expected that the weakness of a lack of information relating to
human capital, except the percentage of female workers in the establishment, might be
mitigated by adding the average values of education, experience, and tenure for 169
three-digit industries from the Basic Survey on Wage Structure (BSWS) in relevant years
and information regarding employee wages into the model in this study. The second
limitation of the data is that the data do not include information on trade unions in
establishments. The process of adding union-related variables into the ELCS data will be
explained in the relevant chapter later.

Because no establishment-identifying information for data from 1987-1994 is available
unfortunately, the chapter uses only data from 1995 to 1999. The total number of

establishments used in the analysis is 8,479.

56

f. lllll'thlUL'li

The Chop

the results an,

3' Wild and

like r

“Platoon

CHAPTER FOUR

UNION EFFECT ON WAGES

1. Introduction

The chapter is organized as follows: Section 2 describes the estimation methods,
especially quantile regression (QR), and then describes the data set. Section 3 introduces

the results and explains the implications from the results. Finally, section 4 concludes.
2. Model and Data

ModeLmd Estimation

Like most of studies on the union wage effect, this chapter adopts a Mincer-type
wage equation (Mincer, 1974) and union status, U,, is added to the equation as one of the
explanatory variables. Other covariates, X,-, are a vector of individual’s years of education,
sex, marital status, experience, squared experience, tenure, and squared tenure and the
characteristics of establishments such as firm size, industry, and region. The dependent
variable, In W,-,, is the logarithm of monthly gross wages divided by hours worked. The
main goal is to estimate the parameter associated with union status. Thus, our wage
equation is

In W" = Xnflr+ UNION,-,,Br + 8,1. (4.1)

57

 

If?
Lil-.1

re;

hie:

Cr] i| -

The model of the ordinary least squares (OLS) assumes that the regression coefficients
regarding union status are constant across the entire conditional wage distribution.

The quantile regression model, also known as least-absolute value model (LAV) or the
method of minimum absolute deviations (MAD), is aimed to minimize the sum of
absolute residuals rather than the sum of the squares of the residuals as in ordinary least
square regression (STATA, 1999). While the union effect by the OLS is an estimate on
the mean wage, quantile regression does not assume that union effects are equal over all
quantiles of the wage distribution. The quantile regression assumes that the covariate
parameters determining wages may vary over all quantiles of wage distribution.

This characteristic of quantile regression compared to the OLS is more suitable to the
purpose of this study on the union effects on the wage inequality because prior studies
report that unions affect wages for employees in the lower half of the distribution more
than the wages for employees in the upper half of distribution. In addition, quantile
regressions have some other advantages compared to the assumptions behind the OLS.
The quantile regressions relax the assumptions by the OLS, in which the errors are
identically distributed and normally distributed, by admitting that the coefficients of
covariates determining wages may vary over all quantiles of distribution. Also a criterion
minimizing errors quantile regressions, based on the sum of the absolute values of the
errors, is likely to be better than quadratic-error-minimizing criterion of the OLS in that
outlier observations in the quantile regressions are not given as much weight different
from in the OLS and, also, the quantile regressions deals with statistical
heteroscedasticity (Buchinsky, 1998; Schultz & Mwabu, 1998). For these reasons, the

number of studies adopting quantile regressions in the area of wage inequality has been

58

 

or,

(a

uh.

All

ofL
t.“

0D

tesi

increasing (Buchinsky, 1994; Chamberlain, 1994; Poterba & Rueben, 1994; Schultz &
Mwabu, 1998; Pizer, 2000).

From the wage equation (4.1), it is assumed that the qth quantile of the conditional
wage distribution is a linear function of the individual attributes (XH):

Quantq (In W), |)(,-,) = Xuflq, , + UNION), q, r. (4.2)
Quantile regression estimates can be obtained where the slope coefficients B, for the qth
quantile are gained from the following minimizing problem
minl/nZ[ql(y,-2B'x,~) + (1- q)l(y,~<B'x,- )] l y.-|3'x,-| (4.3)

where 1(A) is the indicator firnction that equals one if event A obtains and equals zero
otherwise, and n is the sample size (Koenker & Basset, 1978). For instance, if y,2[3'x,- , the
deviation is positive (at least above the fitted line), and is weighted by q. If y,-<[3'x,-, the
deviation is negative (below the fitted line) and is weighted by 1- q. Thus, the quantiles
other than the median are estimated by weighting the regression residuals and the weight
on positive and negative residuals depends on their location relative to the median
residual. In the process of minimizing the sum of their absolute values, the deviations
above are treated as nonnegative numbers (Schultz & Mwabu, 1998; STATA, 1999). In
the context of this study, the eqation (4.3) can be rewritten with

minl/n2[ql(y,-,2 Xnflq, , + UNION” w) + (1- q)l(y,-,< Xnflq, , + UNION” q,,)]

* | y.,- Jr.,/3,, - UNION.,5.,,,I. (4.4)

In most cases, the standard errors in quantile regression are computed according to the
method by Koenker and Bassett ( 1982). However, it has been pointed out that these
standard errors are downward-biased because of heteroscedasticity of the disturbance

terms. A bootstrap approach to estimate the standard errors, allowing for bootstrap

59

 

“her
from
hoot
USlng

To

sampling from the original sample with a replacement, has been used in previous studies
(Chamberlain, 1994; Buchinsky, 1994; Schultz & Mwabu, 1998; Hartog et al., 2001).
After estimating coefficients at each quantile based on a linear programming algorithm,
the means of the estimated bootstrap coefficients are used to calculate their variance,

V(,6,), and the associated standard errors, using the equation

Vw.)=n/szwr””- MM)- 13.): (4.5)
where B = number of bootstrap replication samples, b = l, ..., ,B/b) = parameter estimate
from the bootstrap replication b; and ,6, = the mean of all parameters obtained from the
bootstrap replication samples. The bootstrap standard errors in this study are calculated
using 20 iterations following Poterba & Rueben (1994) and Hartog et al. (2001).

To discover how the unions affect the entire distribution of wages among union
members compared to the distribution among nonunion members, this study first
estimates the union effects conditional on the expected log wage by the OLS as a
baseline. Then, the estimates of the union effects for the median wage earner and other
deciles in the distribution of wage residuals, such as the 10th and 90‘“ percentile, are
obtained from quantile regressions. Finally, a full quantile regression at every quantile,
which is expected to reveal how the union effects on different deciles vary, will be

conducted.

Data and Variables
As mentioned in the chapter briefly explaining the nature of data earlier, this chapter

uses dataset from the Basic Survey on Wage Structure (BSWS)2°, which has been

k

2° They had been termed the Occupational Wage Survey (OWS).

60

collected 1
Mlllidl‘itll'l
hiring 31 le
police. an
included i
estrhlishm
determined
esoblishm:
approximut
esuhlishmt
CSldl‘llShlllt
emplotees.
employees .

The BS“
heause thr
“Oiled, SC)
“(them all
WI}; fir
lnlllllllSlriaj
lilierangl
Willem
Ofobsm'allt

:t
ills

.. “’9 Joe-
W5, Pt «d

l
l

collected by the Ministry of Labour (MOL) in Korea every year since 1972.21 The
population of the BSWS is all kind of workers working in the private establishment
hiring at least ten or more regular workers. Therefore, civil servants such as armed forces,
police, and employees in public educational institutes, libraries, and hospitals are not
included in the BSWS. In addition, agricultural workers and employees in foreign
establishments are also excluded. Every year the number of target establishment are
determined, based on a stratified random sample method with consideration of industry,
establishment size, and region. The number of establishment participating in the BSWS is
approximately 2,700-5,500 sample establishments (MOLa, various year).22 If an
establishment is determined to participate in the BSWS, the numbers of employees in the
establishment are chosen, based on the establishment size. From wage records for all
employees, which is kept in the establishment, all information on the number of
employees are collected.

The BSWS data have been used by Korean researchers examining union wage effect,
because they contain rich information on individual workers such as wages, hours
worked, sex, martial status, age, tenure, career, occupation, and the level of skill held by
workers, and on the characteristics of firms, which respondents work for, such as region,
industry, firm size, and the existence of unions (Lee & Kwon, 1995). In addition, students
in industrial relations have favored the BSWS data, because the number of observations
is big, ranging from a minimum of 383,000 to a maximum of 585,000. However, it is not
possible to get such full data set. Only a randomly extracted 10% of the original number

of observations from the BSWS is available to researchers.

2' It is reported that Japan has a similar survey used by the Japanese Ministry of Labor (Kim and Topel,
1995).

22 The number of sample establishments depends on years investigated.

61

Some til
the sunei
sized t‘SlJi‘
is that the
rather the
Variable in
inf-071113110
lltll he a 5‘
including 1

“0 matter

Some drawbacks of the BSWS should be acknowledged. First of all, the population of
the survey is an establishment hiring 10 or more employees, and thus excludes small-
sized establishments and their employees. One of the characteristics of the BSWS survey
is that the respondents are asked whether trade unions are organized in their workplace,
rather the respondents are union members (Uh and Lee, 1992). In that case, the union
variable in the BSWS is similar with the Current Population Survey in the U.S. including
information on collective bargaining coverage (Shin, 2001). However, the variable must
not be a serious weakness when measuring union effects because bargaining outcomes
including wages and fiinge benefits affect all workers in a potential union member pool,
no matter how great the percentage of the pool is represented with the union in the
establishment concerned with labor laws (Lee and Nam, 1990).

The main dependent variable here is the logarithm of real hourly wages, nominal
wages deflated by the Customer Price Index (1995=100). The logarithm of real hourly
wages reflects bonus as well as regular payment and overtime payment. Dummy
variables for one-digit industries (omitted category is manufacturing), based on the
Standard Industrial Classification, are included in order to capture any industry difference
in the union effect on wages. Dummy variables for the establishment region (omitted
category is region for Seoul) and the dummy variables for size (omitted category is
establishment hiring less than 30 employees) are added into the regression equation. To
measure human capital of the employees, three variables, education, experience, and
tenure, are considered. The dummy variables (omitted category is middle school
including elementary school) are used for education. Experience is measured by “age-

total years of education-6.” Also squared experience and squared tenure are included.

62

 

lit.

re;

1101

tori

Orly

Ԥ\.
\.

 

... \il

Two personal characteristics, sex (1 if female, 0 otherwise) and marital status (1 if
married, 0 if unmarried) are also measured and included into the equation. A variable
regarding union status as a form of the dummy variable (1 if union exists, 0 if union does
not exit) is put into the equation. Again, an important thing to remember is that the union
variable is not the answer if individual employees are or not union members. Thus, the
union variable is expected to capture the effect of unions on wages when a union is
organized within an establishment.

This study selects and uses only regular employees ages 19-65. Also the data used are
limited to employees who have non-professional and non-managerial occupations, since
labor laws prevent them from joining unions.23

This study uses data for 1987-1999. Most previous studies in Korea examined data
for only one or a few years and analyzed the period of the late-19805 and early-19905.

None, except Shin (2001), use data covering a long time-span as this study does. Thus, it

might be meaningful to examine a long time-span of 1987-1999 including recent periods.

 

23 The original sample of the BSWS in this study included professional and managerial employees of
12.7% (73,365 for 1987-1999), 7.4% in 1987 at a minimum and 18.3% in 1998 at a maximum. Article Two
in the Trade Union and Industrial Relations Adjustment Act requires that managerial employees “who
always act in their employer’s interests are not allowed to join” trade unions. Thus, there appears to be no
need to exclude professional and managerial occupation from the sample for this study, because the union
dummy for the occupation should always be zero in the wage equation if employees in the occupation
cannot be union members. However, the expectation may not be true. I ran the wage equations for only
professional and managerial occupations. Among the 13 years (1987-1999) in this study, union coefficient
for professional and managerial employees was not statistically different from zero at the 5% level in eight
years. However, the union coefficients for the employees are significantly positive at the 5% or less level in
five years; .060 in 1988, .024 in 1990, .031 in 1992, .044 in 1998, .051 in 1999 7.4% in 1987, 18.3% in
1998. It appears to a surprising that professional and managerial employees, who cannot be union
members, can have positive union wage premium in the five years. However, it may make sense,
considering that, according to DiNardo, Hallock, and Pischke (1997), the union can indirectly affect wages
for managerial employees not under union coverage in unionized establishment, even though they reported
opposite direction of union effect. How and why union can affect salaries for professional and managerial
employees may be an interesting another research question, which cannot be dealt with herein.

If the professional and managerial employees are included in the sample for the study herein the indirect
effect of unions on professional and managerial employees will bias the results regarding the direct effect
of unions on potential union membership. For this reason, professional and managerial employees are
excluded from this study.

63

l-l

Ct

fr;

.\t

Table 4-1. Descriptive statistics for all variables in all, non-unionized, and unionized

sectors for 1987-1999 (unweighted)

 

 

Variables Explanations A11 Nonunion Union
Lwage Logarithm of hourly real wage 8.345 8.105 8.477
(.774) (.831) (.706)
Busan 1= Busan .098 .095 .099
(.297) (.293) (.299)
Kyungki 1= Seoul suburbs .215 .214 .216
(.411) (.410) (.412)
Joongbu l= Kangwon, Choongchung, & Cholla .175 .135 .198
(.380) (.341) (.398)
Kyungsang l= Kyungsang .214 .195 .225
(.410) (.396) (.418)
Mining 1= Mining .018 .005 .025
(.132) (.071) (.155)
Construction 1= Construcrion .010 .001 .015
(.101) (.035) (.122)
Transportation 1= Transportation & Public Utilities .024 .041 .014
(.152) (. 198) (.119)
Wholesale 1= Wholesale and retail Trade .069 .104 .049
(.253) (.305) (.216)
Finance 1= Finance insurance & Real estate .090 .031 .123
(.286) (.173) (.328)
Services l= Services .104 .122 .095
(.306) (.327) (.293)
Public 1= Public services .079 .084 .075
(.269) (.278) (.264)
30$Ees<100 1= 30-99 employees .096 .201 .039
(.295) (.400) (.193)
1003535600 l= 100-299 employees .202 .264 .168
(.401) (.441) (.374)
300$Ees<500 1= 300-499 employees .195 .165 .314
(.396) (.371) (.212)
EesZSOO 1= 500 or more employees .478 .300 .576
(.500) (.458) (.494)
Sex 1= female .322 .388 .287
(.467) (.487) (.452)
Marital Status I: married .616 .563 .645
(.486) (.496) (.479)
Highschool 1: high school .528 .512 .537
(.499) (.500) (.499)
College 1= 2-year college .075 .081 .072
(.263) (.272) (.258)
University 1= 4-year university .130 .134 .127
(.336) (.341) (.333)
Expe year of (age-educated years-6) 15.509 14.861 15.865
(10.955) (1 1.747) (10.476)
SQexpe Squared term of Expe 360.536 358.861 361.459
(513.857) (513.858) (421.552)
Year of Tenure 5.496 3.748 6.459

 

 

64

 

5m or
ton] n
110nm
0f 01w
Olsen
311d u
resi-‘k'e
.QUlle
[hf Cot

A Q
are the
“(flier

“‘agcs‘

 

Tenure (5.332) (3.930) (5.743)
SQtenure Squared term of Tenure 58.639 29.496 74.702
(1 10.800) (68.029) (125.564)
Union 1= unionized establishment .645 - -
(.479) - -
N = 486,685 172,929 313,756

 

 

Note: mean without parentheses and standard deviation in parentheses

Table 4-1 reports the descriptive statistics for all (column 3), non-unionized (column 4),
and unionized samples (column 5) from a pooled cross-sectional data for 1987-1999. The
total number of workers used in the analysis is 486,685 including 172,929 (35.53%) in
non-unionized and 313,757 (64.47%) in unionized establishments. The average number
of observations for all samples per year is 37,437 and the smallest number of
observations for all samples is 26,700. Also the average number of observations for men
and women samples per year is 25,363 (total 329,730) and 12,073 (total 156,955),
respectively.24 Considering that the coefficients estimated by quantile regression become
‘quite accurate’ (Buchinsky, 1998; 91) in the large sample, it is expected that the size of
the coefficients in this study will be stable.

A comparison of the means of a number of characteristics indicates that union workers
are more experienced by one year, and longer tenure by 2.7 years than non-unionized
workers. Also they have a higher probability of being married and male, earn more

wages, and are concentrated in larger-sized firms.

 

2‘ The exact number of observations for all, men, and women in each year can be found in Table 4-2, 4-3,
and 4-4, which will be presented later.

65

3. Results

Results from All Sammles

Table 4-2 presents the OLS estimates and estimates of 6,“ from quantile regressions
with q = .10, .50, and .90 for the years 1987-1999. The OLS results in the second row of
Table 4-2 shows that the union wage premium existed for 1987-1999. The size of union
wage premium varies depending on year examined, ranging from .01 (1% = exp [.01]-1)
in 1996 to .074 (7.7% = exp [07414) in 1999.25

Comparing the results from the OLS to the results from the QR at the median in the
fourth row in the table, the estimated unionized sector wage premium, estimated by
median (q= 0.50) quantile regression, is approximately from .005 (0.5 %) in 1999 to .025
(2.5%) in 1987 smaller than that estimated by ordinary least squares, in all cases but
1998. The results from both OLS and QR show that the union wage advantage had
gradually declined from .069 (7.1% = exp [.069]-l) by the OLS and .044 (4.5% = exp
[.044]-1) by the QR in1987 to .022 (2.2% = exp [.022]-1) by the OLS and .005 (0.5% =
exp [.005]-1) by the QR in 1993. Despite the increase of a narrow range, above .03 or
3%, in 1994 and 1995, the union wage premium remained around one or two percent by
both the OLS and QR in 1996 and 1997. Finally, the union wage effects have increased to
above .06 or 6% in 1998 and 1999, as they have faced a serious financial crisis in Korea

since December 1997. The results indicate that there has been a union wage effect in

 

2’ It is easier to understand more the meaning of union wage premiums when percentage is used instead of
the coefficient of union in the wage equation. The coefficient of union in the wage equation can be
transformed by the formula of exponential form, exp(union coefficient) — 1 (Halvorsen & Palmquist, 1980;
Kennedy, 1981). If the size of union coefficient in the wage equation is below 0.1, the size of the
coefficient becomes close to the relevant percentage. For example, the coefficients of union, .01, .05, and
.1, in the wage equation are 1% (.01005), 5% (.05127), and 10.5% (.10517).

66

833:; Eoucoaovfi me :25 can 608:9. 23:2
.0558 695:9. 85593 605598 83.55:. 3:283? 82: £38m .858 £3. @3653 BB EoEsmanmo S8
8.3.8.56 9265 E38 Jug—E6 anew»: Sou 63%.? «concave 8 own? 32 350; do we. muons—oi 83%? 682

35. 82 no can 2: a 285%: 28:. e ”302

€28.32 mac ..8 335583 5 Echo c.8983 83oz ”802

8355.3 E 22823.. 2:53 88 echo c8238 agnoom “802

2228.805 28 80358 95535 28 own 3.». wins—n85 moEEam ”202

 

 

28.8 88.8 8.8 88.8 88.8 88.8 5.9.. 88.8 8.8 83v 8%" 88.8 83m 2

88.8 683 A83 383 A83 83 88.8 383 A83 A83 A83 A83 A83

88. .88. 88. .38. 88. as. :o. 88.- NS. 88. .28. .28. .8. Sta

883 A83 83 83 A83 A83 983 A83 A83 A83 A83 A83 A83

88. .88. 88. .88. .88. .88. 88. 88. LE. :8. 88. 88. .38. Gala

83 883 :83 83 A83 883 A83 A83 83 683 8:: A83 A23

88. .38. 88. 88.- :8. .8. 88. as. 88. L8. 88. 88. :8. Sta
mo

:83 A83 A83 A83 A83 A83 A83 A83 A83 A83 A83 A83 683 $48

88. 88. 88. 88. 48°. 88. .88. 88. 88. L8. :8. 88. 88. as:

83 .83 33 82 83 33 83 83 32 82 82 83 83

 

 

mcofimoumom 2:530 Ea mAO E E20580 coED we aim d-v 2%...

82-82 5 83:8 a 5

67

Korea for the period 1987-1999. However, the size of the union wage premium is not a
large number, .043 (4.4%) by OLS or .032 (3.3%) by QR on average for the thirteen
years. Also, we can describe the changes of union wage advantage as a gradual decline
for the late-19803 and early-19903, some fluctuations of a narrow range in the mid-19905,
and a relative high union wage premium during 1998-1999 during the crisis.’

The results with regard to the size of union wage effects do correspond by and large to
previous studies. For example, the union premium of .053 (5.44%) by OLS in 1988 is
quite close to the estimated union premium (5.50%) as Bai (1991) predicted for the year.
Also the union premium of -1 .6% in the manufacturing sector in 1994 by Lee and Kwon
(1995) does not differ much from our estimated size (-2.7%) when I follow their
estimation procedures.

The estimates of the union wage effect from the quantile regression with q = .10, .50,
and .90 assure the assumption behind quantile regression that the regression union wage
effect may not be constant across the entire conditional wage distribution. In every year,
the size of the union wage effects at q = .10, .50, and .90 is different.26 More importantly,
a quantile producing the largest size of union effect among q = .10, .50, and .90 is not
consistent. For the year 1987-1993, the estimate of union status at q = .10 is larger than
those at q = .50 and .90. For instance, unionized workers at q = .10 in 1987 began to earn
larger wage increase (8.7%) three times greater than workers at q = .90. For the years
1994, 1995 and 1998, the union wage effect at q = .50 is larger than those at q = .10 and

.90. In addition, the union effect at q = .90 is larger than those at q = .10 and .50 for the

 

26 F-tests comparing coefficients of union variable in between two quantile regressions at q = .10 and at q =
.90 are conducted, though not reported here. The null hypothesis that the size of union coefficients at q =
.10 and at q = .90 are equal is rejected at the 5% significance level in every year, but the year of 1995 ( F (
1, 32,637) = 0.1], Prob > F = 0.740) and 1998 ( F ( l, 30,880) = 3.65, Prob > F = 0.056).

68

years 1996, 1997, and 1999. This result means that unions had protected the best the
interests of workers at the bottom of wage distribution for 1987-1993 and workers at the
top of wage distribution for the year of 1996, 1997, and 1999, while the median workers
for the yearsl994, 1995, and 1998 had better benefits from a wage increase by collective
bargaining, compared to the workers in the lower and upper tail of wage distribution.

As the absolute size of union wage effect, estimated by either OLS or median QR,
has declined during the period 1987-1993, the difference between the largest and smallest
union effects from q = .10, .50, and .90 also have decreased. This result indicates that
wage policies of the trade unions have changed from favoring only a limited number of
workers in the lower end of wage distribution to balancing various interests of different

workers across the wage distribution, during the period.

 

Figure 5-1. Union Wage Premiums for All sample
in 1987 and 1999

.0
N
J

 

union coefficient
0
O L.
A 01
l

. m '\ fl‘:
0.05 = ‘

1112131415161718191

quantile in the wage distribution

 

 

- - - -1999 —1987

 

 

 

 

 

 

69

To confirm the findings that the quantile producing larger union wage effect varies
among q = .10, .50, and .90, full quantile regressions ranging from q = .01 to q '= .99 were
run for 1987 and 1999. Figure 4-1 presents how the unions wage premium changes in
each quantile in the distribution of wages for the two years. The results from Figure 4-1
were surprising in that the size of union premium for employees, opposite of our
expectation, was smaller in the lower part than in the higher part of the wage distribution
in 1999, while the union premium was larger in the lower part than in the higher part in
1987 as expected. This finding means that higher-wage workers, rather than lower-wage
workers, benefit more from trade union and collective bargaining by trade unions at least
in 1999. This result was not found through the OLS reporting union wage premium at
mean and through quantile regressions at q = .10, .50, and .90 as already seen in Table 4-
2.27

The argument that the Korean financial crisis may have caused the exceptional
phenomenon in 1999 could be raised. To answer the question, I ran full quantile
regressions at every quantile for every year from 1987-1999 for all workers, as seen in
Figure 4-2. A graph of union coefficients for each year was drawn from 99 quantile
regressions ranging from q = .01 to q = .99. Even though some variations in details are
evident, it is clear that the size of union coefficient was larger for workers at the lower
end of wage distribution than for workers at the higher end of the wage distribution from

1987-1995. In 1987, it is clear that workers ranging from q = .01 to q = .11 have much

 

27 To date, most studies adopting quantile regressions report only the results at some quantiles such as q =
.10, .25, .50, .75, and .90 (Chamberlian, 1994; Poterba and Rueben, 1994). The comparison between Table
4-2 and Figure 4-1 implies that the traditional reporting practice might be incomplete and even arbitrary in
a sense. One of the reason explaining why the reporting practice has existed might be the time in running
quantile regressions ranging from q=.Ol to q = .99. For example, it takes about an hour to run a quantile
regression with bootstrap replication of 20 iterations at a quantile in this study. When increasing the
number of bootstrap replication from 20 to 500 repetitions by using the data here, the time spent to a run a
quantile regression increases to a whole day.

70

Figure 4-2. Union Wage Premiums for a Samples for 1987-1999

 

union coefficients for all in 1987

 

0.2
0.16 -
0.12 A
0.08 \‘\

0.04 M 4— 4"“

 

 

 

 

 

 

 

quantiles

 

 

 

Note: all quantiles significant at the 5% or less level

 

union coefficients for all in 1988

 

0.08
0.06

0.04 \x i
0.02 \\ A I

 

 

 

 

 

 

quantiles

 

 

 

Note: quantiles from q=.95 to q=.98 i_nsignificant at the 5% level

71

 

 

union coefficients for all in 1989

0.1

 

0.075 “A

 

005 X

 

 

 

 

 

 

 

 

 

 

 

 

o 1 1 1 L 1 1 1 1 1 1 1 1 1 1 1 1 1
1 11 21 31 41 51 61 71 81 91
quantiles
Note: all quantiles except q=.9l significant at the 5% or less level
union coefficients for all in 1990
0.06
0.04 W
0.02
0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
1 1 1 21 31 41 51 61 71 81 91

quantiles

 

Note: all quantiles except q=.97, .98, and .99 significant at the 5% or less level

72

 

 

 

 

 

 

union coefficients for all in 1991

0.05

 

 

0.03 v

 

0.01

 

 

-0.01

 

quantiles

 

Note: all quantiles except q=.01, .02, .67, .74, .85, .87, .88, .89, .92, .93, .94,
.95, .96, .97, and .99 significant at the 5% or less level

 

 

union coefficients for all in 1992

 

 

 

0.02 ~~ - ‘
0.01
l
o q 1 1 1 1 1
1 11 21
-0.01 ~

quantiles

 

 

Note: only quantiles at q=..5~.l3, q=.IS~.26, q=.29~.32, q=.36, q=.48, q=.50
and q=.60 significant at the 5% or less level

73

 

 

 

 

me
006
err

ma
I UL

0C1

NEW-

 

 

union coefficients for all in 1993

 

0.08

 

0.06 \
004

 

0.02 \\

 

 

 

 

 

-0.02
quantiles

 

Note: quantiles at q=.Ol ~ q=. l7, q=.23, q=.27, q=.4l, q=.58,
q=.6l , and q=.70 ~ q=.84 significant at the 5% or less level

 

 

union coefficients for all in 1994

 

 

 

 

 

 

 

 

0.045
0.015
0 1 1 1 L 1 1 L 1 L 1 1
1 11 21 31 41 51
-0.015

quantiles

 

Note: quantiles except q=.92, .93, .95, .96, .97, .98 and .99 significant
at the 5% or less level

74

 

 

 

union coefficients for all in 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.1 K
O. 075
\1
o 1 1 1 1 1 1 1 1 1 1 1 1 1
1 11 21 31 41 51 61 71 81 91
quantiles
Note: all quantiles significant at the 5% or less level
[\ union coefficients for all in 1996
0-025
0.02 /A
0,015 [AVAV
0'01 IA“! L r/
I m
0.005 \ ‘r/AA
o 1 1 4.1 1 1 1 1 1 1 1 1 1 1 4 1 1 1
41005 41 V11! 21 31 41 51 61 71 81 91
-0-01
quantiles

 

K

Note: quantiles at q=.37, .40, .4], .42, .43, .45, .55, .59, .60, .68, .69, .70,

.79, .85 ~ .95, and .97 significant at the 5% or less level

75

 

 

 

 

 

0C

0C

\ 01

\ k
\\.

 

union coefficients for all in 1997

 

 

 

 

 

 

 

 

 

0.05
0134 [‘51
0.03 rJ\
0.02 N \
0.01 M \.

 

 

 

 

 

 

 

 

 

 

 

1 1 1 21 31 41 51 61 71 81 91
quantiles
Note: quantiles except q=.Ol ~ q=.20, q=.23, q=.98 and q=.99 significant
at the 5% or less level
union coefficients for all in 1998
0.12
0.09 //
0.03
0 1 1 1 1 1 1 4L 1 L 1 1 1 1 L 1 1 L 1
1 1 1 21 31 41 51 61 71 81 91
quanfihs

 

Note: all quantiles significant at the 5% or less level

76

 

 

 

union coeffcients for all in 1999

0.1

0.075 A
i W \

0.025

 

 

 

 

 

 

 

1 11 21 31 41 51 61 71 81 91
quantiles

 

 

 

Note: all quantiles except q=.02 significant at the 5% or less level

higher union wage premium, compared to workers in the remaining quantiles. The
situation was similar in 1993 and 1995. However, in the years 1988-1992, it is clear that
the workers at the lower half of wage distribution have a larger union premium than the
workers at the upper half of the distribution. Furthermore, it appears no significant
difference in union premium among workers at the lower half of wage distribution
existed in 1994.

The finding that the size of union premium was larger for workers in the lower half
than workers in the upper half in wage distribution was not repeated during the period of
1996-1999. In 1996, the lower half of the wage distribution showed a smaller size of

union premium, compared to the upper half of the wage distribution. Even the union

77

coefficients at from q = .01 to q = .36 are not significant statistically at the 5% level.28

Again, the union coefficients from q = .01 to q = .20 in 1997 were not significant at the
5% level. This finding means that workers in the quantiles during 1996 and 1997 did not
have extra wages through the unions, compared to the comparable workers in the non-
unionized sector. In the period of 1996-1999, the union premium for the workers was
smaller at the lower half than at the upper half of the wage distribution. In particular,
workers ranging from q = .01 to q = .20 suffered the lowest size of the union premium,
while well-paid workers ranging from q = .80 to q = .99 enjoy the highest amount of the
union premium for the period. Furthermore, Figure 4-2 implies that the result from
quantile regression for 1999 is not a more exceptional phenomenon because the
phenomenon that the size of union premium is larger for the well-paid workers than for
the lesser-paid workers started in 1996, two years before the Korean financial crisis

occurred in December, 1997.

Qgrr_1p_arison between Men and Women

Table 4-3 and Table 4-4 present union coefficients from the OLS and the quantile
regressions for 1987-1999 for men and for women, respectively. Except for the period of
1987-1989 and 1999, the size of the union coefficient from the OLS is larger for women
than for men. This finding is as expected in that previous studies argue that the union
wage premium for women is likely to be larger than for men because of lower wages for
women workers.

The results from the quantile regressions at q = .10, .50, and .90 imply again that the

size of the union coefficients varies among the three quantiles for both men and women.

 

28 I cannot report results in detail here due to space limitations.

78

h

333-1013~ Lg: 97:22:... 11:32 Ln: 7.2:..2210LwUu4 D:::w:arv 3:8 V11~AV a... sameUmthzfv :3.:: wk.» nv\.~f1 .~_11\1 .d\n-~\

m03~t¢> E02385 Q :25. Ba 608:3 23:2 65:8 825:8 85:38 .oocotoaxu .mo_EE:n Econ—8:3 09:: .388 REE:
825:5“. "Em EoEsmanmo .58 .mflEEnv .99-GE co>um .mu_EE=u Beam“: 58 638:? Eugene“. an own; .3. .259— mo mo. 625.05 3338.» #4 .9. m4. 232- -_£ 982

.96— 82 .8 8m 2: a EnoEcma 232: a u: 8 m4. 28¢ 28 202
8355.38 E 20 .5: 82.5 285% qfifioom 28 61—0 .5: Echo Emu—am 7453— 51. Q m4 2%:- 88 802

 

 

 

 

 

 

88.: 48.8 38.8 8:: 28.8 88.8 88.8. 88.2 88.2 88.: 8:8 82: 88.: 1 2
53 A23 A23 8:: 53 88.8 8:: A83 2 :3 A83 8:: A83 2 :3
88. 88. 88. v8. 88. 88. 88.- 88.- So.- 88. 88.- 88.- 8 _ o. 8.1 8
8:: 83 A83 83 A83 83 A83 A83 83 A83 8:: A83 8::
88. 88. ms. 8 8. . _ 8. . _ 8. 8 s. 88. 88. 88. 88. . _ 8. 88.- 8.1 8
53 E3 53 8:: A83 2 8.3 : a.v 83 83 :83 8:: 83 9:3
.88. 88. 88. 88. 88. 88. 88. 88. 88. 88. 88. 88. 88. 2.1 8
A83 A83 83 A83 A83 A83 83 A83 A83 83 A83 883 88.5 $qu
88. 88. 88. .88. 88. L8. 88. 88. 88. 88. :8. 88. 88. :82
83 83 83 83 83 83 83 83 83 83 8.3 8.3 803
aaofiihwmfi pom 038mm 2889* How mfiommmo-Hmvm 2tch 98 WAG 5 EBQEQOU :033 m0 0N5 .v-v USN-H
88.8 818 ”8.8 5.8 3.8.8 8.8 88.8 88.8 8.8 88.8 88.2 88.: 88.8 1 2
A83 A83 83 A83 83 A83 883 c 8.8 88.5 83 88.8 83 8::
88. 88. 88. _ 8. 88. 88.- _ 8. 88.- 88. 88. . _ 8. L8. 88. 8.1 8
A83 :83 A83 A83 83 A83 88... 88 V 83 A83 A83 A83 83
88. 88. 88. 88.. 88. 88. 88.- 8 8.- .8. 88. 88. .38. 88. 8.1 8
83 8:: 83 88.8 83 8:: A83 88 v c .3 A83 53 83 53
88. 88. 88. 88.- 88. 88. 88. S 8. 88. 88. 88. 88. L8. 2.1 8
A83 88 v 83 A83 A83 A83 A83 A83 A83 83 :83 A83 83 A39
.88. 88. :8. 88.- 88. 88. 88. 88. 88. 88 88. L8. 88. :8:
83 83 83 83 83 83 83 83 83 83 8.3 83 8.3

 

 

83-83 5 288m 2“: 88 223288 28530 85 So s 820880 :25 .8 £5 .3 038

79

For men, the union coefficient at q = .10 is larger than at q = .50 and q = .90 from 1987-
1994. Since then, the union coefficient at q = .10 is not the highest among the three
quantiles any longer. Rather the union coefficient at q = .90 for 1996-1999 is larger than
at q = .10 or q = .50. This result is similar with what we found in the analysis for all
samples as seen in Table 4-2, Figure 4-1, and Figure 4-2.

In Table 4-4, the situation appears more complicated for women. The union coefficient
at q = .10 is larger than those at q = .50 and q = .90 for the years 1987, 1988, 1989, 1991,
and 1993.29 In the years 1990, 1992, and 1994, the union coefficient at q = .50 is larger
than that at q = .10 or q = .90, while the union coefficient at q = .90 is larger than those at
q = .10 and q = .50 for 1995, 1997, 1998, and 1999. Also, we can state that the union
coefficient at q = .10, q = .50 and q = .90 is larger in the late-19803, in the early-19903,
and in the late-19908, respectively.

For a better understanding of how unions affect wages differently for men and women
workers at each quantile in the distribution, I ran quantile regressions at every quantile
for the 13 years for both sexes.30 Figure 4-3 shows the results. In 1987, union coefficient
for men moved in a relatively narrow boundary, ranging from .05 to .01. Until q = .80,
we can see that the union coefficient for workers at the bottom of the wage distribution
showed a larger size and then decreased slowly, when ignoring the last a few quantile
showing a slope upward in the upper end of the distribution. This fact supports the
argument that low-wage workers benefit the best from the trade unions, rejecting the OLS

assumption that the union wage premium will be constant, regardless of the quantile in

 

29 In 1996, union coefficient at q = .10 is .025 and that at q = .90 is .024. Because the difference in union
gpefficient between the two quantiles is too small, we will not consider it further.
It requires 1,287 quantile regressions (99quantiles * 13 years) for men and women workers, respectively.

80

Figure 4-3. Union Wage Premiums for men and women for 1987-1999

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    

 

 

 

 

 

 

 

 

 

 

(a) 1987
0.21
0.18%
E 0.15 ,
g 0.12 + M
8 0.09-Ne- 6"
0 W " ' " Women
8 0.06 .fl
g 0.03 ‘
0- U u
-o.03 11 21 31 41 54. 31 :11 31 91
quantile
(b)1988
0.09
."N
i; 0.06~' \
Q
E 0.03 v Me”
0 \ ' " " Women
C
Q 0 4L4L11111111L1111
C
3 11121314151617181 .,
-0.03
quantile

 

81

 

 

 

 

 

Figure 4-3. If:

~C®.0.t®00 CO..C3

0. m

“CQ‘UFuvaO CO‘CJ

Figure 4-3. Union Wage Premiums for men and women for 1987-1999

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

    

 

 

 

 

 

 

 

 

 

 

(a) 1987
0.21
I
0.18 \
E, 0.15 .
g 0.12 . M
8 0.09-3% e"
0 006 W - - - Women
.5 ' ‘fl'K. W
g 0.03 ‘
O_W
\I
-0.03 11 2% 314‘. 5‘. 3‘. 71 3‘. 91
quantile
(b)1988
0.09
1“
,5 0.06-' “
E
8 0.03 v Me"
0 - - - Women
-§011J111111111;L1_111
C
3 I1121314151617181,
-0.03
quantile

 

81

 

 

 

 

 

 

union coefficient

0.12

0.1
0.08
0.06
0.04
0.02

-0.02
-0.04

 

 

(c) 1989

quantile

 

 

 

Men
- - - Women

 

 

 

 

union coefficient

0.08
0.06
0.04
0.02

(d)1990

 

 

 

 

 

 

 

Men
- - - Women

 

 

 

‘
l
111111111 111111

 

 

1

11 21 31 41 51 61 71 81 91
quantile

 

82

 

 

 

 

 

 

(e) 1991

 

  

 

  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

quantile

 

 

 

 

0.06
E
'g 0.03
E Men
0 : - - - Women
5 0 .L 1 1 1 1 1 1 1 1
c n \‘
3 1|. 11 21 31 41 51 61 71 81 91m
’ 'l
I
-0.03
quantile
(f) 1992
0.03
0.02 ‘ A» . .
- w "'
E2 0.01 L " “Li
2 '. a
g '0-01 ‘ ' ‘ W' ~°a-‘ " ' " Women
.g -0.02 '1'
3 -o.os
-0.04 "

 

83

 

 

 

 

 

(9)1993

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

  

 

  

 

 

 

 

 

 

 

 

quantile

0.09
‘E
.9
2
E Men
° - - - Women
C
.9
C
3

quantile
(h) 1994

0.06
E 0.04 ————»\—————.-~—
.8 ' w [WV] I‘
E 0.02 — =4 Men
2 0141111111141 _L_\‘---Women
.2
g_0.02I11213141516171819‘l

-0.04

 

84

 

 

 

 

~C¢..C..t@00 CC..C3 ,, ”CWRVVCQQU 20‘23

 

 

(i) 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.09 1
E \
.9 ‘ J“.
9 0.06 4 ‘ “
E '\ WWW/Y \ : Men
0 a. ' .. - - - Women
5 0.03 \ J a.
E W
3
0111111111111111 11
1 11 21 31 41 51 61 71 81 91
quantile
(j)1996
0.08
"E 0.06 '| :1
(D l‘ l
2: 0.04 "‘1 r.
:5 002 . LAW AA. Jae/.1. ' Men
8 ' I V VI - - - Women
5 0111111111111141
C
3 -0.02 -
-0.04
quantile

 

85

 

 

 

union coefficient

0.09

0.06

0.03

0

-0.03

(k)1997

 

 

 

"11 21 3141 51 61 71 81 91

-
-. . ‘

I'M"

I

 

 

 

Men
- - - Women

 

 

 

quantile

 

 

 

union coefficient

0.15

0.1

0.05 ~

 

a)1998

 

 

 

Men

. - - -Women

 

 

l l l l L l l 1 l l l l l l l 1 1 I l

 

1

11 21 31 41 51 61 71 81 91
quantile

 

86

 

 

 

 

 

 

union coefficient

0.09

0.06

0.03

-0.03 r

(k)1997

 

 

  

 

l'112131415161718191

 

   

 

quantile

 

 

 

Men
- - - Women

 

 

 

 

union coefficient

0.15

U)1998

 

 

 

 

 

Men

. - - -Women

 

 

llllllllllllJllllJl

 

1

11 21 31 41 51 61 71 81 91
quantile

 

86

 

 

 

 

 

(m) 1999

 

 

 

 

 

 

 

 

 

 

 

 

 

0.1 !
'V‘II'
88 0.05 ”44’ Me"
2 VM - - - Women
0
'5 0.025
:3
OlllllllLlLlllLlJJll
1 11 21 31 41 51 61 71 81 91
quantile

 

 

 

the wage distribution. In the case of women, the shape of the union coefficient depending
on quantile is more extreme, compared to men. With the highest number of .195 at q =
.01, a series of statistically significant union coefficients in 1987 stop at q = .21.31 This
finding means that only female workers in 21% in the wage distribution earn more
Wages, when compared to female workers in the non-unionized sector. Therefore, when
We view only the union coefficient estimated by the OLS as seen in Table 4-4, we may
have made an error by stating that all female workers will consistantly earn more wages
by 3.05% (=exp(.03)-1), due to the unions. The fact that the significant and positive union

Premium for female workers is found for some workers who are in the lower 21% of

 

 

3' At q = .79, .80, and .8], the union coefficient becomes significant again at the 5% level, in fact.

87

11ng Jim
mmmum
hmcfi
premium
appucm
The\1
lesspait
Can be s
1903 a:
1988 u
“in“ <
BiFOir
Am i
11:111.
“gym

in [in

As
UPP:

lhc}

wage distribution may suggest that female workers suffer from structural discrimination
within unions.

Since the year of 1987, the percentage of female workers showing a positive union
premium increase year by year as seen in Figure 4-3, for the period of 1990-1998, it is
apparent visually that the union wage premium is larger for women than for men.

The view that the size of union wage premium, expressed by percentage of wages, for
less-paid workers is larger than the size of union wage premium for better-paid workers
can be supported to some degree by the case for females in 1988, 1989, 1990, 1991, and
1993 and the case for males in 1987 and 1994. A good example is female workers in
l 988 where the lowest 10% of wage distribution shows the highest wages and then the

union coefficient decreases at a constant rate as the quantile increases. Finally, the upper
3 0% in the wage distribution appears to suffer from wage losses due to the unions.

An interesting finding is that a U-shape distribution of union coefficients varied by
Quantile is found in the case of men in the period between 1989-1993 and in the case of
Women in 1996. This finding means that the union premium was larger for both workers
in the lowest part and workers in the highest part and smaller for workers in the middle
Wage distribution. For instance, workers in the interval ranging from q = .38 to q = .84 in
1 991 did not show significant union coefficient, while other workers did. So far, we do
not know how this phenomenon occurred and which factors in the unions result from it.

As a whole, a new trend that the union premium was low at the bottom and high at the
upper half in the wage distribution becomes clear in 1996. The trend was already found in

the Figure 4-2 dealing with all samples.

88

Sensitivity Tests
The robustness of the results above should be double-checked because a possibility
exists that an important additional or omitted variable may cause very different results
than those we observed prior. To answer this question, I added more variables when they
are available in a specific year. For example, I ran additional quantile regressions
including four occupational dummy variables as well as the already-used variables for
three years, 1987, 1993, and 1999, for all men, and women samples. The partial results
are shown in Figure 44.32 It suggests that adding four occupational variables do not
affect the results without occupational variables at all.

Usually 20 replications with regard to bootstrap standard errors are generally sufficient
even in the case of heteroscedastic errors (STATA, 1999). In fact, Poterba & Rueben
(1994) and Hartog et a1. (2001) calculated bootstrap standard errors by using 20 iterations
in their study. However, Chamberlain (1994) conducted bootstrapping with 500
replications in order to obtain more efficient bootstrap standard errors. In order to assure
if Significant differences occur between 20 replications and 500 replications done by
Cl'la-Inberlain (1994), I repeated the same quantile regression with 500 replications for the
three years, 1987, and 1999, again. However, I did not found any significant difference.

In sum, we may say that the results from quantile regression listed prior are quite robust
beCaux no changes occurred afier adding occupational variables and checking different

Starldard errors as well as the large number of samples in each year.

32\
The results for males and females in l987, I993, and 1999 are available.

89

 

Figur
Occu
In W.

Figure 4-4. Comparison of Union Coefficients in between the equation including
Occupational variables and the equation without the variables for men and women
In 1987, 1993 and 1999

(a) All samples in 1987

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.18
.. 0.15 -
5 Mo occupational
E: 0-12 variables
a:
0.09 ~
3 \ - - - with
,5 0.06 occupational
5 variables
0.03 ~
0 lllLJlllllllllllllI
1112131415161718191
quantile
(b) All samples in 1993
F
E w/o occupational
'5'; variables
%
8 - - - with
g I occupational
g 0 111LL11114111111L1 variables
-0.01 ‘ 11 21 31. 41.5161. .11. 81. 91.
-0.02
-0.03
quantile
L.~

 

 

 

 

 

90

(c) All samples in 1999

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.12

.. 1‘
g 0.09 w/o occupational
“g variables
’13
O 0.06 ‘l‘v .
0 - - - With
E: occupational
g 0.03 variables

0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

1112131415161718191
quantile

 

 

 

I£131¢ations and Interpretamon of Results

The purpose of interference by the government into labor market in Korea was to erase
the effects of institutions such as trade unions on wages for the period prior to 1987
(Song, 1991). The oppressions by the government in the 19905 shared the same purpose.
If the purpose of interference by the government is true, it means that no significant union
Wage premiums should be seen in the period post 1987. In fact, our results implying
Signi ficant union wage premium in Korea indicates that the purpose of interference by the
government into the labor market was not fulfilled. Also it means that trade unions can be
Vi(“Wed as institutions offering economic incentives to which workers are likely to be

attt‘acted.

9l

The 11
111163 1
01.13.3111:
1011111

In at

1111101] 1
1115111111

Tin
11m
ii‘ilfim
distribl
.1150 l:
ligard
bten h
0r SCH

“1‘ch

filed.a

The results from the quantile regression indicate that the OLS model based on mean
values may miss important features of the union wage effects on the wage structure.
Quantile regression reveals that the union wage effect is not constant across the
conditional wage distribution.

In addition, the union wage effect across the conditional wage distribution varies
greatly for the period of 1987-1999. For 1987-1995, in detail, the largest union wage

effect is located at the bottom of the conditional wage distribution. However, the largest
union wage effect moves from the bottom to the middle and to the top of the wage
distribution.

The results from quantile regressions are not surprising in that Korean unions have
prOtected the interest of the workers better for the years 1987-1993 who remained at the
bOttom of wage distribution. It is thought that workers at the bottom end of the wage
diStribution gain the largest wage advantage from unionization (Farber and Saks, 1980).
Also the finding is understandable when considering the situations for 1987-1993 with
regard to who initiated Great Labor Offensive in 1987. The period of 1987-1993 had
been heavily shadowed by the influences of the Great Labor Offensive in 1987. Unskilled
or Semi-skilled male blue-collar workers aged from the late-twenties to thirties, who are
expected to be at the lower half in the wage distribution with shorter tenure, initiated the
waVes of union—organizing drive at that time. They took most of the presidential positions
in the new enterprise unions that were created after the Great Labor Offensive. Then, as a
reasonable result, demands for wage increase every year for the period reflected the
interests of the workers at the lower half of wage distribution in most cases by strict

fixttd-amounts of wage-increases definitely favorable to workers who were paid less,

92

131

rather than the fixed-rate of wage-increase that might be favorable to high-paid workers

(Jung, 1992). Table 4-5 presents how unions determine wage increases in 1990. It shows

that fixed amounts of wage-increase are the most favored way by unions in all three types

of firm size. When regarding mainly fixed amounts and partly fixed rates of wage-

increase as a variant of a fixed amount of wage-increase, the importance of a fixed

amount of wage-increase appears dominant in all firm sizes. Another finding from the

table is that the importance of a fixed amount of wage-increase is larger in the relatively

small-sized firms hiring less than 300 employees, compared to the larger-sized firms

hiring 500 or more employees.

Table 4-5. Ways to Determine Wage Increases by Unions across firm Size in 1990

—_

 

 

 

 

Firms hiring Firms hiring Firms hiring Total
100-299 300-499 500 or more
\ Workers Workers Workers
Fixed Amount Increase 64 (54.2%) 18 (39. 1%) 31 (32.3%) 1 13 (43.5%)
Mainly Fixed Amount and 16 (13.6%) 7 (15.2%) 15 (15.6%) 38 (100%)
Partly Fixed Rate
Fixed Rate Increase 15 (12.7%) 8 (17.4%) 19 (19.8%) 42 (100)
Mainly Fixed Rate and 14 (1 1.9%) 8 (18.4%) 18 (18.8%) 40 (100%)
Partly Fixed Amount
Other 9 (7.6%) 5 (10.9%) 13 (13.5%) 27 (100%)
. Total 1 18 (100%) 46 (100%) 96 (100%) 260 (100%)

 

Source: Jung (1992)

The results for 1996-1999 are surprising in that Korean unions have protected the

interests better of workers who occupy the top part of the wage distribution from 1996-

93

1990, P1
111111):
the 13311
11111111
1991 J.
lilt gm
1111111
1111111
AI
511111

110111

1999. Prior the thought was the workers at the top gained the least wage advantage from
unionization (Farber and Saks, 1980). Thus, this finding is contrary to the literature on
the issue, and the well-paid and skilled workers may benefit only from the craft unions or
political co-optation by the authoritarian corporatist government (Freeman, 1980; Song,
1991). However, no significant emergence of craft unionism or political co-optation by
the government occurred in the mid-19905 in Korea. Also two-tiered wage agreements,
which are believed to push down wages for newcomer workers and to cause union
premium effects favored to existing workers, are not found in Korea (K00, 2001).

At least three explanations may account for the unexpected finding. The first
explanation unraveling the unexpected finding regarding unions protecting well-paid
workers in the late 1990 comes from employers and the government. Korean employers
in the 19905 introduced many schemes that may have affected wage increases, such as
pay-for-performance and team-based pay system. In addition, the government supported
those employers’ changes. However, the ways to increase wages for well-paid workers
better applied to mainly white-collar workers in large companies in the 19905. It is
unlikely that the schemes changed how union premiums are distributed across wage
distribution. In addition, the finding of unions protecting well-paid workers from 1996-
1999 should not be interpreted as a simple sign showing that unions had lost their original
spirit and had declined as a result in the mid and late 19905. Even though union density
has slowly declined since 1990, the Korean unions in individual firms showed that they
have sufficient bargaining power to determine how to retain schemes to increase wages
favorable to them by themselves as seen in the general strike of 1997 and 1998 during the

Korean financial crisis (Choi, 1997; Kuruvillar and Erickson, 1999; Koo, 2001).

94

The second explanation is from internal union matters. Even though ten or more years
have passed since the Great Labor Offensive in 1987, the workers who built new unions
in the late 19805 still run most of the unions at present. They are not workers anymore
who occupy the lower end of wage distribution and those they represented in the late
19805. Their personal needs regarding wages may have moved from the contents of wage
increase from the workers at the lower end to workers at the middle or upper half of wage
distribution. Also the unions have faced accumulated complaints from workers at the
upper end who have gained the least advantage from collective bargaining until the mid-
19905 (Jung, 1992). As a result, the number of bargaining outcomes to increase wages
according to the fixed-rate method rather than the fixed-amount method has increased in
the 19905 (Jung, 1992).

The explanation focusing on union leaders assumes that there has been no change in the
composition of union leaders, but their location in the wage distribution has moved from
the bottom to the median or the upper part of wage distribution. However, a different
explanation should be considered, too. That is, the emergence of well-paid workers with a
higher union wage premium may be a result of their entrance into unions. Many report
that wage differentials between high school graduates and university graduates and
between blue-collar and white-collar workers decreased considerably since 1987.33
Before 1987, university graduates and white-collars had been given large educational or

occupational wage differentials due to the employers wage policy to protect them from

 

33 As will be explained in chapter later, the return rate of university graduates relative to high school
graduates (100%) in the wage equation has decreased from 152% in 1988 to 130% in 1999 in non-
unionized sector and from 144% in 1988 to 121% in 1999 in unionized sector, respectively, when estimated
by OLS. Regarding the wage gap between blue- and white-collar workers, average wages for blue-collar
manufacturing workers have increased from 49.0% in 1981 to 60.1% in 1987 to 80.5% in 1996 of the
average wages for white-collar workers (Chang, 1998).

95

the influence of unions. As new unions reduced the wage differentials and a more
favorable political environment to establish or join unions was created since 1987, many
workers had incentives to join unions. In fact, a large portion of new unions, established
in the 19905, were white-collar unions (Jeong, 1993). It was not difficult even to find
foremen, who expected to be at the upper part of wage distribution, joining already-
established unions in manufacturing firms. In addition, a decline in the manufacturing
sector in the 19905 needs to be considered. The phenomenon, which occurred from 1996-
1999 was a result of an important change in the driving-force from the less-paid workers
at the bottom to the well-paid workers at the upper of wage distribution due to the
emergence of new well-paid union members. Also the change may imply that the
phenomenon of a larger union wage premium for well-paid workers in the late 1990
would be continued in the near future as long as no other change regarding composition
of union membership occurred.

This line of explanation is related to the criticism of labor-autocracy and trade unions
for well-paid workers two segmented labor market theory. The high cost of organizing
for temporary workers, part-timers, and workers hired on small-sized firms resulted in
extremely low union density in the marginal sector and unions only for well-paid
permanent workers.

Another possibility to explain the unexpected higher union wage premium for better-
paid workers who are located in the upper tail of wage distribution comes from whether a
specific ‘wage norm’ existing in Korea produce the result. While wage norms are
generally defined as a nominal wage series, wage norm is about “the direction of the

(union wage) premium (Wachter and Carter, 1989; 234).” In the context of this study,

96

wage norm is defined as an implicit collective norm, shared by most of employees,
regarding how to distribute union wage premium across different groups of workers
differently locating in the wage distribution and which types of order regarding the
distribution should be chosen by trade unions. Assuming that wages are very limited
resources to be distributed among employees, egalitarian institutions like trade unions
may decide to boost wages for less-paid workers first, and to increase wages for better-
paid workers later. The situation in Asian countries such as Korea and Japan might be
true. In fact, all government, employers, and employees had accepted the principle of
‘generous to the less-paid and strict to well-paid’ for a long time (Song, 1991; Bae,
1995). According to the wage norm argument, higher union wage premium for well-paid
workers in 1996-1999 become a phase of the process realizing the norm, existing after
higher premium for the less-paid in 1987-1993 and for the median workers in 1994-1995.
The prevalence of relatively short-term labor contract between employers and unions in
Korea may contribute to preventing internal conflicts within unions, regarding which
group of workers will benefit first and which group will be next, from developing to
serious organizational contention, through adjusting different interests from various
workers group every year.

So far we have no empirical data to solve the question of why the phenomenon
occurred. Therefore, the three explanations listed above remain merely speculation.

The finding that the quantile benefiting from bargaining varies depending on the year
investigated also implies that median voter model cannot always hold. This study appears
to support the finding by Parsons (1992), arguing that the egalitarian wage policies,

benefiting low-skilled workers, are also found in the bargaining units, which high-skilled

97

 

majorities dominate. The median voter model assumes the important role of the median
voter is favorable to less-paid workers than well-paid workers under the condition of
wage rigidity resulting from long-term bargaining contracts and their high level of risk
aversion (Booth, 1984). Wage rigidity means that long-term union contracts make it
possible for union wages to be insensitive to factors regarding aggregate demand
conditions in the economy such as unemployment and inflation (Hendricks, 1981). While
unexpected unemployment pull down wages in unorganized firms, for example, wages in
unionized firms can be protected by long-term contracts. Also the introduction of cost-
living adjustment (COLA) clauses into bargaining contracts between the unions and
employers guarantee union wages helped to keep pace with inflation. It has been said that
the relatively constant union-nonunion wage differential in the U.S. since 19705 is due to
the long-term contracts and the COLA clauses (Kaufinan & Hotchkiss, 1999; 614-615).
Different from the U.S., collective bargaining in Korea can be characterized by the
lack of long-term contracts and the COLA clauses. Since 1987, bargaining over wages
and fringe benefits between employers and unions in most bargaining units has been
repeated every year (Bae, 1995). In Korea, it is difficult to find long-term contracts such
as three years or five years, which are common in the U.S. As a result, no need for the
CC)LA clauses exists in Korea in that if unexpected inflation occurs in a year, the
inflation could be reflected in the bargaining table next year. This situation in Korea may
underrnine the role of the median voter in the Korean unions and distort the distribution
of a union premium in wage distribution. A U-shape union coefficient varied by quantile,
found in the case of men in the period of 1989-1993 and in the case of women in 1996 in

Figure 4-3, may be a sign of the weak role of median voter in unions. Also relative higher

98

union premiums for well-paid workers compared to less-paid or median workers for
1 996-1999 may result from the weak role of the median voter.

The findings with regard to yearly changes of union wage effect, ‘a gradual decline for
the late-19805 and early-1990, some fluctuations of a narrow range in the mid-19905, and
the relative high union wage premium during 1998-1999 under the crisis,’ may give us a
hint regarding the changes in union density. The economic perspective regarding unions
by Dunlop (1944) implies that union wage effect will determine how workers view trade
unions as attractive and become union members. In other words, the size of the union
wage effect impacts union density. In fact, it appears that the changes in the union wage
effect have kept pace with changes in union density. While the period of a declining
union wage effect for 1987-1993 is the time revealing such a decline, union density

increased as the relative large union wage effect occurred in 1998 and 1999.

4. Conclusion

This chapter shows that the level of the estimated union wage effect is sensitive to the
Choice of the quantile in the distribution of wages, when using quantile regression. The
number of studies on the union wage effect by using quantile regression is still extremely
limited.34 Furthermore, most studies adopt quantile regression in the context of wage
inequality (Buchinsky, 1994 and 1998; Chamberlain, 1994). In the context of wage

inequality, the focus analysis would be on the return to education, experience, or tenure,

x

34 Within my best knowledge, there has been no study has been done based on the quantile regression,
eJ'Kamining union wage effect in Korea.

99

not the union itself.35 By connecting the union premium in the distribution of wages with
union governance, it is expected that this study may contribute to the reversal of some
important theoretical issues in industrial relations, such as which group of workers
governs trade unions, and how the group which benefits from unions the most changes,
and why.

I found that the focus of the union wage effect moves across the entire conditional
distribution of wages and across years. In most years, it is shown that a significant
difference among union premiums for workers at the bottom, median, and upper part of
wage distribution is evident. Therefore, it is interpreted that the OLS estimation assuming
equal size of union wage premium regardless of quantile in wage distribution cannot
hold.

Also traditional wisdom in industrial relations claims that workers at the bottom of
wage distribution are likely to benefit the most from unions. The wisdom is revealed to
be true for the period of 1987-1995. However, a surprise surfaced, for the period of 1996-
1999, well-paid workers at the upper of wage distribution enjoyed more union premium,
while less-paid workers at the lowest of wage distribution faced less union premium,
Contrary to traditional wisdom. Assuming that the size of a union premium is a mirror of
Who governs a trade union, the wisdom cannot be effective at least in the period of 1996-
1999 in Korea. The full quantile regressions at every quantile in every year for all, men,
and women samples seem to support the findings.

I examined four possibilities regarding employers’ offensive demands supported by the

government, wage norm argument, and changes by the unions themselves, in order to

X

35 . . . .
Another study needs to focus the union effects on return to education, experience, and tenure In the
c0nditional wage distribution.

100

explain the seemingly exceptional phenomenon. To date, it is unclear why this
phenomenon occurred. Doeringer (1984) introduces a similar case that the emergence of
industrial unions in the 19405 and 19505 in the U.S. negotiated equal absolute pay
increases between skill levels, resulting in a narrowed wage difference; a wage gap
between skilled and semi- or non-skilled workers widened again due to union policy
resulting from a conflict between skilled and semi-skilled workers. One clear thing is that
the finding in this study also may be found in other countries at other times.

This study implies that the size of the union premium at different quantiles of wage
distribution and the issue of which group of workers will benefit the most from unions is
to be found empirically, and not to be assumed. Therefore, this study calls for more
studies in the near future on schemes for wage increases adopted by trade unions, in order
to more understand better why the size of the union premium varies in wage distribution
and across the years and how unions can change groups of workers, and who benefits
from wage bargaining the most.

This paper is not free from limitations, however. At least, two weaknesses are worth
recognizing. First, this paper assumes that the union wage premium is the only reason for
workers to join and participate in the decision-making process within unions. However,
the assumption in this study is too limited because many other issues such as
employment, working conditions including health and safety, autonomy in the workplace,
and so on may become incentives for workers to be union members. When considering
those issues together such as employment, working conditions, as well wages, we could

be more confident in judging which group of workers benefit the most from unions.

101

The second limitation is related to sample-selection bias. Many such as Lee (1978),
Duncan and Leigh (1980), and Abowd and F arber (1982), to name a few, pointed out that
the potential selection biases should be considered in the research on the union wage
effect. However, this chapter has not explored the problem because it is difficult to find
appropriate variables that are likely to affect the probability of unions, but not affect
wages, and consequently could be used to identify selection models. All studies in fact
(Poterba & Rueben, 1994; Schultz & Mwabu, 1998) using quantile regression (QR) on
union wage effect also confronted the same methodological problems. No econometric
method, unfortunately, allows us to explore union effects on various quantiles in the
distribution of wages and at the same time to overcome selectivity bias problem
(Buchinsky, 2001). Overall, it is undesirable, but inescapable to present results from
quantile regressions without a correction of selection bias (Montenegro, 2001).
Therefore, the direction for future analysis should be to solve the selection bias problem

under quantile regression.

102

CHAPTER FIVE

UNION EFFECT ON WAGE DISPERSION

1. Introduction

The purpose of this chapter is to answer the questions, whether any union wage-
equalizing effect is evident in Korea and if so how large is the union effect to decrease
wage dispersion. To answer the questions, establishment-level data will be used to see
whether unionized establishments show lower wage dispersion than unorganized
establishments, after controlling wages, personal characteristics, occupational
composition within establishments, and organizational characteristics such as industry,
region, and establishment size. For the estimation method, the OLS and ZSLS will be
adopted for establishment-level data.

The most important finding is that union wage-equalizing effects have existed in
Korea for the period of 1987-1999. With respect to the size of union wage-equalizing
effect, Korean trade unions reduced the wage dispersion by 7%-25% when using variance
of logarithm of wages in the 19905, compared to the unorganized sector, after controlling
f0r other variables. Also, the size of union wage-equalizing effects is revealed to be 7%-

1 5%, when using the standard deviation of logarithm of wages, in the 19905.
The chapter is organized as follows: Section 2 describes the estimation methods.
Section 3 describes the data set. Section 4 introduces the results and explains the

implications from the results. Finally, section 5 concludes.

103

2. Model & Estimation Method

The literature on union effect on wage dispersion includes several observational units
regarding wage dispersion: establishment level (Freeman, 1980 and 1982), intermediate
level such as occupation-industry cell level (Kahn and Curme, 1987; Asher and DeFina,
1997), industry level (Belman and Heywood, 1990; Budd and McCall, 2001), state or
SMSA (Flaherty and Caniglia, 1992). It might be reasonable to focus on the industry
level where industrial unions dominate and where the union’s role to reduce wage
dispersion within and between industries becomes important.

In Korea, the enterprise unions have predominated for the last decades and unions are
expected to affect mainly wage dispersion within and between establishments. In
addition, prior studies on the topic indicate that the effects of unions on wage dispersion
Within and between industries are varied (Bae, 1990; Jeong, 1991, Uh and Lee, 1992).
This study assumes that the establishment level is more important than the industry level,
Considering the form of unionism and the results from prior studies.

One of the methodologically important ideas to be kept in mind is that any differences
in characteristics regarding workers between unionized and nonunionized sectors should
be controlled because they may affect wage dispersion between the two sectors,
independent of union effects on wage dispersion. The predicted average log wage is
added into the equation because many point out that the average wages may be highly
Correlated with wage dispersion (Hirsh, 1982; Belman and Heywood, 1990). The use of

the predicted average of log wages makes it possible to use an instrumental variable

104

approach in order to reduce any bias resulting from the correlation between average
wages and wage dispersion.

To prevent an endogeneity problem between wages and variance of wages, a two-stage
least square estimation will be adopted. By using 2SLS, squared experience will be
employed as instrumental variables, which will be included in the first-stage but not
included in the second-stage equation, following Belman and Heywood (1990).

In this chapter, I will use two dependent variables: a variance of logarithm of real
hourly wages and standard deviation of logarithm of real hourly wages per establishment.
The independent variables are the variance (or standard deviation) of education, tenure,
experience, and the percentage of female workers, skilled workers, married workers, and
salesmen. Also the dummy variables representing seven one-digit industries, nine
regions, and three establishment sizes will be included. Finally, the wages will be
included into the equation to control any correlations between wages and the variance of

wages.

With respect to the sign of the independent variables, I assume that the variance of
average wages in the establishments will be larger, when the establishment is located in
Seoul. These establishments are mostly larger-sized firms hiring more employees, who
have greater variances in education, experience, and tenure. It is expected that variance of
human capital such as education, experience, and tenure, is positively associated with
Variance of wages. Also the percentage of female workers in an establishment is expected
to be negatively associated with wage dispersions because the average wages of females
are lower than those of males. In addition, the percentage of skilled, sales—, and technician

Workers within these establishments is expected to be related to the variance of wages

105

positively in that the average level of wages for them is higher than that of unskilled
workers. Furthermore, the wages are more likely to be applied by using the concept of the
pay for performance, which is likely to increase the variance of wages. Finally, I expect
that the average level of logarithm of wages within these establishments is associated

positively with the variance of wages.

3. Data

Data here come from the Basic Survey on Wage Structure (BSWS)36 for 1987-1999,
which is conducted by the Ministry of Labor (MOL) in Korea every year. The Survey
contains information on individual workers such as sex, martial status, age, tenure,
career, occupation, and level of skill held by workers, the characteristics of firms,
respondent’s work for, such as region, industry, firm size, and existence of unions, and
the working conditions for individual workers such as wages and hours worked (Lee &
Kwon, 1995). Agricultural workers, civil servants, and those serving in the armed forces
are not included in the data source.

Some drawbacks of the BSWS should be remembered. First of all, the population of

the survey is an establishment hiring 10 or more employees, and thus excludes small-
Sized establishments and employees hired in the establishments, estimated as about one-
third of the nonagricultural labor force (Kim and Topel, 1995). It is reasonable to expect
that wage dispersion in the wage distribution for all employees including the employees
hired in the small-sized firms is likely to be larger than that for employees working for
the establishment that hire 10 or more employees.

\

36 More details for the BSWS can be found in the previous chapter.

106

The data used in this chapter are establishment-level data, constructed from the BSWS.
I selected the establishments where at least five employees are found in the BSWS. Then,
I calculated the means of log wages, average years of education, experience, and tenure
within the establishments. Also I calculated the variance of log wages, of education, of
experience, and tenure within the establishment. To control other personal characteristics
and occupational composition, the percentage of female workers, married workers,
skilled workers, salesman, and technicians within establishments is calculated.

What should be kept in mind is that, even though the population of the BSWS i5
establishments hiring at least ten employees, the sample establishments are biased
towards larger establishments due to the selection process of establishments having at
least five respondents. As a consequence, the number of establishments hiring 10-29
employees decreases, compared to the composition of the individual-level data from the
BSWS.

It would be more desirable to construct establishment-level data every year for the
entire period of this study, 1987-1999, to discover how union effects on variance of log
wages change year by year. However, the job would be beyond my ability because the
Sample size would be over 20,000 for the entire period. Thus, I have constructed the

establishment-level data for five years, 1987, 1990, 1994, 1997, and 1999.37

Descriptive Characteristics of the Datg
Table 5-1 represents the descriptive characteristics of the data for 1987, 1990, 1994,

and 1999. The number of establishments used is 2,251 in 1987, 2015 in 1990, 1,680 in

\

37 The each year was selected due to the Great Labor Offensive in 1987, the resumed oppression towards
collective bargaining and unions in 1990, general changes in wage inequality shown in Table 5—3 in 1994,
the condition of pre-crisis in 1997, and the Korean financial crisis in 1999.

107

1994, and 2,096 in 1999. The percentage of establishments in the manufacturing industry
is quite high, for example, 65.6% in 1987 and 46.7% in 1999. Among the eight industries
including the manufacturing industry as an omitted reference group, the share of the trade
(Dindus_4), transportation (Dindus_5), finance (Dindus_6), and service (Dindus_7)
industry is larger than 5% as a whole, while the number of establishments in the mining
(Dindus_1), electricity, gas, and public utility industry (Dindus_2), and construction
industry (Dindus_3) remains small. The number of unionized establishments is 1,180
(52.4%) in 1987, 1,197 (59.4%) in 1990, 1,065 (63.4%) in 1994, and 1,135 (54.1%) in
1999.

The variances of the average logarithm of wages within establishments (hereafter
variance of wages), a main dependent variable in this chapter, is presented at the top of
the table. While the variance of wages in the unorganized sector is lower than in the
organized sector in 1990, the variance of wages in the unionized sector is smaller than
that in the non-unionized sector in 1987, 1994, and 1999. Especially, the variance of
wages in the unionized sector (.125) in 1987 is much smaller by 60% than that in the non-
unionized sector (.314). It implies that an unfair sized possibility of union to decrease
wage dispersions within establishments exists, although not controlling for the effects of
other factors on wage dispersions.

Regarding the size of the establishments, the table indicates that a unionized
establishment is likely to be larger than a non-unionized establishment. The percentage of
establishments (Dsize_3) hiring 500 employees or more is larger in the unionized sector
than in the non-unionized sector for all four years. Except for 1999, the situation is same

as the percentage of establishments (Dsize_2) hiring 300-499 employees. The finding

108

 

 

109

 

.83 .2... .83 .23 6.8... 3... .83 .3.... .83 88..
8.. .8... .. .. 8..... 8.... 8... ..... 8... ...... .8. 8.3....
.83 ....m. .83 .83 .83 .83 .8... ...-3 ...... .83
8... 8.. 8.. .... 8... 8.. .8. 8... ..8. 8... 0.80.2.3
.8... .83 .8... .8... ...... ..8. ...... .8... .8... .2...
8... 8... 8... ...... ..8. ..8. . .... a8. 8... 8... ...-.53
.8... .8... ....3 G8. .8... .8... .8... .8... .8... ...... a
8... 8... 8. .8. 2... 8... .8. 8... N... 8... 2.2588
...m... .8... .8... .8... .8... .8... .8... 8...... .8... .8...
...... 8... ..8. 8... t... . .... 2... 8... ..8. 8... 8...:
.83 .83 .83 .83 .....3 .83 .83 .....3 ...8. .83
..8. 8... ...... 8... ...... 8... 8... ..8. 8.. 8... 588.5.
.83 .23 .83 .8... ....3 ..8. .83 .88.. .83 .....3
8.. N... 8... ...... 8.. .... 8.. 8.. .8. 8.. 8.8:...
.83 ....3 ....3 ...... .. 8.. ....3 .. .3 ...: . .83 ..N. ..
...... ..8. 8... 8... .8. 8..... ...... 8... ...... 2... .55....
.8... .8... .83 .8... .8... .83 8.... .2... .8... .8...
...... 8... ...... 8... .8. 8..... ..8. 8... 8... 8... 82....
.83 ...8. .. 8.. .83 .83 .83 .. . 3 .8 . . .. .3 .8 . ..
8... v8. .8. 8... 8... 8... ..8. 8... ...... 8... .5856
.83 .2... .8... .8... .8... .8... .8... ...m... ...... .8...
8... 8... 8... 8... 8... ..8. 8... t... 8... 2... 3.3820
.23 ...... .83 .8... .8... ....3 .8... ...»... .83 .8...
...... ..8. 8..... 8... 8... 8..... 8... t... 8... 8... 8383.
.8... .8... .2... .. .... .8... .83 .. .... .2... .8... ......
88. .8. .8. 28. 88. 8.. .8. .8. ..8. ..8. 8.8.5.
.83 .....3 .....3 .83 .83 .83 .83 .83 ..8. .83
8... 8... 8... E... .8. 8... 8.. 8 .. 2 .. 8.. 58m
.8... .8... G8. .8... ...8. .8... ...... .8... ...... .83....
8... 8.. 8.. N... 8.. 8.. ..... 8.. 8.. ...... 083$
3 3-82 3 3-82 3 3-82 3 3-82 3 3-82
8.8.81.6 88.12. 98.12. 3.812. 88.812.
8.: 8... .8. 8... .8.

 

 

 

 

 

 

....a...w.aa§. 8... ...... ...8. ...8. .8. a. 8.08 .8885 ..8 8.8.5.82 838m 8.88% 83.8.29... ...m 8......

.o. .85.... vac-.0280 2.2.2.8388 ... .823 .98-.6... o... ... 35.888 .... ...... owns... .... E... uogcmuoaxocm ...... .89.? ..o

E:._..wo_uowm3.-_ 8.58.8388 5......» 8.5.... .... oo§tm>flo§c3> 8555.388 $.23 oocotoaxo .... 88.. 09.82. .0 oo§§>uoqxo> ...:ofinmznsmm

5......» 5.82.... ..o ...-...... emu-.3.. .... 855F382 wow...» ...-So: .3 «Straw... .... 855508.35 6.02
...-cough... 23.8.. 08 mow-05:28. ... 823 o... .082

 

 

 

 

 

8... .... ..8 N8 ...... ... 8... .... ...... 8.... n...
.888. .88.. .N...8N. ..N..N... 8.8. ......8. ......N. .....N.N. .28.. .88..
8...... .88. 8.88 ......8. N88. ....N... 88.8 ..N..8N 8.8 8.8 8......
..8. ..8. ..8. .83 ..Nm. .8... ..8. ..8. .8... ...8.

.8... 8..... 8..... 8.... .8... 8..... N8... 8..... 8.... 8.....- .83..
..8. ...N. .83 ...NN. .8... .8... .83 .83 .83 .83

...N. .NN. ...... 8.. 8.. 8. .. .8. ..8. ...N. .8. 8.0.5.3
...-N. .....3 .83 .83 ..8. .83 .8... .83 .8... ......

8.. .8. ..8. .8. 8.. .8. ..8. .. .. N... 8... 558.8
.83 .83 .83 .8. .....N. .83 ..8. ...... ..8. ..8.

8.. 8.. 8.. .... .... 8.. .8. ...N. .8. N8. 8......
...8. .83 ...3 .83 ..8. .83 .....N. .83 .83 .83

.... 8.. .8. N... ...... .N.. 8.. 8.. N... .... ...-.52
....3 .83 .83 ...N. .N..N.. .83 .83 .83 ....3 .83

.8. 8.. ...N. .8. ...N. ..8. . .8. 8... 8.. 8... 2.5...
......N. .8... .88. ...-8.. .88.. .......... .8... 8...... .8... .. ...... ..

8.8 8..... . N88 ...... . .88 ..8... .8... .8..- 88... .8..- 2....3>
.....8. .88. .N..... .88. ...... .88. .8... .88. .88. .28.

.....8 88.8 8..... .88 8...... ..88. 8.8 ..N....... ...-...N... 8...... 88>
...8.N. ..8... .888. ......3 ...8.N. .8... .8..N. ...8... .N....N. ..N...N.

8.... N8...- 8..... .8... 8..... 8.... .8... .8... 8..... 83. 80>
.8... .8... .8... .83 .8... ...8. .8... .83 .8... ...8...

8.. ...N. 8.. 2 .. .8. 8.. .8. ...... m... 8 .. 8.8...
...... .8... .8... .....3 .8... .....3 ...... ...... ...8. ...8. ....

..8. 8... ...N. 8.. .8. 8.. .8. N. .. 8.. 8.. 98.88
.. .3 .8... .8... .8... ...... .8... .8... .8... .. . ... .8... 8

8.. 8.. N8. .8. 8.. .8. 8.. .8. .8. SN. Van-.8...
..8. .83 .83 ...... .83 .83 ...8. .8... .83 ...N.

8.. 8.. 8.. .. .. 8.. ..8. ..N.. 8.. 8... ..8. 2......
...8. .83 .83 .83 ..8. ..8. .8... ....N. .83 ....N.

8.. .... .. .. 8.. .N.. 8.. 8.. .8. .8. 8... 8.58

 

 

I]

110

means that the percentage of unionization is high in the larger establishments hiring 300
employees or more, but is low in the smaller establishments hiring less than 300
employees.

With regard to the variances of three human capitals, the variance of education is
higher in the non-unionized sector than in the unionized sector in 1987 and 1990, while
the relationship is reversed in 1994 and 1999. The variance of tenure in the unionized
establishments is shown to be consistently higher than that in the non-unionized
establishments. However, opposite to the case of the variance of tenure, the variance of
experience is higher in the non-unionized establishments than in the unionized
establishments, except in 1987. All three variances of human capital are expected to
increase the variance of wages positively. However, only with the descriptive statistics
between the two sectors regarding human capital, we cannot yet determine what is the
relationship between union and wage dispersions.

The descriptive statistics between the two sectors also imply that the percentage of
females in the non-unionized sector is higher than in the unionized sector, while the
percentage of married workers is higher in the unionized establishments than in the non-
unionized establishments, for all four years. Regarding occupations, the percentage of
skilled workers is higher in the non-unionized sector than in the unionized sector for all
four years. The situations regarding salespersons are similar, except for 1999. The
percentage of technician workers varies according to year. Finally, the level of wages is
always higher in the unionized establishments than in the non-unionized establishments.

Overall, some factors, which are expected to increase the variance of wages, including

the size of establishments, the variance of tenure, the percentage of married workers, and

111

the level of wages are found more often in the unionized sector, while the non-unionized
sector shows the high values of other factors such as the variance of experience, and the

percentage of skilled and salespersons, which are expected to increase wage dispersions

within establishments.

4. Result

1 ) General trends of wage inequality indexes between the two sectors

Since the empirical literature implies that union wages are likely to be less affected by
market-related factors than nonunion wages due to the standard-rate wage policy, union
Wages are expected to have a lower dispersion than nonunion wages. In Korea for 1987-
1 999, this is the case. Table 5-2 presents the standard deviation of log wages in every

Year for the period. It shows that the wage dispersion of union workers is smaller than
that of nonunion workers, as seen in the second column. For the whole period, wage
dispersion in the unionized sector is less by 15.7% than in the non-unionized sector.
when focusing on each year, however, the table presents that the difference of wage
dispersion between unionized and nonunionized sector becomes smaller than that in the
Whole period. Except for the year of 1987, the difference in wage dispersion between the
two sectors is no more than 9% in every year. Another finding from the second column of
Table 5-2 is that the difference in wage dispersion between unionized and nonunionized
Sector for 1994-1999 is stabilized at about 7.4% or 7.5%, while there was very small

Wage dispersion between the two sectors for 1988-1992.

112

Table 5-2. Standard Deviation of Log Wage for All, Male, and Female Workers in All
Industries for 1987-1999

 

 

Period All Male Female
Union Nonunion Union Nonunion Union Nonunion

87-99 .75 (84.3%)* .89 .66 (79.5%) .83 .7 1( 86.6%) .82
1987 .72 (87.8%) .82 .65 (83.8%) .78 .57 ( 83.8%) .68
1988 .55 (96.5%) .57 .47 (87.0%) .54 .36 ( 97.3%) .37
1989 .54 (98.2%) .55 .48 (90.6%) .53 .36 (97.3%) .37
1990 .53 (94.6%) .56 .46 (88.5%) .52 .40 (105.3%) .38
1991 .52 (92.9%) .56 .47 (90.4%) .52 .39 (102.6%) .38
1992 .50 (96.2%) .52 .42 (89.4%) .47 .38 (108.6%) .35
1993 .52 (94.5%) .55 .45 (91.8%) .49 .40 (102.6%) .39
1994 .49 (92.5%) .53 .42 (89.4%) .47 .41 (107.9%) .38
1995 .50 (92.6%) .54 .44 (91.7%) .48 .42 (105.0%) .40
1996 .50 (92.6%) .54 .45 (90.0%) .50 .44 (102.3%) .43
1997 .50 (92.6%) .54 .44 (89.8%) .49 .45 (102.3%) .44
1998 .50 (92.6%) .54 .44 (89.8%) .49 .46 (102.2%) .45
1999 .51 (91.1%) .56 .46 (86.8%) .53 .48 (104.3%) .46

 

 

 

 

* % compared to nonunion
Data source: BSWS in each year
Note: Employees included in the calculation are all workers regardless of age and occupations.

A striking discovery found in Table 5-2 is that wage dispersion of women workers in
the 19903 is larger in the organized sector than in the unorganized sector, while wage
dispersion for men workers in the organized sector is always lower than in the
unorganized sector as expected. The results appear to be unexpected in that wage
dispersion literature implies that the union effects on wage dispersion would be stronger
for women workers than for men workers because women whose personal characteristics
Would have placed them in the low end of wage distribution are thought to have benefited

from unions (Farber and Saks, 1980). In fact, many personal characteristics for women

are not as great as the characteristics for men with regard to education, tenure (3.5 years

113

for women and 6.6 for men), age (28 for women and 36 for men), and etc. More
explanation regarding this finding is discussed later in the chapter dealing with retum
rates for gender.

However, the facts that wage dispersion is lower in the unionized sector than in the
non-unionized sector might be caused by the differences in personal characteristics
between the two sectors, not by the direct union effect (Freeman, 1980). The simple
comparison of wage dispersion between the two sectors in Table 5-2 is incomplete and
the comparison should be done after controlling for differences between union and

nonunion workers’ personal characteristics.

Table 5-3. Trends in the Three Wage Inequality Measures between Union and Nonunion
sectors, 1987-1999

 

 

 

 

 

 

 

Standard deviation Gini-coefficient Coefficient of variation

_\ of log wages

\ union nonunion union nonunion union nonunion
1987 .721 .825 .364 .420 .704 .886
1988 .550 .569 .313 .340 .630 .757
1989 .539 .554 .307 .325 .626 .686
1990 .529 .559 .301 .327 .617 .691
1991 .525 .560 .294 .326 .583 .697
1992 .497 .523 .278 .303 .554 .627
1993 .515 .545 .284 .307 .553 .608
1994 .487 .529 .266 .299 .5 14 .590
1995 .496 .537 .271 .301 .517 .599
1996 .502 .541 .275 .305 .527 .612
1997 .503 .541 .270 .304 .51 1 .618
1998 .501 .539 .270 .304 .515 .608
1999 .506 .559 .274 .314 .547 .642

\

 

Data source: BSWS in each year
Note: Employees included in the calculation are all workers regardless of age and occupations.

114

Table 5-3 gives overall changes in three wage inequality measures for 1987-1999, the

standard deviation of the natural logarithms of wages, Gini coefficient, and coefficent of
variation. All three measures show a considerable trend toward a decline in the wage
inequality over the whole period. Despite of the stable wage increase, for example, the
standard deviation of log wages decreased from 0.721 in the unionized and 0.825 in
nonunionized sector in 1987 to 0.506 in the unionized and 0.559 in the nonunionized
sector in 1999. Also the Gini-coefficient of 0.364 in the unionized and 0.420 in the
nonunionized sector in 1987 dropped to 0.274 in the unionized and 0.314 in the
nonunionized sector in 1999, which indicates a decrease of approximately 25 percent in
both sector. The coefficient of variation for the period shows similar pattern. While the
coefficient of variation in 1987 was 0.704 in the unionized and 0.886 in the nonunionized
sector, the values decreased to 0.547 in the unionized and 0.642 in the nonunionized
Sector in 1999. It means the decrease of 22.3% in the unionized and 27.5% in the
nonunionized sector, respectively. As not seen in Table 5-2, the 90/10 log wages
differential, which is a difference of log wages between 10% and 90% in the wage
distribution, has decreased from 2.23 in 1987 to 1.41 in 1999. These findings are
Consistent with the findings of others who have recently examined at wage inequality
(e.g., Jeong and Choi, 2001).

The majority of this big change mitigating wage inequality occurred in the late 19808.
Especially in 1988, just afier a year since Great Labor Offensive from June of 1987, all
three wage inequality measures dropped by over 10 percent. This summarizes how the
Great Labor Offensive in 1987 shocked wage structures as well as general industrial

relations at the time. Considering the low union density in 1987, in addition, massive

115

reduction of wage inequality in the 1987-1988 implies that there must have been a big
spillover effect transferring a relatively large wage boost for employees at the bottom half
of the wage distribution from the union sector to the nonunion sector.

Furthermore, a considerable decrease in wage inequality, another finding that can be
drawn from Table 5-3 is that the degree of wage inequality began to increase from the
mid-19908 as the wage-equalizing trend stopped at that time. While steadily declining
from 1987 to 1994, the standard deviation of log wages began to increase from 1996.
Though it is less clear than the trend of the standard deviation of log wages, the Gini
coefficient also has decreased from 1987 to 1994 and increased from 1995. The
coefficient of variation shows a similar phenomenon revealing in stead a decrease for
1 987-1994 and an increase from 1995. However, it appears that the small size of increase
in wage inequality from the mid-19905 cannot sweep away the striking achievement in

wage equality since 1987. Rather it means a return to the year 1991 or 1992 with regard

to the degree of wage inequality.

2) Results from establishment-level data

Union effects on the variance of wages

From the establishment-level data, Table 5-4 presents the estimates of various
independent variables on the variance of log real hourly wages within establishments,

estimated by the OLS and ZSLS, between non-organized and organized sectors in 1987,

1990, 1994, 1997, and 1999.

116

 

 

383 38¢ 38¢ 2 88¢ 38¢ 2 88¢ 38¢ 38¢ 88¢ 68¢
3.2..th 2.2..qu 3.2.35. emco. Soc ......wooc moor moo. woo. ......Nmor outflow
38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 38¢ 38¢ 38¢ 68¢
.....im—O. “tum—o. ...:vmo. 2.3.39 2.2.58. 3.2.58. :26. sec—o. tin:- «aewho. 355%
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 88¢ 38¢ 38¢
9.539. 2.2.2.35. 5o.- _.:.2.__c.- So. «:20.- ..zmmor :20.- ..ieo3. ...:ch flame—0:3
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 28¢ 38¢ 88¢
25. 2.2.on- _.2.2..:o. 2.2. — S. 2:39. ..iewoo. So. shoe. ...“...VN—r ...:35... imam-£-
38¢ 38¢ 38¢ 38¢ 28¢ 38¢ 38¢ 68¢ 38¢ 88¢
:88.- 38- :88. 2:48. 338.- :18. 38.- 38. ...-83. ...-8..- 88.588
28¢ 28¢ 28¢ 28¢ 38¢ 38¢ 28¢ 38¢ 38¢ 38¢
.....ioNo. ......Nvo. coo. seem—o. .....EMNor 2.2.30... see—mo. «...—mo. {SR-5N- ...:N— _... wEEE
38¢ 38¢ 38¢ 28¢ 38¢ 28¢ 38¢ 38¢ 28¢ 38¢
:88.- :-38.- 5.8.- 2.88. :88: :38: .88.- 38. ...-.2. :88. 885:2
38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 38¢ 38¢ 88¢ 28¢
.....im—o... eeaN—OV 2.2.30.- .Iiw—or ......Lmov ....imNov So. No0... So.- SO. xsnw§¥
38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 28¢ 38¢ 88¢ 68¢
2.2.39- ....ito... tin—o.- »«eN—G... :iw—or see—No... seem—c. tin—o. .....inN—r ...:oNor Eng—:-
83 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 38¢ 38¢
88.- 38.- 88.- 48.- :So- 88.- : _ 8. .. 8. :38.- .8.- V.88:
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 68¢ 38¢ 38¢ 38¢
.88.. _8. m8. ...-8. ...-38.- :88.- :28. 88. ...-8..- 38.- .8885
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 68¢
......wco. «woc. 2.2.2.. 5..- eehcor ...:omor ......v—c... 2.2..— _c. seem—o. ...:n©_. ...VMO. Vang—2:0
38¢ 38¢ 28¢ 38¢ 38¢ 38¢ 38¢ 38¢ 88¢ 88¢
.88. 88.- 88. 38. ...-ms.- :88.- :88.- .:~8.- :82. ...-88. 85:3.
28¢ 28¢ 28¢ 28¢ 28¢ 28¢ 38¢ 38¢ 88¢ 38¢
... :25.- aevoo: 2.2.. S.- ..2: 8.- 2.23m _ ct «aim-co.- Ncc... Noo- onuhwo. 2.2..me 39.va—
38¢ 38¢ 38¢ 38¢ 38¢ 28¢ 28¢ 38¢ 88¢ 28¢
2.2.2.35... 8.2.39.- eeezo... {...—No- ..iewNor :ecmor mcor seem—or 8.2.3 _.- ...:ohor 583m
3% 30 3mm 30 33 30 33 So 33 30
82 82 82 82

 

 

wqmm v.8 qu .3 33 can 33 .33 .223 .33 5

883 2.82 82.. 88.3. 8 28% 8.8: cc 888...”. ...-m 03.8

117

.32 8.2 on. 8 .52-«Em... .22.. .26. .3 on. ... .52-«Em... .2. .26. 9% 05 .a .5853... . .802 3355.8.— 5 mAO .8 98.5 23:88 .328 .082
do. .3 8:222: a.a axoOm .2255 69.3. 53> 8.. Echo 23:8... 98 35.0580 .802

 

 

683 683 86.. 86.. 86.. 86.. m 83 283 .33 .33 2
38.8 2.... 83... 83... - .88 68... 8.... .8... .8... we
68¢ 68¢ 68¢ 68¢ .8.¢ 68¢ 63¢ 68¢ 68¢ 68¢
:88.- 68. :8...- :.8.. :83. .. :8. .. 88. .- .8... :88... ......8. .5280
38¢ 28¢ 28¢ 28¢ 28¢ 28¢ 68¢ 8.: 38¢ 38¢
......mmor ......N—or ......Eor .....hoor ......vmoc .2..— _o.- ooov ooo. ....ivfl. .....mmo. c223
68¢ 38¢ 28¢ 38¢ 38¢ 38¢ 38¢ 8.: 38¢ 38¢
:88. 8... ...-.8... :88.- .:8.. 8...- :6. .. .8.- .: . mm.- :..8. .- 883..
68¢ 38¢ 68¢ 38¢ 68¢ 38¢ 68¢ 38¢ 68¢ 68¢
:88. :88. :88. :88. :88.- ..88. 88. :88. :88. 3.8.. 5.25.88
68¢ 38¢ 68¢ 38¢ 68¢ 38¢ 38¢ 68¢ 2 8¢ 68¢
88.. :_....8. :2...- .:..8. :88: :88. :88. :88. :88.- 38.- 858.8
38¢ 38¢ 38¢ 28¢ 38¢ 28¢ 68¢ 28¢ 68¢ .83
:88.- :88. 2:8.- :8... :88. 8...- ..3.... :88. :88. :. 8.- 8:5
38¢ 38¢ 68¢ 38¢ 8.: 38¢ 2 .3 38¢ 68¢ 68¢
:88.- :88.- ...-E..- :88.- :.._8. .- :88.- ..88.- :88.- :88. :88. 8.5:
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 68¢ 68¢ 38¢ 38¢
.....vho. ......vfio. ......oho. ......vmo. 3...: _. 9.2.0 _ o. ..mho. moo. ......oow... ......mmor 2&th
68¢ 38¢ 68¢ 38¢ 68¢ 8.: 38¢ 68¢ 28¢ 38¢
......8. :88. :88. :88. :88. :88. 88. :8... ...-.8. :82. 8.8585
38¢ 28¢ 38¢ 28¢ 8.: 68¢ 68¢ 38¢ 2 8¢ 68¢
.....ovo. ......oNo. ......omo. .....No. .....mmo. ......hoo. ....oNo. .....ooo. 32.2. — .. woo. oorunxo>
68¢ 68¢ 68¢ 38¢ 68¢ 38¢ 68¢ 38¢ 38¢ 38¢
......hoo. .....m— _. ”no. 8.. 2.2.31. 2.2.21. 2.2.w3. tics. m3. ......hmoA oo__.=vo>
68¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 68¢ 38¢
.88. .8. :88.- ..88. ...: . 8.- .8.- .88- 8... :88. 88. 823.".
6 .3 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢
.38. ...... :88.- :. 8.- :88.- :88.- : . 8.- 88. :8... 88. 89388
6 . 3 38¢ 38¢ 38¢ 38¢ 38¢ 68¢ 38¢ 38¢ 68¢
..nmor VNor ..voo... moo... :2. _ Nor ......ooof woo. ......Ns. ...2. — NM. «......ovo. oomeomwoo_
38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 38¢ 3 .3 38¢
.88.. ..8... :88. :85. :8...- :88.- :83. :88. 2...... .238. 0.3....

 

11"

 

118

First of all, the table shows that the wage dispersions, captured by the variance of
wages, within establishments are much higher in the establishment that is located in
Seoul (omitted reference group) than in the establishments in other regions. This result
makes sense because the establishments in Seoul tend to show a higher level of wages,
resulting from an increase of wage dispersions. Also this result implies that our
expectation regarding the sign of regions in the equation, mentioned earlier, is
reasonable. The differences in wage dispersion between Seoul and other regions are not
big, approximately 1% or 2% in most cases. Also it appears that the differences in wage
dispersion between Seoul and other regions are the least in 1997 among the five years.
The bottom part of the first page of Table 5-4 indicates that the effects of industries on
the variance of wages varies according to years and estimation method, either in the OLS
or ZSLS. Considering that the omitted reference category is the manufacturing industry,
the wage dispersion was lower in the trade (Dindus_4), finance (Dindus_6), and service
(Dindus_7) industries than in the manufacturing industry in 1987, 1990, and 1994, as a
Whole. In 1999, however, the variance of wages in the trade industry and finance industry
rose higher than in the manufacturing industry.

The results regarding the size of establishments are mixed. The belief is that the
establishment size is associated with the variance of wages. In 1987 and 1999, the
variance of wages for establishment (Dsize_3) hiring 500 or more employees is higher
than that for smaller-sized establishments. In 1999, for instance, the variance of wages for
establishments hiring 500 or more employees is more by .045 than that for the

establishments hiring less than 100 employees as a reference category, when estimated by

119

the ZSLS. However, the effects of the establishment size on the wage dispersion proved
the opposite in 1990 and 1994, when estimated by ZSLS.

All three human-capital-related variables, the variance of education, the variance of
experience, and tenure, are found to affect wage dispersion positively, except the effect of
tenure in 1987. The positive effects of the three variables can be seen in the results of the
four years except 1997, both by OLS and ZSLS. This finding means that an increase in
heterogeneity of human capital for workers is likely to result in an increase in wage
dispersion, as we expected.

The table also implies that the portion of female workers within establishments is
positively related to the variance of logarithm of hourly wages in 1990, 1994, 1997, and
1 999, against our expectations. I set an expectation earlier that the variance of wages and

the percentage of females are negatively related to each other, because most female
Workers are paid less, due to less education, experience, and tenure relative to male
workers in the sense and structural discrimination existing in firms regarding wages. In
addition, the percentage of married workers is shown to be negatively associated with the
Wage dispersion in all four years, both by the OLS and ZSLS. This finding may be
connected to the characteristics of the sample. In the sample, the average percentage of
married workers ranges from about 50% to over 70%. In other words, married workers
are close to the mean value and an increase in married workers receiving wages close to
mean wages is likely to result in decrease of wage dispersion. Currently, I have no exact
answer about the unexpected findings.

Also the results from the table show that the increase in the percentage of skilled

workers and technician workers is related to the increase in wage dispersion within

120

establishments, in 1987 and 1990. These results fit previous studies on this issue in that
skilled and technician workers tend to earn larger wages than unskilled workers, resulting
in an increase in wage inequality. An increase in the percentage of technician workers
within establishments is shown to affect positively wage dispersion more than an increase
in the percentage of skilled workers. Thus, I can infer from the result that the level of
technical workers would be higher than that of skilled workers.

The effects of salesmen on wage dispersion are not clear for the four years. In the
19903, the introduction of pay for performance in Korea was applied more to the
occupation of salesman than in any other occupation within firms (KLSI, 1998). If so, the
increase in the percentage of salesmen should be related to an increase in wage
dispersion. However, the results are not consistent with this prediction. From this
inconsistent result, we may conclude that the introduction of pay for performance for
salesman has not been great enough to make a significant and consistent difference in
wage dispersion yet. In fact, a survey conducted by the Ministry of Labor (MOL) in 1999
reported that only 27.1% of 4,998 firms that answered already had introduced pay for
performance (Don-A Ilbo, March 3, 2000). Furthermore, the portion of wages variable to
performance is only 30 % or less in 78.6% of the firms endorsing pay for performance.

The level of the logarithm of hourly wages is found to be associated with an increase in
wage dispersion, when estimated by the 2SLS in 1990, 1994, and 1999. The finding fits
what we anticipated earlier.

Different from some independent variables listed above that appear to be unclear, the
results regarding the effects of the unions on the wage dispersion seem to be clear. In

1987, the unions are shown to have reduced the variance of wages by 8.8% (OLS) and

121

7.4% (ZSLS). However, the union effects decreased to .011 (1.1%) by the OLS or .024
(2.5%) by the ZSLS in 1994 and .012 (1.2%) by the OLS or .025 (2.5%) by the 2SLS in
1999,38 even though the effects of unions on the variance of wages become insignificant
in 1990, the coefficient for union (t=l.65) is close to the 5% significant level when
estimated by the 2SLS. We found that the results for unions, estimated by the OLS, are
not much different from the results estimated by the ZSLS. This finding implies that the
true size of union estimate is likely to be close to those estimated by OLS and 2SLS.
Table 5-5 presents the results from first-stage estimation regressing the observed
logarithm of hourly wages on other variables including the squared experience as an
instrumental variable, which is not included in second-stage estimation in Table 5-4. The
coefficients of the squared experience are significant at the 0.1% significance level in
1987, 1990, 1994, 1997, and 1999. This fact indicates that the use of the instrumental

variable works and the estimation by the ZSLS are likely to be justifiable.

Union effects on the staflard deviation of wages

From Table 5-4, I found that the unions are associated with the decrease in the
variance of wages within establishments in 1994, 1997, and 1999. However, some
unexpected results are found such as the percentage of females and the married and
inconsistent results across years. In addition, the value of R-squared in the equation that is

estimated by the OLS in 1987 remains low (0.087), although the values of R—squared for

 

3‘8 When using ZSLS, the value of R" does not have same meaning as in the OLS. Thus, no value of R2 in
ZSLS in 1994 or very low value of R2 in 2SLS in 1999 is not problematic.

122

Table 5-5. Results from first-stage in ZSLS in equation with variance of logarithm of
wages as dependent variable in 1987, 1990, 1994, 1997 and 1999

 

 

 

1987 1990 1994 1997 1999
Busan -.030*** -.065*" -.019"* -.084“* -.054***
(.007) (.004) (.005) (.006) (.007)
Kyungki 046*" 019*" 035*" -003 .003
(.006) (.004) (.004) (.005) (.005)
Kangwon 247*" 042*" .004 -.047*** -.039***
(.014) (.009) (.008) (.009) (.010)
Chungbuk .103*** .001 .024M .028” -.051***
(.014) (.008) (.008) (.008) (.009)
Chungnam -.135*** -.085*“ -.003 -.003 -.007
(.010) (.007) (.007) (.007) (.007)
Junbuk -.072m -.071*** .01 l -.01 1* -.013
(.012) (.010) (.008) (.008) (.009)
Junnam -.081"* -.030*** -.044*" 034*" .004
(.012) (.006) (.006) (.007) (.007)
Kyungbuk -.015* -.027*" 030*" -.005 -0128
(.007) (.004) (.005) (.006) (.006)
Kyungnam .lll*" 053*" 098*" 061*" .012
(.007) (.005) (.005) (.005) (.006)
Mining -.106""' 067*" 083*" .184“" .224***
(.014) (.013) (.012) (.016) (.016)
Construction 264*" .001 ..124"* -.l48*" 044*"
(.025) (.018) (.013) (.01) (.013)
Transportation ..049"* 056*” .004 -.065"“ -.062m
(.015) (.010) (.008) (.009) (.009)
Wholesale 074*" 081*" ..073"* -.08l*" -.019**
(.012) (.007) (.006) (.006) (.006)
Finance 082*" -.0l6*" -.037m ..04"* .017"
(.007) (.005) (.005) (.006) (.006)
Services .045m 071*" .001 -.021*" -.017"
(.010) (.006) (.005) (.005) (.005)
Public 076*" .181"* .010 06"" .119***
(.009) (.004) (.006) (.006) (.006)
10031365600 .224m 062*" 087*" .024m .127m
(.009) (.005) (.006) (.006) (.023)
3009365600 .314m .159m 087*" 09*" .152***
(.009) (.005) (.006) (.006) (.023)
5652500 471*" 259*" .164*** .162"* 245*"
(.008) (.005) (.005) (.006) (.023)
Veduqoo -.017 366“" 262“" -.097 .630*“
(.099) (.056) (.056) (.063) (.067)
Vexpe*100 -.144*** -.104*" -.096"* -.125*** -.187"*
(.005) (.003) (.003) (.003) (.004)
vtenm-eqoo 1242*" 605*" 360*" 332*" 355*"
(.018) (.010) (.008) (.008) (.009)
Female -.643"" -.508*" -.556*" -.62"* -.586*"
(.007) (.006) (.006) (.007) (.007)
Married 002*" 371*" 441*" .463*" 340*"
(.000) (.009) (.006) (.009) (.009)

 

 

Skilled 038*" .089m 088*" .141*** .116***
(.007) (.005) (.006) (.007) (.007)
Salesman -.244*** -.076*** 229*" 396*" 229*"
(.012) (.009) (.007) (.008) (.009)
Technician 289*" 338*" 410*" 403*" .315"!M
(.010) (.006) (.009) (.009) (.009)
SQexpe*100 -.014*** -.032*** -.023m .023m -.028*"
(4.34e-06) (.001) (.001) (.7.69e-06) (.001)
Union 083“" 062*" 068*" -.074*** .117*"
(.004) (.003) (.003) (.004) (.004)
Contant 6561*" 7997*" 8241*" 8505*" 8400*"
(.01 1) (.008) (.009) (.01) (.024)
R’ 0.529 0.636 0.597 0.596 0.547
N 2,251 2,015 1,680 1,672 2,096

 

 

Note: coefficients and standard errors for Vedu, Vexpe, Vtenure, SQexpe are multiplied by 100.
Note: robust standard errors for OLS in parentheses
Note: " significant at the 5% level " Significant at the 1% level ““ significant at the 0.1% level

equations, estimated by the OLS, in 1990 (0.139), 1994 (0.237), 1997 (0.232), and 1999
(0.135) appear to be fairly acceptable.39
To double-check the results from the equation with the variance of wages as a
dependent variable in Table 5-4, I ran another equation with the standard deviation of
logarithm of wages within establishments in all the four years both by the OLS and the
2SLS. The results are presented in Table 5-6.
The results from Table 5-6 are not considerably different from the findings from Table
5-4. The unexpected positive effects of females earlier are repeated in Table 5-4,
becoming stronger due to the positive effects in 1987. Also the negative effects of

married workers are consistently repeated in all five years, both by the OLS and the

 

39 The low value of R-Squared in the equation for 1987 is not usual. Thus, I examined the data and checked
whether an important variable is excluded from the equation. However, 1 have found nothing wrong so far.

124

ZSLS. All three kinds of human-capital-related variables are found to be positively
significant in most cases.

Our interest is in the coefficients for the unions. Table 5-6 indicates that the effects of
union on wage dispersion still exist, even though a different dependent variable is used.
Except for 1987 estimated by the OLS, all equations show that unions tend to decrease
wage dispersions within establishments, captured by the standard deviation of wages. The
Size of the coefficients for the unions in Table 5-6 is slightly different from Table 5-4.
The larger effects of unions decreasing the wage dispersions in Table 5-4 become smaller
in Table 5-6. However, the Size of the union effects to decreasing the wage dispersions is
not much changed in 1990, 1994, 1997, and 1999. Again, it is shown that unions are
associated with a reduction in wage dispersion by 0.004 (.4%) by the ZSLS in 1987, by
0.005 (0.5%) by the OLS or 0.026 (2.6%) by the 2SLS in 1990, by 0.018 (1.8%) by the
OLS or 0.041 (4.2%) by the 2SLS in 1994, by .013 (1.3%) by the OLS or 0.038 (3.9%)
by the 2SLS, and by .025 (2.5%) by the OLS or .043 (4.4%) by the ZSLS.

When using the standard deviation of wages as a dependent variable, the Size of the
effects of unions on decreasing wage dispersions is larger when estimated by the ZSLS
than when estimated by the OLS. Even though the value of R-squared in the OLS
estimation remains quite low (.059) in 1987 as in Table 5-4, the values of R-Squared in
other years are acceptable. Table 5-7 presents the results from first-stage estimation
regressing observed logarithm of hourly wages on other variables including the squared
experience as an instrumental variable, which is not included in second-stage estimation

in Table 5-4. The coefficients of squared experience are significant at the 0.1%

125

 

 

 

 

 

.....inmor ......o _ of Moo. :woo. .....mmor :emmor woo. nil-oo- _.:.:..oho. ...:ooo. 0:95
68¢ .88.. .88.. .88.. 68¢ .88.. 8.... 68¢ 68¢ 68..
2.595. 2.336. 2:..— _o- _oo. moor 8:: or aim—o.- ahoo. ......ovor ....iovo... cuff—om
.88.. .88.. .88.. .88.. 68.. .88.. 8.... 68¢ 68¢ .88..
......ooo. .....iwoo. «...—mo. ...:MNo. :Lmo. ...:N—o. 1.1.—o. :woo. ...—mo. ...:mmo. 3.85..—
.8.... .8.... 68¢ .88.. 8.... .88.. 2.8.. 28¢ 2 8.. 68¢
......uSo- etuo—o. moor :3” _o.- moo. ......vmor ...:ono-t :— _o-t ......ovo. ......oNo. 230—0;
68¢ 68.. .88.. 6...... 68¢ 68¢ 68¢ .8.... 68¢ 28..
8... :8...- :_..:... 8... .88. 88. 28....- .8. :88.- :88.- .885
68.. 2.8.. 68¢ 2.8.. 68¢ 68.. 2...... 28.... 2...... 68¢
moor voo. .....ivoo. ...:omo. ...:nho. ......wmo. Moo. :2 _ o. 9:25.- tim— _.t comuozbmaonv
68.. 6...... 6...... 68.. 2.8.. 28¢ 68¢ 6...... 68¢ 68..
...: _ no. timo- nm _ o. 5*ano. ......ovor ...:mNor 2.5qu ...:owo. ...:wo _ .t ......mmo... 925—2
.88.. .88.. .88.. .88.. 2.8.. .88.. 68.. .88.. 6...... 68¢
...:oNo... :43 o... ...:omor ...:omo... .3... _ For ...:mmor «to—or .....iooo. 2.....me ......»oo. Eacwczhvm
.88.. .88.. .88.. .8.... 68¢ .88.. .88.. .8.... 68¢ 68¢
.13” _o... :1. o o.. 2.....MMor ......Nmor «equof ...:Nmor .....i— _o. c'choo. noo- §ooo. xsanSM
68¢ .8.... 68¢ 68¢ 68¢ .88.. 8.... 68¢ 68.. 68¢
iimmor :23 _ or ...:mNor ...:o _ o... ......o _ or ......wmor ...:omo. BEL-No- ...ivvo-t .....inor 55:5,:-
68¢ 6...... 68¢ 68¢ 28.... .8... 68¢ 68¢ 28.. 68¢
8...- .8. 8...- 8...- :2...- 8...- :88. :2... 28....- 8...- 2.58.
6...... 68¢ 68¢ 68¢ 8.... 68¢ 68¢ 68¢ 68¢ 68¢
{boo-.- moor woo. :woo. :50 _ or {tum—or 3.336. 2.16 _ o. ...:nNor Noo. Emcmcszo
68¢ 68¢ 8.... 68¢ .88.. 6...... 68¢ 28¢ 2...... 68¢
:88. :88. :88.- 8...- :88: :88.- :88. :88. 8... 6....- .8885
.88.. 2.8.. 68¢ 68¢ 68¢ .8... 68¢ .88.. .88.. .88..
8...- 8...- 8... 8... :3...- 8...- :88.- :8....- :88. 8.... 8.8.5.
.88.. .88.. .88.. .88.. .88.. .88.. .88.. .8.... 28¢ 68¢
88.- 8...- :8....- :88.- :88: :88.- .8.- :88. :88. :88. .855.
68¢ .8.... 68¢ .8... 68¢ 8.... 68¢ 8.... 68¢ 68¢
...:N—or ...:oNor «sin—o.- Szomor «:35... 8:36... _oo.t :«w_o-t ...—mo... :ghhor 583m
8.8 So 8.8.. 8.0 8.8 30 8.8 So 38 m...
888. .88. 888. 888. .88.
3mm ......

m... ... 88. ...a .88. .88. 8. 8..... 8.5.3 8863 8.58. we. .6 8.8.82 2858 eo soc-.8 8...: .6 8.8.8.8.”. 6-8 6.88

126

 

.32 88. .o 85 .8 ..88.-:38 ...: .32 88— o... 8 .58c_=m_8 .. .96. ..\..n on. .8 2.85%... 8 .302 880558.. ... moo 8,. 8.9:... 8385.8 .839. .802

 

coogu
vvo..

.88..
88*hm—g-

.888.
§§gmvor
.m_o.
«8*wn_.
ahooo
***mmo.
Aoooo
woo.
.888.
88*m_of
Avoo.
88*woor
.8.8.
5.85...
._oo.
«8*ooo.
..88.
88*m_o.
..ooo
**moo.
.8.8.
*iovor
.8.8.
oNor
.8.8.
vmor
Avooo

coca“
mm...

Aamo.
***hv_.

A_ooo
*88mNov
Amoco
8*ooo.
Amoco
§§fiooo.
Amoco
88*omo.
.888.
*iiooo.
Amooo
***obor
Amooo
8*umNo.
.oooo
***N_o.
.888.
«auhoo.
._ooo
888voo.
.8.8.
m_o(
.8.8.
hoof
.8.8.
o_or
Amoco

Nhe;

.88..
§u*_mwr

Amooo
*885Nor
.8.8.
885nm.-
.888.
888mmo.
Awooo
m_oc
Avooo
88*h_or
Aoooo
***_m_r
..8.
88*wN_.
A_ooo
88*moo.
A_ooo
{8*o_o.
A_ooo
ooo.
Avooo
«.m—or
Amooo
8’9omor
Amooo
moor
Amooo

use;
hom..

.888.
§§poom.

A—ooo
...m_or
Amoco
.§§h_of
.888.
§**hm—.
Amoco
«88vwo.
.888.
888ooo.
Amooo
*885wor
Amooo
8*8nmo.
Aoooo
u**o_o.
Aoooo
«8*hoo.
A_ooo
*Noor
Amoco
..8N_o.
.888.
«.ooor
Amoco
~oo.
Amoco

owoJ.

.ow..
...:oovN-

Amoco
88.—vov
Ammoo
«88omm.
...8.
...—nor
Aoooo
.88Nvor
Avooo
..8Nmor
Awooo
88.no_r
.8.8.
§«§M_N.
A_ooo
ooor
..88.
88*N_o.
A_ooo
88*moo.
Amoco
8.5mmor
.888.
u§oovor
.888.
ua§omor
.888.

owe.—
thA.

..88.
u§nme.

..oo.
888w~or
Amoco
.oo.
.888.
§*¢vwo.
Amoco
{8*VBo.
Amoco
.oo.
Amooo
88*coor
Amoco
«8*NNo.
.888.
..8N_o.
Aoooo
§§§voo-
..88.
88.0oo.
.888.
moo:
.888.
88¢m—or
Amooo
*o.o_or
Amoco

m.an

AoMNo
......med-

.888.
8880~or
.888.
88*wmm.
...8.
8mmo.
Ahooo
88*hmo.
Avooo
«woo.
Awooo
{*fimo.r
.8.8.
88*Mh_.
..oo.
«moor
..88.
8.8%—o.
..oc.
Noo-
Awooo
*.aMbo(
Amooo
...—cor
Avooo
moor
Ahooo

n_oJu
ohmgo

Avmco
8*cov_.

..88.
88¢moor
.888.
*hoo.
.888.
**§_v_.
Amoco
88*mvo.
Amooo
«8*omo.
Amooo
a¢§vo_r
Amoco
moo.
Aoooo
88*v_o.
Aoooo
«iimoo.
..coo
..uhoo.
Amooo
8*woo.
.888.
«.woo.
Amoco
.«*v_o.
Amooo

.mflm

.wmmo
......oonfi

Amooo
«..omo.
.888.
*¢§h_mr
.8.8.
§§§QNN.
.m.oo
*8vvo:
Amoco
58*w_or
.888.
888Noo.
.888.
88*ov_r
.voo.
*«umvo.
..88.
~oor
Amooo
«*hoo.
.w_oo
88*hm—.
Am_oo
*fiuhoo.
A__oo
oocwwo.
Anooo

.mNJn
anoA.

Ahmoo
58*m5w.

Amoco
88*mNo.
.vooo
8*qowof
Amoco
§§*_o_.
.oooo
8*h_o.
Amooo
ufiuvmor
Aoooo
*88Noo.
Amooo
Ǥo_o{
..ooo
§8§o_o.
..88.
{*umoo.
.888.
a**__o.
Amooo
8*aomo.
.888.
§§§mmo.
Amoco
..8hmo.
.mcco

 

>.
8.

“Swan—00

=25.

awakq

888.8888.

Sena—mm

.xzzxm
nothsz
2888...
22.8mm
2.50m
stun—m

oowNmom

oomeomwoom

oomvmomwg—

 

127

Table 5-7. Results from first-stage in ZSLS in equation with standard deviation of

logarithm of wages as dependent variable in 1987, 1990, 1994, and 1999

 

 

 

1987 1990 I994 I997 1999

Busan -.017* -.057*** -.011* -.082*** -.055"
(.007) (.004) (.005) (.006) (.007)

Kyungki 053*" 020*" 036*" -.00l .004
(.006) (.004) (.004) (.005) (.005)

Kangwon 240*” .036*" .006 -.O43“* -.032“
(CM) (.009) (.008) (.009) (.010)
Chungbuk .l01*** -.005 .022" .030" * -.048***
(.014) (.008) (.008) (.008) (.009)

Chungnam -.125** * -.085*** -.008 -.001 -.008
(.010) (.007) (.007) (.007 (.007)

Junbuk -.060*** -.075“* .012 -.012 -.012
(.012) (.009) (.007) (.008) (.009)

Junnam -.07 l ** " -.034*" -.048*" 034*" -.003
(.Ol2) (.006) (.006) (.007) (.007)

Kyungbuk -.01 l -.021*" 029*" -.005 -.015*
(.007) (.004) (.005) (.006) (.006)

Kyungnam .1 l2*" 055*" 096*“ 060*" .014*
(.007) (.005) (.005) (.005) (.006)

Mining -.077"* .077"* .088‘" .l90*" 222*“
(.014) (.012) (.012) (.016) (.0l6)

Construction .298‘" .022 -.128*" -.l45"* 053*“
(.024) (.017) (.013) (CID (.013)
Transportation .043" 062*" .003 —.058""‘ -.054***
(.0 l 5) (.010) (.008) (.009) (.009)

Wholesale 072‘" 083*" -.069*" -.078"'" -.013*
(.012) (.007) (.006) (.006) (.006)

Finance 099*" -.019*** -.032*" -.036"* 022""
(.007) (.005) (.005) (.006 (.006)

Services 049*" 080*" .0] 1" .015" -.0l2‘
(.010) (.006) (.005) (.005) (.005)

Public 076"" .l76*** 027*" 066*" 133*"
(.009) (.004) (.006) (.006) (.007)

lOOSEes<3OO 220*“ 055*" 076*" .0l9" .l18*"
(.008) (.005) (.006) (.006) (.023)

3003535600 .30l"* .l4l*" 076“" 084“" .143“*
(.009) (.005) (.006) (.006) (.023)

5652500 447*" 238*" .l47*** .153“: 230*"
(.008) (.005) (.005) (.006) (.022)

sped“ -l .83e-06 .019'" .01 l"* -.002 024*"
(.004) (.002) (.002) (.002) (.003)
SDexpe -.028"" -.023‘" -.023*" -.026*" -.036"*
(.00l) (ON) (.001) (.001) (.001)

SDtenm-e .108""' 050*" 039*" 036*" 040*”
(.001) (.001) (GM) (.001) (.001)
Female -.607"‘“ -.486*" -.548"* -.612"* -.579"*
(.007) (.006) (.006) (.007) (.007)

Married 002*" 378*" 437*" .46?" 339*"
(.000) (.009) (.009) (.009) (.009)

Skilled 029*" 080*“ 089*" .14)!" .108***
(.007) (.005) (.006) (.007) (.007)

 

128

 

Salesman -.253*** -.077*** 281*" 385*" 216*"
(.012) (.009) (.007) (.008) (.009)

Technician 262*" 307*" 332*" 392*" 285*"
(.010 (.006) (.009) (.009) (.009)

SQexpe -.000*** -.000*** -.000*** -.000"* -.002***
(4.3le-06) (9.53e-O6) (8.16e-06) (7.63e-06) (8.62e-O6)

Union .071 m 052*" 059*" 07*" .109***
(.004) (.003) (.003) (.004) (.004)

Contant 6502*" 8010*" 8256*" 8.548*** 8.467“*
(.013) (.003) (.010) (.010) (.025)

R2 0.531 0.645 0.602 0.593 0.552

N 2,251 2,015 1,680 1,672 2,096

 

 

significance level in all five years. Again, this fact assures that the instrumental variable

works and the estimation by the 2SLS are likely to be reliable.

Discussion

Our expectation was to find a considerable size of union effect reducing wage
inequality just after the Great Labor Offensive in 1987. The expectation appears to be
supported by the fact from Table 5-2 and 5-3 that the various values of wage inequality in
the both unionized and nonunionized sector dropped from 1987 to 1988. However, our
regression results shows that there was no statistically significant union wage-equalizing
effect at least in 1987, regardless of the dependent variable used, standard deviation of
log wages or variance of logarithm of wages. Rather, we found a significant union efi‘ect
increasing wage inequality. How can we interpret the unexpected result? I think that the
finding occurred, due to union threat effect toward nonunionized employers. Facing
massive labor offensive in 1987, employers without unions in their establishments had a

strong incentive to prevent their employees from establishing new unions, through

129

providing a wage increase more generous to potential workers who are likely to involve
in the process of unionization. Thus, the unexpected finding should not be read a sign
showing that unions increased wage inequality at that time.

The incentive held by employers became smaller as the government began to repress
union movement from 1990. Also, employers in the nonunionized sector started to
introduce pay-for-performance wage schemes in the 19905. As a result, union wage
equalizing effects become visible in the 19905.

Many trade unions have resisted many employers’ will to introduce performance-
based wage system, resulting in increased wage dispersion, in the 19905 and a large
portion of the unions succeeded in deterring the employers from introducing the system
(KSLI, 1998). It is very likely that the union, in which blue-collar workers, high school
graduates, and workers with longer tenures are a majority, can succeed in deterring their
employers from introducing the pay for performance (KSLI, 1998). For example, 79.3%
of unionized firm can introduce a change in wage payment only after being agreed upon
by unions, according to a survey conducted by the KSLI (1998).

This study supports the role of trade unions by finding union diminishing wage
dispersion within establishments. However, the size of union effects on wage dispersion
is smaller than that of the United States ranging 10-25% (Freeman, 1980 & 1982). It
appears that the small size of union wage-equalizing effect comes from the relative
higher level of wage equality in Korea. Egalitarian societies like Asian countries like
Korean and Japan have emphasized the symbolic role of pay equity. The Korean
government also has valued equality in wages until at least the early part of the 19905 in

order to minimize workers’ complaints about low wages resulting from export-driven

130

economic development and about chaebols dominating the economy. In this situation, the
gap in wage dispersion between unorganized and organized sectors has been lower than
that in the western countries. The partial introduction of pay for performance in the 19903
appears not to change the wage dispersion completely within firms. The relative smaller
size of the union effect on wage dispersion found here should be attributed to the

egalitarian wage structure, which has existed for a long time.

5. Conclusion

This study does not emphasize greately how differences in bargaining structures are
related to the different levels of wage dispersion. It is argued that the bargaining
structures have a significant influence on the level of wage dispersion (Rowthom, 1992).
For example, wage dispersions among workers are much higher in the countries where
bargaining is decentralized, as in the U.S. and Japan, than in the centralized economies of
Scandinavia (Rowthom, 1992). Korea seems to be an exception to the argument in that
the Korean economy with a decentralized enterprise union system has shown a extremely
low level of wage dispersion, compared to the U.S. and Japan with a similar
decentralized bargaining structure. However, we already know that the level of wage
inequality in economy in Korea is much smaller than that in United States. While greater
wage inequality in U.S. may allow a larger union wage-equalizing effect to exist, small
wage inequality may not allow for big union wage-inequality in Korea.

Therefore, an interesting question for future research is to investigate how different

levels of wage dispersion are found across the three countries, where they share similar

131

decentralized bargaining structures. In addition, it would be useful to examine to see if
different levels of decentralization of bargaining structures across industries produce
different levels of wage dispersion in Korea.

The findings in this chapter have indicated that the trade unions in Korea have
decreased wage dispersion within establishments for the period examined. Prior research
on wage dispersion, however, has not answered the question of what amount of wage
dispersion is optimal (Bloom & Michel, 2002). It is difficult to tell whether union effects
on wage dispersion so far are close to an optimal level or already beyond a desirable
level. The question depends on which criterion would be adopted to judge the optimum
of degree of wage dispersion. In addition, the judgment will be varied according to the
organizational characteristics (e.g., manufacturing versus non-manufacturing), the subject
of the judgment (employers, unions, average workers, and a few employees at the top of
the wage structure, and the conflicting goals of the subject (e.g., competition versus
collaboration). Thus, most of the research on the union wage-equality effect including
this study has remained focused on the existence of the effect and has not gone further in
understanding whether the size of the effects is desirable or how the reduced wage
dispersion affect outcomes such as job-related satisfaction, organizational commitment,
productivity, product quality, and profit level. Hopefully, future study on union wage-
equality effect will answer the questions.

As expected, wage inequality will continue to increase for the time being when
considering the expectations from management, labor, and government. The principle of
‘generosity to the less-paid and lesser generosity to the better-paid before 1987 is not

supported by the government currently. However, the implicit policy on low wage

132

providing a wage increase more generous to potential workers who are likely to involve
in the process of unionization. Thus, the unexpected finding should not be read a sign
showing that unions increased wage inequality at that time.

The incentive held by employers became smaller as the government began to repress
union movement from 1990. Also, employers in the nonunionized sector started to
introduce pay-for-performance wage schemes in the 1990s. As a result, union wage
equalizing effects become visible in the 19905.

Many trade unions have resisted many employers’ will to introduce performance-
based wage system, resulting in increased wage dispersion, in the 19903 and a large
portion of the unions succeeded in deterring the employers from introducing the system
(KSLI, 1998). It is very likely that the union, in which blue-collar workers, high school
graduates, and workers with longer tenures are a majority, can succeed in deterring their
employers from introducing the pay for performance (KSLI, 1998). For example, 79.3%
of unionized firm can introduce a change in wage payment only after being agreed upon
by unions, according to a survey conducted by the KSLI (1998).

This study supports the role of trade unions by finding union diminishing wage
dispersion within establishments. However, the size of union effects on wage dispersion
is smaller than that of the United States ranging lO-25% (Freeman, 1980 & 1982). It
appears that the small size of union wage-equalizing effect comes from the relative
higher level of wage equality in Korea. Egalitarian societies like Asian countries like
Korean and Japan have emphasized the symbolic role of pay equity. The Korean
government also has valued equality in wages until at least the early part of the 19905 in

order to minimize workers’ complaints about low wages resulting from export-driven

133

inequality in tandem with low wages cannot be sustained as real wages have increased
rapidly. Overall, increase of wage dispersion during the Korean financial crisis cannot be

viewed as a short-lived phenomenon.

134

CHAPTER SIX

UNION EFFECT ON FOUR WAGE DETERMINANTS: EDUCATION,

EXPERIENCE, TENURE, ANS GENDER

1. Introduction

In the previous chapter, it is stated that the Korean unions had narrowed wage
dispersion in the organized sector, which was greater than in the unorganized sector
during the period of 1987-1999. Therefore, the question here is whether trade unions
affect the rates of returns for human capital for workers regarding education, experience,
and tenure, and the effects of personal characteristics such as gender, regarding wages,
when compared to the unorganized sector. The question is accompanied by another
related question, how significantly unions reduce return rates for education, experience,
and tenure and the coefficient of gender. Regarding the whole period concerned in this
study 1987-1999, whether any significant changes in the union effects on the coefficients
for the four variables within the period are evident should be one of our interests, too.

To answer these questions, this chapter will use individual-level data from the BSWS
that were used in the previous chapter dealing with the union wage premium in the
distribution of wages. Individual level-data will be utilized to determine if significant
negative effects of unions on return rates for education, experience, tenure, and sex are
present. The results from individual-level data will be estimated by OLS and quantile

regression.

135

The results, estimated by OLS and quantile regression, from individual-level data show
that trade unions really did reduce the return rates for education and tenure in most of the

years between 1987-1999 and experience and gender in many years during the 19905.

2. Model and Data

The data used here are individual-level data from the BSWS, which are the same as
the data in the previous chapter, which examines the union wage effect on the distribution
of wages. This individual-level data are used to see if significant differences exist in
return rates for education, experience, and tenure and in the coefficient of sex between
unionized and non-unionized sector.

The model that will be estimated in this chapter is

an,-= a+X,- 1+8; (6.1)
whether In W,- represents the natural logarithm of monthly gross wages, divided by hours
worked, for individual i, X,- represents a vector of individual characteristics and
organizational characteristics regarding the establishments which the individual works
for, and e,- is a normally distributed error term. When estimating in unionized sector, the
equation of 7.1 will be as follow,

In Win = a- + Xmflr + 8m (62)
whether In W,-,, represents the natural logarithm of monthly gross wages, divided by hours
worked, for individual i, whose establishment is unionized. Also, the wage equation
estimated in the non-unionized sector would be as follow,

In Wa- = a,- +Xinflt+ 8m (63)

136

whether In Wm represents the natural logarithm of monthly gross wages, divided by hours
worked, for individual i, whose establishment is not unionized.

Due to space limitation, I do not present descriptive statistics for all variables of all,
unionized, and non-unionized workers, in each year in the period 1987-1999. Therefore,
instead of those statistics, the descriptive statistics for the period 1987-1999 are presented

in Table 4-1 in the previous chapter examining union effect on wages.

3. Results

Results from OLS estim_ation

Figure 6—1 shows the graphs for the coefficients of four major wage determinants,
estimated by the OLS, between unionized and non-unionized sectors for 1987-1999. The
logarithm of hourly real wage is used as the dependent variable in the wage equation. The
covariates of the four major wage determinants are and four regional dummies, seven
industry dummies, four establishment dummies, marital status, experience squared,
tenure squared, and a remaining education-related variable for two-year college
graduates.

Figure 6-1, (a), shows the effect size of education, which is represented by high
school and university graduates compared to elementary and middle school graduates,
between the unionized and non-unionized sectors during the period 1987-1999. With an
exception of 1990, the sizes of coefficients for high school and university graduates are
larger in the non-unionized sector than in the unionized sector in all years. It appears that

the unions are associated with the reduced effects of education on wages in every year.

137

Figure 6-1. Return rates for the four determinants of wages, estimated by the OLS,
between non-unionized and unionized sector for 1987-1999.

(a) education (university and high school graduates)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6

3 o 5 /\
D ' /
3.; 0.4 /\>\ w’ ------ Non-TU(High Sc.)
2 O3 / \\ _ ---TU(High SC.)
.3 ' "’ \ —Non-TU(Univ.)
E 0-2 , —TU(Univ.)
g ., \/
0 0-1 f7 «....» 3 "“-=?r"“"""~-......—/‘“‘\...

O 1 1 1 1 1 1 1 1 1 1 1 1

1987 1990 1993 1996 1999
year
(b) experience

0.03
Ill
n.(1025
X
I“ 0 02 \ A
3 0 015 \ /—/A\\/\/7 —Non-TU
o _--
g 0.01
8 0.005

O 1 1 1 1 1 1 1 1 1 1 1 1
1987 1989 1991 1993 1995 1997 1999
year

 

 

 

138

(c) tenure

 

 

coefficient of TENURE

0.15
0.12
0.09
0.06
0.03

O

 

 

 

 

— Non-TU
—TU

 

 

 

 

 

 

 

 

1987 1989 1991 1993 1995 1997 1999
year

 

((1) sex

 

 

coefficient of SEX

—0.05
-O.1
-O.15
-0.2
-0.25
-O.3
-0.35
-0.4
-O.45

year

 

 

 

— Non-TU
——TU

 

 

 

 

139

 

The gap in the size of coefficients between the unionized and non-unionized sectors
becomes wider in the 19908, for both high school and university graduates. Although it
might be an extreme case, the difference in the size‘of the coefficients between the two
sectors is .09 (9.4%) for high school graduates and .152 (16.4%) for university graduates
in 1999. It means that the wage difference between workers who graduated elementary or
middle schools and workers who graduated high schools is larger in the non-unionized
sector by 9.4% than in the unionized sector. Furthermore, it means that the wage
difference between workers who graduated elementary or middle schools and workers
who graduated universities is greater in the non-unionized sector by 16.4% than in the
unionized sector. The amounts of difference regarding education between the two sectors
are quite large, assuring our prediction with respect to the impact of unions on human
capital.

The results relating to experience, (b) in Figure 6-1, are confusing in that the size of a
year of experience is higher in the unionized than non-unionized sector for 1988-1993
and lower in the unionized than the non-unionized sector for 1994-1998. Our expectation
that size of experience would be lower in the unionized than the non-unionized sector fits
the fact only for 1987 and 1994-1998.

The Figure showing the size of the coefficient for a year of tenure for 1987-1999, (c) in
Figure 6-1, appears simple. In every year, the size of coefficient for tenure is lower in the
unionized sector than in the unorganized sector. Except for a considerable decrease in the
effect of tenure for 1987-1988, the absolute sizes of the coefficient are stable in both the
unionized and the non-unionized sector. For example, the difference in the effect size of

tenure between the unionized and non-unionized sector is .014 (.077 for the non-

140

unionized and .063 for the unionized sector) and the mean of tenure for the entire sample
is 6.62 in 1997. Thus, the difference between the two sectors results in the decrease of
wages for average workers with mean values of other characteristics by 9.27% in the
unionized sector, relative to comparable workers in the non-unionized sector. Figure 6-1
supports the notion that trade unions are likely to reduce the effect of human capital on
wages.

In 1987, the size of the coefficient for sex is -.366 in the unorganized sector and -.392
in the unionized sector, as seen in (d) in Figure 6-1, implying that female workers earned
smaller wages by 44.2% than male workers in the unorganized sector and by 48% than
male workers in the organized sector. We may find that a huge gender discrimination
toward female workers with regard to wages is evident. Also a seemingly interesting fact
is that the size of the coefficient for females is lower in the organized sector than in the
unorganized sector for 1987-1989 and 1999. However, the differences in the coefficients
for female workers between unionized and non-unionized sectors for the period 1987-
1989 are not statistically significant at the 5% level.40 Even though unions are supposed
to protect marginal workers like female workers more, in reality structural discrimination
toward female workers is also severe in the organized sector as in the unorganized sector
for 1987-1989.

From 1990, the trade unions appear to do what they are supposed to do, protecting
marginal workers like female workers. Even though the results are not presented here, the
size of the coefficient for interaction of the unions with gender in the wage equation

having full interaction terms of the unions with independent variables is .038 in 1990,

 

‘0 It means that, in the wage equation with full interactions of unions with other independent variables, the
coefficient of ‘union‘gender’ for 1987-1989 was not significant at the 5% level. The result for this fact is
available upon request.

141

.030 in 1991, .027 in 1992, .029 in 1993, .044 in 1994, .055 in 1995, .040 in 1996, .022 in
1997, and .045 in 1998, respectively. The interaction terms are all significant statistically
at the 1% level. Thus, the implication is that, due to trade unions, the wages for female
workers in the unionized sector is larger by 2.7 % (= exp[.027]-1) in 1992 or 5.7% (=
exp[.055]-1) in 1995 than in the non-unionized sector.

The difference in the coefficient for female workers between the unionized and non-
unionized sectors becomes negative (-.023) again and it is statistically significant at the
5% level.

Figure 6-1 shows the decrease in the size of the coefficient for gender for 1987-1999.
The size of the coefficient for female decreases from -.366 in 1987 to -.259 in the 1999 in
the unorganized sector and from -.392 in 1987 to -.289 in 1999 in the organized sector.
For the thirteen years, the size of the wage gap between female and male workers
decreased 14.6% in the unorganized sector and 14.5% in the organized sector for the
entire period of 1987-1999. For the period of 1990-1998, the size of the coefficient for
female workers is higher in the organized sector than in the unorganized sector, meaning
the union’s contribution to reduces the gender wage gap.

In sum, the results from OLS implies that the unions for 1987-1999 clearly reduced the
effects of human capital such as education and a year of tenure on wages, except
experience. Also trade unions for 1990-1998 increased the size of coefficient for female

workers.

142

Figure 6-2. Return rates for high-school graduates and university graduates, estimated by
quantile regressions, between non-unionized and unionized sector for 1987-1999. (HS
and U means high-school and university, respectively. Non-TU and TU means non-
unionized sector and unionized sector, respectively. Reference group is workers
graduated middle-school or less.)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Return rates for education in 1987

0.6

0.5 W/
'5 0.4
.3 0.3 EMAE " "AH -- HSin Non-TU
E 02 4" J ----HSinTU
§ . RA! __,,-. -----;.-—./” U in Non-TU

0.1 .'V“" -:;;.‘J__,,...-r~.-_v-_ 7:11,..‘2’... “.1,..- ”W _ _ _ _ U inTU

Oi‘gf’fn111g111141111411111
-0.1 —
quantile
Return rates for education in 1988

1
. 0.8
g , —~--—--~-HS in Non-TU
'§ 0-5 _,. ~---HSinTU
:5 0.4 Uin Non-TU
3 - - - -u in TU

0.2 — —:— VMWMgJfi-
01111111111111111111
1 11 21 3141 51 61 71 81 91
quantile

 

 

 

143

 

 

Return rates for education in 1989

 

 

 

 

 

 

 

 

 

0

 

 

0.8
0.7
§ 3': —-Hs in Non-TU
'1} 0'4 ----HSinTU
55-", 0'3 0 in Non-TU
o . - . - - .
o 02 U In TU
0.1
O
1 1121 314151617181 91
quantile
Return rates for education in 1990
1
- 0.8
5 -—--—--HS in Non-TU
"g 0.6 ..... HSinTU
:5. 0.4 U in Non-TU
o - - - -u in TU
° 0.2

 

 

 

 

1112131415161718191
quantile

 

 

144

 

 

 

 

Return rates for education in 1991

 

 

 

 

 

 

 

 

 

 

 

 

 

 

g -~ HS in Non-TU
'§ - - - - HS in TU
% U in Non-TU
8 - - - - U in TU
1 112131415161718191
quantile
Return rates for education in 1992

0.7

0.6
g 0.5 ~——---———-~ HS in Non-TU
'§ 0.4 ----HSinTU
’5 0.3 U in Non-TU
8 0.2 ----UinTU

O

 

1112131415161718191
quantile

 

145

 

 

coefficient

 

Return rates for education in 1993

 

 

 

0.6

0.5

0.4

0.3 ~-~~-------~~ HS in Non-TU

02 ,_ A . ~ -HS in TU

0,1 ':,;.1:.:;;:;;;:_ ” U i" NWT”
0 - - - - U in TU

 

 

1112131415161718191
quantile

 

 

coefficient

 

Return rates for education in 1994

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6
0.5 f
0.4 / -~-~-HSin Non-TU
03 W - - - -HSinTU
' ‘fi— 1 - [A U in Non-TU
0-2 W ‘ 1.133,. - - - - U in TU
01 i" " W “ml/w”
”...”...
O 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 11 1
1 1121314151617181 91
quantile

 

146

 

 

 

 

Return rates for education in 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6
0.5 l/
'5' 0.4 ‘ /' HS in Non-TU
'§ 03 W ~--—-—-HSInTu
5‘5 ' \_ v fl -——Uin Non-TU
30.2 ~ ..— ----UinTU
0.1 1:; """" .. ...» we
0 l l 1 1 i l 1 1 l l l 1 l L J i l i J
1 11 21 31 41 51 61 71 81 91
quantile
Return rates for education in 1996
0.5
E 0'4 w HS' N TU
ewe-e In on-
d)
E 0'3 M M - - - -HSinTU
% 0.2 ' ,_.---—.. W W U in Non-TU
3 m e’ - - - - U in TU
0.1 1 u“ 'Wwi..,...wr~"*“fi‘ “
Pmmowww‘w
OJlljlllJlJiJliLllll
1 11 21 314151617181 91
quantile

 

 

147

 

 

 

Return rates for education in 1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1112131415161718191
quantile

 

148

 

 

 

0.5
.. 0.4 ' 4N .
.3 03 WA” MHWJU
g ' W - - - - HS In TU
15 0.2 7 Uin Non-TU
8 ~, - ”7,47,; - - - -U in TU

0 . 1 “if: :iT :::;W“W*° -.....» ‘L———‘

o 1 1 L 1 1 1 1 1 1 1 L 1 1 L441 1 1
1 1121 3141 5161 7181 91
quantile
Return rates for education in 1998

0.6
a 0.5 I
g 0.4 M ""’_—"'~ HS in Non-TU
'8 _ --—--HSinTU
3;, 0'3 ' 77 " —Uin Non-TU
8 0.2 \ WWW“ _r7777,_,---_..-‘.----77 ,1 - - - -Uin TU

0. 1 ff, 4 7,...fl:-4::.:;,T.::;::P“=‘“‘T

V
04114111111111111111

 

 

 

 

 

 

coefficient

Return rates for education in 1997

 

 

 

 

 

 

 

 

 

 

 

 

 

0.5

‘ -~—--—~—- in on-
0.3 W A” - ~ - - HS in TU
02 77,1" U in Non-TU
0.1 e+:,::.~... 77773.17»- -- - we”

0 1141111L11111111111
1 11 21 3141 5161 71 81 91
quantile

 

 

 

coefficient

Return rates for education in 1998

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.6
0.5 I
0.4 M; —--—-—-- HS in Non-TU
W — - - - HS in TU
0'3 ’ “v ' Uin Non-TU
0-2 \ 7 7-.-... 7_7,-7-_-7__,,_-._W . - - - - U in TU
0.1 73.1.7777“; ..... www.cwwmme-‘w ..
01111L11111111111111
1 112131415161718191
quantile

 

 

148

 

 

 

 

 

Return rates for education in 1999

0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1

O

 

—~—--~---—~ HS in Non-TU
- - - HS in TU

U in Non-TU

- - - - U in TU

 

coefficient

 

 

 

 

1112131415161718191
quantile

 

 

 

Results from Quantile Regression

Figures from Figure 6-2 to Figure 6-5 displays the difference in the return rates for
education (Figure 6-2), a year of experience (Figure 6-3), and a year of tenure (Figure, 6-
4) and in the coefficient of sex (Figure 6-5) between the unionized and non-unionized
sectors. These figures are drawn from individual level data from the BSWS. Also the
figures are estimated by quantile regressions, which was used in the chapter 4. However,
quantile regression is conducted on each sector separately, while it was done for the
whole sample in the previous chapter.

The comparison of the return rates for high school graduates and university graduates
between the unionized and non-unionized sectors for 1987-1999 in Figure 6-2 is evident.
For example, it is clear that the return rates for university graduates are larger in the non-
unionized sector than the unionized sector in 1987, implying the union wage-equalizing
effect. However, the difference in the return rates for high school graduates between the

unionized and non-unionized sectors is much smaller than a comparable difference for

149

university graduates between the two sectors.41 Also a large difference in return rates is
evident for high school and university graduates between the two sectors. The trend
continues until 1993. The extreme case would be 1990, where it appears that only two
lines are seen, one for high school graduates and another for university graduates,
because no visual difference is evident in the return rates for both high school and
university graduates between the unionized and non-unionized sectors. This finding
indicates that little union effect on reducing return rates for education in that year is
evident. However, the trend regarding the return rates for education changes are clear in
1994. The trade unions in that year clearly reduced the coefficient for high school
graduates as well as university graduates, compared to the non-unionized sector. For
example, the return rates for high school graduates in the unionized sector is below half
of the rates in the non-unionized sector at the lower half of the wage distribution in 1999.
Also the return rates for university graduates in the unionized sector are in the middle
between the return rates for high school graduates for the non-unionized and university
graduates for the non-unionized.

Figure 6-3 shows yearly changes in the return rates for a year of experience between
the unionized and non-unionized sectors for 1987-1999. The figures show that the slope
of the return rates for experience is upward, meaning that the workers at the top of the
wage distribution are likely to benefit more from an added year of experience than do the
workers at bottom or middle of the wage distribution. It is clear that the difference of the

return rates between the two sectors over many years are not large, even though the

 

4' More precise comparison of coefficient size in between unionized and non-unionized sector should be
done by statistical tests. It is possible to add a variable showing interactions between education and union
in the equation. Because of time limit resulting from the nature of quantile regressions with 20 bootstrap
repetitions, the job is not done in this study. The situation is the same for experience, tenure, and sex.

150

Figure 6-3. Return rates for a year of experience, estimated by quantile regressions,
between non-unionized and unionized sector for 1987-1999.

 

Return rates for a year of experience in 1987

0.05

 

 

 

 

 

 

Non-TU
- - - -TU

 

 

 

coefficient

 

 

 

o lLJllLllllllJ_LLllll

1112131415161718191
quantile

 

 

 

 

 

Return rates for a year of experience in 1988

 

0.04
0.03 I

M Non-TU
0.02 I if ____TU
0.01 ~—

OlJlJJJlllllllJJiLJl

1112131415161718191
quantile

 

 

 

 

 

 

 

coefficient

 

 

 

 

 

 

151

 

 

coefficient

0.06
0.05
0.04
0.03
0.02
0.01

-0.01

Return rates for a year of experience in 1989

 

 

l

 

 

.o'l

 

 

 

 

 

 

.I —— Non-TU
Jim—“v "

'/
i_IAJ l l l 1 l__l 1474 l J l 1 J

-W4—94—
quantile

 

 

 

 

 

 

 

coefficient

0.04
0.03
0.02
0.01

0

Return rates for a year of experience in 1990

 

 

 

 

—l-' —— Non-TU

 

 

 

 

W ""T”

 

V7

J 1 l l l l l l J l l l L 1 1 L l J l

 

 

 

1112131415161718191
quantile

 

 

 

152

 

 

 

coefficient

Return rates for a year of experience in 1991

0.035
0.03
0.025
0.02
0.015
0.01
0.005
O

1112131415161718191
quantile

 

 

— Non-TU
----TU

 

 

 

 

 

coeffcient

Return rates for a year of experience in 1992

0.035
0.03
0.025
0.02
0.015
0.01
0.005

 

1112131415161718191
quantile

 

153

 

 

—Non-TU
- - - -TU

 

 

 

 

 

 

 

coefficient

Return rates for a year of experience in 1993

0.07
0.06
0.05
0.04
0.03
0.02
0.01

0

1112131415161718191

quantile

 

 

—Non-TU
- - - -TU

 

 

 

 

 

 

coefficient

Return rates for a year of experience in 1994

0.06

 

 

0.05
0.04

 

 

 

0.03
0.02 54—4-95

 

— Non-TU
----TU

 

 

 

0.01 -M

 

 

 

 

0111111111111111111

1112131415161718191
quantile

 

 

154

 

 

 

Return rates for a year of experience in 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.04
‘5 0.03 j
o
'8 ———Non-TU
% 0.02 _ _ _ -TU
o
0 0.01 —\
0 1 1 1 1 1 1_L 1 1 1 1 J 1 1 1 1 1 1 m
1 11 21 31 41 51 61 71 81 91
quantile
Return rates for a year of experience in 1996
0.04 [’r"
g 0.03 /_/J",\'.
'§ 7 7,.” ' Non-TU
% 0.02 MI - - ' 'TU
3 0.01 [fw—
0 “71,-: 1 #4 1 1 1 1 1 1 1 L4 1 1 14 1

 

 

1112131415161718191

quantile

 

155

 

 

 

coefficient

Return rates for a year of experience in 1997

 

0.05
0.04

 

 

 

 

 

 

 

0.03

—Non-TU
0‘02: ff"? ----Tu
0.01m
07 1411111411L111

4,701 11 21 31 41 51 a1 71 a1 91

 

 

 

 

 

quantile

 

 

 

coeffcient

Return rates for a year of experience in 1998

0.05
0.04

 

 

 

 

 

 

 

0.03 _ : Non‘TU
0.02 W - - - 'TU

 

 

 

 

0.01 _H

q
\

0‘ 11111111111111111

1112131415161718191
quantile

 

 

 

156

 

 

Return rates for a year of experience in 1999

 

 

 

 

 

 

 

 

 

 

0.05

0.04 J!
E”, 0.03 ..
'5 ‘ ———Non-TU
15 0.02 ----TU
3 o.o1~

0 1 1 1 1 1 1 1 1 1 1 1 1 [L1 1 1 14
43.01 1 11 21 31 41 51 6111 131 91

 

 

 

quantile

 

 

 

absolute size of the coefficient ranging from 0 to .04 is smaller than other coefficients
such as education, tenure, and sex. Until 1995, we cannot determine that the return rates
for a year of experience in the unionized sector are lower than in the non-unionized
sector. Even the return rates for experience in the unionized sector are apparently larger
than in the non-unionized sector in 1990 and 1993. It is the period of 1996-1998 when the
return rates for a year of experience in the unionized sector exceed that in the non-
unionized sector.

The comparison of the return rates for a year of tenure between the unionized and non-
unionized sector for 1987-1999 is shown in Figure 6-4. First of all, it is evident that the
slope of the coefficients for tenure is downward at a decreasing rate as a quantile in the
distribution of wage increases. It implies that workers at lower end of wage distribution
are likely to benefit less from an increase of a year of tenure than are workers at the upper

part of the wage distribution. This finding cannot be obtained through the OLS estimation

157

Figure 6-4. Return rates for a year of tenure, estimated by quantile regressions, between
non-unionized and unionized sector for 1987-1999.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Return rates for a year of a year of tenure in 1987
0.25
.. 0.2
12:} 0-15 M —-—Non-TU
8‘; 0.1 \ ----TU
° 0.05 ‘
0111L1L1114141111111
1112131415161718191
quantile
‘1
Return rates for a year of tenure in 1988
0.1
.. 0.08 M
5
.5 0.06 Non-TU
E 0.04 - - - -TU
0.02
01111111111111111111
1112131415161718191
quantile

 

 

 

158

 

 

Return rates for a year of tenure in 1989

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0.12
.1 0.1
E ' N ----TU
0.02
01111nm1111111111111
1 11 21 31 41 51 61 71 81 91
quantile
Return rates for a year of tenure in 1990
0.1
.. 0.08 ‘-
5
'§ 0.06 ‘ —'—NOfl-TU
gm —w
" 0.02
01111111111111111111
1 11 21 31 41 51 61 71 81 91

quantile

 

159

 

 

 

Return rates for a year of tenure in 1991

0.1
0.08 M

0.06 V w

 

 

 

w —Non-TU
0.04 - - - 'TU

 

 

 

 

coefficient

011111114144114mn111

1112131415161718191
quantile

 

 

 

 

 

 

 

Return rates for a year of tenure in 1992

0.1
0.08 W

0.06 k

1. ——Non-TU
0.04 N . . . 'TU

0.02

 

 

 

 

 

 

 

coefficient

 

 

 

o 11111111111111#L11

1112131415161718191
quantile

 

 

 

160

 

 

 

Return rates for a year of tenure in 1993

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1112131415161718191
quantile

 

 

 

 

0.12
.. 0.1 M\
C
m 0.08 [A
'§ 006 \ \ Non-TU
a“; . M ""TU
3 0.04

0.02

0 l I J l l l 1 l l l l l l 14 l l l 1
1112131415161718191
quantile
Return rates for a year of tenure in 1994

0.1
.. 0.08
E 006 r\ X
g - Non-TU
t N ----TU
41 0.04
3 a

0.02

0 111111111111111111

 

161

 

 

 

 

 

Return rates for a year of tenure in 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1

11 21 31 41 51 61 71 81 91
quantile

0.12
..- 0.1 “’\
r:
'5 0.08 \ N ——Non-TU
E 0.06 \ ----TU
§ 0.04 M
0.02
01111111111111111111
1112131415161718191
quantile
Return rates for a year of 1996
0.12
.... 0.1
§ coagxsx N TU
1%: 0.06 \fi ”~13"
3 0.04 V~
0.02
01111111111111111411

 

 

162

 

 

 

Return rates for a year of tenure in

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1997
0.1
:1; WW
'5 0.06 Non-TU
€0.04 \V ----TU
o 0.02
OJJIJIIILJJlilllllJl
1112131415161718191
quantile
Return rates for a year of tenure in
1998
0.1
La; 0.08 "m
'8 0.06 Non-TU
E w ...-
8 0.04 \7 TU
o 0.02
0111111111111L411111

 

 

 

1112131415161718191
quantile

 

163

 

 

 

 

Return rates for a year of tenure in
1999

0.12
0.1
0.08
0.06
0.04
0.02
0

 

Non-TU
----TU

 

 

 

 

coefficient

 

1112131415161718191
quantile

 

 

 

in the (c) in Figure 6-1. Second, the size of return rates for a year of tenure is less in the
unionized sector than in the non-unionized sector in almost every quantile in every year
for 1987-1999. It reveals clearly the union effects on reducing return rates for tenure.
Third, the size of coefficient for a year of tenure is stable in both unionized and non-
unionized sectors in the 19903, as also seen in the results from the OLS in the (c) of
Figure 6-1. It may mean that neither employers nor trade unions attempted to foster
internal labor market by increasing the return rate of tenure in the 19903.

Figure 6-5 shows the relative loss of wages for female workers in the unionized and
non-unionized sectors, compared to the male workers (reference group). First of all, the
fact is discovered that the size of the negative coefficients for female workers was
reduced throughout the late 19803 and 19903. In the late 19803 and early 19903, the
coefficients for female workers were less than -.3 in most quantiles. However, from 1994

or 1995, not many quantiles show coefficients for female workers larger than -.3 in both

164

the non-unionized and especially unionized sectors. The result means that the serious
issue of gender discrimination regarding wages lessened in the late 19903. The finding
was already predicted from the OLS estimation, as seen in (d) of Figure 6-1 presented
earlier.

In 1987, the average coefficient for female workers for all 99 quantiles in wage
distribution is -.402 in the unionized sector and —.366 in the non-unionized sector. This
finding means that female workers earned less wages by 49.5% in the unionized sector
and 44.2% in unorganized sector than comparable male workers having same
characteristics. An interesting point is that the relative loss of all female workers in wage
distribution except the workers in the lowest 6% is larger in the unionized sector than that
in the non-unionized sector. This pattern of coefficients for female workers in unionized
and non-unionized sector continued until 1989. From 1990, the return rates for female
workers in the unionized sector rose above those in the non-unionized sector. The
difference in coefficients for female workers, between the unionized and non-unionized 4
sectors, is the largest in 1995. Then, the average coefficient for female workers for all 99
quantiles in the wage distribution is -.274 in the unionized sector and -.311 in the non-
unionized sector. This fact means that the relative loss of wages for female workers in the
unionized sector is less by 3.8% than that for comparable female workers in the
unorganized sector. The result is what the literature expected regarding unions to protect
less—paid or marginal workers. However, the trend stops in 1999 and the coefficient for
female workers became less in the unionized sector than in the non-unionized sector. In
sum, the union effect reducing the wage gap between male workers and female workers

appears unstable for the entire period of 1987-1999.

165

Figure 6-5. The coefficients for sex (female workers), estimated by quantile regressions,
between non-unionized and unionized sector for 1987-1999.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Return rates for sex in 1987
O fif I I I I l I l l I I I I I I Tfi T
_0.1 11 21 3L41 SL 61 71 81 91
E -o.2
'6 1. ——Non-TU
if; .03 MW ' ' ' 'TU
3 -0.4 w— __\_\
-0.5 1
-0.6
quantile
Return rates for sex in 1988
o I I 1 I I I T I I I I Ifi I I I l I l
_0.1 I 11 21 31 41 51 61 71 81 91
‘5' .32
'8 ' - Non-TU
'1. .13 as»
-o.4 N
-0.5
quantile

 

 

 

I66

 

 

Return rates for sex in 1989

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

quantile

 

 

 

 

 

0 I I I I Hi I j I I I I I I I I T I I
_0.1 I 11 21 31 41 51 61 71 81 91
i—i
E1 '02 ——Non-TU
g -o.3 - ' ' ' ‘TU
0
-0.4 ‘7
-0.5
quantile
Return rates for sex in 1990
o I I I I I I I I I T I I I I I f I I I
I 112131415161718191
.- -0.1 *-
C
0
'§ Non-TU
15 ----TU
o
0

 

 

167

 

 

 

Return rates for sex in 1991

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 I I I I I I T I I I I I I T I I I 1 I
—0.1 I 11 21 31 41 51 61 71 81 91
i=2
.§ '02 K Non-TU
g -O.3 W ...-TU
0
-0.4 '
-0.5
quantile
Return rates for sex in 1992
O I I l T I I I I I I I I I I I I I I I
‘I 11 21 31 41 51 61 71 81 91
.... —0.1
c
'2 Non TU
u ..
I; '02 ‘ ----TU
8 '8‘
'0'3_— W
-0.4
quantile

 

168

 

 

 

 

 

 

Return rates for sex in 1993

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 I I I I I I I I I I I I I I I I I I I
I 11 21 31 41 51 61 71 81 91
... -0.1
C
'2 Non TU
u —— ..
a; —0.2 _ _ _ -TU
' W“
-0.4
quantile
Return rates for sex in 1994
O I I I I I II T I I I H I I I I I I I I
I 112131415161718191
..- -0.1
C
'2 Non TU
u -
$0.2 :----TU
8 _03 _CA\‘\ M
’ W
-0.4
quantile

 

169

 

 

 

 

Return rates for sex in 1995

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

0 I I I I I I I I I I I I Ifi I T I I I
-0.05 +——1—1—21——34——41—51——61—71—81—94—
g -0.1
E '0°15 ——Non-TU
if; -072 x.“ ----TU
o -0.25 " a
W.
-0.35
quantile
Return rates for sex in 1996
o I I I I I I I I I I I I I I I T I I I
.o,05 -I—1~1—24—31~—41—54—64—7—1—81—94——
E -0.1
3;, -0.15 ——Non-TU
if. -o.2 - - - -TU
3 -o.25 “L ’V
-073 _W
-0.35

 

 

 

quantile

 

170

 

 

 

Return rates for sex in 1997

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

o 1,7r,¥rrrTeTT.,Tfi.IT
-o.05 —I—44—24—34—44—54—6+—74—8~1—9+—
g -o.1
'5 4115 Non-TU
'{E -o.2 , - - - - TU
8 .025 1"
-o.3 ~
-o.35
quantile
Return rates for sex in 1998
o
41.05 +—4—1—24——34—44—54—64—7—1—84—94~—
g -o.1
-§ -o.15 Non-TU
E -o.2 - - - - TU
3 43.25 ”V
-o.3 W
-o.35

 

 

 

quantile

 

 

171

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Return rates for sex in 1999
0 I I I I I I I I I I I I I I I I I I T
-o.05 —‘l——4—1——24—3~1—44—54—6~1—174—84—91—
*5 411
43
E -o.15 j Non-TU
-o.2 . - - - - TU
§ -o.25 ‘\\— W
-o.3 -
p.35
quantile

 

 

 

 

Also, Figure 6-5 shows that the size of the coefficients varies across quantiles and across
years concerned. In the case of 1987, it appears that the size of coefficients for female
workers is stable across quantile, except for the bottom and top 10%. However, the graph
for size of the coefficients in 1988 and, especially, 1989 shows a slope downward trend
in both unionized and non-unionized sector, as the quantile concerned increases. It means
that the disadvantage of wages for female workers, compared to male workers, increases
as the quantile increases. Again, the relative loss in wages for female workers is larger in
the upper part of wage distribution than in the lower part of wage distribution. However,
the situation changed from 1995, showing that the disadvantage of wages for female
workers became larger in the median than in the lower or upper part of wage distribution.

Once again, the distribution of the coefficients for female workers changes in 1998 and

172

1999, implying that the relative loss in wages for female workers became less in the
upper part than in other part of wage distribution. It is especially evident, when
comparing the figure for 1988 or 1989 to equivalent figure for 1998. While it shows a
slope downward in 1988 or 1989, the graph for size of the coefficient turns to a upward
sloping in both unionized and non-unionized sectors in 1998.

From Figure 6-5 through the quantile regressions, we can conclude that gender
discrimination regarding wages was more severe for female workers in the upper part of
wage distribution for the period 1987-1995, while gender discrimination becomes more
serious for female workers located in the lower part of wage distribution in the late
1990s. Thus, it appears that no fixed pattern in the graph appears showing the size of the
coefficients for female workers across quantiles and years and graphs showing the size of
coefficients for female workers in the wage distribution varies, depending on quantiles
and years examined. This finding cannot be obtained, when examining sizes of
coefficients for female workers in the wage equation, which are estimated by the OLS, in

(d) of Figure 6-1 shown earlier.

Discussion

In the 19903, many trade unions showed their efforts to eliminate wage discrimination
by gender, especially in the finance industry employing a large portion of the female
workers (Jeong, 1993). This study reveals that the efforts by unions, especially in the
finance and banking industry, resulted in reducing the wage gap between male and
female workers with comparable characteristics. Also as expected, the Equal
Employment Act, enacted in 1987, which required employers to pay the equal wage for

work of equal value, contributed to the reduction of the wage gap between male and

173

female workers to some degree. However, it is unlikely that the wage gap between female
and male workers today is optimal excluding wage discrimination by gender in reality.
The wage schemes explicitly discriminating against female workers are found still in
many firms (Hankyeorae, October 22, 2002).42 Structural discrimination towards female
workers can be found even within trade unions. According to a survey by the Ministry of
Labour in 1995, the ratio of female workers is 9 % among the union leaders and 4% in
the presidents of unions, while the ratio of female workers is 22% of all the union
membership (Kwon, 1997).43

The yearly changes in return rates for education, experience, and sex between the
unionized and non-unionized sectors may imply that the role of unions reducing return
rates for those variables cannot be realized in a short time. In 1994, the union effects on
reducing return rates for high school and university graduates, compared to the non-
unionized sector, are visible. Also in 1990, the difference in return rates for female
workers between the two sectors becomes clear. It takes time for free collective
bargaining to reduce the return rates for much human capital for workers in the unionized
sector.

One of the most unexpected findings in this chapter is that the return rate for an
additional year of tenure is lower in the unionized sector than in non-unionized sector for
the entire period, 1987-1999. It is expected that trade unions tend to increase the tenure

for their workers by increasing the return rate for tenure (Freeman and Medoff, 1984).

 

‘2 In October 2002, the Ministry of Gender Equality issued a ministerial order requiring the Hanyang
University, one of the big universities in Korea, to revise the existing wage scheme discriminating against
female workers. So far the wage scheme forced female workers to earn less wages than comparable male
workers. Also, compared to males, females who graduated universities and who graduated high school had
to wait for three more years vs. two more years to be promoted, respectively (Hankyeorae, October 22,
2002).

‘3 I assume that no significant difference between male and female workers is present in the level of
intention to undertake the job of union leaders or union president.

174

However, the finding regarding tenure in Korea is contrary to the expectation. The result
that the return rate for tenure is lower in the unionized than non-unionized sector also
implies that a trade union in Korea is not a institution protecting and giving weight to the
seniority-based wage scheme, which is still strong in Japan, even though the Korean
industrial relations system are based on the enterprise union system as in Japan. It might
be read as a sign showing that the Korean unions do not contribute to the increase in
commitment of workers attached to their firms.

In fact, the average years of tenure for workers in the sample has changed a great deal
in between the two sectors. The average year of tenure for workers in the non-unionized
sector increased slowly from 3.75 in 1987 to 4.8 in 1999, while the number increased
rapidly from 4.59 years in 1987 to 8.24 years in 1999. Thus, the difference of the average
year of tenure for regular workers between the two sectors increased for 1987-1999,
though the return rate for tenure in the unionized sector has been always lower than in the
non-unionized sector. Thus, we can conclude that the increase in tenure in the unionized
sector, compared to that in non-unionized sector, does not result from the increase in the
return rate for tenure in the sector. Rather, the reason why the increase in the tenure
occurred more visibly in unionized than in the non-unionized sector should be sought in
the union wage premiums or other factors.

Why did the unexpected result of smaller return rate for a year of tenure in the
unionized than non-unionized sector occur? The first probable explanation for the result
is because younger workers, aged twenties or thirties with less years of tenure, have led

trade unions from 1987.

175

Finally, the significant decrease in the return rate for education in the period of the
1980s and 19908 appears to have affected white-collar unionism. The 19905 witnessed
the upsurge of white-collar unionism facing reduced wage differentials between blue- and
white-collar workers. Therefore, we can imagine that the decreased return rates of
education and other human capital could be the source of white-collar unionism to protect
white-collar personnel facing decreasing return rates for education, experience, tenure,

etc.

4. Conclusion

On the one hand, trade unions as institutions in the labor market affect the level of
wages throughout union wage premiums. The variable of unions in the wage equation,
indicating the direct effect of unions on wages, is shown to be significantly positive, as in
the chapter 4. On the other hand, trade unions affect wages indirectly throughout
changing the size of coefficients for some variables regarding human capital or personal
characteristics such as education, experience, tenure, and sex, in the wage equation. This
chapter shows the indirect effects of unions on wages, via education, experience, tenure,

and sex.

176

CHAPTER SEVEN

UNION EFFECT ON FRINGE BENEFITS

1. Introduction

This study examines the union effects on fringe benefits, specifically voluntary benefits44
which are not required by labor law, since 1987. The first question raised in this study is
whether union effects on benefits are present in Korea. When union wage effects exist,
the traditional explanation in industrial relations, assuming that unions are monopolistic
organizations with bargaining power to increase benefits as well as wages, expect union
effects on benefits exist, also. In addition, the greater importance of fringe benefits within
firms due to a weak social welfare system and ‘one-digit wage guideline,’ recommending
employers and unions to transfer wages to benefits, recommended by the government of
Korea in the 19905 also support the existence of union benefit effects. However, the
compensating wage theory assumes that no union benefit effects are present because of
negative tradeoffs between wages and benefits. In addition, excluding major benefits
such as pensions and health insurance from negotiations between employers and unions
because of statutory benefits may imply no union benefit effects exist in Korea.

By using data from the Enterprise Labor Cost Survey (ELCS), this study examines

both union effects on the level of total voluntary benefits expended by employers and on

 

‘4 Examples of required (or statutory) benefits are national pension, medical insurance, employment
insurance, and workers’ compensation. Voluntary benefits are defined as benefits which individual
employers can determine the level and whether they are provided. For examples of voluntary benefits in
Korea, refer to Table 2-3.

177

the provision of major benefits item. In addition, this study considers two additional
bargaining-power-related variables, the union organizing rate and the affiliation of
individual enterprise unions with either of the two national union confederations (the
FKTU and the KCTU), on both the level of benefits and provision of benefits. Using the
two variables will allow us to find any differences within the unionized sector regarding
union benefit effects. From a methodological perspective, this study adopts the two-stage
least square (ZSLS) to minimize a possible endogeneity problem between benefits and
wages.

The chapter is organized as follows: Section 2 describes Model to be estimated and
data used. Section 3 introduces the results and explains the implications from the results.

Finally, section VII concludes.

Model and Data

When we follow the traditional human capital earnings function (Mincer, 1974),
introduced earlier in chapter 3, the basic OLS model for union effect on the level of
benefits is as follows:

In B,= Xiflr+ UNION,- M + g, (7.1)
where B, is the amount of benefits, X,- is a vector of independent variables, which may
affect benefits, including personal characteristics such as sex, education, tenure, and
experience and establishment characteristics such as region, industry, and establishment

size, UNION,- is union status, and 5,- is a random error term.

178

However, the equation (7.1) is not complete because it is recommended in the
literature dealing with the union benefit effect that a trade union may indirectly affect
fringe benefits via earnings or wages (Freeman, 1991; Belman & Heywood, 1991). It is
likely that the level of benefits is positively correlated with the level of wages. Also it is
reported that the preferential tax treatment45 given to fiinge benefits grows with the level
of wages (Schiller and Weiss, 1980). Therefore, if the level of wages is placed into the

equation (7.1), the OLS model for union wage effect will be as follows:

In B,- = )Gflfk WAGEjflfi] + UNIONjflfi2 + 8] (7.2)

where WAGE,- is the observed level of wages and B,- , Xi, UNION, and 8, are same as in
(7.1).

This basic model using a union dummy variable as one of independent variables
assumes that the size of the coefficients for other independent variables except the union
variable is constant across unionized/nonunionized firms. However, union wage
differential studies report that many differences in characteristics regarding human capital
exist (Bloch & Kuskin, 1978). In fact, Even and Macpherson (1991) report that the
characteristics of unionized workers explain between one- and two-thirds of their higher
benefit coverage. Also, the process of determining shares and levels of fringe benefits
may be different in organized firms than in unorganized firms (Belman & Heywood,
1990). If so, a possibility exists that the size of coefficients for establishment

characteristics such as size, industry, and region, and for personal characteristics such as

 

‘5 Group health insurance and life insurance are examples of some non-wage benefits, which are excluded
fi'om the individual income tax base by the preferential tax treatment in the U.S. (Gentry and Peress, 1994).

179

education, experience, and tenure, are not the same between the two sectors. As a result,
the basic model having union dummy variable may not capture the true union effects on
shares and/or levels of benefits in total compensation (Fosu, 1984).

When the level of wages including bonuses and overtime payments is added into
equation in order to control any potential relationship between benefits and wages
(Mitchell and Smith, 1992), the equation (7.2) might still remain incomplete because
OLS may not work to estimate correctly the union effect on benefits, due to possible
correlated errors in wages and benefits (Belman & Heywood, 1991). To prevent the
possibility from occurring, a two-stage least square (2SLS) estimation is recommended in
the literature on union benefit effect.

The ZSLS estimation consists of two stages, first stage estimation using so-called
‘reduced form’ and second stage estimating the dependent variable in the OLS. First
stage regresses the endogenous independent variable, the observed wages, on variables in
X; and (a) instrumental variable(s). Thus, the first stage equation will be expressed as
follows:

In WAGE,- = X,,B,+ Instru mm + UNIONiflfiz + u, (7.3)

where WAGE,- is the level of observed wages, X,» is a vector of independent variables,
which used in the equation (7.1), InstruVi is a instrumental variable, which should not be
included in the second-stage equation, UNION; is union status, and u,- is a random error
term.

The two-stage estimation using the predicted wages obtained from the first stage

estimation, the equation (7.3), instead of actual wages in our study will be as follows:

180

In B,- = mm PredWAGEflw + UNIONUBW + 5] (7.4)

where PredWAGE, is the predicted level of wages rather than the observed level of
wages, WAGE,- in the equation (7.2), and 8,, X,, UNION), and 31 are same as in (7.1). This
study will estimate the equation (7.2) by using the OLS and also the equation (8.4) after
estimating the equation (7.3) by using the ZSLS.

By using the two-stage least square estimation, an identification problem arises in that
at least one of the independent variables, which is not included in the second stage
estimation process, should be used in the first stage estimation. Assume that the z is the
instrumental variable included in first stage, but excluded in second stage estimation. The
instrumental variable, 2, should satisfy two assumptions to obtain consistent estimators of

,6, and ,6” I in the equation above,

Cov(z, a) = 0 and Cov(z, WA GE) at O.

The key in conducting ZSLS estimation is how to obtain an appropriate instrumental
variable (Wooldridge, 2000; Olsen, 2002). This study uses squared experience as a
instrumental variable, which is expected to affect wages, but assumed to have no great
influence over benefit, following Belman and Heywood (1991). According to them,
experience squared is known to affect wages, it has been seldom used in the benefit

equation as one of independent variable in the literature.

181

In addition to the justification by Belman and Heywood (1991), another context-
specific justification exists for using experience squared as an instrumental variable. In
the U.S., where health insurance and pension share the majority of benefits, of which
collective bargaining between employers and unions can determine the level, it is likely
that it is married workers who have concerns about family health insurance and older
workers who have to prepare for their life after retirement will prefer benefits to wages
(Olsen, 2002). If so, experience squared might be related to benefits and, therefore, one
of the assumptions above, Cov(z, a) = 0, cannot hold. However, health insurance and
pensions in Korea are not issues that can be negotiated between employers and unions
because health insurance and pensions are categories of required benefits and the level of
the two is set by law as a fixed rate of the level of wages. Thus, in the context, it is

scarcely possible that experience squared affects benefits in Korea.

DLta

Which data are appropriate and available for study on union effect on fringe benefits
in Korea? Freeman and Medoff (1984) used two types of data in order to find the union
effect on fringe benefits. One is the Survey on Expenditures for Employee Compensation
(ECC) built by Bureau of Labor Statistics since 1982, reporting labor costs by individual
firms (Choi, 1992). The other is various surveys for individual workers such as Current
Population Survey where each individual worker is asked to answer questions about
whether he or she is covered by specific fringe benefits and how much the benefits are.
While surveys by firms make it possible to compare types and amounts of benefits across

firms, surveys by individual workers allow us to see the difference across individuals

182

after controlling for demographics (Freeman and Medoff, 1984). Most previous studies
on the union effect of fi'inge benefits are based on individual level data (Belman &
Heywood, 1990; Even & Macpherson, 1991) because many of them depend on Current
Population Surveys (CPS). However, this study uses establishment level data regarding
employers’ expenditures, the Enterprise Labor Cost Survey conducted by the Ministry of
Labor.46
The survey has been conducted annually since early 19805 and contains items asking
about the existence of unions, and the amounts of obligatory and non-obligatory benefits
paid by individual firms, etc. The population of the survey is all firms hiring at least 30
permanent employees (10 employees from 1997) and the size of the sample for the
survey is 2,406 in 1997, for example. The data from the Enterprise Labor Cost Survey
(ELCS) have plural strengths, compared to other data set used in other countries. First,
the data are not limited to specific types of benefits such as pension and health insurance
or a part of the benefits which employers offer to employees, but cover comprehensively
all expenditures regarding benefits including statutory benefits, voluntary benefits, and
even training-related expenditures and recruitment expenditures. Therefore, we can
probably observe the union effect on the whole features of the benefits, not on a specific
benefit, and not overlooking any misunderstanding resulting from the potential
relationship between a specific benefits examined and other benefits, when focusing on a
part of the benefits.
However, two important limitations of the data exist, too. Compared to individual-

level data like the CPS data, one of limitations of ELCS data in this paper is that the data

 

‘6 Data on employers’ expenditures are frequently used in the literature on union benefit effects (e.g.,
Freeman, 1981; Leibowitz, 1983; Fichtenbaum and Olson, 2002).

183

in the sample lack information on human capital such as experience, education, and
tenure.47 Thus, it is expected that the weakness of lack of information relating human
capital, except the percentage of female workers in the establishment, might be mitigated
by adding average values of education, experience, and tenure for 169 three-digit
industries from the Basic Survey on Wage Structure (BSWS)48 in relevant years and
information about employee wages into model in this study. A second limitation of the
data is that the data do not include information on trade unions in establishments.

For the purpose of this study, I needed to find relevant information on unions and
combine the information with the main data. Fortunately, the names of establishments for

1995-1999 could be identified.49

Variables

The dependent variables in this study are logarithm of the total amount of voluntary
benefits per employee and logarithm of total compensation per employee, divided by the
CPI deflator (1995=100).50 Also whether employers provide a specific form of voluntary
benefits or provision of voluntary benefits is used as another dependent variable. Because

the Enterprise Labor Cost Survey (ELCS) contains information on employers’

 

’7 The Employer Cost Index by the Bureau of Labor Statistics (BLS) in U.S. since 1982 also offers
comprehensive information on various benefit expenditures by employers. Also it lacks information on
human capital of employees such as education, experience, and tenure (Poterba and Rueben, 1994;
Carrington, McCue, and Pierce, 2002). It seems that ECI in U.S. and ELCS in the Korea are similar to each
other in this sense.

48 The BSWS, which is conducted by the Ministry of Labor (MOL) every year, is already used in the
chapter examining union wage premium in the distribution of wage. Refer to the relevant chapter.

’9 Names of establishment for data of 1994 also could be identified, based on regional codes, establishment
identification number, and information of various benefit costs in previous year included in the data of
1995.

5° When it is not possible to obtain data expended by employers for benefits, data on perceived value of
benefits to employees might be arguably used (e.g., Leigh, 1981; Olson, 1994).

184

expenditures for a specific form of benefits, a dummy value for a benefit equals 1 when
employers expended for the benefits, 0 otherwise.

Most studies focused on one or two forms of benefits to find any effects of unions on
benefits such as pensions and health insurance (Freeman & Medoff, I984; Belman &
Heywood, 1990; Even & Macpherson, 1991). In addition, a number of studies regarding
union effects on benefits are to focus exclusively on the provision of the particular benefit
and to examine whether any significant difference of probability exists in receiving the
particular benefit between union and non-union sector (Freeman, 1981; Fosu, 1984;
Belman and Heywood, 1990; Even and Macpherson, 1991).

However, these studies might be not adequate to view one of the two benefits to judge
the union effect on benefits. In addition, the approach to focus on the level of one of the
two forms of benefits cannot rule out a possibility of trade-off relations between each
form of benefits. Focusing only on the provision of benefits might not capture any
differences in the generosity of benefits. This study will consider the union effects on the
level of benefits and the provision of the particular benefits together, hoping a more
thorough picture regarding union effects on benefits will appear. Also, another dependent
variable, the share of benefits (and voluntary benefits) in total compensation, will be
considered if needed to discover any trade-off relationship between wages and benefits.

The two kinds of fringe benefits are divided between the voluntary, which are offered
voluntarily by employers, and the enforced voluntary fringe benefits, which the
employers are required to pay by law or government regulations (for example, Social
Security and unemployment compensation premiums). Little difference exists between

the unionized and unorganized firms in the portion of employee benefits with regards to

185

required fringes (Woodbury, 1983; Kaufman & Hotchkiss, 1999). In order to see union
differentials in fiinge benefits, it is necessary to focus on part of the voluntary fringe
benefits rather than the enforced fringe benefits.

Dummy variables for one-digit industries, based on the Standard Industrial
Classification code, and dummy variables for region (omitted category is Seoul) are
included in order to capture any difference in the union effect on benefits. Dummy
variables for establishment size (omitted category is establishments hiring less than 30
employees) are added into the regression equation because it is expected that
establishment size may be related to some benefits such as meals, employee housing, and
facilities for employee vacation and recreation, due to the economies of scale (Lester,
1967; Feuille et al., 1981). Moreover, establishment size may affect the share of benefits
positively because larger establishments, due to the possibilities of within-firm mobility,
likely indicate workers who have a longer tenure, which may result in a larger fraction of
benefits (Freeman, 1981). Also economies of scale regarding benefit purchase and
administration costs are expected to produce positive relations between size and benefits
(Fosu, 1979).

Regarding the characteristics of individual workers affecting organizational
productivity and collective preferences towards benefits, four variables are considered:
the percentage of female workers in the establishment, the average years of education,
experience, and tenure. In the context of the U.S., it is reported that the difference in sex
may result in different preference in the way that while males prefer wages, females have
greater interest in non-wage benefits (Ashenfelter and Johnson, 1972; Feuille et al, 1981).

However, a counterargument is presented in that females may have more interest in

186

 

wages rather than males because of relatively shorter tenure. Even though childcare
among the various items of benefits can be affected by sex because females have much
more interest in it, the portion of childcare in the total benefit is negligible as seen in the
table 2-3. This study does not have any presumed expectation about the sign of the
variable. The average years of education, experience, and tenure for 169 three-digit
industries from the Basic Survey on Wage Structure (BSWS) are added into the equation.
Previous studies suggest that well-paid workers, who are likely to have more education,
experience, and tenure, prefer benefits to wages. Thus, we expect that education, tenure,
and experience will be positively related to benefits.

Logarithm of monthly real wages per employee is included in the regression to
distinguish between the union effects on benefits via the level of compensation and via
the share of compensation (Freeman, 1981; Alpert, 1082; Fosu, 1984; Belman &
Heywood, 1991; Swindinsky and Kupferschrnidt, 1991; Even & Macpherson, 1991).

Three kinds of information of union are used in this study. For 1995-1999, the first
union-related variable, whether or not union is organized within the establishment, for
10,389 establishments has been checked through Annual Directory of Trade Unions in
Korea, published by the Ministry of Labor. Also four other union directories, provided by
the FKTU, by the KCTU, by the Labomews (http://www.labomews.co.kr) and by the
Laborworld (http://www.laborw.com), are used. Second are the union-organizing rates
within the establishment, calculated from the ratio of number of union membership

divided by total regular workers in the establishment.51 Third is the information regarding

 

5‘ It might be true that organizing rates, calculated from number of union membership divided by number
of bargaining boundary within establishments, is a better indicator of union density than the organizing
rates based on union membership divided by number of regular employees within establishments.

187

affiliation with the national union confederation (FKTU and KCTU).52 To identity
organizing rates and the affiliation with national union organizations, annual activity
reports by total 33 industrial union federations under the KCTU and F KTU and internal
unpublished directory by the KCTU have been utilized as well as the four union
directories listed. To minimize potential errors in calculating organizing rates,
establishments having an organizing rate of over 100%, which occurred due to original
incorrectness in the referred data or mismatch of survey between the ELCS and the five
directories, have been excluded from this analysis. To make sure of the organizing rates
and affiliation with either of two national unions, telephone interviews of 178 with at
least union full-time officers, who belong to 148 unions, have been done from November
2001, to April 2002.

This study focuses on the period of 1995-1999. Due to the nature of data as a pooled
cross-section data, year-dummy variables are used. Previous studies predict that
employees want the higher share/level of benefits in periods of relative price stability
(Mabry, 1973). Also employers are likely to cut benefits first before reducing wages in
the period of economic recession. Then, we expect negative effects of the year-dummy

variables during the Korean financial crisis, 1998-1999.

 

However, organizing boundary varies with each establishment and no data for it is available so far in
Korea.

52 Until 1997, the government had not authorized the KCTU. Thus, all government records and union
directory published by the FKTU until then argue that all individual unions are affiliated with the FKTU.
To solve the obscure affiliation problem regarding national organizations, I mainly relied on the union
directory published by the KCTU. When facing arguable cases, I sorted establishments according to two
criterion: which national union could be given union dues from individual unions and which national union
was ‘practically’ partner to individual unions in usual aspects, regardless of government records.

188

 

 

Table 7-1. Descriptive statistics for all variables in all, non-unionized, and unionized
sector for 1995-1999
Variables Explanation All Non-union Union
Lvolunp Logarithm of voluntary benefits 6.621 6.487 6.885“
per employee (1.099) (1.145) (.949)
Lbenep Logarithm of total benefits per 13.053 12.341 14.457 "‘
employee (1.855) (1.563) (1.561)
Rvolunta Ratio of voluntary benefits in 6.283 6.31 6.229
total compensation (%) (4.821) (5.009) (4.428)
Rbene Ratio of total benefits in total 20.214 19.124 22.659"
compensation (%) (9.966) (9.391) (10.633)
Lcashp Logarithm of wages per 6.894 6.908 7.135“
employee (.434) (.421) (.421)
Union 0=nonunion, l=union .337 - -
(.473) - -
Fpercent percent of female (%) 27.326 29.858“ 22.342
(21.909) (22.76) (19.177)
Busan 1= Busan .086 .09“ .078
(.281) (.286) (.268)
Kyungki l= Seoul suburbs .231 .238“ .216
(.421) (.426) (.412)
Joongbu l= Kangwon, Choongchung, & .207 .217" .185
Cholla (.405) (.413) (.339)
Kyungsang 1= Kyungsang .19 .193 .185
(.392) (.394) (.388)
Mining l= Mining .017 .018 .014
(.127) (.132) (.119)
Construction 1= Construction .014 .01 .022‘
(.1 16) (.097) (.146)
Transportation l= Transportation & Public .051 .066" .021
Utilities (.221) (.249) (.1 15)
Wholesale 1= Wholesale and retail Trade .079 .096" .046
(.27) (.295) (.209)
Finance 1= Finance insurance & Real .092 .053 .167"
estate (.289) (.225) (.373)
Services 1= Services .183 .207“ .136
(.387) (.405) (.342)
Public l= Public services .095 .099“ .087
(.293) (.299) (.281)
30$Ees<100 l= 30-99 employees .269 .365“ .08
(.443) (.255) (.272)
lOOSEes<3OO 1= 100-299 employees .265 .255 .283“
(.441) (.436) (.451)
300$Ees<500 l= 300-499 employees .104 .071 .17“
(.306) (.257) (.376)
SOOSECS<I ,000 I: 500-999 employees .I l I .059 215*
(.315) (.235) (.41 1)
1,000_<_Ees l= 1,000 or more employees .11 .042 .243“
(.313) (.201) (.429)
Yearl996 l=year ofl996 .161 .152 .179“
(.368) (.359) (.384)

 

 

189

 

Yearl997 1= year of 1997 .188 .17 223*
(.391) (.376) (.416)

Year1998 1: year of 1998 .248 .261 r .221
(.432) (.439) (.415)

Year] 999 1: year of 1999 .238 254* .207
(.426) (.436) (.405)
Samples (N) 8,479 5,625 2,854

 

 

Note: mean without parentheses and standard deviation in parentheses

Note: "‘ means significant difference between the two sector at the 5% level

Note: The contents of table is limited to the period 1995-1999 because the period 1987-1993 lacks
information on union density and national union affiliation and uses estimated probability of union to be
existed instead of union dummy variable.

Descriptive Statistics

Table 7-1 presents descriptive statistics of all variables used for all, non-unionized,
and unionized sectors for 1995-1999. Among 8,479 establishments used in analysis, the
number of non-unionized and unionized establishments is 5,625 (66.3%) and 2,845
(33.7%), respectively. T-test results imply that unionized establishments provide more
wages, voluntary benefits, total benefits than non-organized establishments. In addition,
the share of total benefits in total compensation is significantly higher in unionized sector
than in the non-unionized sector, while no significant difference is seen in the share of
voluntary benefits in total compensation between the two sectors. Non-unionized
establishments have shown a higher ratio of female employee in total regular employees
and can be easily found in small-sized establishments hiring less than 100 employees and

in every industry except construction and the finance industry.

Results and Implication

190

Union effect on the level of voluntary benefits

Table 7-2 presents the effects of unions on the level of voluntary benefits per
employee, measured by logarithm of voluntary benefits per employee, and estimated by
OLS and ZSLS for all samples for 1995-1999. Column 1 shows the results from the
equation without logarithm of real wages per employee as one of the independent
variables, that is, equation (7.1). Column 2 presents results from the equation, (7.2), using
logarithm of wages per employee as one of independent variable, to control probable
relationship between the level of voluntary benefits and wages, along with other variables
in column 1. When running the regression equation excluding logarithm of wages per
employee, the results in column 1 presents a modest size, 0.043 (4.4% = exp (.O43)-1), of
a positive union effect on the level of voluntary benefits. When controlling information
on wages, the size of the union effect on benefits decreased to a much smaller size (-
0.007) and the sign of effect changed to negative. The fact of a significant effect of wages
on the voluntary benefits in column 2 and an increase of R2 from 0.298 in column 1 to
0.355 in column 2 assures the role of wages in benefit equation and better performance of
the benefit equation (7.2) than equation (7.1). This finding also supports the indirect
effect of unions on benefits through affecting log wages (Belman and Heywood, 1991).

With respect to the region dummies, the level of voluntary benefits in suburb on Seoul
and Kyungsang area is higher than in Seoul, while voluntary benefits in Busan and other

L53

areas are lower than in Seou Differences in voluntary benefits across industries are

also found. The level of voluntary benefits is lower in the manufacturing industry,

 

’3 Suburban Seoul suburb and Kyungsang areas are the most important industrial areas where the majority
of manufacturing unions are located. Busan is the second largest city in Korea. Other areas including
Choongchung and Cholla are where the share of manufacturing or industrial activities are not developed.

191

Table 7-2. Union Effects on the Level of Voluntary Benefits

for 1995-1999 for All Samples (weighted)

 

 

 

OLS 2SLS
Model6
(DV=% of
Model 1 Model 2 Model 3 ‘ Model 4 Model 5 benefit)
Yearl996 .103"‘" 036*" 075““ .416‘" -.070*" -.825"*
(.001) (.001) (.001) (.012) (.003) (.018)
Yearl 997 . 140"" .066" * 099*" 489*" -.062*" -.736"*
(.001) (.001) (.001) (.013) (.003) (.018)
Yearl998 -.007**"‘ -.215"‘ -.205"“ 974‘" -.618"* 5074""
. (.001) (.001) (.001) (.037) (.006) (.039)
Yearl999 078“" -.236"‘" -.238"* 1.559‘" -.837*" 6398*“
(.001) (.001) (.001) (.056) (.009) (.057)
Busan -.007" .002 059*" -.048"* 040"“ -.432"‘"
(.002) (.002) (.002) (.005) (.003) (.019)
Kyungki .l47"“ .144*" .117“" .162‘" 086“" .019“
(.001) (.001) (.001) (.002) (.001) (.008)
Joongbu .l36*" .172*“ 216*" -.031"* 257*" 1590""
(.002) (.002) (.002) (.007) (.002) (.014)
Kyungsang 272“" 218*“ 209"" 526*" 064*" -.278"“
(.001) (.001) (.001) (.010) (.002) (.016)
Mining .736"* 293*" 222"" 2827*" -.510"‘" -3.552*“
(.006) (.006) (.008) (.077) (.001) (.077)
Construction .442‘" .443"* 214“" 439*" .141‘”'“" 1278*”
(.001) (.002) (.002) (.003) (.001) (.016)
Transportation -.366"* -.403*“ 051"“ -.191‘“ .029‘" .064“
(.003) (.003) (.003) (.008) (.003) (.028)
Wholesale 040*“ -.013"* 056"“ .288‘" -.089""‘ -.962"*
(.002) (.002) (.002) (.010) (.003) (.019)
Finance .l73"" 224*" .221‘" -.069"* 375“" 3399*"
(.002) (.001) (.002) (.010) (.003) (.020)
Services -.307*" -.497"* -.407"* 580*" -.823""‘ 4798*"
(.002) (.002) (.003) (.034) (.007) (.043)
Public -.491""' -.813*" -.767"* 1.027‘" -1.309"“ 6942“"
(.003) (.003) (.003) (.057) (.009) (.055)
30$Ees<100 .018“ -.042"* 1813*" 302"“ 4344"" 22.38?“
(.008) (.008) (.041) (.023) (.1 16) (.705)
100$Ees<300 .005 -.146*" 1655*" 716‘“ 3967*" 19.517““
(.008) (.008) (.040) (.033) (.1 15) (.696)
300$Ees<500 .198““' -.020* 1819*“ 1225*“ 4025*“ 19.526"*
(.008) (.008) (.040) (.043) (.l 15) (.693)
500$Ees<1,000 233*“ -.021** 1147”“ 1434*“ 3912"“ 18.935"‘M
9.008) (.008) (.040) (.049) (.1 15) (.692)
1,0005Ees 380““ .008 1754*“ 2134*" 3710*" 16.979‘"
(.008) (.008) (.040) (.069) (.l 14) (.687)
Fpercent -.008"* -.001"‘" -.001““ -.044"* 012*" .l l 1“"
(.000) (.000) (.000) (.001) (.000) (.001)
Edu .119“" 049“" 065*" 452*" -.082*** -.487*"
(.001) 4.001) (.001) (.013) (.002) (.015)

 

 

 

 

We can assume that the relationship between the voluntary benefits and regions is related to the role of

unions.

192

 

Expe -.056*" -.035*** -.030*** -.155*** -.008*** .118*"
(.000) (.000) (.000) (.004) (.000) (.003)
Tenure .090m 047*" .061 WM 293*" 009*" -.042***
(.000) (.000) (.000) (.008) (.001) (.006)
Lwage 1008*" 1104*" 4751*" 2.952m 15.2111"M
(.002) (.003) (.179) (.026) (.171)
Union 043*" -.007*** 278*"
(.001) (.001) (.009)
Density 003*" 004*M 031*"
(.000) (.000) (.001)
SQdensity 1.6e-06* -.000m -.000***
(7336-07) (8.68e-07) (5.67606)
KCTU -.286*“ -.446*" 274*"
(.001) (.002) (.015)
constant 5550*" -.542*M -3.512*" 34.271*** -l6.954“* -114.89***
(.017) (.022) (.049) (1.079) (.221) (1.408)
R’ 0.298 0.355 0.338 - 0.146 -
N 8,479 8,479 2,845 8,479 2,845 2,845

 

 

 

Note: All equations are weighted by number of employees. Logarithm of wages in model 4, model
5, and model 6 is predicted values, resulting from first stage equation, instead of observed
values. Robust standard errors are in parentheses. The dependent variable in Model 6 is the

percentage of voluntary benefits in wages plus voluntary benefits, multiplied by 100. *, ",

and **‘“ means significant at the 5%, 1%, and 0.1% level, respectively.

omitted industry in the equation, than in mining, construction, wholesale & retail, service,
and public services industries. However, it is shown that establishments in the
manufacturing industry provide more generous benefits than in the transportation &
public utilities industry and finance industry.

The 2SLS results also reports the positive effect of size of establishments on voluntary
benefits. The level of benefits increases as establishment size increases. For example, the
level of voluntary benefits in the largest firms hiring 1,000 or more employees is
expected to be extremely higher by over 7 times, 7.45 = exp (2.134)-l, than the benefit
level in the smallest firms hiring 29 or less employees. When changing mid-size

establishments hiring 300-500 employees as the reference group, the coefficients for

193

establishments (100-299), establishments (500-999), and establishments (1,000 or more)
are -.509, .209, and .909, respectively. It implies that establishments hiring 1,000 or more
employees offer larger voluntary benefits by one and half times, 1.482 = exp (.909)-1,
than establishments hired 300-499 employees, while establishments having 100-299
employees provide much less generous benefits by 66.4% than establishments hiring 300-
499 employees. It is certain that size of establishment heavily affects the level of
voluntary benefits in Korea.

Regarding personal characteristics, the percentage of female workers is negatively
associated with voluntary benefits. Previous studies in the U.S. argue that, while males
prefer wages, females have greater interest in non-wage benefits (Ashenfelter and
Johnson, 1972; Feuille et al, 1981). However, this is not the case in Korea. As lower
wages, less attachment to the establishment is due to shorter tenure of females than
males, and other factors appears to urge female workers to prefer benefits to wages. As
expected, education and tenure are positively associated with the level of benefits from
results in OLS (column 2) and 2SLS (column 4). However, experience is negatively
related to benefits against expectation.

Finally, the size of the union effect on voluntary benefits, estimated by 2SLS“, jumps
to .278 (32% = exp (.278)-l) from -.007 in column 2 of equation (8.2). These percentage
indicate that unionized establishments provide much more voluntary benefits by 32%
than unorganized establishments, other things being equal.

The results regarding first-stage for column 4 are shown in the column 1 of Table 8-3,

meaning that the squared year of experience, assumed to affect observed wages, works as

 

5’ In the results of ZSLS, the R2 is none in column 4. All results regarding first-stage for 2SLS will not be
presented here due to space limitations.

194

expected. Also, it indicates that 2SLS is likely to be free from the identification problem.

To ensure whether the specification error regarding ZSLS occurs, the Hausman test is

conducted. The Hausman’s specification error test is broadly used for testing the

hypothesis of no misspecification in the model (Maddala, 1992). The results from the

Hausman test implies no specification error at the 0.1% significant level. 55

Table 7-3 Wage-related equations for unionized samples for 1995-1999 (weighted)

 

 

Model 1 Model 2 Model 3 Model 4
[DV=lwage] [DV=log of [DV=log of [DV=lwage
(wages+voluntary (wages+statutory+
benefits)] voluntary benefits]
Fpercent -.007""' -.007"""" -.007"“"" -.006“'"
(8.56e-06) (8.85e-O6) (8.75e-06) (8.34e-06)
Edu 080*" 093*" .091 "“" 067‘"
(.000) (.000) (.000) (.000)
Expe -.01 1*" -.011"* -.010"* -.008"“
(.000) (.000) (.000) (.000)
Tenure 028*" 029"" 029*" .020‘"
(.000) (.000) (.000) (.000)
Density -.001*" -.001"* -.001*" -.001““
(.000) (.000) (.000) (.000)
SQdensity 000*" 000*" 000"" 9.4%-06*“
(2.00e-07) (2.09e-07) (2.07e-07) (1 .92e-07)
KCTU 087*“ 064*" 068‘“ .103"*
(.000) (.000) (.000) (.000)
Log of benefit 085*"
(.000)
Constant 7273*" 8.151 **"‘ 8213"" 6888*"
(.043) (.021) (.021) (.042)
R2 0.653 0.634 .0644 0.686
N 2,845 2,845 2,845 2,845

 

 

Note: All equations are weighted by number of employees. Five year dummies, five region dummies, eight
industry dummies, six size dummies are included into all equations above along with independent variables
reported. Robust standard errors are in parentheses. ‘, *"', and "" means significant at the 5%, 1%, and

0.1% level, respectively.

 

55 The details are not presented here. However, they are available upon request.

195

Effects of baggining—related variables on the level of benefits

The result from column 5 in Table 7-2 indicates that unions affiliated with the militant
union federation, the KCTU, have a much lower level of voluntary benefits by .446
(56.2% = exp(-.446)-1) than unions affiliated with the moderate, or rather docile, FKTU.
The OLS results in column 3 report similar but a smaller (-.286 and 33.1%) effect size of
the KCTU.

This is puzzling because it would be expected that the unions with stronger bargaining
power to show a higher level of voluntary benefits. To unveil this puzzling result, first, I
examined the percentage of voluntary benefits in wages plus the voluntary benefits
between the unions under the two national union confederations. The weighted sample
mean of the share of voluntary benefits is 7.93% for all unionized samples. The weighted
sample mean of the percentage of voluntary benefits is 8.24% for the FKTU—affiliated
unions and 7.57% for the KCTU-affiliated unions. The difference of 0.67% indicates that
the share of voluntary benefits is higher in the FKTU-affiliated unions than in the KCTU-
affiliated unions. Then to find the difference after controlling the influences of other
variables including wages, I ran anther two-stage regression with the percentage of
voluntary benefits in wages plus voluntary benefits, multiplied by 100, as the dependent
variable. Column 6 in Table 7-2 is the result. According to column 6, the difference in the
share of voluntary benefits in wages plus the voluntary benefits between unions under the
KCTU and unions under the FKTU jumped to 2.74% after controlling other independent -
variables. Recall that the sample mean of the share of voluntary benefit in wages plus

voluntary benefits is 7.93% for all unionized samples. Therefore the result from column 6

196

implies that unions affiliated with the KCTU have a smaller share of voluntary benefits
by 34.7% (=2.74/ 7.93) than what the unions under the FKTU indicate.

Then to know whether the unexpected difference in the level of benefits is related to the
level of wages between the KCTU-affiliated unions and the FKTU-affiliated unions, I ran
four types of wage equation for the unionized samples. Table 7-3 presents a reduced
form of the results. Due to the difficulty in identifying appropriate instrumental variables,
which affect the selection of a national union federation and do not affect wages, I
estimated the equations by OLS. When dependent variable is logarithm of wages in
column 1, the coefficient of peakorg is .087, suggesting that unions affiliated with the
KCTU have more wages by 9.1% that unions affiliated with the FKTU. When the
dependent variable becomes a logarithm of (wages+voluntary benefits) in column 2 or a
logarithm of (wages+statutory+voluntary benefits) in column 3, the wage differential
between KCTU- and FKTU-affiliated unions decreased to .064 in column 2 or .068 in
column 3. The decreased size of the coefficient of peakorg is that benefits, favored by the
unions under FKTU as seen in column 3 and 5 of Table 7-3, are included into the
dependent variable. When adding a logarithm of voluntary benefits into the equation of
column 1, the size of peakorg in column 4 changed from .087 in column 1 to .085 as
expected. To assure whether the wage differential between the unions under the KCTU
and FKTU is accurate, I ran other wage-related equations by adding squared experience,
squared tenure, and working hours into the lists of independent variables. However, the
size of the wage differential between the F KTU-affiliated and KCTU-affiliated unions

did not decrease but rather increased to .103 (10.8%=exp(.103 )-1).56

 

5" The results from the equation are not reported here.

197

In sum, the KCTU-affiliated unions show a lower level of voluntary benefits by 33.1%
(OLS) or 56.2% (2SLS) than the FKTU-affiliated unions, while the KCTU-affiliated
unions benefit from the higher level of wages by 9.1% or 10.3% than the F KTU-affiliated
unions. Even though they could increase the level of benefits by transferring wages to
benefits if they wanted to, why do the KCTU-affiliated unions face fewer benefits than
the FKTU-affiliated unions? Is that a simple reflection of wage-favored preference they
have regarding wage-benefit mix? It appears that it is not case. I believe that no the
benefit effect of the KCTU-affiliated unions is not the result of weak bargaining power
but the result of militant bargaining power. Facing employers’ resistance and so-called
‘one-digit wage guideline’ by the government in the 1990s, the KCTU-affiliated unions
with stronger bargaining power have been more successful in increasing wages and, thus,
the FKTU-affiliated unions have consented to transferring wages to benefits as the
government recommended.

This seemingly puzzling finding, however, can be supported by small number studies,
arguing that union and union bargaining power tilts compensation packages toward
wages relative to non-wage benefits (Donsimoni and Shakotko, 1979; Feuille et al.,
1981). Feuille and his colleagues (1981) found that union strength is likely to increase the
share of wages rather than non—wage benefits regarding working conditions. They
attributed the phenomenon to the “diminishing marginal utility of improvements in
working conditions relative to that of wages (p. 47).”

These results regarding the differences in wages and benefits between the KCTU-
affiliated unions and the F KTU-affiliated unions cannot be found through the Basic

Survey of Wage Structure (BSWS), which was used in chapter 5 examining the union

198

wage effect in the distribution of wages, because the BSWS does not include information
on peakorg (KCTU/FKTU) and union density. When controlling the level of wages as
one of the independent variables as in column 3 and column 5 of Table 7-2, we could not
find any systemic differences regarding wages between the KCTU and FKTU. Also it
leads us to overlook the fact that the wages are higher for the KCTU by over 10% than

the FKTU.

Provision of four nflm voluntary benefits
Is there any consistent difference in probability between the unionized and non-
unionized establishments that employers provide for each of the sub-categories of
voluntary benefits? If so, what are differences in the provision of sub-categories in
voluntary benefits between unionized and non-unionized establishments? Among the
twelve sub-categories as mentioned earlier in Table 2-3, I selected four important types of
voluntary benefits, considering the relative share of each in the total voluntary benefits in
1999; meals, schooling, housing, and the intra-firm benefit fund.57
Table 7-4 presents the results of logistic regression on the provision of each of the
types of benefits among the four between the unionized and non-unionized sectors for
1995-1999. The dependent variable in each equation is whether establishments provide
relevant types of benefits (1 if employer provides, 0 otherwise). In column 1 and 2,
unions are shown not to affect the provision for meals and housing. However, column 3
and column 4 show that the union increases the probability of employers providing
schooling and intra-firm benefit fund to their employees significantly. In order to see any

impact of bargaining-related variables on the provision of benefits, I repeated the same

 

’7 A brief introduction on intra-firm benefit fund was given in the footnote of no. 12.

199

logistic regression for only the unionized sample. The results in column 5 of Table 7-4
present that unions affiliated with the KCTU are related to the increased probability of
employers to provide a benefit of schooling expenditures at the 5% significance level.
Also, the density and squared density become jointly significant at the 0.1% level (chiz =
15.12, Prob> chi) =0.0005) in column 6 with intra—firm benefit fund as the dependent
variable for the unionized samples. However, no effect of bargaining-related variables on
meals and housing is seen between the FKTU-affiliated and the KCTU-affiliated
unions.58
What do the results from table 7-4 imply with regard to the impact of unions and
bargaining-related variables on the provision of voluntary benefits? The two benefits,
meals and housing, on which unions do not impact, are older and more basic benefits
than other voluntary benefits. In fact, employer expenditures on meals and housing for
employees had already developed in the 1950s and 19608 in Korea (An, 1989). However,
employer expenditures on schooling began in the late 1970s or early 1980s and increased
in the 19903 by the demands from employees. Also, for the first time, the intra-firm
benefits fund came about in 1984 by the recommendation of the Ministry of Labor.
Thus, it might not be an exaggeration that, while voluntary benefits in non-union
establishments focus on more basic benefits and remain old-style, union establishments

pay more attention to diversified benefits beyond basic benefits. This result supports the

role of unions assumed in industrial relations literature that trade unions will understand

 

58 The results are not shown in Table 7-4, due to space limitation.

200

Table 7-4. Logistic multiple-regression for union effects
on the provision of various voluntary benefits for 1995-1999.

 

 

 

All Samples Unionized Samples
Meals Housing School Benefit School Benefit
Fund Fund
Yearl 996 .996 .984 1.1 12 .840 1.291 .998
(.120) (.084) (.108) (.124) (.247) (.186)
Yearl997 .773" .910 .800* .896 .862 .801
(.088) (.075) (.075) (. 125) (.154) (.144)
Yearl998 .900 .838“ .439"* 599“" 530"" .506‘“
.102 .070 .041 .093 .095 .104
Yearl 999 (1 .03 1) $7552" $395 '2" " .(673 '2 $5441?“I 2342’
(.121) (.067) (.038) (.108) (.103) (.1122
Busan 1.396“ .953 1.307“ .812 1.381 .512
(.190) (.104) (.141) (.164) (.299) (.151)
Kyungki 1.409*" 1299*“ 1.102 .841 1.061 .795
(.143) (.099) (.091) (.1 10) (.173) (.138)
Joongbu 1.058 1434’" 1.091 .842 .786 .749
(.099) (.1 12) (.097) (.1 18) (.125) (.140)
Kyungsang 1.289“ 1.477‘“ 1.034 .681 " .987 .635“I
(. 135) (.1 18) (.092) (.102) (.169) (.127)
Mining .164‘" 1.451 .963 .320 .369" .329
(.037) (.286) (.227) (.193) (. 141) (.256)
Construction .555" .725 .784 3.851 "'" .770 4.296‘”
.155 .157 .225 .998 .384 1.435
Transportation $391)" $4322" .(362*)" $650) $4022 $778 )
(.070) (.065) (.055) (.213) (.169) (.404)
Wholesale .441"* .434‘" 1.149 1.310 1.516 2.107"
(.070) (.050) (.137) (.269) (.488) (.610)
Finance .455‘" .491 "" 1.082 1 .541“ .624“ 1.670“
(.066) (.056) (. 130) (.286) (. 130) (.3 88)
Services 228"" 494"“ .867 1.519“ .709 2340""I
(.033) (.055) (.108) (.303) (.198) (.672)
Public 202*" 257"“ .988 1.024 .881 1.051
(.029) (.035) (.127) (.236) (.223) (.345)
30$Ees<100 .827 1424*" 2243*" 1.661 3.670“ 4.575
(.097) (.146) (.233) (.537) (2.090) (8.052)
1005Bes<300 .702" 2266*" 5.137‘” 2885“" 7.458‘" 5.158
(.084) (.234) (.553) (.887) (4.166) (8.893)
300$Ees<500 1.034 3196*" 10.927‘" 3389‘“ 17.288‘" 7.390
(.160) (.385) (1.441) (1.075) (9.804) (12.718)
500$Ees<1,000 .901 3.879‘” 14.97?" 5161‘" 21.752‘" 10.387
(.131) (.465) (2.070) (1.610) (12.312) (17.887)
1 OOOSEes 1.302 7.122‘" 36.318‘“ 6.515‘” 51.176‘" 11.508
’ (.206) (.892) (6.665) (2.034) (30.21 1) (19.686)
Fpercent .997 1.001 .991 ”" .993“ 985*" .989“
(.002) (.001) (.002) (.003) (.003) (.005)
Edu .724""' .907 1410*” .761 * 1.300 .798
(.052) (.055) (.093) (.083) (.181) (.1 l9)
Expe 938"" .943 *" 1.035“ .934“ 1.032 .946
(.016) (.013) (.015) (.025) (.034) (.034)
Tenure .913‘" .992 1.156”‘" 1.061“ 1.110“ 1.073
(.016) (.016) (.022) L030) (.043) (04 1)

 

 

 

201

 

LWages .510*** 1.786*** 4.461*** 3.634*** 2.499*** 3.594***
(.056) (.158) (.527) (.619) (.461) (1.065)
Density 1.009 1.019
(.009) (.012)
Sqdensity 1.000 1.000
(.000) (.000)
KCTU .742* .901
(.1 13) (.133)
Union 1.098 .987 2.027*** 1.688***
(.092) (.063) (.148) (.194)
Log likelihood -3058.846 4729.906 3950.721 4780.215 -1 145.133 -968.064
Chi-square 907.97*** 982.63*** 1841.58*** 488.23*** 357.44*** 243.78***
N 8,479 8,479 8,479 8,479 2,845 2,845

 

 

 

Note: * significant at the 5% level, ** significant at the 1% level, and *** significant at the 0.1% level
Note: density and Sqdensity jointly significant at the 0.1% level (chi2 = 15.12, Prob> chi2 =0.0005) in the
equation with benefit firnd as dependent variable for unionized samples

needs of employees better with regard to benefits and contribute to develop more
diversified benefits throughout collective bargaining (Mabry, 1973: Freeman and Medoff,
1984). The ‘voice’ aspect of trade unions to monitor the preferences of union members
appears to contribute to maximize the utility of employees by diversifying the types of
voluntary benefits. As a consequence, trade unions have pioneered new kinds of benefits,
such as dental insurance, in the United States (Antos, 1983; Freeman & Medoff, 1984).

The same role done by trade unions regarding fiinge benefits is found in Korea, also.

Discussion
When using the results from ZSLS, we found that trade unions for 1995-1999 in Korea
increase the level of voluntary benefits by 32%, compared to an unorganized

establishment, other things being equal. This result means that the exclusion of pensions

202

and health insurance from negotiation between employers and unions because statutory
benefits do not affect the existence of union effects on benefits.

In addition, the size of union effect on voluntary benefits, 32%, is much larger than the
size of union wage premium. Even though simple comparison is impossible because of
different data, we have already seen in Table 4-3 in Chapter 4 dealing with union wage
effect in the distribution of wage that the union wage effect, estimated by OLS, for 1995-
1999 ranges from 1% (= exp (.01)-1) in 1996 at the minimum to 7.7% (= exp (.074)-1) in
1999 at the maximum. No study reports such a large union wage premium in Korea like
the union benefit effect of 32% mentioned above. The size of the union benefit effect in
Korea for 1995-1999 is higher somewhat than the size of the union benefit effect of 20%-
30% in the U.S., estimated by Freeman and Medoff (1984), and lower than the size of
45.5% in Canada, found by Renaud (1998). This finding means that the union wage
effect in Korea is likely to underestimate the whole union effect on total compensation as
in the U.S. and Canada. A lesson from Korea for 1995-1999 is that the proportional
assumption behind most studies on the union wage effect, assuming that unions would
affect on fiinge benefits proportionally as on wages appears to be false (Alpert, 1982)

Also, the existence of the union benefit effect of 32% in Korea suggests that
compensating wage theory predicting no union benefit effect and a negative trade-off
between wages and benefits is unlikely to be true. Of course, an accurate test of the
compensating wage theory cannot be accomplished in this study (Smith and Ehrenberg,
1983; Olsen, 2002). Nonetheless, the result from this study seems to support the
monopoly model of the union in industrial relations assuming that unions enhance

benefits as well as wages by using bargaining power.

203

The results regarding bargaining-related variables such as the union organizing rate
and characteristics of national union confederation present more complicated
implications.

It was discovered that the two separate national union confederations, the F KTU and
the KCTU, represent the different preferences of workers towards wages and benefits.
That result indicates that the FKTU-affiliated unions show a relative lower level of wages
and higher level of benefits, while the KCTU-affiliated unions enjoy a considerably
higher level of wages but a lower level of benefits. With respect to the level of benefits,
the difference between the FKTU and the KCTU is understandable, at least partially,
because the rank-and-file union members and union leaders in the FKTU-affiliated
unions place more emphasis on the importance of benefits, compared to those in the
KCTU-affiliated unions (Kwon, 1997). In addition, the union leaders and officers value
the importance of increasing benefits more than the rank-and-file union members do in
the FKTU-affiliated unions, while union leaders place less emphasis on the value of
benefits than rank-and-file union members do in the KCTU-affiliated unions (Kwon,
1997)

Regarding the higher level of wages in the KCTU-affiliated unions, compared to those
in the F KTU-affiliated unions, I assume this fact might be the result of a different level of
bargaining power between the two national camps resulting from a difference in
democracy and/or militancy. In regard to democracy, the direct election for president of a
union, less discretion from the union presidents, and the ratification ballots by the rank-
and-file union membership are more easily found in the KCTU-affiliated unions than in

the KCTU-affiliated unions (Kwon, 1997). Also, the KCTU-affiliated unions show a

204

higher level of bargaining duration, strike propensity, and participation in political
activities than the F KTU-affiliated unions, regarding militancy. The fact of a higher level
of total compensation for the KCTU-affiliated unions supports the argument that
militancy in bargaining (Fiorito and Hendricks, 1987) and internal democracy within
unions (Levine & Feuille, 1983) is positively related to better outcomes.59

Regarding the mix of wages/benefits in between the KCTU- and FKTU-affiliated
unions, it would be desirable if the phenomenon of a higher share of wages for the
KCTU-affiliated unions and a higher share of benefits for the FKTU-affiliated unions
reflects simply different preferences of the rank-and-file union members between the two
kinds of unions. However, another possibility exists in that all union members prefer
wages to benefits, regardless of the KCTU or FKTU, and also the lower level of
democracy and militancy in the F KTU-affiliated unions prevent the unions and the
unions’ leaders from concentrating on wages, as the rank-and-file union members want to
do. So far, it is not clear whether either possibility is true.

The compensating wage theory explains that workers having different preferences
towards the wage-benefit mix sort themselves across firms offering different wage-
benefit mix (Schiller and Weiss, 1980; Smith and Ehrenberg, 1983; Rosen, 1986; Olsen,
1994 & 2002). In this theory, the preferences of workers and employers’ different
strategy to attract newcomer result in different wage-benefit mix across firms. The

monopoly model of unions and the results in this study imply that collective preferences

of workers towards a wage-benefit mix may be a result of a combination of situations, the

 

’9 If it is true in reality that the level of total compensation is higher in the KCTU-affiliated unions than the
FKTU-affiliated unions, the believed observation that the FKTU-affiliated unions is closer to the business
unionism than the KCTU-affiliated unions might be false. As a result, the presumption based on the
observation that the level of total compensation would be higher in the F KTU-afiiliated unions than in the
KCTU-affiliated unions also might be not true.

205

existence of a union, the variation of bargaining power within unions, and possibly other
factors. We cannot conclude which causal direction between preferences and structures
including situation is more plausible.

The dominant influence of establishment size on benefits is worth drawing our
attention. Many in Korea are concerned about the dominant effect of firm size on wages
(Song, 1991 & 1994). Our study shows that the effect of establishment size is much
greater on benefits than on wages. As the differences in wages and benefits are increasing
between a limited number of large firms and small- or medium-sized firms, in fact dual
labor markets, which consist of core and peripheral market, are concerned. Therefore, this
study appears to call for active efforts by the government, employers, and unions to

narrow the gap in wages and benefits between the two sectors.

4. Conclusion

Where governments do not provide good social welfare systems for workers, the roles
of corporate welfare system within firms increase. Fringe benefits offered by employers
become one of the conditions under which workers show a high level of productivity
(Davis, 1986). Also individual workers and union leaders want the shares and types of the
benefits to be increased, for various reasons. The case of Korea is not an exception for the
phenomenon. Overall this study examines the effects of union - and union-related
variables — on the benefits in Korea for 1995-1999.

One of the main features of this study is that the effects of the trade unions on the level

of benefits and the provision of individual benefits are examined together. Another

206

feature of this study is to use union organizing rate in the establishment and the
individual union’s affiliation with either two of the national union confederations
(FKTU/KCTU) as the main independent variables. Most studies on union effect on
benefits assume equal effects of unions on benefits. In reality, there appears to be
significant differences in union bargaining power and preferences of union membership
across unions.

Some limitations in this study need to be kept in mind. First, this study did not examine
the period of 1987-1994 due to a lack of union-related information. Thus, it is not
possible to say that the union benefit effect had existed for the entire period of 1987-
1999. Also, we cannot express anything about the yearly or periodical changes in the
union benefit effect for the entire period. Second, some incomplete or unmeasured
variables are presented, which are likely to affect the union benefit effect. Because of the
lack of information, the average of education, experience, and tenure in 169 three-digit
industries from the Basic Survey on Wage Structure (BSWS) are used as proxies for the
average year of education, experience, and tenure in establishments. Also, unmeasured
are the composition of occupations in an establishment, the percentage of married
workers, and the preferential tax treatment for voluntary benefits that may impact on the
union benefit effect.

Even though the union benefit effect in Korea for 1995-1999 is as high as 32%,
estimated by 2SLS, it is unlikely to expect union benefit effect to decrease in the near
future, due to the trend of an increasing share of benefits in total compensation,
diversified preferences among workers reflecting an increase of education levels and

economic development, and the recent increase of various private insurances including

207

health insurance, life insurance, and pensions (Choi, 1992; Bang, 2002). To employers,
existing types of benefits are too basic and undifferentiated to retain their skilled workers
and to attract new able workers. Employees’ commitment and involvement to their
organizations are unlikely to be secured through existing levels and content of benefits. In
recent years, many insurance companies have developed and started to sell private
pensions and various insurances including life insurance and health insurance. Though
remaining small, the number of companies providing private pensions and insurances is
increasing as a result of employees’ demand, as seen in Table 2-3 earlier.

In addition to the tradeoffs between wages and benefits, other tradeoffs between
wages (or total compensation) and other contract provisions are possible (Feuille et al.,
1981). In fact, about a half of collective bargaining in Korea deals with contract
provisions including working conditions, the scope of union activities, the composition of
disciplinary committee, the procedures for layoffs or dismissal, etc., as well as wages and
benefits. Because bargaining outcomes are not restricted to wages and benefits (Kochan
and Block, 1977; Feuille et al., 1981; Fiorito and Hendricks, 1987), a study examining
union effects on total compensation should take into account the potential tradeoffs
between total compensation and other contract provisions in order to be more complete.

For this reason, the results in this study should be read carefully and studied thoroughly.

208

CHAPTER EIGHT

CONCLUSION

The purpose of this study is to examine how the Korean trade unions have acted with
regarding the economic function involving wages since 1987, when political
democratization came about in June and the ensuing the Great Labor Offensive in July,
August, and September brought free collective bargaining into the Korean industrial
relations.

Considering many factors such as the long history of low wages combined with the
export-driven strategy of economic development set by the government for the past 30
years, the nature of the enterprise union system, and the first steps in free collective
bargaining, union activities aiming to enhance wages for their membership are expected
to be the best way to examine how trade unions have affected the labor market and how
unions have been viewed by their membership.

In order to examine the economic effects of trade unions since 1987, I divided the
possible research areas into four parts: the union wage premium in the distribution of
wages, the union effects on wage dispersion and wage equality, the union effects on the
influences of human capital and personal characteristics, such as education, experience,
tenure, and sex, over wages, and the union effects on benefits.

This concluding chapter consists of two parts, the one to summarize the major results

from the four empirical previous chapters and the other to mention some important points

209

needed to predict Korean industrial relations in the near future. In the former, the
strengths and limitations of this study will be mentioned, along with a summary. Also the
possible theoretical contributions to the area of industrial relations and policy
implications from this study for government, employers, and unions including workers
will be explored. In the latter, the emerging public sector, a possibility of a merge
between the FKTU and the KCTU, the scheduled organizational changes from the
enterprise union towards the industrial union system, and the political parties created by

the two national labor camps will be explored.

1. Summary of Four Empirical Chapters

Reviewing some studies on the union effects on wages, wage dispersion, four wage
determinants, and fringe benefits in Korea revealed that the union effect on wages, wage
inequality, total compensation in the U.S. is also found in Korea. Therefore, this
discovery means that the theory and empirical work, which have been conducted in the
U.S., can be applied to Korea without major changes.

In detail, unions enhance the level of earnings and compress the distribution of
earnings in the organized sector.

First of all, union wage premiums, estimated by the OLS and QR, are found in the
period of 1987-1999, even though the results are not free from the sample selection bias.
The existence of union wage premiums in Korea for 1987-1999 implies that the political
oppression of trade unions by the government, aimed at deleting union wage effects in

the labor market, was not firlfilled. Furthermore, trade unions succeeded in offering

210

economic benefits to their rank-and-file membership for the period, also. The absolute
size of union wage premiums are shown to be moderate in that they are below 7% in
most years and quantiles, when compared with about 10-15% (Hirsch & Addison, 1986)
or even 20-25% (Freeman & Medoff, 1981; Lewis, 1986) in the United States and 10-
20% in Canada (Benjamin et al., 1998; Kuhn, 1998) in the 1990s. The moderate size of a
union wage premium may result from many factors such as a short history of free
collective bargaining and a need to secure viability of union itself rather than higher
wages due to continued strong opposition.

Regarding the size of union wage effect, the OLS estimation method assumes that the
union wage effect is the same across the distribution of wages and the traditional theory
in industrial relations assumes larger union wage effect in the lower of wage distribution
and less in the upper of wage distribution. Our results using the data from the BSWS for
1987-1999 indicate that the OLS estimation assuming constant size of union coefficient
in the equation is not correct. Rather the size of union coefficient varies considerably,
depending on the focus on the wage distribution. In addition, the traditional theory
assuming larger size of union coefficient for the lower-wage workers turned out to be
only partially true for 1987-1995 and false for the period 1996-1999. Thus, our
comprehensive QR results examining every quantile in the wage distribution imply that
the issue of which group among lower-wage, average-wage, and higher-wage workers,
will be benefited from wage bargaining led by unions is time-variant. To understand the
unexpected findings, this study considered four factors, changes in the ways to determine
wage increase by unions, fixed amount of wage increase and fixed-rate of wage increase,

a possibility that the voice of higher-wage earners becomes important in the 19905 due to

211

increased complaints resulting from the fixed amount of wage increase in the late 19803,
wage norm argument regarding which group of workers benefit from union wage
premium first and which group will do next, and any changes in the union leadership.
However, we do not find any critical factors explaining the results because discovering
determinants of the phenomenon is beyond the boundary of this study. It appears that the
influences of employers and government policy on the issue as well as the three potential
factors must be considered from now on.
Thus, the voice of the unions contributes to reducing wage inequality in Korea.
Throughout the establishment-level data, which are constructed from the individual-
level BSWS data, the chapter analyzing union wage equalizing effect shows that Korean
trade unions narrowed the wage dispersion within establishments, resulting in the
increase of wage equality. In detail, the size of the union effects decreasing wage
dispersion is 8.8% in 1987. However, the size decreased to 1.1% (OLS) or 2.5% (ZSLS)
in 1994 and to 1.2% (OLS) or 2.5% (2SLS) in 1999. Despite the variations according to
the year examined, it can be said that the existence of unions have had negative effects on
wage dispersion in Korea for the past 15 years. It is also mentioned that the relatively
smaller size of the union wage equalizing effect may be attributed to the prevailing norm
of the wage equality in the society and/or the spillover effect of the unions decreasing

wage inequality into the non-unionized sectors.

By employing never-used data from the ELCS by the MOL, this study also found that

the positive union benefit effects had existed in the late-19903, depending on the

estimation by the 2SLS. In addition, the bargaining-related variables such as the rate of

212

union organization within establishments and union’s affiliation with national union
federations affected the level of benefits were discovered. While the union organization
rate is found to increase the level of benefits, the effects are not monotonically increasing,
showing an inversed-U shape pattern. Also, the results indicate that the two existing
union confederations affect differently the preferences of the unions regarding the
wage/benefit mix in the total compensation. Therefore, specifically saying, the individual
unions, which are affiliated with the FKTU, a moderate and pro-government federation,
showed a relatively higher level of benefits, while the KCTU, a militant federation,
achieved a higher level of total compensation, but a lower level of benefits. It was
already known that the union activities regarding wage policy were not uniform during
the entire period of 1987-1999. In the period of 1987—1989, the union activities appears to
focus on increasing the absolute size of wages, resulting in higher rates of wage increase
and ignoring the importance of benefits or gender wage gap. From 1990, an increase in
the portion of benefits in total compensation and a decrease in the wage gap between

male and female workers is witnessed.

Evalpation of Union Activities

The results that the Korean trade unions enhance wages, wage equality, and voluntary
benefits for their union membership relative to comparable workers in non-unionized
establishment and reduce the return rates for education, experience, tenure, and sex on
wages imply that the unions have acted as the rank-and-file union membership wants and
have fulfilled to some degree what union membership expects unions to do for

themselves. Thus, it is unlikely that most of rank-and-file union members, driven by

213

economic utilities, will secede from unions in the near future as long as the trade unions
sustain in securing economic results through collective bargaining, such as more wages,
more wage equality, and more benefits, and less effects of human capital on
compensation. In addition, these economic accomplishments achieved by the trade unions
for the past fifieen years are expected to affect workers in the non-organized firms either
through a direct extension of the union base to the unorganized sector or through a threat

effect increasing wages for workers in non-unionized firms by employers.

Implications to the nature of unions

With regard to the debate between John Dunlop (1950) and Arthur Ross (1956), it can
be said that this study focuses exclusively on the economic side of unions, examining
union wages- or compensation-related effects including wage, wage inequality, and
benefits. Of course, wages or total compensations are not all of the economic facets of
trade unions. In many cases, employment is another aspect of economic side of unions
and becomes a trade-off relationship with wages. Also the study here pays little attention
to the political side of unions such as preferences of leaders and other political activities
of unions. Furthermore, an apparent limitation of this study is not to examine these
aspects.

Despite the limitations, the results from this study show that the political aspects of
unions are instilled into largely the economic side of unions (Kaufman, 2002). For
example, the facts state that the union wage effects are conditional on the distribution of
wages, which implies that union membership consists of different factions with different

preferences. Also the fact that groups for which unions offer the largest wage effects

214

depends on the year concerned needs some help from political explanations toward
unions. Thus, this study urges further thinking about the implications of the nature of
trade unions.

It might be true that the process of how union maximizes membership’s wage bills can
be well explained by the economic perspective toward trade unions researched by Dunlop
(1950) and the classic monopoly model of unions (Farber, 1986). However, the issues of
how pools of wage bills, secured from the employers by unions, should be distributed to
the union members is a realm of unions as political organizations, which consist of
various groups of workers and union leaders having separate preferences and are operated
by the median-voter model. In addition, another issue of how unions adjust the
wage/benefit mix and which benefits should be increased does involve the conflicts and
compromises among various actors such as union leaders, older workers, marginal
workers, as well as employers and external environments regulating the workplace such
as labor laws. At this point, this study shows how economic and political aspects of
unions in nature are closely interconnected to each other.

This study reveals some obvious gaps in the theoretical explanations on the nature of
unions described in the literature. First, a lack of clarity is evident in the median-voter
model. The model regarding union effect on wages and wage dispersion predicts that the
lesser paid workers will benefit the most from union wage bargaining and the preferences
of the median voter will lean towards workers at the bottom of the wage distribution
because the mean wage within the unionized firm exceeds the median. With regard to
fringe benefits, however, the median-voter model is used because it expected that unions

will favor older workers who have stronger preferences for larger benefits. In other

215

words, the median-voter model, regarding wages and wage dispersion, expects unions
will benefit marginal workers at the bottom of the wage distribution, while the model
regarding fiinge benefits expects the union to follow the preferences of older workers,
who are likely to be at the upper part of wage distribution. The median-voter model

appears not to answer the question of how trade unions deal with the wage/benefit mix.

Limitations

From chapter 4 to chapter 7, this study pays minimal attention to the roles of
government and employers, even though the two are the most important actors, along
with trade unions, in industrial relations and the labor market. By limiting within the
establishment level regarding the union wage-equalizing effect and the union benefit
effect and focusing on the personal characteristics and organizational characteristics, the
strategic aspects of government and employers are not captured. Especially, the role of
government in Korea has been enormous in every aspect of industrial relations (Shin,
2002). To date, despite the role of government and employers, few variables are available
for consideration in the many empirical models, as seen in this study. Based on a
theoretical description and prediction, the year-dummy variables are frequently used to
capture the roles of the government and employers. However, the year-dummy variables
used may capture only the partial effects of the government and leave exact effects of the
government on the labor market and outcomes of industrial relations. Therefore, this
study did not attempt to examine empirically the role of government and employers, but

the task should be performed in firture study.

216

Another limitation of this study is that it did not considered potential tradeoff
relationship between total compensation and other outcomes including employment,
working conditions, and number of full-time union officers paid by employers, which are
produced by trade unions. Focusing on the important roles of wages and fringe benefits,
this study does not give much attention to the relationship between unions and
employment. Many assume that wages and employment are two essential parts consisting
of the economic utilities of union membership in general and both can be a substitute for
each other. However, no appropriate data are available to allow for examination the union
effects on both wages and employment in Korea. In addition, most unions and union
leaders did not devote much attention to employment because there had not been any
massive layoffs before the Korean financial crisis.

More important is that the trade-off between gains from total compensation and losses
from employment are not always found. While trade unions seek to gains from both total
compensation and employment, employers also try to keep their power to set the level of
both compensation and employment (Block, 1995). Furthermore, Korea has been away
from competitive market where the trade-off between compensation and employment is
likely to be found. Rather, Korea has been a type of a less competitive market, due to
authoritarian government over all aspects of employment relations, internal labor market
resulting from enterprise union system, and monopsonistic employers in many industries.
Therefore, this situation makes it difficult to expect the clear trade-off relationship
between compensation and employment and an examination of union effect on only total
compensation here might be tolerable. Overall, a more balanced study on union effects on

both compensation and employment in Korea should be done in the near future. Finally,

217

before seeing more balanced studies, the results from this study should be considered

with great caution.

Policy Implications

With respect to fiinge benefits, the previous chapter shows the effects of unions (and
union density and nature of national centers) on the benefits. Two ways for unions to
secure the various types of benefits for their members exist: through nationally legislated
social welfare programs and through individual collective bargaining with their
employers within firms (Wallerstein, 1985; Goldfield, 1987). Due to the weakness of
labor in national politics and the lack of resources in national union confederations owing
to the enterprise union system, the Korean unions had to choose an alternative way to
secure benefits through individual collective bargaining with their employers since 1987.
The significant positive union effects on benefits are a result of the situation. However, a
second way to achieve benefits through collective bargaining is less desirable than the
first way through social legislation in national politics in many regards, for example, a
limited number of workers are eligible for benefits due to the low union density,
variations across unionized firms, and a close attachment of benefits to the specific firms
and no applicability to the unemployed (Goldfield, 1987). For these reasons, we can
expect that Korean unions will change the way to secure benefits for their members,
when they can concentrate resources into national union confederations and arm with a
close relationship with political parties. As collective bargaining in individual firms

becomes stable, unions also are likely to move their concerns from individual bargaining

218

to national legislation in national politics. In the near future, the focus of conflicts

regarding benefits will be relocated from the boundaries of firms to national politics.

2. Prospects in the Near Future for Korean Industrial Relations

For the past 15 years since 1987, a compromise secured by employers, unions, and
government together is that free collective bargaining and the right to organize workers is
guaranteed. As a result of the compromise this study becomes possible. However, it does
not mean that the existing Korean industrial relations are sufficiently and stably rooted in
the soil of the Korean economy and society. Rather the Korean industrial relations
continue to suffer from the lack of more compromise over the major issues of industrial
relations. Therefore the stability of American industrial relations resulting from the
Wagner Act in 1935 cannot be found in Korea yet.

In the mid-19903, a pessimistic perspective towards unionism arose, as unions faced
the decrease in union density and strikes, government oppression, etc (Kim and Song,
1998; Kim, 1995). This current study indicates the existence of union wages and benefit
premiums and the union wage-equalizing effect, which implies that to date the Korean
trade unions appears to be doing what they are supposed to do as the rank-and-file union
membership expect them to do. However, it is too soon to mention the decline of
organized labor in Korea, considering what trade unions have achieved for their
memberships, plus no further decline in union density, and the still viable bargaining

power of trade unions.

219

However, the near future of Korean industrial relations will depend on some aspects
such as the emergence of public sector unionism, organizational transformation by the
union itself fiom the enterprise union system to an industrial union system, a possibility
of a merger between the F KTU and KCTU, the existing two national union

confederations, plus the political parties created by the two union confederations.

Public Sector as An Unopened Space

The importance of collective bargaining in the public sector has increased rapidly in
the United States since the early 19603. It is recognized that the factors supporting the
emergence of collective bargaining in the public sector are the grth in government
budgets, civil rights and other social movements, and the passage of laws favorable to
collective bargaining in the public sector (Katz and Kochan, 1992; 359). Trade Unions in
the public sector are said to allow their memberships to benefit from considerably less
wage premiums (Moore and Raisian, 1987) and face constraints regarding strikes.
Compared to the precipitating union density in the private sectors during the past four
decades and, in spite of these limitations, the union density in the public sector in
developed countries is still very high and stable and thus the number of union
memberships in the public sector in the total union membership is increasing.

While public sector unions have made it possible for American unions to survive thus
far, collective bargaining in the public sector in Korea has not yet developed. In 1998, the
total number of teachers in Korea was 428,257 who were allowed to organize unions
using the revised labor laws. Prior to that date, the Korean Trade Union for Teachers

(KTUT), affiliated with the KCTU since 1989, remained unrecognized. Then from 1998,

220

another national union federation, the Federation of Korean Trade Unions (F KTU) started
an organizing drive among teachers. Also the revised laws contained a plan to allow
lower—level government workers including those in the local units to have the right to
organize unions starting in 2002. The total numbers of government officers was 886, 582
in 1998 including 553,561 in the central government and 315,370 in local government
(National Statistical Office, 2000).

Already two existing unions for civil servants, the National Trade Unions for Civil
Servants (NTUCS) keep a close relationship with the KCTU and the Korean Federation
of Trade Unions for Civil Servants (KFTUC), which is supported by the FKTU, and are
said to have secured approximately 70,000 and 20,000 union memberships, respectively
thus far (Bae, 2002).60

Considering the union density of 12.6% and the total union membership of 1,401,940
in 1998 (Korea Labor Institute, 2000), it is apparent that the union organizing drives in
the public sector will change the union density and the number of union membership
substantially in the near future.“

Moreover, the emergence of unions in any seemingly-good occupation or industry
may mean that the relative wages and working conditions for workers in the occupation
or industry, compared to those for ordinary workers in an average industry, has been
eroded. The emergence of white-collar unionism in the 1990 can be interpreted as a sign

indicating that the severe division between while- and blue-collar regarding wages and

 

6° Up to now, the two unions for civil employees have not attained legal status. Discussions for the legality
of unions for civil employees are ongoing. Government tends to recognize the right to organize and a
partial right to negotiate, declining the right to act collectively by union members, while the two
unrecognized unions demand the full right to organize, to negotiate, and act collectively.

6' However, no realistic possibility exists that police (total number of 94,819 in 1998) can be organized by
trade unions in the foreseeable future.

221

working conditions in the 19703 had been eroded since 1987. In a similar manner, the
emergence of public sector unionism in Korea in the 21St century is a result of consistent
erosion in the supremacy of the government since the political democratization in 1987.
Also public sector unionism can be viewed as a phenomenon representing the
disappearance of privileges for public employees regarding wages and other working
conditions, relative to other workers in the private sectors. For these reasons, the
emergence of public sector unionism will substantially affect the future industrial

relations in Korea.

An Intended chflge from Entegprise Unionism to Industrial Unionis_r_n_

In contrast to the case of the U.S. where unions have shown a highly decentralized
bargaining structure in recent years (Katz and Kochan, 1992), a vehement interest of
labor to move an organizational focus on from the enterprise unions to the national union
federations and industrial unions (or federations) is increasing in Korea due to the
accumulated weaknesses of the enterprise unions and the new bargaining area at the
industrial and national level. Therefore, we may find the opposite direction of a shift in
unions in the U.S. and Korea. In the 19903, unionism felt the weaknesses of enterprise
unionism, which produced a weakening of national and industrial union organizations,
the inability of organizing new union members in the service sector, etc. The Korean
Union Federation of Hospitals transformed the system of unions from enterprise to
industrial unions in 1998 and bargained with the representatives of employers at the
industry level. In the textile industry, bargaining at the industry level between the

representatives of unions and employers covers some large companies, but not all.

222

Table 8-1. Progresses in Changing from Enterprise Unions to Industrial Unions within
the F KTU and KCTU

 

 

Federation of Korean Trade Unions Korean Confederation of Trade Unions
(FKTU) (KCTU)
Already- Bank and Financial (Mar. 2000) Hospital (Feb. 1998)
changed Taxi (April 2000) University Workers (Nov. 1998)
Industries Railway Teachers (1989)
Post Cargo Transportation (Feb. 1999)
Tobacco and Ginseng Construction (Dec. 1999)

Metal (Sept. 2000)
Press (Sept. 2000)

Industries Chemical Clerical and Financial
scheduled to Metal Public Service
change soon Textile Commercial

Tourist

Chemical-Textile

 

Source: Lee (2000)

To Korean unionists’ point of view, a high union wage premium should not be in part
attributed to the enterprise union system in itself. It is assumed that an institutional
change from the enterprise union to the industrial union system will boost union density
and bargaining power as well as curing the present weaknesses of the enterprise union
system such as the difficulty in achieving solidarity beyond the boundary of enterprise,
the lack of financial and personnel resources on the national level, and the difficulty of
organizing new members in the unorganized sector. In addition, the form of the industrial
union is expected to reduce wage dispersion between establishments, which enterprise
unionism has widened (Koo, 1996). While the degree of decentralization of the
bargaining structure has increased over the last few decades in the developed countries,
an opposite phenomenon, an institutional change from enterprise unionism to industrial

unionism intended by labor, is occurring in Korea. Even though employers appear to take

223

a hostile stand toward such a change, the change by labor will likely affect firture
industrial relations if realized.

As a result of the efforts to transform enterprise- to industry- or occupation-based
union system by the two national union confederations over the past few years,
considerable changes have been made. For example, industrial or occupational unions
covered 41.1% (247,458) of 602,339 union memberships for the KCTU in June 2002
(Nodong Ilbo, Dec. 31, 2001; Bae, 2002). Also about 130,000 union members under the
F KTU were reported to belong to the industrial or occupational unions (Bae, 2002). In
the process of the transformation, conflicts over between new organized industrial unions
and employers raised questions as to whether or not employers can agree with new forms

of unions and to form a joint bargaining table (Bae, 2002).

Toward Unification between the FKTU and KCTU

The 19903 witnessed the continuous decrease of union membership for the F KTU and
the continuous increase for the KCTU. When the National Congress of Trade Unions
(NCTU), a predecessor of the KCTU, was formed in 1990, it had only approximately
160,000 union members, 8.6% of the total union members (Kim, 1995). The number of
union members for the KCTU increased to 410,000 in 1995 and to 610,000, 41.4% of
total union membership in Korea in 2000 (Nodong Ilbo, Nov. 1, 2001).62 With the rapid
increase in union membership for the KCTU and the decline in union membership for the
FKTU, conflicts and competitions between the two national union confederations became

severe.

 

62 For 1993-1994, many enterprise unions withdrew from the F KTU and joined the KCTU because of
dissatisfaction about the wage agreement between the FKTU and Korea Employers Federation (KEF) in
(Kim, 1995; Choi, Y.K., 1999).

224

Facing intensified Opposition by employers and anti-unionism of worldwide
dimensions including the neo-liberalism expansion of capitalism since the 19803, a
merger between unions occurred in many developed countries. The merger of unions is
said to be good for union membership in that mergers between unions increase
operational efficiency and solidarity among workers across occupations and firms. In
Korea, the two national unions themselves repeatedly raise the possibility of a merger
between the FKTU and the KCTU.63 Also observers raised and recommended the
possibility (Kim, 2000). Following the track of a merger between the two national union
confederations, the results from this study imply that the two confederations should be
aware of the different preferences between the two toward wages and benefits.

Many observe that there have been noticeable differences in the internal governance
structure, composition of union leaders, style of bargaining mobilization of and rank-and-
file membership between the FKTU and the KCTU. In addition to the differences, a part
of this study dealing with the union effect on fiinge benefits reveals that the two national
union confederations, the F KTU and the KCTU, have separate preferences toward the
wage/benefit mix and each of the sub-categories of voluntary benefits. In addition, the
results in chapter 4 imply indirectly the differences in the size of the union wage
premium between the two confederations. Some say that a need exists to unite the two
national union confederations and make a stronger union confederation in the near future.
In fact, both union confederations have suggested a merger of the two unions. Regarding
the merger of the two, this study implies that the two union confederations should

consider the differences including the wage/benefit mix and the size of the union wage

 

63 For example, in 2000, the President of the FKTU, Namsoon Lee, suggested forming a joint committee to
discuss a merger between the FKTU and the KCTU (Nodong Ilbo, May 21, 2000).

225

premium as well as others before a merger. The different preferences held by union
membership regarding wages and bargaining are unlikely to go away anytime soon and

will continue to be a source of internal conflict.

A Political Ergeriment: Establishments of Political Parties by Labor

For the presidential election of December 2002, the Democratic Labor Party (DLP),
which was created mainly by the KCTU in January 2000, selected Youngkil Kwon, a
former president of the KCTU, as its presidential candidate. The DLP had already earned
over 8% of voters in the previous election for local self-governance in June 2002,
emerging the 3rd party.

The other national confederation of unions, the F KTU, formed a separate political
party, the Democratic Society Party (DSP), in November 3,2002, and announced their
candidate to run for the presidential election of 2002. Namsoon Lee is now president of
the FKTU and was selected as a presidential candidate for the DSP. To date, the FKTU
has consistently taken the position to support presidential candidates for the ruling party,
expecting favorable labor policies in exchange for its political support. However, the
creation of the DSP by the FKTU means a departure from its traditional political position.

It appears that the creations of political parties by the two national confederations of
unions have resulted from the experiences of the past fifteen years. To the KCTU, the
strong need for political party representing the interests of labor has been felt in that it has
no connection with any political parties including ruling and opposition parties. Visible
oppressions by the government and its illegal status also prevent the KCTU from

expressing its interests efficiently in national politics. As a result, the situations which

226

occurred where successful strikes and militancy led by the KCTU resulted in vague
compromises done by the existing political parties, was clearly seen in the general strike
in early 1997. To the FKTU, the traditional support towards the ruling parties became one
of the reasons for them to be blamed for a govemment-controlled union, causing a
decline in union membership and damage to its legitimacy. The gains in exchange for
political support have been little, relative to the losses.

Obviously, the prospects for the two newly-born labor parties do not look bright for
the near firture, because so far the electorate tends to give their votes to either
conservative or mid-liberal party. Furthermore, the past fifteen years spent doing
collective bargaining within individual firms has stabilized industrial relationships in the
workplace level, leaving national industrial relations untouched. Therefore, we can
expect that the Korean trade unions will spend more resources on the national industrial
relationships and the political aspects of the unions in the near future, as they recognized

the lack of political power in fulfilling their interests in national politics.

227

 

 

APPENDIX

228

< Appendix: Telephone Interview Protocol Regarding Labor Cost Survey>
Hello? Is that **** Trade Union?

(If yes) How do you do? This is Jeonghyun Lee, I am a Ph.D. student in the School of
Labor/Industrial Relations at Michigan State University. While preparing my doctoral
dissertation, I am making a data-base for enterprise trade unions by collecting
information on unions. May I talk to the secretary of your union or one of full-time union
leaders, regarding the exact number of union membership and your union’s affiliation
with respect to national unions?

(If the wanted union leaders come on the phone)

Hello? (Job Title in the union?) How do you do? This is Jeonghyun Lee, I am a Ph.D.
student in the School of Labor/Industrial Relations at Michigan State University. While
preparing my doctoral dissertation, I am making a data-base for enterprise trade unions
by collecting information on unions. For that job, I am referring the Annual Directory of
Trade Unions, which is published by the Ministry of Labor in Korea. However, I have
found some unclear points or seemingly mismatched facts regarding your unions. So I am
calling you to make them clear. From my experiences working for trade unions, I know
very well that you are very busy, please may I have ten or twenty minutes of your time?

(If yes) Thank you very much. What I would like to know is
1. Regarding Union/Non-union

According to the Annual Directory of Trade Unions by the Ministry of Labor, your union
did not exist in the year of 19". Is that true?

(If yes)
(If no) 13 that so? The content in the Directory was not true.

2. Regarding the number of union membership

The Directory by the Ministry of Labor states that the number of your union membership
exceeds the number of regular employees, who, that is, are not temporary workers, part-
time workers, workers in the period of apprenticeship, in the end of 199*. I guess that the
number provided by the Directory is not correct. Would you check the exact number of
union membership in the end of 199*?

I know very well that it is not easy to check the fact, because the fact is from a few years
ago and/or probably other union leaders were in charge at that time. Other unions have
provided me the correct number by looking up records for check-off of union dues, which
were made by enterprise-side. If available, would you check them for me? Otherwise,
other official records that were prepared by your union are also good enough.

229

(If not available) If so, would you mind checking the fact through newsletters or meeting
minutes in your union?

(If all records mentioned above are not available) Could you or your colleagues in
your office remember the approximate number of union membership in the year of 199*?

(If the number of union members from him/her still exceeds the number of regular
workers) If all records were not available, would you find the exact number of regular
workers in your enterprise in the end of 199*? There is a possibility that the inaccurate
number of regular workers at that time is the cause of the problem, even though the
number of union members in the Directory was correct. If you do not have relevant
records for the fact, could you call anyone in your personnel or similar department in
your enterprise and obtain the true number for me? If needed, I can wait or call you back
at your convenience.

(If the number wanted is obtained) Thank you very much.
(If failed to get the number) I know you have done all you can do. I truly appreciate
your time. I will handle the problem, some other way.

3. Regarding the affiliation of your union to national union confederations
According to the Directory by the Ministry of Labor, your union was belonged to the
Federation of the Korean Trade Unions (or the Korean Confederation of Trade Unions) in
the year of 199*. Is that true? When using the term, “affiliation,” the criterion is whether
your union pays a part of union dues to the national confederation and/or your union

leaders participated in the formal conferences held by the national confederation.

Also the Directory says that your union had joined the F KTU (or KCTU) in the 199* and
later changed the affiliation to the KCTU (or FKTU) in the 199*. Is that true?

4. Regarding other plants or headquarter in other area
I would like to know if your enterprise where your union is located has headquarters or
plants in other areas. If so, does your union cover workers in the headquarters or plants or
do separate unions cover them?

5. Regarding miscellaneous questions
Please let me know if you know of any different enterprises with same name as yours.

6. Closing greeting
Thank you very much for your invaluable help. Thanks to you, I can supplement the

inaccuracies found in the Annual Directory of Trade Unions published by the Ministry of
Labor. Again thank you. Good bye.

230

BIBLIOGRAPHY

231

Abowd, J. M. and H. S. Farber. 1982. “Job Queues and the Union Status of Workers.”
Industrial and Labor Relations Review, Vol. 35, No. 3, 354-67.

Adams, J. Tracy. 1964. “Inequity in social exchange.” in Leonard Berkowitz (ed.),
Advances in Experimental Social Psychology, Vol. 2, 267-299, New York: Academic
Press.

Adler, Nancy J. 1997. International Dimensions of Organizational Behavior, Cincinnati:
South-Western College Publishing.

Agell, Jonas and Kjell Erik Lommerud. 1992. “Union Egalitarianism as Income
Insurance.” Economica, Vol. 59, Issue 235, August, 292-310.

Alpert, William. “Unions and Private Wage Supplements.” Journal of Labor Research 3,
Spring 1982, 179-190.

Alpert, William T. and Stephen Woodbury (Ed.). 2000. Employee Benefits and Labor
Markets in Canada and the United States, MI: W.E. Upjohn Institute for Employment
Research.

Allen, Steven G. and Robert L. Clark. 1986. “Unions, Pension Wealth, and Age-
Compensation Profiles.” Industrial and Labor Relations Review, Vol. 39, No. 4, July,
502-517.

Amsden, Alice H. 1989. Asia ’s Nest Giant: South Korea and Late Industrialization,
Oxford University Press.

An, Chun-sik. 1989. “Personnel Management.” In Koean Employers Federation, A
History of Labor Economy for the Past 40 Years, Seoul: KEF.

Antos, Joseph R. 1983. “Union Effects on White-Collar Compensation.” Industrial and
Labor Relations Review, Vol. 36, No. 3, April, 461-479.

Arbache, Jorge Saba, “Do unions always decrease wage dispersion? The case of Brazilian
manufacturing,” Journal of Labor Research, Summer 1999, Vol. 20, Issue 3, 425-436.

Ashenfelter, Orley and George E. Johnson. 1969. “Bargaining Theory, Trade Unions, and
Industrial Strike Activity.” American Economic Revew, Vol. 59, No. 2, 35-49.

Ashenfelter, Orley and George E. Johnson. 1972. “Unionism, Relative Wages, and Labor
Quality in U.S. Manufacturing Industries.” International Economic Review, Vol. 13, No.
3, October, 488-508.

Bae, Kiusik. 2002. “The Evaluation of Industrial Relations in the First half and Prospects

for the Seond Half.” Monthly Labor Trend, August, Seoul: Korea Labor Institute.
<http://www.kli.re.kr>

232

Bae, moo-gi. 1990. “Relative Wage Effects of Trade Unions.” Journal of Korean Labour,
Vol. 1,No. 1, 1-25.

. 1995. Labor Economics, Seoul: Kyungrnoonsa.

 

Bang, Hanam. 2002. “Selective benefit system and the role of private insurance.”
presented in the conference of corporate welfare and the role of insurance by Korean
academy of insurance in May 10, 2000.

Beam, Burton T. and John J. McFadden. 2001. Employee Benefits (sixth Edition),
Chicago, IL: Real Estate Education Company.

Becker, Gary S. 1964. Human Capital, New York: Columbia University Press.

Bellante and Long. 1981. “The Political Economy of Rent-seeking Society: The Case of
Public Employees.” Journal of Labor research, Vol. 2, Spring, 4-12.

Belman, Dale and John S. Heywood. 1990. “Union Membership, Union Organization and
The Dispersion of Wages.” Review of Economics and Statistics, Vol. 72, No. 1, 148-153.

. 1990. “Application of the ‘Oaxaca Decomposition’ to Probit Estimates: The
Case of Unions and Fringe Benefits Provision.” Economic Letters 32, 101-104.

 

Benjamin, Dwayne, Morley Gunderson, and Craig Riddell. 1998. Labour Market
Economics: Theory, Evidence, and Policy in Canada, Toronto: McGraw-Hill.

Berg, Peter and Eunmi Chang. 1999. “Comparative Industrial Relations.” In Managing
Human Resources in the 21‘" century, Ellen Kossek and Richard Block (ed.), Cincinnati,
Ohio: South-Western, 26.2-26.20.

Blanchflower, David G. and Richard B. Freeman. 1992. “Unionism in the United States
and Other Advanced OECD Countries.” Industrial Relations, Vol. 31, No. 1, Winter, 56-
79.

Bloch, Farrell E. and Mark S. Kuskin. 1978. “Wage determination in the union and non-
union sectors.” Industrial and Labor Relations Review, Vol. 31, No. 2, January, 183-192.

Block, Richard N. 1995. “Labor Law, Economics, and Industrial Democracy: A
Reconciliation.” Industrial Relations, Vol. 34, No. 3, July, 402-416.

Block, Richard N., Jeonghyun Lee, and Eunjong Shin. 2000. “The Korean Industrial
Relations System: From Post-Independence To Post-IMF.” Presented at the Conference
Transforming Korean Business and Management Culture, Sep. 19-20, 2000, Michigan
State University, East Lansing, Michigan.

233

Bloom, Matt & John G. Michel, “The Relationships among organizational context, pay
dispersion, and managerial turnover.” Academy of Management Journal, 2002, Vol. 45,
No. 1, 33-42.

Booth, Alison. 1984. “A Public Choice Model of Trade Union Behavior and
Membership.” Economic Journal, December, Vol. 94, Issue 376, 883-898.

Brown, William, Simon Deakin, David Nash and Sarah Oxenbridge. 2000. “The
Employment Contract: From Collective Procedures to Individual Rights.” British Journal
of Industrial Relations, Vol. 38, No. 4, October, 611-629.

Buchinsky, Moshe. 1994. “Changes in the US wage structure 1963-1987: application of
quantile regression.” Econometrica, Vol. 62, 405-458.

. 1998. “Recent Advances in Quantile Regression Models: A Practical

Guideline for Empirical Research.” Journal of Human Resources, winter, Vol. 33, No. 1,
88-126.

 

 

. 1998. “Quantile Regression with Sample Selection: Estimating Women’s
Return to Education in the U.S.” Empirical Economics, Vol. 26, 87-113.

Buchmueller, Thomas C., John DiNardo, and Robert G. Valletta. 2002. “Union Effects
on Health Insurance provision and Coverage in the United States.” Industrial and Labor
Relations Review, Vol. 55, No. 4, July, 610-627.

Budd, John W. 1998. “Fringe Benefits and Union Status: Members versus Covered
Nonmembers.” Industrial and Labor Relations Association Series, Proceedings of the
F ifiieth Annual Meeting, Vol. 1, 250-258.

Budd, John W. and Brian P. McCall. 2001. “The grocery stores wage distribution: A
semi-parametric analysis of the role of retailing and labor market institutions. ” Industrial
and Labor Relations Review, Vol. 54, No. 2A, 484-401 .

Card, David. 1992. “The Effect of Unions on the Distribution of Wages: Redistribution or
Relabelling?” Working Paper No. 4195, Cambridge, MA: National Bureau of Economic
Research, October.

. 1996. “The effect of unions on the structure of wages: A longitudinal
analysis.” Econometrica, July, Vol. 64, No. 4, 957-979.

 

Carrington, William J ., Kristin McCue, and Brooks Pierce. 2002. “Nondiscrimination
Rules and the Distribution of Fringe Benefits.” Journal of Labor Economics, Vol. 20, No.
2, 5-33.

Chae, Changkwoon. 1992. Relative Wage Effects of Trade Unions in Monopolized and
Nonmonopolized Sectors, unpublished doctorial dissertation, Seoul National University.

234

Chamberlain, Gary. 1994. “Quantile regression, censoring and the structure of wages, in
Proceedings of the Sixth World Congress of the Econometrics Society, Barcelona, Spain
(eds.) C. Sims and J. J. Laffort, Cambridge Univesity Press, New York, 171-209.

Chang, Hong-Geun. 1999. The Transformation of ‘Labor Regime’ in Korea, 1987-1997,
unpublished Ph.D dissertation, Seoul National University.

Cho, Woo-Hyun and Sec, Kyeong Joon. 1992. “Determinants of Propensity to Join
Unions and Unions’ Relative Wage Effects,” Economic Journal, Vol. 45, No. 3, 99-127.
(in Korean)

Choi, HyoChul. 1992. A Study on the Economic Rationale and Effects of Non- Wage
Labor Costs. Unpublished Ph.D. Dissertation, Seoul: Seoul National University.

Choi, J ang-jip. 1997. Labor Movement and the State in Korea. Seoul: Nanarn Publishing
House.

Choi, Jong-Tae. 1996. Stategic Industrial Relations, Seoul: Kyungrnunsa Publishing
Company.

Choi, Kyun. 1992. Socio-economic Characteristics of Corporate Welfare in Korea,
unpublished Ph.D Dissertation, Seoul National University.

Choi, Seun-No. 1998. The Korean Big Business Groups, Seoul: Nanam.

Choi, Yooseok. 2000. “The influences of market structure on the level of corporate
welfare.” Donghyangkou Chunmang, Vol. 45, summer, Seoul: Korea Social Science
Institute.

Choi, Young-Ki. 1999. “The Tradition of Social Pact and the Tripartite Committee in
Korea,” paper presented in the Joint Conference by the Korean Academy Association of
Industrial Relations, Korean Academy Association of Labor Economics, and Korean
Academy Association of Labor Law, January, Seoul.

Clark, Robert L. and M. Melinda Pitts. 1999. “Faculty Choice of a Pension Plan: Defined
Benefit versus Defined Contribution.” Industrial Relations, Vol. 38, No. 1, January, 18-
45.

Cobble, Dorothy S. and Hae-Jin Kim. 1998. “Work and Employment relations: A
Doctoral Student’s View.” Perspectives on Work, Industrial Relations Research
Asociation, Vol. 2, No. 1, 47-50.

Cohen, Murray E. and Cynthia F. Cohen. 1998. “The impact of union status on general
employee municipal pension plans in Florida.” Journal of Collective Negotiations, Vol.
27, No. 2, 121-129.

235

Commons, John R. 1909. “American Shoemakers, 1648-1895.” The Quarterly Journal of
Economics, Vol. 24, No. 2, 39-84.

Cowherd, Douglas M. and David I. Levine. 1992. “Product Quality and Pay Equity

between Lower-level Employees and Top Management: An Investigation of Distributive
Justice Theory.” Administrative Science Quarterly, Vol. 32, 302-320.

Davis, Mike. 1986. Prisoners of the American Dream: Politics and Economy in the
History of the US Working Class, London: Verso.

Deyo, Fred. 1989. Beneath the Miracle: Labor Substitution in the New Asian
Industrialism, Berkeley, CA: University of California Press.

DiNardo, John and Thomas Lemieux. 1997. “Diverging male wage inequality in the
United States and Canada, 1981-1988: Do institutions explain the difference?” Industrial
and Labor Relations Review, Vol. 50, No. 4, July, 629-651.

DiNardo, John, Kevin Hallock, and Jorn-Steffen Pischke. 1997. “Unions and Managerial
Pay.” NBER Working Paper No. w6318.

Doeringer, Peter B. 1984. “Unions: Economic Performance and Labor Market Structure.”
in The Economics of Trade Unions: New Directions, Jean-Jacques Rosa (ed.), 1984,
Kluwer-Nijhoff Publishing, 315-338.

Donohue, Susan M. and John S. Heywood. 2000. “Unionization and Nonunion Wage
Patterns: Do Low-Wage Workers Gain the Most?” Journal of Labor Research, Summer,
Vol. 21, No. 3, 489-502.

Doiron, Denise and Riddell W. Craig. 1994. “The Impact of Unionization on Male-
Female Earnings Differences in Canada.” Journal of Human Resources, Vol. 29, Spring,
386-395.

Donsimoni, Marie-Paule and Robert Shakotko. 1979. “Unionism and the Structure of
Total Compensation.” Columbia University Department of Economics Discussion Paper
No. 21, August.

Dubofsky, Melvyn and Foster R. Dulles. 1999. Labor in America: a history, Wheeling,
111.: Harlan Davidson.

Duncan, Greg M. and Duane E. Leigh. 1980. “Wage Determination in the Union and
Nonunion Sectors: A Sample Selectivity Approach.” Industrial and Labor Relations
Review, Vol. 34, No. 1, October, 24-35.

Duncan, Greg and Frank Stafford. 1980. “Do Union Receive Compensating Wage
Differentials?” American Economic Review 70, June, 355-371.

236

Dunlop, John T. 1944. Wage Determination Under Trade Union, New York: Augustus
Kelley.

. 1993. Industrial Relations Systems, Boston, M.A.: Harvard University

 

Press.

Escobar, Modesto. 1995. “Spain: Works Councils or Unions?” in Roger, J. and W. Streek
(ed.). Work Councils: Consultation, Representation and Cooperation in Industrial
Relations. Chicago: University of Chicago Press, 153-189.

Even, William E. and David A. Macpherson. 1991. “The impact of unionism on fiinge
benefit coverage.” Economic Letters, Vol. 36, 87-91.

Farber, Henry S. 1978. “Individual Preferences and Union Wage Determination: The
Case of the United Mine Workers.” Journal of Political Economy, Vol. 86, No. 5, 923-
942.

. 1986. “The Analysis of Union Behavior.” In 0. Ashenfelter and R. Layard,
eds., Handbook of Labor Economics, Vol. 2. Amsterdam: North-Holland.

 

Farber, Henry S. and Daniel H. Saks. 1980. “Why Workers Want Unions: The Role of
Relative Wages and Job Characteristics.” Journal of Political Economy, April, Vol. 88,
No. 2, 349-369.

Feuille, Peter, Wallace E. Hendricks, and Lawrence M. Kahn. 1991. “Wage and
Nonwage Outcomes in Collective Bargaining: Determinants and Tradeoffs.” Journal of
Labor Research 2, No. 1, Spring, 39-54.

Fichtenbaum, Rudy and Paulette Olson. 2002. “The Impact of Unionization On Health
Insurance Benefits.” Journal of Economic Issue, Vol. 36, No. 2, 223-230.

Filer, Randall K., Daniel S. Hamermesh, and Albert E. Rees. 1996. The Economics of
Work and Pay (sixth edition), New York: HarperCollins College Publishers.

Fiorito, Jack and Wallace E. Hendricks. 1987. “Union Characteristics and Bargaining
Outcomes.” Industrial and Labor Relations Review, Vol. 40, No. 4, July, 569-584.

Fosu, Augustin K. 1979. “Choice of Fringe Benefits As a form of Labor Earnings — A
Theoretical and Empirical Analysis.” Ph.D. Dissertation, Northwestern University.

. 1984. “Unions and Fringe Benefits: Additional Evidence.” Journal of
Labor Research, Vol. 5, No. 3, Summer, 247-254.

 

. 1993. Nonwage benefits as a limited dependent variable: Implications for
the impact of unions.” Journal of Labor Research, Vol. 14, No. 1, Winter, 29-43.

 

237

Freeman, Richard B. 1980. “Unionism and the dispersion of wages.” Industrial and
Labor Relations Review, October, Vol. 34, No. 1, 3-23.

. 1981. “The Effect of Unionism on Fringe Benefits.” Industrial and Labor
Relations Review, Vol. 34, No. 1, October, 3-23.

 

. 1982. “Union Wage Practices and Wage Dispersion within Establishment.”
Industrial and Labor Relations Review, Vol. 36, No. 1, October, 3-21.

 

. 1994. “American Exceptionalism in the Labor Market: Union-Nonunion
Differentials in the United States and Other Countries.” In Clark Kerr and Paul D.
Staudohar (edit), Labor Economics and Industrial Relations: Markets and Institutions,
MA: Harvard University Press, 272-299.

 

Freeman, Richard B. and James L. Medoff. 1984. What Do Unions Do?, New York:
Basic Books.

Frenkel, Stephen J. and David Peetz. 1998. “Globalization and Industrial Relations in
East Asia: A Three-Country Comparison.” Industrial Relations, Vol. 37, No.3, July, 282-
310.

Friedman, Milton. 1962. Capitalism and Freedom, Chicago: University of Chicago Press,
124.

Gahan, Peter G. 2002. “What do unions maximize? Evidence from survey data.”
Cambridge Journal of Economics, Vol. 26, 279-298.

Gentrty, William M. and Eric Peress. 1994. “Taxes and Fringe Benefits Offered By
Employers,” National Bureau of Economic Research Working Paper #4764, Cambridge,
MA.

Gerschenkron, A. 1962. “Economic Backwardness in Historical Perspective.” In A.
Gerschenkron (ed.), Economic Backwardness in Historical Perspective and Other Essay,
Cambridge, M.A.: Harvard University Press.

Ghilarducci, Teresa and Kevin Terry. 1999. “Scale Economics in Union Pension Plan
Administration: 1981-1993.” Industrial Relations, Vol. 38, No. 1, January, 11-17.

Goetschy, Janine and Annette Jobert. 1998. “Employment relations in France.” In Greg J.
Bamber and Russell D. Lansbury. International and Comparative Employment Relations:
A Study of Industrialized Market Economies. Allen and Unwin. 169-199.

Godard, John. 2000. Industrial Relations, the Economy, and Society, Toronto: Captus
Press

238

Goldfield, Michael. 1987. The Decline of Organized Labor in the United States, Chicago:
University of Chicago Press.

Gordon, David M., Richard Edward, and Michael Reich. 1982. Segmented Work and
Divided Workers: The Historical Transformation of Labor in the United States,
Cambridge: Cambridge University Press.

Gordon, Cleveland, Morley Gunderson, and Douglas Hyatt. 2003. “Union effects in low-
wage services: Evidence from Canadian childcare.” Industrial and Labor Relations
Review, Vol. 56, No. 2, January, 295-305.

Greene, William H. 2000. Econometric Analysis, Upper Saddle River, NJ: Prentice Hall.

Halvorsen, Robert and Raymond Palmquist. 1980. “The Interpretation of Dummy

Variables in Semilogarithmic Equations.” American Economic Review, Vol. 70, Issue 3,
January, 474-475.

Hart, Robert A. and Seiichi Kawasaki. 1999. Work and Pay in Japan, Cambridge:
Cambridge University Press.

Hartog, Joop, Pedro T. Pereira, and José A. C. Viera, “Changing returns to education in
Portugal during the 19803 and early 19903: OLS and quantile regression estimators.”
Applied Economics, 2001, Vol. 33, 1021-1037.

Heckman, James J. 1979. “Sample Selection Bias as a Specification Error,”
Econometrica, Vol. 47, 153-161.

Hendricks, Wallace. 1981.“Unionism, Oligopoly, and Rigid Wages.” Review of
Economics and Statistics 63, No. 2, May, 198-205.

Hildreth, Andrew G. 2000. “Union wage differentials for covered members and
nonmembers in Great Britain.” Journal of Labor Research, Winter, Vol. 21, Issue 1, 133-
147.

Hill, Richard C. 1974. “Unionization and Racial Income Inequality in the Metropolis.”
American Sociological Review, Vol. 1974, No. 4, August, 507-522.

Hirsch, Barry T. and John T. Addison. 1986. The Economic Analysis of Unions, Boston:
Allen & Unwin.

Hofstede, Geert. 1980. Cultural Consequences: International Differences in Work-
Related Values, Beverly Hills, CA: Sage Publications.

Hong, Kyungzoon. 1996. Determinants of the Company Welfare: The Organizational

Characteristics of Manufacturing Companies in Korea. Unpublished Ph.D. Dissertation,
Seoul: Seoul National University.

239

Hoxie, R. [1915] 1966. Trade Unionism in theUnited States, New York: Russell and
Russell.

Hyclak, Thomas. 1979. “The Effect of Unions on Earnings Inequality in Local Labor
Markets.” Industrial and Labor Relations Review, Vol. 33, No. 1, October, 77-84.

Hyman, Richard. 1975. Industrial Relations: A Marxist Introduction, London:
Macmillan.

lchniowski, Casey, Richard Freeman, and Harrison Laue. 1989. “Collective Bargaining
Laws, Threat Effects, and the Determinants of Police Compensation.” Journal of Labor
Economics, Vol. 7, No. 2, April, 191-209.

ILO (lntemational Labor Organization). Various years. Yearbook of Labor Statistics,
Geneva: ILO.

Im, Youngil. 1997. Labor Movement and Class Politics in Korea (1987-1995): Struggles
for Changes and Changes for Bargaining, unpublished Ph.D. dissertation, Busan
University.

Jarrell, Stephen B. and T. D. Stanley. 1990. “ A meta-analysis of the union-nonunion
wage gap.” Industrial and Labor Relations Review, Vol. 44, No. 1, October, 54-67.

Jeong, Insoo. 1991. Structures of Wages in Korea, Korean Labor Institute: Seoul.

Jeong, Jinho and Kangsik Choi. 2001. An Analysis on the Determinants of Increasing
Inequality of Wage Income. Seoul: Korea Labor Institute.

Jeong, Seongkee. 1993. A Study on the Ideal and Reality of Labor Union Movement
Between Blue-Collar and White-Collar Workers in Korea. Unpublished Ph.D.

dissertation, Seoul: Seoul National University.

Johnson, George E. 1984. “Changes over Time in the Union-Nonunion Wage Differential
in the United States,” in The Economics of Trade Unions: New Directions, Jean-Jacques
Rosa (Eds), Kluwer-Nijhoff Publishing.

Johnson, George E. 1975. “Economic Analysis of Trade Unionism.” American Economic
Review, May, Vol. 45, No. 2, 23-28.

Johnson, George and Kenwood C. Youmans. 1971. “Union Relative Wage Effects by
Age and Education.” Industrial and Labor Relations Review, Vol. 24, No. 2, 171-179.

Jorgensen, Helene J. 1998. “What Has Happened to the Union Wage Gap in Times of
Deunionization? Trends in the Union Wage Gap, 1984-1994.” Politics and Society, Vol.
26, No. 2, June, 257-272.

240

Juhn, Chinhui, Kevin M. Murphy, and Brooks Pierce. 1993. “Wage Inequality and the
Rise in Returns to Skill.” Journal of Political Economy, Vol. 101, No. 3, 410-442.

Jung, Ih-hwan. 1992. Changes and Industrial Relations in Internal Labor Market in
Manufacturing Sector. Unpublished Ph. D. Dissertation, Seoul National University.

Kahn, Lawrence M. 1979. “Unionism and Relative Wages: Direct and Indirect Effects.”
Industrial and Labor Relations Review, Vol. 32, 520-532.

. 2000. “Wage inequality, collective bargaining, and relative employment
from 1985 to 1994: evidence from fifteen OECD countries.” Review of Economics and
Statistics, November, Vol. 82, No. 4, 564-579.

 

Kahn, Lawrence M. and Michael Curme. 1987. “Unions and Nonunion Wage
Dispersion.” Review of Economics and Statistics, Vol. 69, No. 4, November, 600-607.

Kahn, Shulamit, “What occupational safety tells us about political power in union firms?”
RAND Journal of Economics, Vol. 21, No. 3, Autumn 1990, 481-496.

Kang, Chulkyu, Jeongpyo Choi, and J isang Jang. 1991. Chaebol: Sungiang jooyeokinga
Tamyeokye Hwasininga (Hero of Development or a Perfect Picture of A varice). Beebong
Publishers. Seoul.

Katz, Harry C. and Thomas A. Kochan. 1992. An Introduction to Collective Bargaining
and Industrial Relations. New York: McGraw-Hill, Inc.

Kaufman, Bruce E. and Julie L. Hotchkiss. 1999. The Economics of Labor Markets, FL:
Dryden.

Kennedy, Peter E. 1981. “Estimation with Correctly Interpreted Dummy Variables in
Semilogarithmic Equations.” American Economic Review, Vol. 71, Issue 4, September,
801.

Korean Confederation of Trade Unions (KCTU). 2001. Democratic Unions: History of
Struggle and Being Oppressed (1 970-2000), Seoul: Future In The Workplace.

Kerr, G, J. T. Dunlop, F. Harbison, S. Meyers. 1960. Industrialism and Industrial Man,
Cambridge: Harvard University Press.

Kelly, John. 1998. Rethinking Industrial Relations: Mobilization, collectivism, and long
waves, London: Routledge.

Kim, Dae-Il and Robert Topel. 1995. “Labor Markets and Economic Growth: Lessons
from Korea’s Industrialization, 1970-1990.” In Richard Freeman and Lawrence Katz,
eds., Diflerence and Chagnes in Wage Structure, University of Chicago Press, 227-264.

241

Kim, Dong-Chun. 1995. A Study of Working-Class in Korea, Seoul: Yeoksabeepyongsa.

Kim, Eun Mee. 1987. From Dominance to Symbiosis: State and Chaebol in the Korean
Economy, 1960-1985, Unpublished Ph. D. Dissertation, Brown University.

Kim, Jang-Ho. 1991. “Wage Effects of Trade Unions: Differences in Wage
Determination Mechanism in Korean Manufacturing Sectors.” Journal of Economics,
Vol. 39, No. 1, 25-49.

Kim, Jeong-Han. 1997. A Study on Present Collective Bargaining in Korea, Seoul: Korea
Labor Institute.

Kim, Kyong-Dong. 1984. “Socio-Economic Changes and Political Selectivity in the
Development of Industrial Democracy in the Republic of Korea.” Economic and
Industrial Democracy, Vol. 5, 445-467.

Kim, Sang-Kon. 1996. “A Historical Study on Organizational Form of Industrial Unions
in Korea.” Management Research, Vol. 3, Hansin University.

Kim, Se-Kyun. 2001. “Reforms in the Dae-Joong Kim Regime.” In Democratic Unions:
History of Struggle and Being Oppressed( I 970-2000), KCTU, Seoul: Future In The
Workplace, 306-319.

Kim, Sikhyun and Jaehoon Jeong. 1995. Industrial Relations, Seoul: Hakhyunsa. Seoul.

Kim, Sookon. 1983. “Characteristics of Korean Industrial Relations and Policies on
Income Equalization.” Journal of Korea Development, August, Vol. 5, No. 3, 2-18.

. 2000. “Korean Trade Union Movement, Manpower Policy and Labor
Market.” In Proceedings of the Twelfth World Congress of the International Industrial
Relations Association (IIRA), Track 3, May 29-June 2, Japan, 64-73.

 

Kim, Younghak and Ho-keun Song. 1998. Networks of The Korean Trade Unions. Seoul:
Central Institute of the Federation of Korean Trade Unions (F KTU).

Kleiner, Morris M. and Young-Myon Lee. 1997. “Works Councils and Unionization:
Lessons from South Korea,” Industrial Relations, Vol. 36, No. 1, January, 1-16.

KLI (Korea Labor Institute). 1995. A Study on Efi’icient Working Methods for Existing
Business and Workers' Compensation Due to An Organizational Change fi'om in Korea
Labor Welfare Corporation, Seoul: Korea Labor Institute.

. various years. KLI Labor Statistics: Data Available for Collective
Bargaining. Seoul. (in Korean)

 

242

KLSI. 1998. Trade Unions and Wage System. Seoul: Korean Labor and Society Institute.

Kochan, Thomas A., Harry C. Katz, and Robert B. McKersie. 1986. The Transformation
of American Industrial Relations, New York: Basic Books.

Koenker, Roger, and Gilbert Bassett. 1978. “Regression Quantiles.” Econometrica, Vol.
46, 33-50.

Koenker, Roger, and Gilbert Bassett. 1982. “Robust Tests for Heteroscedasticity Based
on Regression Quantiles.” Econometrica, Vol. 50, 43-61.

Koike, Kazuo. 1988. Understanding Industrial Relations in Modern Japan. Mary Saso,
trans, New York: St. Martin’s.

Koo, Hagen. 1996. Work, Culture, and Consciousness of the Korean Working Class.” In
Elizabeth J. Perry (ed.), Putting Class in Its Place: Worker Identities in East Asia,
Berkeley, Calif: Institute of East Asian Studies, University of California, 53-76.

. 2001. Korean Workers: The Culture and Politics of Class Formation, Itaca:
Cornell University Press.

 

Koo, In-Hwoi. 1996. A Study on the Eflects of Trade Unions upon Degree of Wage
Inequlity, unpublished Master dissertation in Seoul National University.

Kornfield, Robert. 1993. “The effects of union membership on wages and employee
benefits: The case of Australia.” Industrial and Labor Relations Review, Vol. 47, No. 1,
October, 114-128.

Kuhn, Peter. 1998. “Unions and the Economy: What We Know and What We Should
Know.” Canadian Journal of Economics, Vol. 31, No. 5, November, 1033-1056.

Kuruvilla, Sarosh an Cristopher L. Erickson. 1999. “Industrial Relations Change in Asia:
A Comparative Review” June, UCLA Center for lntemational Business Education and
Research Working Paper Series, 99-6.

Kwon, Hyun-Ji. 1997. Administration of Trade Unions Today and Tasks, Seoul: Central
Institute of the Federation of Korean Trade Unions (F KTU).

Lanot, Gauthier and Ian Walker. 1998. “The union/non-union wage differential: An
application of semi-parametric methods.” Journal of Econometrics, Vol. 84, Issue 2,
June, 328-349.

Lansbury, Russell D., Byong-Hoon Lee, and Seoghun Woo. 2002. “The reform of
production system and employment relations in the Korean Auto Industry: The
experience of Kia Motors in the 19903.” New Zealand Journal of Industrial Relations,

February, Vol. 27, No. 1, 79-91.

243

Lazear, Edward P. and Shewin Rosen. 1981. “Rank-order tournament as Optimum Labor
Contracts.” Journal of Political Economy, Vol. 89, No. 5, 841-864.

Lee, Byunghoon, Kwangpyo Noh, Kunho Oh, and Soobum In. 2001. A Study on F u]!-
time Union Leaders. Seoul: Korea Labour & Society Institute.

Lee, Jae-Yeol. 1993. “A study on mechanisms of wage determination with a perspective
of organizational theory. ” Journal of Korean Sociology, Vol. 27, Summer, 87-116.

Lee, Jae-Yeol and Hyun-Ji Kwon. 1995. Korean Trade Unions in l990s: A Search for
Roles of Trade Unions and Industrial Democracy, Institute of the Federation of Korean
Trade Unions (FKTU).

Lee, Jeongwoo and Sangseup Nam. 1990. “The Effects of Trade Unions on Wage
Distribution in Korea.” Research on Economics, Vol. 41, No. 3, 251-277.

Lee, Joohee. 2000. “The Financial Crisis, Search for Flexibility and Changes in Industrial
Relations: The South Korea Case.” Presented for the international conference on

Transforming Korean Business and Management Culture at Michigan State University,
September 19-20.

Lee, Lung-Fei. 1978. “Unionism and Wage Rates: A Simultaneous Equations Model with

Qualitative and Limited Dependent Variables.” International Economic Review, Vol. 19,
No. 2, June, 415-433.

Leibowitz, Arleen. 1983. “Fringe Benefits in Employee Compensation.” In J .E. Triplett,
(ed.), Measurement of Labor Costs, Chicago, IL: University of Chicago Press, 371-394.

Leigh, D. E. 1981. “The Effect of Unionism on Workers’ Valuation of Future Contract
Coverage in the United States, 1983-88.” Industrial and Labor Relations Review,
October, Vol. 44, No. 1, 510-521.

Leigh, J. Paul. 1982. “Are Unionized Blue-Collar Jobs More Hazardous Than
Nonunionized Blue-Collar Jobs?” Journal of Labor Research, Vol. 3, Summer, 349—357.

Lemieux, Thomas. 1993. “Unions and Wage Inequality in Canada and the United States,”
in David Card and Richard B. Freeman, (ed.), Small Difi’erence that Matter: Labor
Markets and Income Maintenance in Canada and the United States, Chicago and
London: University of Chicago Press, 69-107.

. 1998. “Estimating the Effects of Unions on Wage Inequality in a Panel
Data Model with Comparative Advantage and Nonrandom Selection.” Journal of Labor
Economics, Vol. 16, No. 2, 261-291.

 

244

Leslie, Derek and Yonghao Pu. 1996. “What Caused Rising Earnings Inequality in
Britain? Evidence from Time Series, 1970-1993.” British Journal of Industrial Relations,
Vol. 34, No. 1, March, 111-130.

Lester, Richard A. 1967. “Benefits as a Preferred Form of Compensation.” Southern
Economic Journal, Vol. 33, April, 488-495.

Levine, David I. 1993. “What Do Wages Buy?” Administrative Science Quarterly, Vol.
38, Issue 3, September, 462-483.

Levine, David and Peter Feuille. 1983. “Behavioral Research in Industrial Relations.”
Industrial and Labor Relations Review, Vol. 36, No. 3, April, 341-361 .

Lewis, H. G. 1963. Unionism and Relative Wages in the United States, Chicago:
University of Chicago Press.

. 1986. Union Relative Wage Effect: A Survey, Chicago: Chicago University

 

Press.

Lie, John. 1991. “The Prospect for Economic Democracy in South Korea.” Economic and
Industrial Democracy, Vol. 12, 501-513.

Mabry, Bevars. 1973. “The Economics of Fringe Benefits.” Industrial Relations, Vol. 12,
No. 1, February, 95-106.

Maddala, G. S. 1983. Limited Dependent and Qualitative Variables in Econometrics.
Cambridge: Cambridge University Press.

Maddala, G. S. 1992. Introduction to Econometrics (2” Edition), New York: Macmillian
Publishing Company.

Maranto, Cheryl L. 1985. “Union Effects on Human Capital Investments and Return.”
Journal of Human Resources, Vol. 20, Issue 3, Summer, 453-462.

Martin, Donald L. 1980. An Ownership Theory of the Trade Union, Berkeley: University
of California press.

Martin, Ross M. 1989. Trade Unionism: Purposes and Forms, Oxford: Clarendon Press.

Martocchio, Joseph J. 2001. Strategic Compensation: A Human Resource Management
Approach, Upper Saddle River, New Jersey: Prentice Hall.

Marx, Karl. 1947. The German Ideology, lntemational Publishers.

McNamara, Dennis. 1996. “Corporatism and Cooperation among Japanese Labor.”
Comparative Politics, Vol. 28, No. 4, July, 379-397.

245

Menezes-Filho, Naercio, David Ulph, and John Van Reenen. 1998. “The determination of
R&D: Empirical evidence on the role of unions.” European Economic Review, Vol. 42,
Issue 3-5, 919-930.

Mershon, Carol A. 1989. “Between Workers and Union: Factory Councils in Italy.”
Comparative Politics, Vol. 21, No. 2, January, 215—235.

Milkovich, George T. and Jerry M. Newman. 1999. Compensation, New York: McGraw-
Hill.

Miller, Ronald L. and Matthew Matsukichi Amano. 1995. “Trade Unions in Japan: The
Consequences of Enterprise Unionism.” New Zealand Journal of Industrial Relations,
Vol. 20, No. 1, April, 35-48.

Mincer, J. 1974. Schooling, Experience and Earnings. New York: Chapman and Hall.

Ministry of Labour (MOLa), various years. The Report on Basic Survey on Wage
Structure, Seoul.

. 1998, Survey Report on Labour Costs of Enterprises, Seoul.

 

. various years, Directory of Trade Unions in Korea, Seoul.

 

Mitchell, Olivia S. and Robert S. Smith. 1992. “Public Sector Pension: Benefits, Funding
and Unionization.” Proceedings of the Industrial Relations Research Association,
January, 126-133.

Montenegro, Claudio. 2001. “Wage Distribution in Chile: Does Gender Matter? A
Quantile regression Approach.” Policy Research Report on Gender and Development,
World Bank Working Paper Series No. 20.
Grttp://www.worldbank.org/html/prmge/prr/wp20.pdf)

Moore, William and John Raisian. 1987. “Union-Nonunion Wage Differentials in the
Public Administration, Educational and Private Sectors.” Review of Economics and
Statistics 69, November, 608-615.

Nahm, Pyeong Tak. 1997. “Changes in Wage Inequality and the Effect of Human Capital
in Korea: 1972-1989.” Journal of Korean Labor Economics, Vol. 20, No. 1, July, 211-
231.

National Statistical Office (N SO). Korean Statistical Information System.
(http://n30.go.kr)

NCTU (National Congress of Trade Unions). 1990. An Analysis on Standard Agreements
of Collective Bargaining, Seoul: Dolbegae.

246

. 1992. An Analysis on Standard Agreements of Collective Bargaining,
Seoul: Dolbegae.

 

Ng, Ignace and Dennis Maki. 1994. “Trade Union Influence on Human Resource
Management Practices.” Industrial Relations, Vol. 33, No. 1, January, 121-135

Noh, Joong-Ki. 2001. “Changes in the Korean Labor Politics Regime.” In Democratic
Unions: History of Struggle and Being Oppressed (I970-2000), KCTU, Seoul: Future In
The Workplace, 282-305.

Olsen, Craig A. 1994. “The Value of Health and Pension Benefits to Workers.” In
Proceedings of the F orty-Sixth Annual Meeting of the Industrial Relations Research
Association, edited by Paula Voos, 37-47, Madison, WI: Industrial Relations Research
Association.

. 2002. “Do workers accept lower wages in exchange for health benefits?”
Journal of Labor Economics, Vol. 20, No. 2, April, 91-114.

 

Olsen, Randall J. 1980. “A Least Squares Correction for Selectivity Bias.” Econometrica,
Vol. 48, No. 7, 1815-1820.

Olson, Mancur. 1971. The Logic of Collective Action, Mass.: Harvard University Press.

Oaxaca, Ronald. 1973. “Male-Female Wage Differentials in Urban Labor Market.”
International Economic Review, Vol. 14, No. 3, October, 693-709.

O’Neill, Kathleen B. 1991. “Industrial Relations in Korea: Will Korea Become Another
Japan?” Comparative Labor Law Journal, Spring, Vol. 12, No. 3, 360-392.

Organization for Economic Co-operation and Development (OECD). 1998. OECD
Economic Surveys: Korea (Special Features: Financial Sector Reform Corporate
Restructuring The Labor Market), August, No. 1, Paris.

. 1999. OECD Economic Outlook, December, No. 66, Paris.

 

Park, Charnim. 1996. “A Study on the Determining Factors of the Company Welfare in
Korea (1985-1994).” unpublished Ph.D. dissertation, Yonsei University.

Park, Fungoo. 1984. “Changes in the wage distribution in Korea in the 19703.” Journal of
Korea Development, Vol. 6, No. 2, 2-17.

Park, Joon-sik. 2002. “A study on structures and functions of corporate welfare in Korea”
(http://www.hallym.ac.kr/~jsp/36.html)

247

 

Park, Joonsung. 1995. Problems in managing Female Workers. Seoul: Korean Women’s
Development Institute. (http://www.sungshin.ac.kr/~parkcom/riba/fl 6.html)

Park, Se-Il. 1984. Wage Structures In Korea, Seoul: Korean Development Institute.

. 1993. “The role of the state in industrial relations: The case of Korea.”
Comparative Labor Law Journal, Vol. 14, No. 3, Spring, 321-338.

 

Park, Woo-Sung and Gyu-Chag Yu. 2000. “Transformation and New Patterns of HRM in
Korea.” Presented in the Conference titled of Transforming Korean Business and

Management Culture at Michigan State University, September 19-20.

Park, Young-Bum, 1993, Trends and Characteristics of Wage Structures in Korea in the
19803, Journal of Korean Labor, Vol. 4, No. 14, 27-45.

Park, Young-bum and Michael Byungnam Lee. 1995. “Economic development,
globalization, and practices in industrial rations and human resource management in
Korea.” In Anil Venna, Thomas A. Kochan and Russell D. Lansbury (ed.) Employment
Relations in the Growing Asian Economies. London: Loutledge.

Parsons, Donald 0. 1992. “The Intemal Distribution of Union Rents: An Empirical Test
of the Voting Power Model.” Review of Economics and Statistics, Vol. 74, Issue 3,
August, 439-445.

Pemberton, James. 1988. “A ‘managerial’ model of the trade union.” Economic Journal,
Vol. 98, September, 751-771.

Pencavel, John. 1991. Labor Markets Under Trade Unionism, Cambridge, Mass: Basil
Blackwell.

Perlman, Selig. 1928. A Theory of the Labor Movement, Reissued, New York: Augustus.

Petersen, Jeffrey S. 2000. “Health Care and Pension Benefits for Construction Workers:
The Role of Prevailing Wage Laws.” Industrial Relations, Vol. 39, No. 2, April, 246-264.

Pfeffer, Jeffrey. 1998. The Human Equation: Building Profits By Putting People First,
Boston, Mass.: Harvard Business School Press.

Pfeffer, Jeffrey and Jerry Ross. 1980. “Union-Nonunion Effects on Wage and Status
Attainment.” Industrial Relations, Vol. 19, No. 2, Spring, 140-151.

Pfeffer, Jeffrey and Alison Davis-Blake. 1992. “Salary dispersion, location in the salary

distribution, and turnover among college administrators,” Industrial and Labor Relations
Review, Vol. 45, No. 4, July, 753-763.

248

 

Piore, Michael J. and Charles F. Sable. 1984. The Second Industrial Divide: Possibilities
for Prosperity, New York: Basic Books.

Pizer, Steven D. 2000. “Does international competition undermine wage differentials and
increase wage inequality?” Journal of International Economics, Vol. 52, No. 2,
December, 259-282.

Podgursky, Michael. 1983. “Unions and Family Income Inequality.” Journal of Human
Resources, Vol. 18, Issue 4, Autumn, 574-591.

Porges, Jennifer L. 1991. “The Development of Korean Labor Law and the Impact of the
American System.” Comparative Labor Law Journal, Vol. 12, No. 3, Spring, 335-359.

Poterba, James M. and Kim S. Rueben. 1994. “The distribution of public sector wage
premia: New evidence using quantile regression methods,” Working Paper No. 4734,
Cambridge, MA: National Bureau of Economic Research.

Ramsay, Harvie. 1997. “Fool’s Good? European Work Councils and Workplace
Democracy.” Industrial Relations Journal, Vol. 28, No. 4, 314-322.

Renaud, Stephane. 1998. “Unions, Wages, and Total Compensation in Canada: An
Empirical Study, Relations Industrielles, Vol. 53, No. 4, 710-730.

Rodriguez-Gutierrez, Ce’sar. 2001. “Wage dispersion within firms and collective
bargaining in Spain.” Economic Letters, Vol. 72, 381-3 86.

Roger, J. and W. Streek (eds). 1995. Work Councils: Consultation, Representation and
Cooperation in Industrial Relations. Chicago: University of Chicago Press.

Rosen, Sherwin. 1969. “Trade Union Power, Threat Effects, and the Extent of
Organization.” Review of Economic Studies, Vol. 36, No. 2, April, 185-196.

Rosen, Sherwin. 1986. “The Theory of Equalizing Differences.” In Orley C. Ashenfelter
and Richard Layard (eds.), Handbook of Labor Economics, Vol. 1. Amsterdam: North-
Holland.

Ross, Arthur. 1956. Trade Union Policy, Berkeley: University of California Press.

Rowthom, R. E. 1992. “Centralization, Employment and Wage Dispersion.” Economic
Journal, May, Vol. 102, No. 412, May, 506-523.

Schiller, Bradley R. and Randall D. Weiss. 1980. “Pensions and Wages: A Test for

Equalizing Difference.” Review of Economics and Statistics, Vol. 62, Issue 4, November,
529-538.

249

Schultz, T. Paul and Gerrnano Mwabu. 1998. “Labor Unions and the Distribution of
Wages and Employment in South Africa,” Industrial and Labor Relations Review, Vol.
51, No. 4, July, 680-703.

Schumacher, Edward J. 1999. “What explains wage differences between union members
and covered nonmembers?” Southern Economic Journal, Vol. 65, Issue 3, January, 493-
512.

Seifert, Wolfgang. 1988. “Some Thoughts on the Problem of Intemal Union Democracy
in Japan.” Economic and Industrial Democracy, Vol. 9, No. 3, August, 373-395.

Shafer, Michael. “The Political Economy of Sectors and Sectoral Change: Korea Then
and Now” in Sylvia Maxfield and Ben R. Schneider, (ed.), 1997. Business and the State
in Developing Countries. Cornell University Press.

Shin, Eunjong. 2001. “Unions, government, and the politics of industrial relations in
Korea: Union bargaining power and labor control policy from democratization to post
IMF -intervention,” unpublished doctorial dissertation, Michigan State University.

Shirai, Taishiro. 1983. “A Theory of Enterprise Unionism.” In Taishiro Shirai (ed.)
Contemporary Industrial Relations in Japan. WI: University of Wisconsin Press. 117-
143.

Smith, Robert S. and Ronald G. Ehrenberg. 1983. “Estimating Wage-Fringe Trade-Offs:
Some Data Problems.” In J .E. Triplett (ed.), Measurement of Labor Costs, Chicago, IL:
University of Chicago Press, 347-369.

Song, Ho-Keun. 1991. Labor Politics and Market in Korea, Seoul: Nanam.
. 1994. “Economic Effects of Trade Unions: Do Trade Unions Contribute to

enhance Working Conditions?” Open Market, Closed Politics: Democratization and
Labor Regime in South Korea, Nanam Publishing House: Seoul, 297-318.

 

. 1995. A Study on Corporate Welfare, Seoul: Korean Labor Institute.

 

Song, Jangjoon. 2002. “Social and Economic Problems in Employment Permission
Policy.” National Economy, Vol. 116, July, Seoul: KDI Economic Information Center.

STATA. 1999. Reference Manual (Stata release 6.0). Stata Corporation, University
Drive, East College Station, Texas.

Suh, Jae Jean. 1991. The Capitalist-Class in Korea, Seoul: Nanam.
Swidinsky, R. and M. Kupferschmidt. 1991. “Longitudinal Estimates of the Union

Effects on Wages, Wage Dispersion and Pension Fringe Benefits.” Relations
Industrielles, Vol. 46, No. 4, 819-837.

250

 

Tilly, Charles. 1985. “War Making and State-making as Organized Crime.” In Peter
Evans, Dietrich Rueschemeyer and Theda Skocpol (ed.) Bringing the State Back In,
Cambridge University Press.

Traxler, Franz. 1994. “Austria: Still the country of corporatism?” in Hyman, Richard and
Anthony Fermer. (ed.) New Frontiers in European Industrial Relations. London:
Blackwell Publishers. 270-297.

Triplett, J.E. (ed.). 1983. Measurement of Labor Costs, Chicago, IL: University of
Chicago Press.

Uh, Soo-Bong and Tae-Hyun Lee. 1992. “Trade Unions’ Wage-Equalizing Effects:
Focused on the Assembled Metal Industry.” Korea Labor Study, Vol. 3, 27-76. (in
Korean)

United Nations (UN). 1999. Economic and Social Survey of Asia and Pacific Asia (Asia
and the Pacific into the T wenty-first Century: Information Technology, Globalization,
Economic Security and Development), New York.

Van de Meer, Marc. 1996. “Aspiring Corporatism? Industrial Relations in Spain” in Van
Ruysseveldt, J. and Visser, J. (ed.). Industrial Relations In Europe; Traditions and
Transitions. London: Sage Publication. 310-336.

Wachter, Michael L. and William H. Carter. 1989. “Norm Shifts in Union Wages: Will
1989 be a Replay of 1969?” Brookings Papers on Economic Activity, Vol. 1989, Issue 2,
233-264.

Wade, Robert. 1992. “East Asia’s Economic Success: Conflicing Perspectives, partial
Insights, Shaky Evidence.” World Politics, Vol.44, No. 2, January, 270-320.

Wallerstein, Machael. 1985. Working-Class Solidarity and Rational Behavior,
unpublished Ph.D. dissertation, University of Chicago.

Walton, Richard E. and Robert B. McKersie. 1965. A Behavioral Theory of Labor
Negotiations: An Analysis of a Social Interaction System, Itaca, New York: ILR Press.

Webb, Sidney and Beatrice Webb. 1894. The History of Trade Unionism, London:
Longmans.

. 1897. The Industrial Democracy, London: Longmans.

 

White, Michael D. 1982. “The Intra-Unit Wage Structure and Unions: A Median Voter
Model.” Industrial and Labor Relations Review, Vol. 35, Issue 4, July, 565-577.

251

World Bank. 1993. The East Asian Miracle: Economic Growth and Public Policy, New
York: Oxford University Press.

. 1999. Global Economic Prospects and the Developing Countries: Beyond
the Financial Crisis, 1 998/1999, Washington, DC.

 

Woodbury, Stephen A. 1983. “Substitution between Wage and Nonwage Benefits.”
American Economic Review, Vol. 73. Issue 1, March, 166-182.

Wooldridge, Jeffrey M. 2000. Introductory Econometrics: A Modern Approach.
Cincinnati, OH: South-Westem College.

Wunnava, Phanindra V. and Bradley T Ewing. “Union-nonunion differentials and
establishment size: Evidence from the NLSY.” Journal of Labor Research, Spring 1999,
Vol. 20, Issue 2, 177-183.

Yoo, Jang-Soo. 1987. “A Study on Wage Differentials in terms of Skills in Korea:
Focused on Male Manufacturing workers.” Journal of Economics, V0. 35, No. 1, 83-102.

(in Korean)

Yoo, Gyeongjoon. 1999. An Analysis and Decomposition of Wage Inequality in Korea.
Seoul: Korea Development Institute.

252