t. l.
(n... V): . .D:
..I ‘zx.£ly}.a.

3. .
.5. .. .
Ways: aft...

Erna; ,1

h:
.2 xl’ 5...: ..
. .155...

usu‘nnvl.

.. :.

£5.59“. «32
.41»: .uvulwa Ersnwr.
di.nu¢..€ﬂl.§ (-

4’41. ‘
!(|

.12
51.9.1:
:I. .5). 3....“
Ezaliii.
5.5

 

 

 

 

«#wa rt,

”L mm. _ ‘ “mm...”mmgﬁﬁgpeﬁﬁwnémawﬁ _

 

a... : :31
.1 .

“WEB

4001’

This is to certify that the

dissertation entitled

AN ANALYSIS OF ELECTORAL PARTISAN FRAGMENTATION
IN SINGLE-MEMBER SIMPLE PLURAILTY RARE'IAMENTARY
ELECTIONS

presented by

Myoung-Ho Park

has been accepted towards fulﬁllment
of the requirements for

 

Eh. D. degree in Political Science

W

Major professor

Date Ofﬂééz

MSU is an Afﬁrmative Action/Equal Opportunity Institution 0-12771

 

 

LIBRARY

Michigan State

University

 

 

PMCE 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

 

AFR 282303

 

m i. i 2.003

 

 

 

 

 

 

 

 

 

 

 

 

6/01 c:/CIRC/DateDua.p65-p.15

 

AN ANALYSIS OF ELECTORAL PARTISAN FRAGMENTATION IN SINGLE-
MEMBER SIMPLE PLURAILTY PARLIAMENTARY ELECTIONS:
A MACRO AND MICRO ANALYSIS OF CANADA AND KOREA
BY

Myoung-Ho Park

A DISSERTATION

Submitted to
Michigan State University
in partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Political Science

2002

ABSTRACT
AN ANALYSIS OF ELECTORAL PARTISAN FRAGMENTATION IN SINGLE-
MEMBER SIMPLE PLURAILTY PARLIAMENTARY ELECTIONS:
A MACRO AND MICRO ANALYSIS OF CANADA AND KOREA
By
Myoung-Ho Park
The current understanding of party system development within the tradition of the

diversity of political institutions has focused on data from the national level and if any
provincial level. However, the basic unit of electoral competition for legislative seats is
the local electoral district. This study is the ﬁrst attempt to analyze the degree of electoral
partisan fragmentation at the district level in Single-Member Simple Plurality electoral
systems under different institutional contexts (decentralized unitary presidential vs.
federalist parliamentary systems). Utilizing district-level data from national, provincial,
and local elections, it examines the sources of partisan fragmentation at the district and
national level in Canadian and Korean parliamentary elections. It then deepens this study
of the sources of political cleavages and partisan fragmentation though the utilization of
micro-level survey data. This study ﬁrst conﬁrms the necessity of sub-national and/or
district-level analysis of the level of electoral partisan fragmentation in legislative
elections, especially under the SMSP. Then, it argues that the local/district partisan
characteristics from local elections are crucial for understanding the degrees and patterns
of electoral partisan competition at the local district level. These points, furthermore, are
supported by the micro-level analysis that in both cases there exists a concentration of

electoral support at the individual level following regional lines.

TABLE OF CONTENTS

LIST OF TABLES .................................................................................... v
INTRODUCTION .................................................................................... 1
CHAPTER 1
THEORETICAL CONSIDERATIONS
Introduction ................................................................................. 1 0
Political Party and Party System Development ....................................... 11
Political Party and Its Origin .................................................... 11
The Development and Shape of Party System ................................ 13
Institutional Explanation on Party System Development ................... 17
Diversity of Political Institutions in Explaining Legislative
Electoral Multipartism ..................................................................... 20
Why Incorporate ‘Other’ Political Institutions ................................ 20
Presidentialism, Presidential Elections, and Electoral Multipartism
in Legislative Elections .......................................................... 21
Nature of Electoral Competition in Legislative Elections ........................... 25
Controversies over the Appropriate Level for Legislative Electoral
Multipartism Explanation ........................................................ 25
Unit of Electoral Partisan Competition in Legislative Elections .......... 26
Local Elections, Local/District Partisan Characteristics, and the Degree
of Electoral Competition at the District Level in Legislative lections ............... 29
Research Design ............................................................................ 34
Selection of Cases ................................................................. 34
Data and Measurement .......................................................... 37
CHAPTER 2
MACRO ANALYSIS OF THE KOREAN CASE
Introduction ............................................................................... 47
Degree of Electoral Multipartism at National and Sub-National Levels ......... 48
Determinants of District-Level Electoral Multipartism in
Legislative Elections ....................................................................... 55
Data and Method .................................................................. 55
Analysis and Discussion ......................................................... 62
CHAPTER 3
MACRO ANALYSIS OF THE CANADIAN CASE
Introduction ................................................................................. 76
Degree of Electoral Multipartism at National and Sub-National Levels ............ 77
Determinants of Riding-Level Electoral Multipartism in
Legislative Elections ....................................................................... 88

iii

Data and Method .................................................................. 88

Analysis and Discussion ......................................................... 93
CHAPTER 4
MICRO ANALYSIS OF THE KOREAN CASE
Introduction ................................................................................. 99
Data and Measurement .................................................................. 100
Analysis and Discussion ................................................................. 113
CHAPTER 5
MICRO ANALYSIS OF THE CANADIAN CASE
Introduction ................................................................................ 13 1
Data and Measurement .................................................................. 132
Analysis and Discussion ................................................................. 140
CHAPTER 6
CONCLUSION .................................................................................... 151
BIBLIOGRAHPY ................................................................................. 163

iv

LIST OF TABLES

Table 2-1 Comparison of the Degree of Electoral Multipartism ............................ 48
Table 2-2 Comparison of Electoral Multipartism at National and District Levels .......... 49

Table 2-3 Provincial Level Comparison of Electoral Multipartism in Recent Legislative
Elections ............................................................................................... 51

Table 2-4 Provincial Level Comparison of Electoral Multipartism in Recent Presidential
Elections .............................................................................................. 52

Table 2-5 Provincial Level Comparison of Electoral Multipartism in Recent Local

Elections .............................................................................................. 54
Table 2-6 Division of Seats in Korean National Assembly Elections ........................ 57
Table 2-7 Speciﬁcs on National Constituency Seats Allocation .............................. 57
Table 2-8 Elections in Korea Since Democratization ........................................... 58

Table 2-9 Local Partisan Characteristics and Legislative Electoral Multipartism Across
Districts ................................................................................................ 62

Table 2-10 District Partisan Characteristics and Legislative Electoral Multipartism
Across Districts ....................................................................................... 69

Table 2-11 District Level Presidential Partisan Strength and Legislative Electoral

Multipartism Across Districts ..................................................................... 73
Table 3—1 Division of Seats and Vote Shares at the National Level in Canada ............ 77
Table 3-2 Comparison of Electoral Multipartism at National and Riding Levels ......... 78

Table 3-3 Provincial Level Comparison of Electoral Multipartism in Canadian
Legislative Elections ................................................................................ 81

Table 3-4 Riding Level Comparison of Electoral Multipartism in Canadian Legislative

Elections .............................................................................................. 83
Table 3-5 Regional Bases of Canadian Parties (Seats) ........................................ 86
Table 3-6 Regional Bases of Canadian Parties (Vote Shares) ................................ 86

Table 3-7 Provincial Elections in Canada ....................................................... 89

Table 3-8 Construction of Riding Diversity Index ............................................. 92
Table 3-9 Local Partisan Characteristics and Legislative Electoral Multipartism Across
Ridings in the 2000 Canadian Elections .......................................................... 93
Table 3-10 Riding Partisan Characteristics and Legislative Electoral Multipartism Across
Districts in the 2000 Canadian Elections (Ontario Only) ..................................... 95
Table 3-11 Different Measures of Degree of Electoral

Competition .......................................................................................... 97
Table 4-1 2000 Legislative Election ............................................................. 101
Table 4-2 1996 Legislative Election ............................................................ 101
Table 4-3 1992 Legislative Election ............................................................. 101
Table 4-4 Categorization of Variable Across Elections in KES ............................ 103
Table 4-5 Determinants of Partisan Choice in 1992 Legislative Election ................. 114
Table 4-6 Multinominal Logistic Estimation of 1992 Election ............................. 117
Table 4-7 Determinants of Partisan Choice in 1996 Legislative Election ................. 121
Table 4-8 Multinominal Logistic Estimation of 1996 Election ............................. 124
Table 4-9 Determinants of Partisan Choice in 2000 Legislative Election ................. 126
Table 5-1 Comparison of Partisan Choices at Individual and Aggregate Levels ......... 131
Table 5-2 Comparison of Survey and Ofﬁcial Data in Comparative Perspective ........ 132
Table 5-3 Categorization of Variables in 2000 CES ......................................... 138
Table 5-4 Determinants of Partisan Choice (Two Major Parties) .......................... 140
Table 5-5 Determinants of Partisan Choice (Two Minor Parties) ......................... 141
Table 5-6 Multinominal Logistic Estimation of Partisan Choice ........................... 146

vi

INTRODUCTION

This study analyzes the degree of electoral competition at the district level in the
Single Member Simple Plurality (SMSP) electoral systems in the comparative
institutional contexts of Canadian and Korean parliamentary elections. Utilizing
legislative electoral district-level data from national, provincial, and local elections in
both countries, it ﬁrst examines the sources of electoral partisan fragmentation at district
vs. national levels. Next, we take a closer look at the source of political cleavage and
partisan fractionalization at the individual level through the utilization of micro-level
survey data.

Existing scholarship suggests that it is important to explain why some polities
have more fragmented electoral (or legislative) party systems than others and what
factors work to determine the level of electoral partisan fragmentation, in our cases, in
Canada and Korea. One of the most interesting and controversial questions in the ﬁeld of
comparative electoral systems is the extent to which, in the words of Siavelis,
“majority/plurality electoral formulae or small magnitude electoral systems have the
capacity to exercise a reductive effect on the number of parties and whether such systems
encourage centripetal drives within party systems” (Siavelis 1997: 652).

The influence exercised by different electoral systems on the degree of electoral

partisan competition in legislative elections become a topic of great debates in the ﬁeld1

 

‘ However, the institutionalist approach does not hold a monopoly of inﬂuence in determining the degree of
electoral partisan fractionalization. Some scholars argue that social cleavage is the primary force that
structures the development of party systems (Rokkan 1970). Furthermore, an interactive approach emerges
in which the inﬂuence of electoral systems is conditioned on the content of social cleavages in a polity
(Jones 2001; Neto and Cox 1997; Ordeshook and Shvetsova 1994; Taagepara 2000). Discussion of the
competing approaches will be detailed in the next chapter of theoretical consideration.

(Duverger 1954; Jones 1995, 1997, 1999; Rae 1971; Sartori 1976; Riker 1982; Lijphart
1984, 1994; Taagepara and Shugart 1989). One line of research in the “institutionalist”
approach to party system development takes into account the variety of political
institutions in the broader context of a political system, in order to understand the level of
electoral partisan fragmentation in legislative elections.

The basic premise behind incorporating the diversity of political institutions of a
polity into a study is to explain the degree of legislative electoral multipartism and
interdependence of political institutions within a given country. In other words, national
legislative elections and legislative institutions do not exist in institutional isolation
within their political systems (Shugart and Carey 1992).

Thus, much research has emerged on the interrelationship between different types
of national/sub—national and legislative elections in determining the patterns of electoral
partisan fragmentation with respect to legislative seats. The degree of electoral partisan
competition for national (as with the presidency) and/or sub-national (as with the
govemorship) political ofﬁces appears to have an important effect on the level of
electoral multipartism in legislative elections. Speciﬁcally, while most studies focus on
the importance of presidential elections (Coppedge 2000; Cox 1997; Jones 1995; Lijphart
1994; Mainwaring and Shugart 1997; Shugart 1995; Shugart and Carey 1992), a small
number of researchers have emphasized the importance of sub-national, local elections
(Jones 1997; Samuels 2000 a/b).

The degree of electoral competition emanating from electoral systems in
presidential and gubernatorial elections and their relative concurrence in legislative

elections, have been found to have a signiﬁcant inﬂuence on the level of electoral

multipartism in legislative elections. The inﬂuence of presidential elections on legislative
electoral multipartism has been analyzed in most cases at the ‘national’ level with respect
to electoral competition for the presidency. In contrast, those studies emphasizing
gubernatorial elections have focused on sub — national (provincial) electoral competition
for govemorship and legislative seats, especially under the “party list PR” electoral
system found in federalist contexts.

However, legislative elections in most cases (especially under the single member
simple plurality electoral system) are basically held at the local electoral district level
across a country. This becomes especially evident when legislative members are elected
under SMSP electoral systems. If there is any inﬂuence of electoral competition patterns
in important national and sub-national local elections on the degree of electoral
multipartism in legislative elections within a broader institutional context, then it is
important to focus on the legislative electoral district in which legislative elections are
held. Since electoral partisan competitions for legislative seats takes place at local
legislative electoral districts, the appropriate unit (or level) of analysis for party system
developments in national legislative elections must be ‘local’ districts (Chhibber and
Kollman 1998; Cox 1997; Gaines 1997, 1999; Jones 1997, 2001).

Most studies on the effects of presidential elections on legislative electoral
multipartism have been conducted at the national level. Furthermore, most research on
the interrelationships between different types of elections is implicitly based on an
assumption that a national party system more or less represents the pattern of electoral
partisan competition for national legislative seats across all local electoral districts. In

other words, explanations of the relationship between electoral systems and structures of

social cleavage in the shape of party system development have mainly focused on the
party system at the ‘national’ level as the dependent variable.

However, given that legislative elections are held in local districts across a
country, electoral partisan competition patterns appear to be different from district to
district, contributing in different ways to the formation of a national-level party system.
Thus, in the words of Jones: “Any analysis of legislative elections must either be
conducted at the district level or else make several assumptions regarding the distribution
of the vote and inﬂuence of electoral laws in aggregate at the national level” (Jones 2001:
4). Our understanding, therefore, of the political consequences of electoral systems (and
social heterogeneity as well) in national legislative elections can proﬁt from analyses at
the electoral district level. As Duverger contended, “The true effects of the simple
majority system [in legislative elections] is limited to local bipartism” (p. 28). In short,
the mechanism at work in legislative elections, given the broad diversity of political
institutions appears to be driven by events at the local electoral district level (Taagepara
2000)

All of this suggests that any attempt to understand the degree of electoral partisan
fragmentation in national legislative elections requires a district-level (or at least sub-
national local level) analysis. Since the basic unit for electoral partisan competition for
national legislative seats is the local electoral district, the appropriate level of research on
the interrelationship between national/sub—national elections and legislative elections
should be at the local district. Previous studies have been concerned with the impact of

national and/or sub-national elections on legislative electoral multipartism and the level

of electoral competition at the national level (presidential elections) and/or at the
provincial level (gubernatorial elections), not that of the legislative electoral district.

Incorporating presidential elections in our explanation of the degree of electoral
partisan fragmentation in legislative elections involves focusing on the effects of patterns
of ‘national-level’ electoral competition for the presidency on legislative electoral party
system developments across a country. In contrast, taking into account sub-national
elections put an emphasis on the inﬂuence of patterns of ‘local-level’ electoral
competition for the govemorship on legislative electoral partisan fragmentation across
districts. Electoral competition for legislative seats in a given local legislative electoral
district is subject not only to the usual institutional or national-level inﬂuences at work,
but also by the local forces like electoral competition patterns in races for prominent sub-
national, provincial political ofﬁces (Kernell and Jacobson 1981).

Presidential elections are basically ‘national—level’ electoral partisan competitions
representing ‘national-level’ electoral partisan strength and fragmentation. However,
local elections are ‘local-level’ electoral partisan competitions more or less representative
of ‘local — level’ partisan strength and fractionalization, presumably different from that of
the national level, across localities. Different types of sub-national local party systems
may coexist, given a certain form of institutional structure (federalism or decentralized
unitary system) and social/political diversity following regional lines across a country.

Electoral competition patterns at the local or district level, therefore, appears
relevant for explaining district-level electoral partisan fragmentation for legislative seats
across a locality. local/district partisan characteristics’, based on the degree of electoral

competition in prominent local elections in a given district, help to shed light on district-

level electoral multipartism in legislative elections in which basic electoral competition
for legislative seats emerges (especially under the SMSP electoral system).

Using the local district in legislative elections as the basic unit of electoral
competition for legislative seats, the degree of electoral partisan competition across a
locality provides different incentives or motives for voters and candidates to associate
with any given party for electoral purposes across a locality. Among these incentives or
motivations are the local strength of a party and the subsequent degree of electoral
partisan fragmentation at the local level that is relevant for voters and politicians for
deciding which party to associate with in a given locality. Thus, sub-national elections
are more reﬂective of local-level (or district —- level) partisan strength and subsequent
electoral partisan fragmentation. Also, local (or district)-level electoral competition
patterns in local (as well as presidential) elections appears most adequate for
understanding the degree of electoral multipartism across local district levels in
legislative elections.

Despite of the possibility that local elections are highly reﬂective of local partisan
characteristics in inﬂuencing the degree of electoral competition across districts in
legislative elections, there have existed to date only ‘two’ studies of the phenomenon
(Jones 1997; Samuels 2000a/b). Jones and Samuels both suggest that federal systems
allow for sub-national political and social diversity and the presence of political power at
multiple levels across a country. Then, institutional structures and political diversity
inﬂuence the level of electoral partisan fragmentation in legislative elections within sub-
national units. Jones and Samuels speciﬁcally argue that electoral competition patterns in

gubernatorial elections are inﬂuential in determining the degree of electoral partisan

fragmentation in a given provincial-level district in legislative elections under the party
list proportional representation electoral system for legislative elections.

However, there is a void in the literature that should be ﬁlled to achieve a more
systematic understanding of electoral multipartism across districts in legislative elections.
There have been few attempts to extend to a district-level analysis of legislative elections
under the single member simple plurality rule electoral system in a ‘decentralized unitary
presidential democracy’ (Korea)’ and in a ‘federalist parliamentary context’ (Canada).
Current district-level studies of electoral partisan fragmentation with local elections focus
on federalist presidential institutions such as in Argentina and Brazil,as well as the party
list PR electoral system for legislative elections. Therefore, this study is the ﬁrst
comparative attempt to analyze the degree of electoral competition at the district level in
legislative elections under the SMSP electoral systems in different institutional contexts.
Our understanding of party system development in legislative elections can be greatly
strengthened by the district-level analyses of party system development for which
adequate district-level data are available.

This research, therefore, is structured as follows: Chapter One reviews the
existing scholarship on party system developments in national legislative elections. Given
the importance of understanding party system fractionalization in a polity, we focus ﬁrst
on the general theoretical development of the determinants of the degree of electoral
multipartism in legislative elections, especially current understandings of the
interrelationship between different types of elections at the national level. Next we
discuss what should be the appropriate unit (or level) of analysis for the degree of

electoral multipartism in legislative elections.

Legislative elections under the SMSP electoral system are basically “district-level
phenomena.” Thus, local and district-level partisan strength and fractionalization is
highly inﬂuential in determining the electoral multipartism in a given district in
legislative elections. If the local (or district)- level pattern of electoral competition is
relevant for explaining the degree of electoral partisan fragmentation in a given district in
legislative elections, then prominent sub-national elections appear to be the best indicator
of local (or district) partisan characteristics. Finally, Chapter One discusses research
design, cases, data, variable descriptions and research methodology for the macro
analysis chapters.

Chapters Two and Three represent macro analyses of the sources of electoral
partisan fragmentation at the district level using electoral data from different types of
sub-national elections while controlling for other district-level factors. The cases of
Canada and Korea represent quite different institutional contexts (a decentralized unitary
presidential and a federalist parliamentary democracy). However, both Canada and Korea
seem to produce a “patchwork” type of party system at the national level, indicating that
electoral competition patterns across localities seem to inﬂuence the degree of electoral
competition in legislative elections at the district (or ridingz) level.

The shape of electoral partisan fragmentation in a given locality is basically based
on the distribution of individual votes cast in elections (i.e., partisan preference). While
the previous two chapters focus on sources of partisan fragmentation at the district and
national levels, Chapters Four and Five represent searches for sources of partisan

fragmentation and political cleavage across a country using survey data from both

 

2 “Riding” is a Canadian term for legislative “constituency” or “district.”

Canada and Korea. Macro as well as micro-level analysis is called for in order to achieve
a balanced understanding of electoral competition patterns in legislative elections in
Canada and Korea.

Chapter Six is the concluding chapter, which recapitulates the major ﬁndings of
the macro- and micro-analyses of Canadian and Korean cases and discusses possible
future arenas for the development of a more systematic understanding of party system
developments in legislative elections, in particular, and comparative political institutions

in general.

CHAPTER ONE

THEORETICAL CONSIDERATIONS

Introduction

Chapter One puts the primary research question raised in the previous chapter in a
larger theoretical context by reviewing the theoretical development on the shape of party
system in legislative elections. First of all, this chapter discusses the existing literature on
what is a political party and how they are developed as a party system in a country. While
there have been two competing arguments on the development of party system:
Institutionalisit vs. Sociological views, a new approach recently emerges advocating the
interactive inﬂuence of electoral systems and underlying social cleavage structure on
party system developments. After admitting the importance of political institutions, this
chapter reviews the current understandings on the interrelationship between different
types of elections (mainly presidential elections) in explaining the party system
fragmentation in legislative elections.

The basic unit of electoral partisan competition for legislative seats is the local
legislative electoral district, especially under the SMSP electoral system. Thus, the
appropriate level of analysis for the electoral multipartism in legislative elections is the
local district across a country. The ‘local/district partisan characteristics’ representing
local (or district) — level electoral competition pattern appears important in understanding
the degree of electoral multipartism across districts level in legislative elections.
Furthermore, sub - national level electoral competitions for local political ofﬁces seem

more adequate in identifying the local/district partisan characteristics that are presumably

10

different from locality to locality under certain institutional contexts, and inﬂuence the
degree of electoral partisan fragmentation across districts in legislative elections.

After discussing the existing scholarships on the sub — national level analysis of
legislative elections with local elections to the legislative electoral multipartism, this
chapter ﬁnds a void in the current literature that should be ﬁlled out for our systematic
understanding of party system development at the district and national level in legislative
elections. Then, a brief discussion of cases, data and measurements will follow for

empirical analysis.

Political Party and Party System Development
Political Party and Its Origin

Political parties are political organizations that are able to participate in elections
as well as to choose do so (Sartori 1976; Mainwaring Scully 1995). Political party has
critical functions despite its declining importance in both established and emerging
democracies (Dalton et a1 1984; Nie et a1 1976; Wattenberg 1990). Political party is
intrinsic to every representative democracy and the study of its function and form is one
of the most dynamic areas within contemporary scholarships of political science.

However, scholars disagree about the origins of political parties. The
modernization schools have argued that political parties emerged when public policy
could no longer be made by a small number of cadres of political elites who were not
concerned with public opinion (LaPalombara and Weiner 1966). Accordingly, the advent
of mass suffrage that was combined with the complexity of social problems after the

Industrial Revolution ﬁnally required political party.

11

Other political scientists focus on the mass suffrage as a key factor in the
development of political parties. Political parties link society and the political world by
mediating the cleavages of a society. This line of argument leads to “the sociological
view of party system development.” According to Lipset and Rokkan (1967), the social
cleavage structure of the 1960’s in Western Europe is virtually identical to those of forty
years of later. More speciﬁcally, Epstein (1980) explicitly denies the institutional
inﬂuence on party system development, referring to party systems’ proximity to civil
service reforms and their relationship to social/cultural and geographical cleavages.
However, the development of party system outside of Western society undermines the
external validity and explanatory power of the argument.

Contrary to those traditional scholars that put an emphasis on the social cleavages
in the deve10pment political party and party systems, researchers in the tradition of the
rational choice suggest that political parties emerge as tools for rational politicians to gain
political ofﬁces and make public policy. According to this view, political parties are
created since politicians consider political party as useful for their electoral and political
successes (Aldrich 1994; Downs 1957; Schlensinger 1991). They argue that political
parties are to address collective action problems in election campaigns and in the
legislative body.

If political party emerges as a response to the social/political changes and to
mediate social conﬂicts, its primary function is to provide voters with a link to the
legislature and government. In contrast, political parties’ communication with the
electorate is to be subsumed under their role as political instruments for taking care of

collective action problem as long as political parties are tools for rational politicians. The

12

underlying assumption of political party and subsequent party system development for
this research is that the main role of political party is to help politicians win elections and

govern, and voters associate with a party for their political representation.

The Development and Shape of Party System

Whenever political party emerges, it cooperates as well as competes with one
another for gaining political ofﬁces or making public policies. According to Sartori
(1976: 44), the “System of interactions resulting from inter — party competition” mainly
constitutes a party system in a country. As the development of individual political parties
respond to the institutional features of electoral systems and social cleavages, so does the
party system.

Political scientists have developed two general measures for deﬁning party
systems. They are ‘the (effective) number of political parties’ and ‘their ideological
distance.’ The classiﬁcation of party systems has focused on the number of parties
(Duverger 1954). Then, scholars expanded the analysis to the number and ideological
distance between parties (Sartori 1976) and the number of parties and majority formation
(Rokkan 1968). As suggested, the (effective) number of parties has consistently played a
signiﬁcant role in understanding party system development.

The (effective) number of parties in a polity is consequential in determining who
attain political power to make public policy. According to Sartori (1976: 120), “It does
matter how many are the parties. For one thing, the number of parties immediately
indicates, albeit roughly, an important feature of the political system: the extent to which

political power is fragmented or non — fragmented, dispersed or concentrated.”

l3

In addition, the comparative studies of regime stability and governmental
performance have suggested that the nature and structure of political party systems do
matter. The higher the degree of partisan fractionalization, the more likely national
governments are to falter (Lijphart 1984, 1994). Furthermore, presidential democracies
tend to experience executive — legislative stalemate (Jones 1995; Linz 1994), if there is a
higher level of legislative partisan fragmentation. Also, structural economic reforms can
be abandoned since there are highly volatile party systems.

Furthermore, if our study of comparative electoral systems is concerned with the
degree of representativeness in representative democracies, it is “expected to be fairly
closely related to the number of parties” (Shugart and Carey 1992: 208). Depending on
the number of political parties in a polity, the extent and quality of representation and the
stability, and the effectiveness of political system do varies. Because of these reasons, it
is very important to understand why some countries have more fragmented electoral or
legislative party systems than others, and what factors work in determining the level of
partisan fragmentation in a polity. In a word, elections lie at the heart of representative
democracy. Since political parties organize electoral competition in most representative
democracies, a vital party system seems to be essential to democratic process (Katz 1997).

Whether a polity contains two or more parties depends on (1) the nature of the
limit on the introduction of new parties and (2) the shape of the distribution of voters.
The ﬁrst indicates how strong or permissive an electoral system the political system
adopts is (Institutionalist view), and the second implies how many social/political
cleavages there are in the polity (Sociological view). While the former approach focuses

on the primary inﬂuence of electoral institutions on the shape of party system, the latter

l4

approach puts an emphasis on the importance of underlying social cleavages in a country
to determine the number of parties. Scholars such as Duverger (1954), Rae (1971),
Sartori (1976), Riker (1982), Lijphart (1984; 1994), Taagepara and Shugart (1989),
Myerson and Weber (1993) and Cox (1997) emphasize the importance of institutions
(electoral systems). In contrast, some researchers like Grumm (1958), Lipset and Rokan
(1967) and F ukui (1988) stress historical and social factors.

Proponents of the sociological approach argue that social cleavages better explain
the development of party system in a country. According to this view, the number of
social/political cleavages in a society ultimately determines the shape of the party system,
number of relevant political parties. According to Evans and Whiteﬁeld (1993) and
Randall (1995), political parties need deep social roots and have to be internally united.
That is, salient social/politicall cleavages form political groups like political parties.

In addition to the challenges to the institutionalist arguments, some scholars assert
that the direction of causal relationship between electoral systems and party system
development is not correct. Electoral systems do not determine the party system; rather,
the existing party system determines the electoral systems. Recent analyses indicate that
existing institutional features structure institutional choices, including electoral systems
(Bawn 1993; Remington and Smith 1996; Fish 1998).

Both electoral systems and social cleavage structures are essential in determining

how many political parties a given democracy will contain when it reaches equilibrium.

 

' Coppedge (1997) uses “political cleavage” rather than “social cleavage,” since “political cleavages are
more speciﬁc than social cleavages and more general than party identiﬁcation. ...... Political cleavages are
the divisions between general political orientations that are derived, in part, from social characteristics but
are mediated by culturally deﬁned identities and perceptions of conﬂict and shared interest. The most
familiar examples of the blocs deﬁned by political cleavages are positions on the left — right dimensions;
however, blocs based on religious, ethnic, or gender identities, or perhaps even personalistic loyalties, are
also possible” (pp. 169-174).

15

Recent research shows that both institutional and social factors affect the development of
party system. Duverger himself indicates that electoral systems “Act as an accelerator or
a brake” (1986: 73), putting an emphasis on national and social factors. It underscores the
inﬂuence of institutional variables as well as sociological factors in the party system
development. The interactive approach suggests that the inﬂuence of electoral systems on
the degree of multipartism is conditional on the extent of social cleavages in a polity2
(Cox 1997; Jones 1997, 2001; Lijphart 1994; Neto and Cox 1997; Ordeshook and
Shvetsova 1994; Taagepara 2000).

Among the approaches on the development of party systems, this research admits
the importance of political institutions in determining the degree of electoral multipartism
in legislative elections. The primary assumption of sociological view on party system
development is that the social cleavages salient in a country will lead to the formation of
political groups — political parties. However, collective action problems may undermine
or retard the translation of social cleavages into political parties. Additionally, it is not
clear what makes a social cleavage salient in a polity. While some social cleavages are
strong enough to help develop parties, others may not be as important. Furthermore,
focusing on institutionalist explanations on party system development does not deny the

important role of social cleavages in the political process (Lijphart 1994).

 

2 They ﬁnd that both electoral systems and social heterogeneity have an effect. On whether the effects of
heterogeneity and electoral systems additive or multiplicative, Powell (1982) used an addition of two
factors, but Ordeshook and Shvetsova (1994), and Amorim Neto and Cox (1997) indicate that a
multiplicative interaction superior. Speciﬁcally, “Low heterogeneity puts a lid on the number of parties
even in the presence of very permissive electoral systems, because there will be no demand for many
parties. Similarly, a low magnitude (single — member districts) tends to put a lid on the number of parties
even in the presence of strong heterogeneity, because few parties can gain representation. This mutual
limitation can be obtained by multiplying the two effects, but not by adding them” (Taagepara 2000: pp.
534-535).

16

Institutional Explanation on Party System Development

Political scientists have sought universal and general theories that explain
individual and group behaviors that are involved in political processes. While required to
develop a research program encompassing and improving upon previous researches,
scholarly works sometimes pursue independent research agendas, and they do not directly
contribute to the development of general and universal theories in political science.
According to Geddes (1991), this kind of tendency in political science research is
particularly evident in the ﬁled of comparative politics, in which the variety of researches
has been historically tied to a researcher’s ideological preference, and theories are
transitory.

However, there exists at least one exemption to the tradition. Riker (1982)
suggests that the study of political institutions and their impact on political party system
development follow a “Lakatosian” path of scientiﬁc research program. Since Duverger’s
theorization of the interrelationship between electoral institutional systems and the degree
of partisan fragmentation, comparative study of electoral and party systems have moved
forward by improving and incorporating previous research for our general understandings.
Speciﬁcally, scholars have investigated how Duverger’s law and hypothesis works in
explaining party system development (Fey 1997; Fedderson et all990; Palfrey 1989;
Taagepara and Shugart 1989). Researchers have also applied the theory of electoral
institutions and party systems development to the variety of countries (Chhibber and
Kollman 1998; Cox 1997; Jones 1995, 1997, 1999; Lijphart 1994; Neto and Cox 1997;

Ordeshook and Shvestova 1994).

17

Political institutions are the array of formal and informal institutional
arrangements that deﬁne the rules of the game in politics and establish how different
elements of the state relate to each other and how individuals attain positions of authority
(North 1990). Political institutions determine who are in a position of policy decision —
making, what views are represented at the highest levels of government and what kinds
of mechanisms are at work to resolve social conﬂicts in a society. Political institutions
have a “genetic code that programs the kinds of bureaucracies, leadership structures,
personal arrangements and the other properties that will ultimately make up its
government” (Moe and Caldwell 1994: 192). Therefore, decisions to select or alter the
existing institutions are not taken lightly and the developments of new political
institutions are based on the desires of competing constituencies (Bawn 1993; Remington
and Smith 1996). .

Among the political institutions, electoral institutions3 have been considered as
the most important in determining the shape of party system in a polity. Electoral
institutions play a signiﬁcant role in structuring partisan electoral competition for
political oﬁices. They directly affect the strategic behavior of political elite and voters by
establishing the rules for political contests. As shown in Duverger’s theorization of the
relationship between electoral systems and party system development, institutional
variables become the primary determinants of the shape of party system. While
sociological and historical factors can also inﬂuence how electoral partisan competition
develops in a country, electoral institutions are primarily determinants of the party system

development.

 

3 It is also referred as “electoral rules,” “electoral laws” or “electoral systems.”

18

It was Duverger that ‘ﬁrst4’ formalized the direct link between electoral
institutions and the number of parties in a polity. Duverger’s Law and Hypothesis assert
that a plurality ballot system tends to produce a two — party system at the district level
and proportional representation leads to multipartism. Speciﬁcally, Duverger’s Law
states that, “[T]he simple —- majority single — ballot system... favors the two — party
system” (Duverger 1954: 217). And Duverger’s Hypothesis indicates that, “[T]he simple
majority system with second ballot and proportional representation favors multipartyism”
(Duverger 1954: 239).

According to Shugart and Carey (1992), “When Riker (1982b) wanted to
demonstrate that political science has a history of cumulation of knowledge like the
“hard science, he returned to the set of reformulations know as Duverger’s law ...... It is
Duverger (1951, 1954) who normally gets the credit for having made questions about the
inﬂuence of electoral rules on party systems a major concern for political science” (p.
206). In short, “Duverger’s law and hypothesis become one of the most famous and cited
generalizations in the ﬁeld of comparative politics” (Gaines 1999: 835). The main
message of Duverger’s theory as an institutionalist understanding of electoral and party
systems is that political institutions do restrict individual voters’ political choice and thus
inﬂuence the structure of party system in a country (Taagepara 2000).

The institutional arguments on the shape of party system in legislative elections

stand up to empirical scrutiny. Scholars ﬁnd that such institutional variables as electoral

 

" Duverger was ‘the ﬁrst’ in the sense that scholars before him based their arguments on logic rather than
empirical evidence. John Stuart Mill suggested that the adoption of proportional representation in the Great
Britain would undermine the two — party system. Henry Droop also posited that two - party electoral
competition would result from the use of a plurality rule for seat allocation in single member district races
(Cox 1997; Riker 1982)

19

formula (majoritarian vs. proportional), district magnitude (effective threshold), assembly
size, apparentment and presidentialism explain most of the variances in party system
fragmentation and proportionality between votes and seats (Jones 1995, 1997, 1999;
Lijphart 1994; Mainwaring and Shugart 1997; Taagepara and Shugart 1989). Among the
factors, district magnitude and electoral formula become evident in their inﬂuence on
party system development. However, the variable of district magnitude has a very modest
impact on average on the level of multipartism in legislative elections especially in the

context of presidential democracies (Jones 1995; Mainwaring and Shugart 1997).

Diversity of Political Institutions in Explaining Legislative Electoral Multipartism
Why Incorporating ‘Other’ Political Institutions

Shugart and Carey (1992) explains why it is necessary to focus on other political
institutions like presidential elections and federalist structure of a country in order to
understand the shape of political party systems in national legislative elections. They
argue, “The usual understanding of Duverger’s rule has suffered from a narrowness of
scope in the sense that studies of electoral systems have tended to assume implicitly that
the only institutions that mattered were electoral rules for [national] assemblies, taken in
isolation from other institutions. Such isolation of assembly from the process of
[national] executive formation may distort our understanding of the number of parties
even in parliamentary systems: it especially is certain to do so in the case of presidential
democracies” (p. 207).

National legislative institutions do not exist in institutional isolation within their

political systems. Speciﬁcally, national legislative members are elected either to form a

20

national executive (in the context of parliamentary system), or to co — exist with a
national executive (in the context of presidential system). Therefore, it appears necessary
to focus on other political institutions such as “presidentialism” and “federalism” to
explain the development of electoral party systems in legislative elections in the broader
context of a political system.

Presidentialism, Presidential Elections and Electoral M ultipartism in Legislative
Elections

Scholars shed light on the presidential electoral systems among the political
institutions that possibly inﬂuence the degree of electoral multimpartism in legislative
elections. For example, Lijphart (1994) introduced the variable of regime type —
presidentialism5 vs. parliamentarism to identify its impact on the degree of partisan
fragmentation in legislative elections. According to him, to be inﬂuential for presidential
elections to electoral multipartism in legislative elections, the presidency must be relevant
political actor.

Discussions of the impact of presidential elections on the legislative electoral
party system fragmentation have been based on the “coattail effects” of presidential
candidates. The “coattail effects” in the context of American politics is deﬁned as “the
ability of a candidate at the top of the ticket to carry into ofﬁce... his party’s candidates
on the same ticket” (Beck 1997: 251). It becomes more evident when it comes to

presidential democracy.6

 

’ Sartori (1994) stipulates that a political system is a “presidential” if “the head of the state (president): (1)
results from popular election, (2) during the pre — established tenure cannot be discharged by a
arliamentary vote, and (3) heads or otherwise directs the governments that he or she appoints” (p. 84).
According to Shugart and Carey (1992), most studies of the interrelationship between vote shares and seat
allocations in legislative elections have not taken into account the feasibility that presidential and
parliamentary democracies must be differ on the relationships that have been investigated.

21

While the legislative elections in the context of parliamentary democracy are to
form a national executive ofﬁce, the congressional elections in presidential democracy
have nothing to do with national executive formation. This fundamental difference in the
ways of constituting a national executive ofﬁce can be expected to inﬂuence the degree
of partisan fragmentation in legislative elections. Therefore, The nature of electoral
competition in the presidential elections has enough potential to inﬂuence the number of
relevant political parties in the legislative body in ‘pure’ presidential systems. The greater
the effective numbers of presidential candidates (parties), the greater the effective
numbers of legislative parties (Cox 1997; Jones 1995; Mainwaring and Shugart 1997). In
turn, the effective number of legislative parties has a strong effect on the level of
presidential support in the legislature. The larger the effective number of parties, the
smaller the level of partisan support7 (Jones 1995, 1999; Linz 1994).

Also, academics generally assume that coattail effects emanate from the more
important elections (president) to the less important ofﬁces (legislature and local ofﬁces),
connecting the electoral fortunes of legislative candidates to their parties’ presidential
candidates (Jones 1999). That is, “Because presidential candidates typically obtain the

lion’s share of campaign ﬁnance and of national media attention and because the national

 

7 Furthermore, Jones (1995) focuses on the interactions between legislative party system fragmentation and
presidential power: he impacts of (l) the rules for executive election in non — established democracies, and
(2) the interplay between executive type and legislative power on the fragmentation of legislative party
system. Then, he prescribes suggestions for the best way of constructing a presidential arrangement. They
are plurality rule for presidential election, unicameral legislative system, PR in moderate sized districts, and
concurrency of legislative and presidential elections. This will more likely to provide a president with a
legislative (or near) majority. While small parties will gain representation, the incentives for party system
fragmentation are constrained. Also, as for the relationship between presidential electoral formula and
partisan fragmentation in legislative elections, he also indicates that “The presence of a majority system
results in a level of multipartism which is 1.331 times the level of multipartism of a plurality system
(holding other factors constant)” (p. 149). Moreover, the non — concurrent runoff combination leads to the
“highest levels of multipartism,” compared to “high levels of multipartism” for non — concurrent plurality
elections (p. 159).

22

party typically organizes presidential nominations, candidates for legislative ofﬁce may
seek organizational and ﬁnancial support from the national party and/or its candidate.
The relative importance of the presidential campaign thus creates strong incentives for
congressional candidates to line up behind the presidential candidate of their party during
the campaign” (Samuels 2000b: pp. 3-4).8

Therefore, how presidents are elected and when presidential elections are held in
relation to legislative elections become consequential in determining the level of partisan
fractionalization in legislative elections. Accordingly, there have emerged many studies
on the interrelationship between presidential and legislative elections (Coppedge 2000;
Cox 1997; Jones 1995, 1999; Lijphart 1994; Mainwaring and Shugart 1997; Shugart and
Carey 1992).

For example, Shugart and Carey (1992) emphasize the role of electoral rules for
president9 and national legislative members in presidential democracies. Speciﬁcally,
they demonstrate the ways in which presidents and legislative members are elected play a
crucial role in determining the shape of party system fragmentation, political agenda, and
the prospects for cooperation and conﬂict between presidents and national legislatures.

That is, Plurality presidential elections and concurrent elections with legislative elections

 

8 In other words, the causality would, if ever, primarily go from presidential elections to congressional or
other elections of lower proﬁle, and not the way around (Cox 1997). Accordingly, the synchronization or
proximity of presidential and congressional elections has little effects on presidential party systems. The
impact of concurrent elections of presidents under plurality with legislative elections is to reduce party
system fragmentation, more speciﬁcally, reduce the number of parties and increase the vote shares of the
largest ones. On the other hand, majority runoff for presidential election tends to have more parties in
legislative elections.

According to Coppedge (2000), the electoral formula for president refers to the “plurality election” in
which the ﬁrst — place candidate wins whether he wins a majority of the votes or not and the “runoff
formula” in which a second round is held between the top ﬁnishers if no one wins a majority in the ﬁrst
round. Other formula includes “electoral college, a runoff in congress, a 40% threshold, the double
simultaneous vote, the double complement rule, or concurrent majorities, which are less common than two
basic options.

23

“allow congressional candidates to enjoy two advantages simultaneously: identiﬁcation
with a national presidential candidate, and the freedom to pursue local particularism”
(Shugart and Carey 1992: 223), reducing the number of effective electoral parties in
legislative elections.10

While presidential electoral formula inﬂuences signiﬁcantly the level of electoral
and legislative party system fragmentation, it also affects the effective number of
presidential candidates itself. Jones (1999) shows the importance of presidential electoral
formula as well as the presence (or absence) of incumbent president in determining the
electoral competition patterns in presidential elections. In addition, the relative timing of
elections for presidents and legislative members, “electoral cycles,” is critical in affecting
the patterns of electoral partisan competition for legislative seats (Coppedge 2000;
Shugart 1995). The closer the presidential election is to the legislative election, the
greater is the coattails effect of the former on the latter, and therefore, the greater its
potential reductive inﬂuence on the degree of electoral partisan fractionalization.

In short, among the political institutions scholars consider to explain the degree of
electoral multipartism in legislative elections, “presidential electoral formula” and
“presidential - legislative election timing (concurrence)” are two main factors of

presidential democracies (Coppedge 2000). The above discussions suggest that the level

of electoral partisan competition in legislative elections can also be inﬂuenced by the

 

'0 On the contrary, the concurrence of relatively high district magnitude proportional representation
legislative elections encourages considerable electoral partisan fragmentation in legislative elections. While
plurality elections of the presidency tends to lead to the formation of two big blocs to contest in presidential
elections, majority run — off presidential elections are oﬁen considerably more fragmented in the effective
number of electoral presidential parties (or candidates).

24

electoral institutions for other political ofﬁces. Taking into consideration of the electoral
competition pattern for the presidency emanating from the speciﬁc presidential electoral
institutions in a broader institutional context of a country contribute to our understanding

of the level of electoral partisan fragmentation in legislative elections.

Nature of Electoral Competition in Legislative Elections
Controversies over the Appropriate Level for Legislative Electoral M ultipartism
Explanation

Political scientists have debated whether the inertia toward two — party electoral
competition in legislative elections under the single member simple plurality electoral
systems is at the national or district level. The main assumption that the institutional
explanations on the shape of party system is a national — level phenomena underlies the
discovery of empirical anomalies in which the SMSP electoral systems do not lead to a
two — party competition at the national level (Rae 1971; Riker 1982; Taagepara and
Shugart 1989).

Thus, those scholars imply that the party aggregation at the national level in
Canada and Indiall to more than two parties suggests that the Duverger’s
Law/Hypothesis malfunction in these cases. Furthermore, Duverger himself indicates that
the exceptions to the institutional arguments can be explained by the evaluation of the
ideological dispositions of political parties. There might exist multiple parties

representing slight variations of the two general ideological positions. Therefore, a kind

 

” Following the Lakkso — Taagepara Index for the effective number of electoral parties, the degree of
multipartism at the national level is “4.11” in the 1997 House of Commons elections for Canada and “7.1 l”
in the 1998 House of the People elections for India.

25

of “Bipolar multipartism” (Duverger 1986) emerges at the national level that
approximates a two party system.

However, the mechanisms at work in the institutional arguments take place at the
district level in which the electoral competition for legislative seats basically exists.
Politicians and candidates reckon the possibility of winning a single seat(s) and base their
strategic entry decisions upon the expected local (or district) performance. Thus, it is
logical to understand that the Duverger’s Law in particular and the institutional
explanations on the shape of party system in general are district -— level phenomena
(Chhibber and Kollman 1998; Leys 1959; Sartori 1986; Cox 1997; Jones 1997). The
relative multipartism at the national level is possibly compatible with the development of
a two — party electoral competition across local districts. The national — level party

system is an aggregation of sub — national (regional and district) party systems.

Unit of Electoral Partisan Competitions in Legislative Elections

As suggested, most current studies on the effects of presidential elections on
legislative electoral multipartism have been conducted at the ‘national level.’ The
electoral competition for the presidency in presidential elections exists at the national
level, indicating that a whole nation is one electoral district. Thus, its impact on the
degree of electoral multipartism in legislative elections has been studied at the national
level - the national level effective number of parties.

In addition, researches on the development of party system is implicitly based on
an assumption that a national party system more or less represents the pattern of electoral
partisan competition for national legislative seats across all local electoral districts within

a country (Chhibber and Kollman 1998; Cox 1997; Gaines 1999). The current

26

explanations seem to assume that there exists one representative pattern of electoral
competition across local districts. Therefore, where data available have been introduced,
they have almost been party system fragmentation at the national level, or if any,
provincial level (Gaines 1999).

According to Cox (1997), the main dependent variable in most of previous
research, “the effective number of electoral parties at the national level,” results from two
— steps processes. First of all, votes by each party (or candidate) at the local electoral
district levels are aggregated and subsequently, there emerges a certain degree of
electoral partisan fragmentation in each local constituency level. Then, those local
electoral district level party systems are aggregated into a ‘national — level’ degree of
electoral partisan fragmentation.

If it is the case, our current understanding of the impact of electoral systems on
party system development may fail to adequately explain the dependent variable, since
the main thrust of the institutional arguments — Duverger’s “mechanical” and
“psychological effects” — operate basically at the local electoral district level (Taagepara
2000). The mechanisms at work in the Duverger’s law appear to take place at the local
district level. That is, the “Duverger’s law and its theoretical descendants essentially deal
with district — level phenomena” (Chhibber and Kollman 1998: 330). Therefore, it is
logical to understand that the Duverger’s Law/Hypothesis is an ‘electoral district — level’
concept and phenomenon.

Legislative elections are held at the local electoral districts ‘in most cases’ within
a polity and the electoral partisan competition patterns appear to be different from district
to district and from province to province, contributing differently to the formation of a

national - level party system. For instance, studies on the US state — level legislative

27

elections have also suggested that the state — level measure of electoral competition
pattern is different from those based on district — level electoral returns in state legislative
elections (Holbrook and Van Dunk 1993; Van Dunk and Weber 1997).12

Therefore, the district — level analysis of electoral competition patterns is
conceptually and empirically distinct from the conventional studies of electoral
multipartism in legislative elections at the national level. Use of the district — level data
(like candidate and district — level conditions) will deﬁnitely add greater precision to our
analysis of electoral competition for national legislative seats across a polity. Such an
approach also makes it possible to identify electoral competition patterns in a manner
consistent with most congressional studies (Hogan 2001; Jacobson 2001).13

Gaines (1997) points out “Oddly, most authors appear to have overlooked a
simple, important point. When elections are fought over several districts, as in most
democracies, such formulae [for measuring the effective number of parties] can be
applied to vote shares at the aggregate level or at the constituency level.” (p. 49). His
argument indicates that the appropriate unit of analysis for the extent of electoral partisan
fractionalization in national legislative elections must be ‘the local electoral district.’ The

resulting ﬁgures for the effective number of parties at both national and local electoral

 

‘2 Furthermore, in trying to explain why midterm losses for the incumbent party in the US House elections
happen, Gaines and Nokken (2001) argue that virtually all the current explanations “seem to imply that
midterm losses are largely independent of members’ actions in ofﬁce. Thus, it has become common to
neglect the electoral consequences of individual - level member roll — call voting behavior when explaining
congressional election results” (p. 2). When one wants to explain why some legislators are re (elected),
while other are not, district - level analysis should be conducted.

'3 In addition, in their efforts to ﬁnd the electoral roots of “conditional party government” across time,
Carson, Jenkins and Schickler (2001) support for the use of constituency — based measures of partisan
preference for party system characteristics at the district level.

28

district levels in a nation can vary depending on the degree of inter — district variances in
vote divisions among parties (or candidates).

Therefore, it becomes clear that any attempt to understand legislative elections
must either be conducted at the district level or else make several assumptions regarding
the aggregate process of local party systems at the national level. Understanding the
political consequences of electoral systems (and social heterogeneity as well) in national
legislative elections can proﬁt from the district — level analyses (Wildavsky 1959).
Focusing on national — level party system fragmentation is to disregard how rich a picture
of partisan electoral competition across local electoral districts in a polity.l4

It becomes clear that the appropriate unit of analysis for the degree of electoral
multipartism in legislative elections is the local electoral district. Thus, a national — level
party system should be understood as an aggregation of sub — national
(provincial/electoral district) party systems. Nevertheless, most previous studies on
electoral systems have focused on a whole nation, rather than on the local electoral
district level. It seems the time to ﬁll “The gap between our electoral theories (mostly
district — level) and data (mostly national — level)” (Cox 1997: 12) with an emphasis on
the electoral district level analysis.

Local Elections, Local/District Partisan Characteristics and the Degree of Electoral
Competition at the District Level in Legislative Elections

Not all the studies taking into consideration of the diversity of political

institutions put an emphasis on the importance of national — level electoral competition

(presidential elections) on the degree of electoral multipartism in legislative elections.

 

'4 For example, Reed (2001) shows that a district — level analysis of the Italian case indicates that the
Duverger’s law works at the district level while it doesn’t work at the national level.

29

While much research has focused on the degree of national — level electoral competition
like presidential elections, only a few, if any, shed light on the effects of sub — national
level — electoral competition like local elections on the level of electoral partisan
fragmentation across districts in legislative elections (Jones 1997; Samuels 2000a/b).

As mentioned earlier, to understand the district — level electoral competition in
legislative elections can proﬁt from the consideration of local (or district) — level
electoral competition. In this sense, the degree of electoral partisan fragmentation at the
local level such important local elections as ‘gubernatorial, mayoral or local council
elections’ appear important to explain the legislative electoral multipartism across
districts in legislative elections, given the necessity of district — level analysis of
legislative elections. Why one should incorporate the electoral competition pattern at the
local (or district) level to understand the degree of electoral partisan fragmentation across
districts in legislative elections can be explained in terms of both institutional and
individual aspects.

The diversity of political institutions in a polity obscures the issue of
interrelationship between the pattern of partisan electoral competition for local ofﬁces
and legislative electoral party system fragmentation. It becomes evident when it comes to
a federalist country, in which there exist multiple electoral arenas with a variety of
electoral systems. Federalism may reﬂect “social diversity of the sort that can support
separate parties” and encourage “party elites to maintain smaller regional parties rather
than fuse with others” (Blais and Carty 1991: 85).15 Therefore, it seems reasonable to

infer that there emerges different types of party systems coexisting within sub — national

 

'5 See Gaines (1999) and Jones (1997).

30

units (Gaines 1999) and it is inﬂuential in determining the variety of electoral partisan
competition at the local (or district) level across a country for national legislature. ‘6

Furthermore, Dahl and Tufte (1973) assume that the size of political entity affects
party system fragmentation. Because a relatively bigger country is likely to adopt
federalist system, the possibility of increased party system fragmentation becomes higher
when it comes to a federalist country. Also, Anckar (2000), the size of a country in terms
of population or area, not only in terms of federalist or unitary system, signiﬁcantly
inﬂuences the degree of electoral multipartism in legislative election while controlling for
such traditional institutional and non — institutional variables as “effective threshold,”
“presidentialism,” and “social diversity.”

In a centralized unitary system in which political power is concentrated at the
national — level govemment, focusing on the national — level party system seems safe in
understanding the nature of party system of the polity under review. However, many
countries have multiple electoral arenas across a country with varied level of partisan
electoral competition for ofﬁces. However, “In a federal (or decentralized unitary) system,
political power exists at both the national and subnational levels. It is likely that these
multiple levels of [electoral partisan] competition inﬂuence each other” (Jones 1997:
540)

The logic behind the importance of sub — national elections in understanding the
degree of electoral partisan fragmentation at the district level in legislative elections is
that a certain type of institutional structure (i.e., federalism or decentralized unitary
system) tends to help social/political diversity following regional lines across a polity

emerge. Then, there may coexist different types of sub — national party systems. It seems

 

'6 In addition, Jones (1997) suggests that “federalism complicates the task of scholars studying the impact
of electoral rules on national party systems” (p. 53 8).

31

to inﬂuence the level of electoral competition across districts in legislative elections.
Therefore, it is the local elections to represent the local/district partisan characteristics
that directly affect the degree of electoral partisan fragmentation at the district level in
legislative elections.

On the other hand, the importance of local -— level electoral competition pattern
(especially that of local elections) for the district — level electoral multipartism in
legislative elections can be explained in terms of individual political actors, rank and ﬁle
of a party and individual voters. If sub — national elections for local political ofﬁces
inﬂuence local political party elites and individual voters, it is expected that the electoral
partisan competition should have signiﬁcant impact on the degree of electoral
multipartism in national legislative elections. The logic behind the importance of local
elections is that afﬁliating with strong ‘local’ politicians, political parties, or factions in a
given locality can bring substantial beneﬁts to legislative candidates running in that local
area.

If we assume that most politicians are motivated both by power seeking like
winning political ofﬁce or spoils of ofﬁce, and by policy goals, to associate with locally
strong parties or politicians will increases their chances of winning elections or getting
ofﬁces in elections that are held in the locality. For example, the great visibility of the
gubernatorial candidate in the race indicates that local voters are likely to focus on the
candidate’s own party as the main vehicle, by which to support their preferred national

legislative candidates. Thus, rational — ofﬁce (or political power) seeking politicians,

32

regardless of their policy goals, will be better off being on the gubernatorial candidates’
slate of proponents in legislative elections.l7

According to Shugart (1995), in the context of presidential democracy, legislative
elections that are held ‘not concurrently’ with presidential elections tend to have a feature
that encourages the locality — oriented voting notwithstanding that what is being elected a
national legislative body. This point suggests that legislative elections especially under
the SMSP electoral system in presidential democracy and not concurrent with
presidential elections are more susceptible to the local — level inﬂuences.

Political parties and party systems are formed largely on the basis of voters’ and
candidates’ incentives to coordinate on common party labels” (Chhibber and Kollman
1998: 329). Cox (1997) also indicates “Candidates value possession of [valuable party]
labels both because they convey a certain number of habitual voters into their camp and
because they publicly certify the candidates’ viability, thus insuring them against
strategic desertion” (p. 152). Therefore, given the institutional context to allow the
coexistence of different types of party systems across a country mentioned above, there
appear different individual incentives for rational politicians to associate with a particular
party depending on the local/district partisan characteristics across localities.

Existing studies help us understand the degree of electoral competition across
districts in legislative elections. However, there is a void in the literature that should be

ﬁlled for a more systematic explanation of electoral partisan fragmentation across local

 

'7 Furthermore, even a party whose chance of winning is very slim is a given local area may present a
candidate in the area in order to institutionalize itself in local voters’ minds as potential future national
legislative seat holders. This logic can be applied to independent candidates for local oﬂices. That is, an
independent candidate may run for a local ofﬁce for the purpose of making him or her known to local
voters so that he or she will have higher probability of winning political ofﬁce in future legislative
elections. They do so in part to present long - term alternative political force in the area, hoping that they
may win elections at some future.

33

districts in legislative elections. Current research focuses on the legislative elections
under the PR electoral systems at the provincial level. As suggested, the appropriate unit
of analysis for the party system development in legislative elections under the SMSP
electoral systems is the local districts. When the basic unit of competition in legislative
elections is the local districts. the local/district partisan characteristics appear to be
relevant to explain the degree of electoral partisan fragmentation across districts in
legislative elections. The sub — national elections among others are more reﬂective of the
local/district partisan characteristics as current studies show.18

We have accumulated important evidence in our understanding of electoral and
party systems that sub — national factors, provincial or local electoral district, affect
national —— level elections for presidency or legislative seats in such countries as Argentina
(Jones 1997), Brazil (Samuels 2000a/b), Canada (Gaines 1999; Shannan 1994), India
(Chhibber and Kollman 1998), Russia (Ordeshook 1996), Spain (Linz and Stepan 1992),
and the US (Chhibber and Kollman 1998). With the district - level analysis of legislative
elections under the SMSP electoral systems in a decentralized unitary presidential and a
federalist parliamentary context, this study would contribute to our understanding of the

source of political cleavages and partisan fragmentation.

Research Design: Cases, Data, and Measurement
Selection of Cases

Inherent in the ﬁeld of comparative politics is the necessity of choosing the

 

'8 While focusing on the importance of sub — national elections, it does not mean that the degree of
electoral competition in national — level elections (like presidential) does not inﬂuence that of legislative
elections. But the impact of presidential elections on legislative elections should be analyzed at the
legislative district level as suggested since what we want to know is the degree of electoral competition
across districts in legislative elections. The legislative district - level degree of electoral competition in
presidential elections will be supplementary used to show the necessity of district — level analysis in the
following chapter.

34

speciﬁc cases for research. It is rarely feasible to incorporate the entire universe of all the
possible cases. Therefore, comparative research is open to bias emanating from the
manner by which cases are selected for testing hypotheses (Geddes 1991). Nevertheless,
it is unavoidable to choose some cases for analysis.

The cases that are selected for this research are Korea and Canada. Both countries
use the SMSP electoral systems in which citizens elect one national legislative member
who gets the most votes cast in the district. It indicates that the basic electoral district for
national legislative seats is sub — provincial (or sub — metropolitan) local district. Thus,
this study is to extend the current understanding of interrelationship between local and
national legislative elections to the legislative elections under different electoral systems
(Party List PR vs. SMSP).

Furthermore, it is unrealistic to presume that the federalist institutional context
‘only’ promotes the sub — national political and social diversity, and allows the presence
of political powers at multiple levels within a country. Focusing on a national — level
pattern of electoral competition to explain the determinants of legislative electoral party
system fragmentation seems safe in a ‘centralized unitary system’ in which political
powers are centralized at the national level (Jones 1997). However, many countries have
multiple electoral arenas with varied level of electoral partisan competition for political
ofﬁces.

The more decentralized political or governmental power in a nation like Korea,

the less complete a research will be if it does not consider the degree of electoral

35

competition in direct elections of local political ofﬁces (governor or mayor)”. Given the
decentralized unitary institutional context and regionalist party politics of the Korean
case (Kang and Jaung 1999; Lee and Brunn 1996; Morris 1996), there might exist
different types of electoral competition along regional lines that inﬂuence the degree of
electoral competition across districts in legislative elections.

Extending the current scholarships on the interrelationship between sub — national
elections and national legislative elections, it is expected that the Korean case study
shows the importance of electoral competition patterns in prominent sub — national (and
sub — provincial) elections in explaining the degree of electoral multipartism across local
electoral districts in legislative elections. That is, the ‘local/district partisan
characteristics’ — different levels of partisan strength and subsequent partisan
ﬁagmentation in a given district — inﬂuence the degree of electoral competition at the
district in legislative elections, while all other district — level factors are held constant.

While the Korean case study appears to be the ﬁrst attempt to extend the current
understandings on the importance of local elections to the legislative elections under the
SMSP electoral system in a ‘decentralized unitary context,’ there are also legislative
elections under the SMSP electoral systems in a different institutional context. Thus, it is
necessary to assess the current theoretical expectations based on the presidential
democracies against the legislative elections under the SMSP electoral systems in a

federal ‘parliamentary’ institutional context.

 

‘9 Even though Korea adopts a unitary system, it allows local self — government by directly electing the
governor and mayor with popular vote. Also, Korean voters elect their local council representatives. In this
sense, Korea is a ‘decentralized’ unitary system with the presence of political powers at multiple levels
within a country.

36

Canada appears to be appropriate in the sense that it adopts the SMSP electoral
system to elect its national legislative members and has a long history of federalism.
Furthermore, it has been suggested that Canadian provinces often feature strikingly
different national and provincial party systems as in Korea. Both Canada and Korea seem
to produce a “patchwork” national — level party system that is an aggregation of strong
regional political parties across local districts. Thus, incorporation of the Canadian case .
would contribute to comparative study of electoral and party systems with the

comparison of cases from advanced and ﬂedging democracies.

Data and Measurement

This research utilizes two sets of data from both cases: Micro and Macro — level
data. The micro —— level analysis utilizes the survey data on individual demographics and
political attitudes/preferences (Korean Election Studies and Canadian Election Studies)20
All the data used for the macro — level investigation are measured at the legislative
districts. The usage of district — level data is appropriate theoretically and empirically
since the appropriate unit of analysis for the extent of electoral partisan fragmentation in
legislative elections is the local legislative electoral districts. Focusing on the district —
level electoral competition for legislative seats seems to help understand how variable a

picture of electoral partisan fractionalization across districts is in a country.”

 

2° While relying on conventional variables, each data set provides country —- speciﬁc factors to identify the
sources of political cleavage and partisan fragmentation at the individual level. Speciﬁcs on data and
measurement will be discussed in each section.

2' The divergence of the effective number of parties at both national and local levels in a polity can vary
depending on the degree of inter — district variances in vote divisions among parties (or candidates) (Gaines
1997)

37

1) Degree of Electoral Partisan Fragmentation in Legislative Elections at the
District Level

The dependent variable for the macro — level analysis is the degree of electoral
multipartism in a given local legislative district. As Sartori (1976) indicates, inter — party
competition constitutes a party system. Among the measures for party systems, the
effective number of parties has been generally used in the literature.22 It refers to how
many politically relevant parties (or candidates) exist in a given unit (country or district).
Also, it represents the extent to which political power is fragmented or not fragmented,
dispersed or not dispersed.23

The number of effective ‘electoral’ parties is the main dependent variable. In this
study, I use ‘the proportion of popular votes cast in an election at the local legislative
district level.’ The dependent variable is operationalized utilizing a measure based on the
percentage of the electoral votes won by the various parties (or candidates) in a given
district in a given legislative election.

Scholars have developed formulae for converting the vote shares won by multiple
political parties into single number of “effective parties” indices, in which each party
weight according to its relative size, the relative electoral strength in the aggregate.
“Effective” refers to the number of parties, which must be reckoned with in electoral

competition. Laakso and Taagepara (1979) proposed an index, which has become the

 

22 The district - level electoral partisan competition for legislative seats is regarded as conceptually and
empirically distinct from the aggregate level (national or provincial/state) (Bibby et al. 1990). Thus, some
scholars of American state legislative elections have developed different measures of electoral competition
at the district level. For example, some scholars use the marginality of state legislative district for
constituency — level electoral competition (Holbrook and Van Dunk 1993). However, Van Dunk and
Weber (1997) utilize the proportion of seats both the Democratic and Republican parties contested in the
state legislative general elections. As explained later in this chapter, these measures are American context -
related ones, indicating that they seem not to be applicable outside of the US.

38

standard measure of “the effective number of parties.” When K parties split the total vote
in the proportions P 1, P2, ...... Pk then:

N = (2 Pt 2)-I

(When P is the proportion of popular votes for party (or candidate) 1')

N is a continuous measure that weights relatively larger parties which gain more
seats or votes signiﬁcantly more than smaller parties. The use of any simple indicator
conceals some information and reveals other. But we maintain that, “N does a reasonably
good job of capturing the trade off between electoral efﬁciency and representativeness”
(Shugart and Carey 1992: 179). This measure is used for the effective number of
‘electoral’ and ‘legislative’ parties. While the former is based on the proportion of votes
cast in an election, the latter depends on the proportion of seats a party held in a
legislative body. The effective number of electoral parties at the district level in
legislative elections for the Korean National Assembly and the Canadian House of

Commons is used in this research.

2) Degree of Electoral Multipartism across Districts

The average of N across local legislative electoral districts provides a measure of
the effective number of parties at the local electoral district level (DN) for legislative
elections. Unlike N, which is based on the vote shares of each party ‘across a nation,’ this
measure shows how the degrees of electoral partisan fractionalizations are distributed

across the local electoral districts. If there are d districts in a given election, the DN for the

 

2’ See Gaines (1997, 1999) and Anckar (2000) for a variety of measures for the degree of electoral
competition and legislative multipartism.

39

election is calculated as follows: DN = (X N,- )/ d (When N is the effective number of

parties at district 1')

3) Local/District Partisan Characteristics

The primary independent variable under consideration is the local/district partisan
characteristics. It refers to the different levels of partisan strength and subsequent
electoral partisan fragmentation in a given local legislative district. The local/district
partisan characteristics pertain to the basic partisan behavior/preference of voters in a
given district: in some districts, voters are overwhelmingly Republican or Democrat, but
in other districts, voters are more evenly balanced between the parties in the US context.

Understanding local or district — level partisan characteristics has been considered
as important in explaining the degree of electoral partisan competition in general and
incumbency advantages in US State legislative and congressional elections in particular.
Scholars of American politics have developed different measures for the local/district
partisan characteristics in a given local legislative electoral district.

Powell (1993) uses the difference between the two party presidential vote shares
in each district to explain congressional election outcomes. Gaines and Nokken (2001)
utilize the president’s share of the two — party vote from the previous presidential election
as a proxy for district partisan characteristics. According to them, the congressional
districts where the president ran strong ought generally to be safe constituencies for that

party in congressional elections.

40

In contrast, some researchers employ such local elections as state legislative
elections to measure the local/district partisan characteristics.24 It is also an outcome —
dominated model to measure incumbency advantage among state legislators. That is,
Berry and his colleagues (1997), and Gierzynski and Breaux (1991) use the previous vote
margin in the last state legislative elections as a measure of district partisan
characteristics in a given district. Hogan (2001) uses the percentage of the two — party
vote shares received by the state legislative candidate or his/her party in the last local
elections.

The above discussions suggest that the local/district partisan characteristics are
important in explaining the incumbency advantage and the emergence of experienced
candidates (Bianco 1984; Canon 1990) in both American congressional and state
legislative elections. While some scholars focus on the importance of national — level
elections, others put an emphasis on the local elections. Since the local/district partisan
characteristics is “viewed as a relatively stable, long — term inﬂuence, researchers
typically average party vote percentages across elections” (Bond et al. 1997: 285).

Whether one regards national elections or different types of elections for the
local/district partisan characteristics, most studies in the US context have used the
measures based on either Democratic or Republican performance in elections.
Speciﬁcally, all the measures used in the previous studies are the Democratic share of the

two party vote shares in the most recent presidential elections (or state — level elections),

 

2’ While focusing on the state legislative elections for explaining the incumbency advantage in state
legislative elections, Carey and his colleagues (2000) use an “improved method of controlling for the
underlying partisan makeup of districts.” According to them, “Directly estimating the likelihood that a
district will elect a candidate of a particular party, given an open - seat contest, is a better way of
controlling for the underlying partisan effects of a district’s population than simply including the
immediately previous partisan vote share” (p. 675).

41

the difference between the two party vote shares in the last presidential election (or state
— level elections), or the average Democratic vote at the constituency level across a
number of different types of elections (for president, governor, senator and state
legislative) over years.

It may be reasonable to focus on one party’s vote share (Democrat or Republican),
or the difference between two parties’ vote shares in either national or local elections to
understand the district - level partisan characteristics given the nature of American party
and electoral systems. With the primary system for electing partisan candidates for
political ofﬁces and the historical tradition of two party competitions in most elections in
the US, using partisan performance from either presidential elections, local elections, or
combining different types of elections appears sufﬁcient for identifying the underlying
partisan makeup at the district level.25

However, it is unrealistic to believe that the measures developed in the American
politics context are applicable under any circumstances. As in most comparative political
research, the same concept that is used in one context can be measured with other
instruments in another case keeping the validity and reliability of the measures. Putting
an emphasis on one party’s electoral performance appears sufﬁcient given the tradition of
relatively stable two party systems in most electoral arenas over time.

In contrast, under the brief history of political parties (Korea) and different

partisan strength along regional lines (Canada and Korea), incorporating all the

competing parties (or candidates) seems necessary for identifying the underlying partisan

 

2’ For the measures of the degree of electoral competition at the district level, electoral marginality or
incumbent vote share are used in the study of American congressional or state legislative elections. See
Koetzle (1998) and Bond et al (2001 ).

42

characteristics at the district level. In other words, if there are more than two parties
competing for political ofﬁces nationally or locally, the effective number of parties based
on the relative proportion of vote share of each party in a given unit is more relevant than
the measure focusing on one party’s performance to measure the local/district partisan
characteristics. Given the sub — national elections as representing the local/district
partisan characteristics, the district — level electoral returns in such local elections as

gubernatorial, mayoral and local council elections are used.

4) District - Level Control Variables

While focusing on the impact of the local/district partisan characteristics, this
study also introduces possible district - level factors that presumably inﬂuence the level
of electoral partisan competition in legislative elections. The ﬁrst variable to consider is
the degree of electoral multipartism at the district level in the last legislative elections.
the inclusion of the previous degree of multipartism in a given district allows us to
control for the enduring and expected level of electoral partisan fragmentation along with
the underlying partisan composition or characteristics of the district (Gierzynski and
Breaux 1991; Hogan 2001). All other things being equal, we would theoretically expect
that the higher the degree of electoral competition in previous legislative elections, the
higher the level of electoral partisan fragmentation in a given district in the next election.

However, Achen (2000) indicates the danger of lagged variables by showing that
lagged dependent variables does bias the substantive coefﬁcients and artiﬁcially inﬂate
the effect of lagged dependent variable. Speciﬁcally, “When one or more lagged
dependent variables are added as explanatory factors, the autoregressive terms take on

strongly signiﬁcant coefﬁcients which improve the ﬁt but squash the effects of the other

43

variables” (p. 21). Therefore, this variable will be complementarily used for additional
support for the main argument of this research.26

Unlike parliamentary systems in which electoral competition between parties
generally takes precedence between individual candidates like the United Kingdom,
electoral competitions for legislative seats between candidates in legislative elections in
presidential democracies take precedence over competition between parties like the US.
(Hickman 1992). In this sense, research has suggested that such individual candidate —
related factors as incumbency of candidates and challenger quality (Brown and Woods
1991) are inﬂuential.

Therefore, this research introduces two more control variables: the
presence/absence of the incumbent legislator in a given district and the quality of
challenger. It is likely that there emerges more intense inter — party (or inter — candidate)
electoral competition for legislative seats if an electoral district has ‘no incumbent
candidate’ running for reelection (open seat). Incumbent candidates in most cases tend to
have well — organized and carefully tended bases of personal electoral support among
local elites and voters. Incumbent candidates have obvious electoral advantages over
challengers. Successes in previous campaigns, previous experience in legislative ofﬁce,
and higher publicity in the district tend to make incumbent candidates stronger

opponents.27

 

26 The same formula utilized for the dependent and the main independent variables is used for this measure.
The effective number of parties across districts in the previous legislative election is used to measure the
district - level previous multipartism in legislative elections.

27 This seems to come from the fundamental difference originates from the different purpose of legislative
elections depending on different governmental types (Shugart and Carey 1992). In parliamentary systems,
legislative elections are to form a national executive branch. On the contrary, the legislative elections of
presidential democracy in most cases are not related to national government formation. While campaigns

44

The difference in electoral competition patterns between parliamentary and
presidential systems suggests that there will be more intense intra — party competition
among candidates for party nomination in a parliamentary system if there is an open seat
in a given electoral district. Also, it is likely that there emerges more intense inter —- party
(or inter - candidates) electoral competition for legislative seats if an electoral district has
no incumbent candidate running for reelection. Incumbent candidates have obvious
electoral advantages over challengers. After all, the presence of incumbents in a given
district tends to have a depressing effect on the degree of electoral multipartism (Jones
1997; Van Dunk and Weber 1997).

In addition, contemporary theories of legislative elections focus on the quality of
challenger. Studies on US. congressional elections (Cox and Katz 1996; Jacobson 1990
and 2001) suggest that challenger quality, measured in terms of possessing previous
elective ofﬁce experience, inﬂuences district — level electoral competition pattern for
legislative seats. Jacobson suggests that US congressional election outcomes are strongly
inﬂuenced by the presence of quality challengers. Especially when it is an open seat,28
experienced candidates are more likely to emerge (Bond, Fleisher and Talbert 1997).

Some district — level control variables are used only in one case depending on the
data availability. Also, the operationalization of the same concept is based on different
instruments depending on data availability. For example, the variable of district nature
(rural vs. urban) is measured by the ofﬁcial administrative standards in the Korean case.

However, the district diversity index will be used for the Canadian case that incorporates

 

are party — centered in parliamentary systems, presidential democracy tends to have “candidate - centered
campaigns,” and thus, “legislative individualism” in legislative behavior aﬁer elections.

Even though there is an incumbent legislator seeking reelection in a given district, it also seems
reasonable to expect that quality challenger(s) increase the level of electoral competition in a given district.
Carson and his colleagues (2001) even ﬁnd that the effects of quality challengers on constituency — level
electoral outcomes are not conﬁned to postwar congressional elections in the US in their analysis of the
1862 congressional elections. Therefore, whether an experienced challenger(s) emerge in the district should
have a positive impact on the degree of electoral multipartism.

45

demographic variables measured at the riding level. Accordingly, the degree of electoral
multipartism at the district level in legislative elections will be expressed with the
following equation: Y = O F(P and Q),

where Y is the effective number of parties at the electoral district level in legislative
elections; P is the local/district partisan characteristics, Q is the district — level control
variables and (9 is a stochastic factor. Function F is assumed to be linear and the above
function can be transformed into a regression equation as speciﬁcally shown in the

subsequent chapters.

46

CHAPTER TWO

MACRO ANALYSIS OF THE KOREAN CASE

Introduction

This chapter is a macro-level analysis of the Korean case. As a decentralized
unitary system, South Korea allows for the presence of political powers at multiple levels
within a federalist-like polity. Korean voters directly elect their governor, mayor, and
local legislative members through popular vote. Furthermore, the regionalist nature of
party politics since the restoration of procedural democratic requirements suggests that
political and social diversity tends to follow along regional lines. Regional inﬂuences,
therefore, have an impact on the level of electoral partisan fragmentation in legislative
elections across local legislative electoral districts.

As stated above, the unit of analysis in the macro analysis entailed in this research
is the local legislative electoral district, since the degree of electoral partisan
fragmentation at the national level represents an aggregate of diverse electoral
multipartism at the local level. If the basic unit of electoral competition for legislative
seats is a local district, any attempt to better understanding the level of electoral
competition should be concentrated at the local district level. Thus, this chapter ﬁrst
provides a preliminary analysis of party system developments at both national and sub-
national (regional and district) levels. Then, it proceeds to an analysis of the determinants
of the degree of electoral partisan fragmentation at the district level with an emphasis on
local/district partisan characteristics, while controlling for other district-level

confounding factors.

47

Degree of Electoral Multipartism at National and Sub -National Levels

The Korean case appears to be an outlier from the dominant or typical form of
electoral system - centered explanations of party system development, reﬂected in the
electoral multipartism of the four legislative elections that have taken place since
democratization. Table 2-1 compares the effective number of electoral parties in the
legislative elections of certain countries that use the SMSP electoral systems.

Table 2-1 Comparison of the Degree of Electoral Multipartisma

 

Country Degree of Electoral Multipartism

 

Canard? 4.09
J amaicac 2.16
Indiad 7. 19
USA" 2.16
Koreaf 3 .83

 

“ These countries under the SMSP electoral systems for legislative elections are selected for
example, for more comparison see Taagepera and Shugart (1989); b House of Commons election
in 1997; c House of Representatives election in 1997; d House of the People election in 1998; °
House of Representatives election in 1998; fthe degree of electoral multipartism in Korea is an
average of four legislative elections under review. Source: calculated from the data provided by
“Political Resources on the Net.”

As Duverger’s Law would suggest, the plurality rule with a single member district
is expected to lead to a two party system. However, the degree of electoral multipartism
at the national level has exceeded theoretical expectation and reached to three or even
four party systems in the Korean legislative elections since 1987. The number of electoral
parties has been 4.22, 3.51, 4.24 and 3.33, respectively. The national-level shape of the
party system in Korean legislative elections has a distinct multiparty character.

However, if we look at the average level of electoral partisan fragmentation
across local districts, it has been relatively closer to a two party system than that of the

national level. Table 2-2 compares the degree of electoral partisan fragmentation at

48

national and district levels in the four legislative elections under consideration. It
suggests a relative tendency towards two party systems at the district level1 (2.87, 2.77,
3.08 and 2.57, respectively), whereas there is a greater likelihood of multiparty systems at
the national level. Therefore, one could say that the Duverger’s Law does work in the

Korean case, but only at the district not the national level

Table 2-2 Comparison of Electoral Multipartism at National and District Levels

 

1988 Election 1992 Election 1996 Election 2000 Election

 

National District National District National District National District

4.22 2.87 3.51 2.77 4.24 3.08 3.33 2.57

 

Note: the district level degree of electoral multipartism is an average of all the local districts following the
formula: DN = (Z N,- )/ d (When N is the effective number of parties at district id is the total number of local
electoral districts)

The two tables indicate that a national-level analysis of thedegree of electoral
partisan fragmentation in legislative elections is likely to be problematic. The comparison
of the degree of electoral multipartism at national and district levels clearly shows what
would be the appropriate unit of analysis for understanding the degree of electoral
competition in legislative elections under the SMSP electoral systems.

The shape of the party system at the national level is an aggregation of local,
district-level party systems. The national party system represents a mixture of dynamic
patterns of electoral partisan competition for legislative seats across localities that are
different from district to district, and, furthermore, province to province (as an

aggregation of local districts), contributing, in turn, to the shape of the party system at the

 

' The average effective number of electoral parties across districts is calculated as follows: l/sum of each
party or candidate ’5 vote share squared at each district and the sum of the eﬂective number of parties of all
the districts is divided by the total number of districts for the election.

49

national level. The degrees of electoral multipartism of some districts or regions are
higher than that of the national district average, and relatively closer to that of the
national level. Multiparty electoral competition is very evident in these cases. The levels
of electoral partisan fragmentation in other districts or regions, however, are lower than
that of the national level. Relatively two-party electoral competitions tend to emerge in
those districts.2

Table 2-3 shows how the degrees of electoral partisan competition for legislative
seats are different from region to region (metropolitan area, provinces, etc.). Relatively
high degrees of electoral multipartism have existed across local districts in some regions
(relatively closer to the national level of electoral partisan fragmentation). In contrast,
certain districts have less than two parties competing in legislative elections (relatively
closer to the district average level of electoral competition).3

Table 2-3 indicates that the degree of electoral multipartism across districts is
closer to a two-party system (district-level electoral competition) or closer to a multiparty
system (national-level electoral multipartism) following regional lines. For example, the
Honam region (Kwangju, Chonrabuk-do and Chonranam-do) is far below the national
average in its degree of electoral partisan competition. This implies that the local strength
of a certain party is dominant in the region. The capital area (Seoul, Inchon and Kyunggi-
do), however, has had more than three parties competing for legislative seats, indicating

that there is no dominant party, but multiple parties vying for partisan support.

 

2 Thus, when we focus on the degree of electoral competition at sub-national levels, we can ﬁnd more
dynamics in the levels of electoral multipartism across a nation (Chibber and Kollman 1998; Cox 1997;
Jones 2001). This point also suggests that there might exist local parties surviving geographically, and thus
different types of party systems following regional lines (Rae 1971; Riker 1986).

3 These points also suggest that the local or district-level partisan strength of a party is different from
region to region. While a party is relatively strong in a region, there are also districts in which more than
two parties compete with each other for legislative seats.

50

Table 2-3 Provincial — Level Comparison of Electoral Multipartism

in Recent Legislative Elections

 

 

Regions/ 1 988 1992 1996 2000 Regional
Elections Election Election Election Election Mean
Seoul 4.01(42) 3.14(44) 3.26(47) 2.5 1 (45) 3 .23
Pusan 2.45 (15) 2.58 (16) 2.53 (21) 2.35 (17) 2.48
Taegu 2.71 (8) 2.66 (11) 3.71 (13) 2.27 (1 l) 2.84
Inchon 3.68 (7) 3.46 (7) 3.52 (11) 2.75 (11) 3.35
Kwangju 1.30 (5) 1.71 (6) 1.30 (6) 1.79 (6) 1.53
Ulsan . 2.44 (5)b

Taejon 3.09 (4) 3.41 (5) 2.97 (7) 3.45 (6) 3.23
Kyunggi-do 3.28 (28) 3.04 (31) 3.62 (38) 2.80 (41) 3.19
Kwangwon-do 2.83 (14) 3.06 (14) 3.61 (13) 3.02 (9) 3.13
Chungchongbuk-do 2.63 (9) 2.72 (9) 3.22 (8) 3.42 (7) 3.0
Chungchongnarn-do 2.65 (14) 2.93 (14) 2.64 (13) 3.01 (11) 2.81
Chonrabuk-do 2.17 (14) 2.35 (14) 2.20 (14) 2.08 (10) 2.2
Chonranam-do 1.79 (18) 2.18 (19) 1.80 (17) 1.93 (13) 1.93
Kyongsangbuk-do 2.42 (21) 2.47 (21) 3.90 (19) 2.48 (16) 2.82
Kyongsangnam-do 2.59 (22) 2.57 (23) 2.90 (23) 2.54 (16) 2.65
Cheju 2.82 (3) 2.51 (3) 3.70 (3) 2.21 (3) 2.81

 

National District Average 2.87 (224) 2.77 (237) 3.08 (253) 2.57 (227) 2.82

 

ITh

e number of districts for each region is in the parentheses.
b Before the election, Ulsan Metropolitan area was included in the Kyongsangnam-do.

Source: calculated from the “Data Book for the 1988, 1992 and 1996 and National Assembly
Elections” and “the Press Releases for the 2000 National Assembly Election” by NBC.

51

Table 2-4 Provincial Comparison of Electoral Multitpartism
in Recent Presidential Elections

 

 

 

Regions/ 1988 1992 1997 Regional
Elections Election b Election Election b Mean
Seoul 3.45 (42) 3.18 (44) 2.57 (47) 3.07
Pusan 2.25 (15) 1.77 (16) 2.45 (21) 2.16
Taegu 1.80 (8) 2.43 (11) 1.80 (13) 2.01
Inchon 3.34 (7) 3.39 (7) 2.91 (1 1) 3.21
Kwangju 1.25 (4) 1.09 (6) 1.06 (6) 1.13
Taejon 3.72 (4) 3.66 (5) 2.88 (7) 3.42
Kyunggi-do 3.16 (28) 3.36 (31) 2.86 (38) 3.13
Kwangwon-do 2.30 (14) 3.08 (14) 2.89 (13) 2.76
Chungchongbuk-do 2.96 (9) 3.48 (9) 3.13 (8) 3.19
Chungchongnam-do 2.90 (14) 3.44 (14) 2.71 (13) 3.02
Chonrabuk-do 1.42 (12) 1.27 (14) 1.18 (14) 1.29
Chonranam-do 1.22 (18) 1.18(l9) 1.12(17) 1.17
Kyongsangbuk-do 1.91 (19) 2.15 (21) 2.26 (19) 2.11
Kyongsangnam-do 2.22 (22) 1.78 (23) 2.45 (23) 2.15
Cheju 2.90 (3) 3.27 (3) 2.94 (3) 3.04
National Average 2.55 (219) 2.56 (237) 2.40 (250) 2.50

 

' Number of ‘legislative districts’ for each region are in the parentheses; l’There is differences between the
number of electoral districts for 13’h Legislative and Presidential elections and 15‘h Legislative and
presidential elections because of administrative boundary changes. Source: calculated from the “Data Book
for the 13’”, 14th and 15th Presidential Elections” by NBC.

52

In 37 out of 61 provinces fall below the national district average in terms of
electoral multipartism across districts, coming closer to that of the national level. The last
column of the table indicates the regional average level of electoral partisan competition
across four legislative elections. This also suggests that the degree of electoral partisan
fragmentation varies along regional lines. Also, regionally distinct patterns of electoral
partisan competition appear across elections. The shape of the development of the party
system at the national level represents a mixture, therefore, of those different party
systems that hold power at regional and district levels.4

The comparative analysis of the degree of electoral multipartism across regions in
legislative elections suggests that different regional patterns of electoral partisan
competition have emerged in Korean legislative elections. Furthermore, if different
degrees of partisan strength run along regional lines within a country, then it should be
evident as well in other types of elections than legislative elections. If indeed Korea has a
region-based party system, we should see the same regional patterns in terms of degree of
electoral partisan fragmentation, regardless of election types.

Table 2-4 shows how the degree of electoral multipartism varies depending on
localities in presidential elections. It indicates that the effective number of parties is
closer to a two-party system or closer to a multiparty system, contingent on regions in
presidential elections. For instance, the Honam regions fall far below the national average

in degree of electoral partisan competition. Among the 45 regions across elections, 20

 

4 While it is beyond the scope of this research, “there is another (little — explored) logic that drives across
district alliance formation, thereby combining a variety of district-conditioned patterns of party competition
into a larger national aggregate” (Cox 1997: 275). That is, why the relative two party system at the district
level leads to the relative multiparty system at the national level, is another research question to be
answered. It is argued in a separate study that a national party system appears to be different from that of
sub-national localities and the geographic variation of electoral partisan competition across a country
contributes to the formation of the national party system, and that the political/social cleavages of a polity
plays a signiﬁcant role in linking the different levels of electoral partisan competition across local districts
and with respect to national-level electoral partisan fragmentation. Irrespective of institutionalist
expectations, the intra-country comparative analysis of party systems suggests the importance of
regionalism or provincial favoritism - geographic concentration of electoral supports — in the aggregation
process of party systems from local to national levels under the same electoral systems.

53

regions fall below the national average, indicating regionally different patterns of partisan

strength or ﬁ'agmentation in the presidential elections.

Table 2-5 Provincial Comparison of Electoral Multipartism

Regions/

in Recent Local Elections

 

 

Elections 1991 Election 1995 Election 8 1998 Election
Seoul 3.20 (44) 2.91 (47) 2.32 (23)
Pusan 3.30 (16) 2.46 (19)b 2.26 (16)
Taegu 3.18(11) 3.10(13) 2.16(8)
Inchon 4.23 (7) 3.31 (11) 2.80 (8)
Kwangju 2.20 (6) 1.41 (6) 1.52 (2)
Taejon 5.43 (5) 2.53 (7) 2.25 (5)
Kyunggi-do 3.49 (31) 3.70 (38) 2.55 (30)
Kwangwon-do 4.32 (14) 4.54 (13) 2.61 (18)
Chungchongbuk-do 3.42 (9) 3.93 (8) 3.36 (10)
Chungchongnam-do 3.86 (14) 2.68 (13) 2.46 (14)
Chonrabuk-do 2.29 (14) 2.59 (14) 2.10 (13)
Chonmam-do 2.50(19) 2.78 (17) 2.13 (18)
Kyongsangbuk-do 2.95 (21) 4.04 (19) 2.35 (18)
Kyongsangnam-do 3.46 (23) 4.04 (23) 2.56 (17)
Cheju 4.38 (3) 3.44 (3) 2.39 (4 )
National Average 32°23.) 322531) b fills)“

 

' Based on municipal elections in 1995 and 1998, and provincial council election in 1991. Both 1991 and
1995 elections are recoded following the 1992 and 1996 National Assembly elections’ legislative electoral
district, respectively; b Two districts with only one candidate in Pusan; ° Among the 232 municipal districts,
there are 23 districts with only one candidate, and in the Kyongsangnam-do province, the metropolitan area
of Ulan is created. It’s regional average is 2.99 with 5 municipal districts. Source: calculated from “Data
Book for 1991, 1995 and 1998 Local Elections” by NBC.

54

These kinds of differences in the patterns of partisan strength across regions and
subsequently different degrees of electoral partisan competition following regional lines
also appears when we see the average effective number of electoral parties at the local
legislative district levels in local Korean elections held since 1987. Local elections are
more reﬂective of the local strength of a party than a national-level election since the
basic electoral district for local elections (especially for mayoral elections) is in most
cases ‘identical’ to that for legislative elections.5 All local elections since the restoration
of procedural democratic requirements in 1987 show almost the same regional patterns in
the degree of electoral partisan fragmentation as shown in Table 2-5.

Therefore, it appears to be clear why we it is important to focus on the district-
level electoral partisan fragmentation in legislative elections. The national-level party
system is a mixed picture since it includes different patterns of local (districts and
regions) electoral competition across the country. A two party system in some electoral
districts, where a certain party has relatively dominant electoral support, is counter-
balanced by a multiparty system in other districts, where no party has sufﬁcient support

to be dominant.

Determinants of District — Level Electoral Multipartism in Legislative Elections
Data and Method

The basic premise underlying a focus on prominent local elections in order to
explain the level of electoral competition across districts in legislative elections is the
presence of political powers at multiple levels and political/social diversity within sub-
national units. However, it would not be reasonable to expect that ‘the federalist structure

of a country’ alone allow for sub-national, social/political diversity and the presence of

 

5 Korea held the 16th National Assembly elections in April 2000. The ‘regionally concentrated political
support of a party’ still seems to exist since the number of nomination applications in each party’s regional
stronghold exceeded its national average. Digital Chosun Daily (Korean), February 8, 2000. Accessed on
February 9, 2000.

55

political powers at multiple levels across a country. Korea is not a federal but a unitary
system reﬂecting the fact that it is a totally homogeneous country in terms of ethnicity
and language. Nevertheless, it allows for local self-government by directly electing the
governor and local administrative manager6 through the popular vote. Also, Korean
voters directly elect the local council representatives of metropolitan areas or provinces,
as well as local administrative districts.

All of this would suggest that Korea is a ‘decentralized’ unitary system in which
political power exists at multiple levels, like federal systems. Thus, the more
decentralized, the higher probability that local/district-level electoral partisan competition
patterns in prominent local elections affect district-level electoral multipartism in
legislative elections. Or, the more decentralized governmental power is in a nation like
Korea, the less complete scholarly research will be if it does not take into consideration
the pattern of electoral partisan competition in elections of local ofﬁces to explain
district-level electoral multipartism in legislative elections (Jones 1997).

Korea is also well known for its stridently “region-based party systems and
regionalist politics” since democratization (Clark 1987; Kang and Jaung 1999; Lee and
Brunn 1996; Morris 1996). Hence, there is relatively less social/political diversity
following regional lines in the country, indicating that there might exist different types of
party systems across sub-national units as in federal systems. With the presence of a
‘decentralized’ unitary system and ‘regionalist’ party politics, ‘the local/district partisan
characteristics’ of the degree of electoral partisan fragmentation in the quest for
prominent local political ofﬁces, appears to be inﬂuential in deciding the level of
electoral multipartism across local districts when electing the members of the Korean

National Assembly.

 

6 The local administrative managers consist of “mayors, ward and county heads” in Korea and will be
referred to as “mayors” hereafter.

56

This study focuses on the four national legislative elections in Korea since 1987.7
There are two reasons to include these four elections (1988, 1992, 1996 and 2000). With
the 13th National Assembly election in 1988, Korea adopted the SMSP electoral system
to elect its national legislators.

Table 2-6 Division of Seats in Korean National Assembly Elections

 

Election/Seats Local District National Constituency Total

 

1988 224 75 299
1992 237 62 299
1996 253 36 299
2000 227 46 273

 

As shown in Table 2-6, the total number of national legislators was 299 for the

1988, 1992 and 1996 elections, and 273 for the 2000 election, including the legislators
from the ‘national constituency.’ District elections are governed by the SMSP, whereas
the national constituency is based on the national vote share of a party with a closed party
list.8 The number of national constituency (PR component) has changed from 75, 62, 36
to 46. The number of local electoral districts has also changed from 224, 237, 253 to 227.
Therefore, local districts have altered from election to election through redistricting and
administrative boundary changes, clearly indicating that local electoral districts are not

identical from election to election. The dependent variable is the effective number of

 

7 It was the ﬁrst election under the new electoral system. See Brady and Mo (1992) and Grofman et al
(1999) about electoral system changes in Korea.
8 However, the speciﬁcs (especially, threshold and 1" party premium) have changed over time as shown
below.

Table 2-7 Speciﬁcs on National Constituency Seats Allocation

 

 

Threshold 1" Party Premium
1988 5 seats from Local Districts '/2 of PR Seats for 1“ Party if not 50% of Districts
1992 Sam as above No Premium; Based on # of Local Districts Won
1996/2000 5 Seats or 5% of Votes No Premium; Based on National Vote Shares

 

57

electoral parties at the district level in those four legislative elections, following the
Laakso- Taagepara Index.

The primary independent variable under consideration is the local/district partisan
characteristics — local or district — level electoral competition patterns that are evident in
local elections.9 This study has a special focus on two types of local elections for the
Korean case study to identify the ‘local/district partisan characteristics’ that inﬂuence the
degree of electoral multipartism in legislative elections across districts: “gubernatorial
and municipal elections.”

Table 2-8 Elections in Korea since Democratization

 

 

Year/Month Types of Elections
1987 December Presidential Election
1988 April National Assembly Election
1991 March Local Elections for City, County and Ward Councilsal
1991 June Local Elections for Metropolitan and Provincial Councils
1992 March National Assembly Election
1992 December Presidential Election
1995 June Local Elections for Councils (citya, county“, warda, metropolitan

and provincial), and Governors, Mayors, Ward and County Head

1996 April National Assembly Election
1997 December Presidential Election

1998 June Local Electionsa (same as 1995)
2000 April National Assembly Election

 

' non-partisan election.

 

9 Local elections consist of four types of ofﬁces: governors for metropolitan areas and provinces, mayors
for cities and counties of a province, wards of metropolitan areas, provincial council members for
metropolitan areas and provinces, and municipal council members for cities, counties and wards. The
elections for governors and mayors are under the simple plurality rule. The single member simple plurality
rule system is used to elect council members and a small portion of them is elected with proportional
representation based on the vote share of parties.

58

The ‘local’ partisan characteristics are measured by the electoral returns from
gubernatorial elections that are disaggregated to match the electoral districts of legislative
elections. In contrast, the ‘district’ partisan characteristics are measured utilizing the
electoral returns from municipal elections. Using the formula utilized for the dependent
variable, the effective number of electoral parties at the legislative electoral district level,
from the ‘closest’ local elections (gubernatorial and municipal), is used for the
local/district partisan characteristics. Table 2-8 shows all the national and local elections
held in Korea since 1987.

The focus on electoral partisan fragmentation in gubernatorial elections helps us
to identify the local (provincial) partisan characteristics that affect the degree of electoral
multipartism across a province in legislative elections. While electoral competition in
gubernatorial elections represents a certain type of local-level party system fragmentation
across a province, it also represents an aggregation of sub-provincial units across the
province. It is not a ‘legislative district-level electoral partisan competition’, but a
province-level electoral competition. Unlike Argentina (Jones 1997) and Brazil (Samuels
2000afb) that use the proportional representation electoral systems at the ‘province level’,
the basic unit for legislative elections in Korea is sub-provincial (and sub-metropolitan)
with local legislative electoral districts under the SMSP electoral system.

Therefore, if there are sub-provincial local elections held at the legislative
electoral district level, this appears to be relevant for understanding district-level electoral
partisan characteristics. The Korean case provides us with an adequate measure for
identifying ‘district-level partisan characteristics’: municipal elections. These elections
are not only sub-national but sub-provincial. That is, the basic electoral district for
electing local administrative managers is in most cases ‘identical’ to those districts for
national legislative elections. Hence, the degree of electoral partisan fragmentation in

municipal elections is reﬂective of the district-level dynamics of electoral partisan

59

competition that affect legislative elections across districts. Introduction of two
complementary measures for local/district partisan characteristics in ‘a decentralized and
regionally heterogeneous case’ can contribute to expanding our understanding of
interrelationships between different types of elections.

This analysis introduces certain control variables to identify the impact of
local/district partisan characteristics on the degree of district-level electoral competition
in legislative elections. The presence of control variables improves the ﬁt of the model.
These variables are ﬁrst, the presence (or absence) of an incumbent legislator and quality
challenger, followed by the distinction between urban vs. rural district. The ﬁrst variable
— presence of incumbent legislator in a given district — controls for whether or not an
incumbent ran in the election under review. Districts where incumbents ran were coded
zero while open seats were coded one. The absence of incumbent(s) is expected to have a
increasing effect on the degree of electoral multipartism across districts. In contrast,
whether or not a quality challenger with previous elective experience in a given district
runs for a legislative seat appears to have an increasing impact on the level of electoral
competition in the district. Thus, districts where a quality challenger ran were coded one
whereas districts without a challenger with previous elective experience were coded zero.

Another possible district-level control variable to take into consideration for the
Korean case is whether a district is in a rural or an urban area. Korea has undergone
rapid economic growth over the last 30 years. Urban areas are mixed with many people
from all over the country, leading to less regional identity than in rural areas. There is
also more mobility among people who live in urban areas. Before the emergence of
regionalist party politics in the 1987 presidential election, urban areas had tended to vote
for opposition parties and rural areas for the incumbent party; this had long been a
dominant voting pattern in Korea (Cho 1998; Kim and Browne 2001; Kim and Khil

1988). This appears to suggest that there would be a relatively small number of parties

60

competing for legislative seats in a rural district, while there would be more parties in
urban districts. Hence, there may be more parties or candidates representing a variety of
interests in an urban district, decreasing regionalist partisan strength but increasing the
degree of electoral partisan competition.IO

These three control variables (urban/rural, open districts, and quality
challengers) should have a positive inﬂuence on the degree of electoral multipartism in
legislative elections at the district level. Accordingly, the effective number of electoral
parties in legislative elections at the district level can be expressed with the following
equation:

Y = O F(P, U/R, I, Q)

where Y is the effective number of parties at the electoral district level in legislative
elections, P represents local/district partisan characteristics, U/R and I stand for whether a
local legislative district is rural or urban, and whether there is an incumbent legislator or
not. Q represents challenger quality and G) is a stochastic factor. Function F is assumed to

be linear and the above ﬁrnction can be transformed into a regression equation as follows:

Y =8] + 32X] + B3X2 + B4X3 +B5X4+u

in which B's are unknown but with ﬁxed parameters, X's 1 through 4 represent chosen
independent variables (local/district partisan characteristics, urban/rural, open seat/or not,

challenger quality), and u represents the stochastic factor, 9.

 

'0 For the analysis of the impact of rural vs. urban districts, the hybrid districts (urban + rural) have been
excluded from the analyses.

61

Analysis and Discussion

Table 2-9 Local Partisan Characteristics and Legislative Electoral Multipartism
across Districts

 

 

l988/
Elections/Variables 1988 1992 1996 2000 33%;
2000
.133 .277*** -.036 .162 .211***
Urban/Rural (.089) (.084) (.164) (109) 0061)
.179** -.278*** .025 -.034 .002
0P6“ Seat 0‘ N01 (.086) (.093) (.117) (.142) (.059)
. .124 -.062 .370*** .270*** .205***
Challenger Quamy (.108) (.089) (.1 13) (.080) (.054)
Local Partisan .765*** .262*** .707*** -.139 .499***
Characteristicsa (.056) (-050) (4)82) (-1 19) (038)
. .202***
1988 Electron (.070)
. 4364***
1992 Electron (.077)
. .250***
1996 Electron (.070)
C .811*** 1.80*** 1.16*** 2.60*** 1.29***
“Siam (.130) (.157) (.203) (.272) (.087)
R2 .6015 .2100 .3542 .0718 .3224
N 196 207 234 203 840

 

* signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; regression coefﬁcients are
unstandarized ones; robust standard error of the estimates is in the parentheses; " Two provinces
had one candidate in gubernatorial elections in 1998, thus, ‘the local partisan fragmentation’
measure is alternatively from local council elections. The 1991 local elections were for electing
‘provincial (metropolitan) council members’ only and it was used for the local partisan
characteristics. Also, this analysis used the 1987 presidential election for the 1988 legislative
election as a proxy measure.

Table 2-9 shows how ‘local’ partisan characteristics affect the degree of electoral

multipartism at the district level in legislative elections, while controlling for other

62

district-level confounding factors. The level of electoral partisan fragmentation across
electoral districts in legislative elections seems to be a function of ‘local partisan
characteristics’ .

The degree of electoral partisan fractionalization from the closest provincial-level
elections (gubernatorial) in a given district is signiﬁcantly inﬂuential in determining the
effective number of parties in the district-level legislative elections. Where the level of
‘local’ electoral partisan fragmentation is relatively high (more parties) in a given district,
the degree of electoral multipartism in the district tends to be high in the subsequent
legislative elections (more parties). Conversely, a district in which the level of electoral
partisan fragmentation is relatively low (fewer parties) tends to have fewer parties in the
following legislative elections (fewer parties).ll

It is important to note that the estimate coefﬁcients for the variable of local
partisan characteristics remain very stable across elections. Local partisan characteristics
have a strong positive impact on the degree of electoral multipartism in a given district in
legislative elections. The estimated regression coefﬁcients for this variable are signiﬁcant
at the .01 levels for a two-tailed t test. For example, the estimated coefﬁcient of .765 for
the 1988 legislative election indicates that in an average district an increase in the local
partisan characteristics from 1 to 2 would, all other factors remaining constant, result in
an increase of approximately .765 of an effective number of electoral parties across

districts in the legislative elections.

 

” Separate analyses were conducted focusing only on the local elections of any type (for 1992, 1996 and
2000 legislative elections) and the gubematorial elections (1996 and 2000 elections) to isolate the
confounding effects of using a proxy measure (legislative district level presidential electoral returns for the
1988 legislative election). Those analyses conﬁrm that local elections are consequential in determining the
nature of district-level electoral competition in legislative elections.

63

Accordingly, the ‘provincial (and metropolitan)-level partisan characteristics’ that
are seemingly different from province to province exerts a signiﬁcant inﬂuence on
candidates’ incentives to coordinate with a political party in a given district in a given
province (or metropolitan area). The degree and shape of electoral competition in the last
gubernatorial elections provided a major motive for rational politicians or candidates to
associate with a party for their electoral successes in the next legislative elections.

The occurrence of a relatively strong party in a province indicates that there are
relatively fewer parties competing in provincial-level local elections. Under these
circumstances, most candidates try to associate with the dominant party, probably
decreasing the degree of electoral partisan fragmentation in the province and increasing
the level of within-party competition for party nominations. On the contrary, where no
parties have strong support, there are relatively more parties competing in provincial-
level local elections. Then, politicians tend to associate with any party they consider to be
the most likely party to serve as a vehicle for their electoral success. This tends to
increase the degree of electoral partisan fragmentation, but to decrease the level of
within-party competition for partisan nominations.

Following Leamer’s logic, the control variables are not considered consequential
to the analysis (Leamer 1990). However, their independent impact on the level of
electoral partisan fragmentation across districts in legislative elections does merit a brief
discussion and it will additionally add support the principal thrust of what is being argued
here. The factors of whether or not a district contains an open seat and the quality of
challenger(s) across districts show mixed results. Challenger quality and the absence of

an incumbent legislator in a given district were statistically signiﬁcant factors in the last

64

two legislative elections (2000 and 1996), as well as in the ﬁrst two elections (1988 and
1992), respectively.

For example, a legislative contest without incumbents in a given district resulted
in a .179 increase in the level of electoral partisan competition for legislative seats,
compared to a legislative election in the district that had incumbents running in the 1988
legislative election. This ﬁnding is statistically signiﬁcant at the .05 level for a two-tailed
t test. Also, the presence of quality challenger(s) in a given district resulted in a .270
increase in the degree of electoral partisan fragmentation compared to a district without
no quality challenger(s) running. This is statistically signiﬁcant at the .01 level for a two-
tailed t test.

The presence of incumbents and a quality challenger(s) in a given district are not
consistently signiﬁcant across elections. This seems to be related to the local partisan
characteristics that are different within sub-national units, and electoral system changes in
1988. As mentioned before, Korea adopted a SMSP electoral system in 1988. Before
democratization, two legislators were elected from each local electoral district. The
change of the electoral system from a ‘two-member district’ to a ‘single-member district’
produced many districts without an incumbent legislator in the ﬁrst election under the
new system, and it seems to have induced relatively higher degrees of electoral partisan
competition for legislative seats in the 1988 election.

With successive elections under the same electoral system, the quality of
challenger(s) becomes signiﬁcant, increasing the degree of electoral multipartism across

districts in legislative elections as expected.12 With the introduction of local elections in

 

'2 This can be also attributed to the changing nature of regionalist party systems since the 1991 merge of
three parties as explained later in this section.

65

1991, candidate(s) emerged with previous experience (to local councils or local
administrative ofﬁces) in legislative elections. Also, politicians who had earlier been
elected to the National Assembly in a given district, but failed to get partisan nominations
from major political parties for an upcoming election, decided to run as independents. All
of this tended to increase the level of electoral partisan competition across districts in the
last two legislative elections.

Interestingly, the factor of whether or not a district contained an open seat for the
1992 election yielded results directly opposed to conventional or theoretical expectations.
A district having an open seat tended to ‘decrease’ the level of electoral multipartism
across districts. The 1992 legislative election, for example, was held after three out of
four major regionally-based political parties had merged into one party in 1991,
complicating party politics still further. The bottom line, of course, for any rational
politician, is to be (re) elected in an election (Mayhew 1974).

Politicians try, therefore, to associate with a party they believe will enhance their
chance of electoral success. Given the highly regionalist character of the party system in
the 1992 election, for example, the only reasonable choice for most politicians was to get
party nominations from the party that was regionally strong in the area. Thus, there were
more intense ‘intraparty’ competitions for party nomination than ‘interparty’ electoral
partisan fragmentation across districts in the 1992 legislative election, indicating that
even a district with an open seat did not experience an increase in the level of electoral
multipartism.

This point is also conﬁrmed when we look at the ‘temporal effects of election

years’ in the pooled analysis section. The inclusion of multiple election years serves to

66

control for any effects that are associated with a speciﬁc election year. The ‘year’
variable helps control for potential temporal effects such as serial correlation. Also, it
controls for events or circumstances in a given election year that might possibly boost or
decrease the level of electoral partisan fragmentation across districts in legislative
elections. Binary election year dummy variables for the elections since 1987 are
employed to control for temporal effects, with all districts during a particular election
year coded one, and all other elections coded zero (Stimson 1985).

Three temporal dummy variables are included in the statistical analyses with the
remaining year, 2000, as a base reference. All the variables are measured in terms of
difference from the 2000 legislative election year. The 1988 legislative election, for
instance, yielded a .202 increase in the degree of electoral multipartism, compared to the
legislative election of 1992, which registered a decrease in the level of electoral partisan
competition across districts by .364 when compared with the election in 2000. The 1988
and 1992 elections registered a signiﬁcantly different degree of electoral partisan
fragmentation than in the 2000 legislative election. The legislative election of 1996, as
well, showed an increase of .250 in the level of electoral competition compared to the
legislative election of 2000.

These points are in need of further substantive historical explanation. There were
four major regionally-based political parties (DJP, UDP, PPD and NDRP) that
participated and survived in the 1988 legislative election. The 1992 legislative election,
however, was held with only the DLP (a merger of the DJP, the UDP and the NDRP) and
the DP (from the PPD). The merger of three parties in 1991 into a regional coalition of

region-based parties was done to achieve a majority in the national legislative body (Kim

67

1997). Excluding one major opposition party (PPD and later DP), the merger of these
three parties changed the number of parties based on regional supports from four to two
in the 1992 legislative election. One faction of the DLP (formerly the NDRP before the
1991 merger) formed a new opposition party (CLD) in the middle of political power
struggles in 1995, and participated in the 1996 legislative election with the NKP
(formerly the DLP) and the NCNP (formerly the DP).

The above explanation of a longitudinal party system change suggests how ‘local’
partisan characteristics within the changing nature of a region-based political party
system, have an inﬂuence, over time, on the degree of electoral partisan fragmentation
across districts in legislative elections. Therefore, the concentration of electoral support
for a party along regional lines and, subsequently, different partisan characteristics within
sub-national units becomes important in determining the degree of electoral multipartism
at the district level in Korean legislative elections.

Table 2-10 shows how the ‘district’ partisan characteristics are inﬂuential in
explaining the degree of electoral multipartism across districts in legislative elections.
District partisan characteristics are based on the ‘legislative district-level’ for the 1996
and 2000 legislative elections. Partisan fragmentation at the district level appears more
reﬂective of legislative district-level partisan dynamics, which affect the district-level
electoral multipartism in legislative elections since the mayoral electoral district is in
most cases ‘identical’ to that of legislative elections. The ‘district-level’ partisan
fragmentation in legislative district-level local elections inﬂuences the degree of electoral
partisan competition in a given district in legislative elections while other factors remain

constant.

68

Table 2-10 District Partisan Characteristics and Legislative Electoral Multipartism

across Districts

 

 

1992/ I988/
Elections/Variables 1996 2000 1996’ 1996/ ‘99”
2000 1996/
2000 2000
286* .259** .282*** .267*** .329***
Urban/Rural (.159) (.107) (.097) (.067) (.056)
-.022 -.O61 -.O38 -.112 .008
Open Seat Or Not (.133) (.131) (.100) (.075) (.061)
Ch 11 1' .350*** .263*** .313*** .209*** .205***
a engerQualty (.126) (.080) (.074) (.060) (.055)
. . . .255*** .169*** .232*** 244*" .343***
District Partisan Characteristics (.052) (.05 6) (.040) (.032) (.030)
. 4453***
1992 Electron (.066)
. .257*** -.202***
1996 Electron (.069) (.077)
. -.355*** -.O89 -.453***
2000 Election (.074) (.064) (.066)
1.90*** 1.83*** 1.99*** 1.76*** 1.82***
Constant (.205) (.187) (. 159) (.111) (.086)
R2 .1450 .1039 .2154 .2005 .2693
N 234 205 439 646 843

 

* signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; regression coefﬁcients are
unstandarized ones; robust standard error of the estimates is in the parentheses.

Put another way, where the level of district-level partisan fragmentation is

relatively low (fewer parties), the degree of electoral multipartism in a given district

tends to decrease in subsequent legislative elections. Conversely, a district in which the

level of electoral partisan competition is relatively high (more parties) seems to have

69

more parties in subsequent legislative elections. Thus, the district-level partisan
characteristics, utilizing the sub-provincial local elections that are held at that unit,
directly inﬂuence candidates’ incentives to coordinate a party for their electoral success
in a given district, increasing or decreasing the degree of electoral multipartism across
districts in legislative elections. ’

The estimated regression coefﬁcients for the variable of ‘district’ partisan
characteristics are statistically signiﬁcant in the hypothesized direction across elections.
For example, the estimated coefﬁcient of .244 for the district partisan characteristics
indicates that an increase in the variable by 1 would, while all other factors being equal,
result in an increase of approximately a quarter of an effective number of electoral parties
across districts in the legislative election of 2000. The ‘local’ partisan characteristics do
not achieve statistical signiﬁcance for the 2000 legislative election. However, district
partisan characteristics are statistically signiﬁcant at the .01 level for a two-tailed t test
for the 2000 election. This suggests that sub-provincial or district-level local elections are
more adequate for understanding the district-level electoral multipartism in national
legislative elections under a SMSP electoral system.

The variable of quality challenger becomes statistically signiﬁcant for the third
legislative elections held under the current electoral system. Table 2-10 conﬁrms that a
district with a quality challenger(s) tended to have a higher degree of electoral
multipartism in the elections of 1996 and 2000. Unlike the previous model focusing on
local partisan characteristics, the distinction between rural vs. urban district is
consistently signiﬁcant across elections. Whether a district is rural or urban is important

in determining the degree of electoral multipartism across districts in legislative elections.

70

 

There tend to be more parties or candidates competing in urban districts with quality
challenger(s). This ﬁnding suggests that the traditional phenomenon of an ‘urban/rural
divide in electoral competition’ has continued to exist even after democratization, and
that it increases the level of electoral partisan fragmentation across districts in legislative
elections.

For instance, a legislative contest without a quality challenger(s) in a given
district results in a .313 drop in the degree of electoral partisan fragmentation compared
to a legislative election in a district that has a quality challenger(s) running for legislative
seats in the pooled analysis of 1996 and 2000 legislative elections. This ﬁnding is
statistically signiﬁcant at the .01 level for a two-tailed t test. Whether a district is urban or
rural also makes a difference in the level of electoral multipartism in legislative elections.
As expected, the urban districts tended to have .286 degree higher electoral partisan
competition for legislative seats than rural districts in the 1996 legislative election. This
ﬁnding is statistically signiﬁcant at the .10 level for a two-tailed t test.

As mentioned above, the legislative electoral district-level presidential election
returns were used to measure the local/district partisan characteristics for the 1988
legislative election, since there were no prior local elections of any kind. This kind of
proxy measure may not be adequate in identifying local/district-level partisan
fragmentation that would serve to explain district-level electoral multipartism in
legislative elections, since the electoral competition takes place at the national level. One
of the main arguments of this study has been that sub-national and sub-provincial local
elections are more meaningful in explaining local or district-level electoral multipartism

in legislative elections. In order to understand the degree of electoral partisan

71

 

fragmentation across districts in legislative elections, therefore, one must understand
electoral partisan competition patterns in ‘sub-national or sub-provincial elections.’

While the fourth column of Table 2-10 focuses on legislative elections where
there had been some type of prior sub-national local election (provincial council elections
for the 1992 legislative election), the third column deals mainly with two legislative
district-level local elections (1996 and 2000). Two models conﬁrm that the degree of
electoral partisan competition in sub-provincial local elections (provincial councils and
mayoral elections) has a strong positive impact on the level of electoral multipartism at
the legislative district level in legislative elections.

The estimated regression coefﬁcient of .232 for the district partisan characteristics
in the 1996 and 2000 legislative elections (.244 for the 1992, 1996, and 2000 legislative
elections) suggests that an increase in the district partisan characteristics from 1 to 2,
holding other factors constant, would result in an increase of approximately one quarter
of an effective number of electoral parties in the legislative elections under consideration.
These ﬁndings are statistically signiﬁcant at the .01 level for two-tailed t tests.

The above discussion suggests that the local/district partisan characteristics from
the level of electoral competition in sub-national and sub-provincial local elections
becomes critical for understanding the district-level electoral partisan fragmentation in
legislative elections. The local/district partisan characteristics are statistically signiﬁcant
in explaining the degree of electoral multipartism across districts in legislative elections

while controlling for other potentially confounding factors at the district level.'3

 

‘3 However, one control variable remains to be considered. This is the degree of electoral partisan
fragmentation in the previous legislative elections. Studies on the constituency- level electoral competition
for US state legislative seats suggest the necessity of including the “electoral multipartism in previous
elections” (Gierzynski and Breaux 1991; Hogan 2001). However, Achen (2001) warns about the danger of

72

Table 2-11 District - Level Presidential Partisan Strength and Legislative Electoral

Multipartism across Districts

 

 

1988/
. . a 1992/
Electrons/Variables 1988 1992 l 996 2000 1 99 6 /
2000
.123 .140* .072 .041 .122**
urban/Rural (.089) (.083) (.162) (.091) (.057)
.188** 4217** 4001 .051 .054
Open seat Or NO‘ (.087) (.074) (.124) (.103) (.056)
. .132 4011 .315** .262*** .217***
Challenger Qual“y (.105) (.081) (.1 17) (.071) (.051)
Presidential Partisan -.3.327*** -1.625*** -2.223*** -1.601*** -2.19l***
Snengdr (210) (184) (266) (203) (119)
. .426***
1988 Electron (.059)
. .289***
1992 Electron (.052)
. .552***
1996 Electron (.061)
Constant 4.571*** 3.639*** 4.158*** 3.290*** 3.563***
(153) (146) (237) (157) (096)
R? .5914 .3371 .2594 .2876 .3815
11 197 207 231 201 836

 

* signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; regression coefﬁcients are unstandarized
ones; robust standard error of the estimates is in the parentheses; ‘ The electoral returns for the 1997
Presidential elections were used as a proxy since the next presidential election will be in 2002.

The analyses so far have focused on the importance of the distinction between

local/district partisan characteristics in affecting the degree of electoral multipartism at

 

using a lagged dependent variable as one of the explanatory variables. They show that controlling for the
effect of previous legislative electoral multipartism does not change the importance of local/district partisan

characteristics.

73

the district level in legislative elections. The variable was measured through the local or
district-level electoral competition (gubernatorial or municipal elections) since these local
elections represent local or district-level partisan electoral characteristics. However, the
preliminary analysis found that different patterns of electoral competition exist along
regional lines regardless of election types. This indicates that the shape of electoral
partisan competition in presidential elections that are measured at the legislative district
level, is similar to that of the local/district partisan characteristics of prominent local
elections.

Table 2-11 shows how presidential partisan strength affects the degree of electoral
multitpartism at the district level in legislative elections, controlling for other district-
level variables. The main independent variable under review has been measured
somewhat differently than has been the case with previous research. District-level
presidential partisan strength is measured through the ﬁrst party’s vote share in'the last
(or upcoming) presidential elections in a given district. Thus, the higher the presidential
partisan strength, the lower the degree of electoral partisan fragmentation in presidential
elections in the district. Or, the lower the level of partisan presidential strength, the higher
the level of electoral competition. It is expected that partisan presidential strength will
have a strong negative impact on the effective number of electoral parties across districts
in legislative elections.l4

As expected, the variable of presidential partisan strength shows a strong
inﬂuence in the negative direction. The estimated regression coefﬁcients for presidential
partisan strength are signiﬁcant at the .01 levels for a two-tailed t test. For instance, the
estimated coefﬁcient of —3.327 for the 1988 legislative election indicates that, on average,

an increase of presidential partisan strength from 1 to 2 would, other factors remaining

 

1’ The use of a traditional measure for the degree of electoral multipartism following the Laakso-Taagepara
Index is not likely to alter the main argument of this analysis.

74

 

constant, results in an decrease of 3.327 of an effective number of electoral parties in the
legislative election of 1988. It is also important to note that the variable of presidential
partisan strength remains very stable across elections.

As for the control variables, the variables of open seat and quality challenger(s)
are statistically signiﬁcant and show the same trends as in the previous models. The
presence of quality challenger(s) is likely to decrease the level of electoral competition.
In addition, the presence of incumbent legislator(s) tends to decrease the level of electoral
partisan fragmentation across districts in legislative elections. That is, a legislative
contest without a quality challenger(s) in a given district results in a .315 drop in the
degree of electoral partisan competition for legislative seats, compared to a legislative
election in a district that had a quality challenger(s) in the 1996 legislative election.
Additionally, the presence of an incumbent(s) results in a decrease of .217 in the level of
electoral competition for legislative seats at the district level, compared to a legislative
contest without reelection-seeking incumbent(s) in the 1992 legislative election.’5

This chapter has sought to understand electoral partisan fragmentation in
legislative elections at the district level under the SMSP electoral system in a
‘decentralized’ unitary context - the case of Korea. It has suggested that local/district
partisan characteristics signiﬁcantly inﬂuence the degree of electoral partisan
fragmentation across districts in legislative elections. As in federalist systems, provincial
and sub-provincial elections have a profound impact on district-level electoral partisan
competition for legislative seats, especially in a decentralized unitary system in which
political powers exists at multiple levels and social/political diversity follows regional

lines.

 

15 As in the previous models emphasizing local elections, the variable of previous electoral multipartism is
also introduced. The inclusion of this previous level of electoral competition does not change the main
argument of the study. See Appendix.

75

CHAPTER THREE

MACRO ANALYSIS OF THE CANADIAN CASE

Introduction

This chapter is a macro-level analysis of the Canadian case. The main assumption
behind incorporating sub-national elections into our exploration of the degree of electoral
partisan fragmentation across districts in legislative elections is based on the nature of
federalist institutions. The particular structure of a country may foster sub-national
diversity, leading to a variety of different types of electoral party systems across the
nation. Canada is an important example for our purposes insofar as it has adopted the
SMSP electoral system to elect members to its national legislature and has a long history
of federalism. Furthermore, it has been suggested that Canadian provinces often feature
strikingly different national and provincial party systems (Gaines 1999; Palfrey 1989;
Taagepara and Grofman 1985) as in Korea. Thus, incorporation of the Canadian case
contributes to our understanding of electoral and party systems from a comparative
perspective.

This Canadian case study utilizes the 37th federal general elections held in 2000
for the House of Commons. The provincial elections of legislative members to provincial
legislatures will be used to identify the local/district partisan characteristics. While
Koreans directly elect representatives to provincial political ofﬁces (governors),
Canadians indirectly choose their provincial premiers by electing provincial legislative
members as with parliamentary systems. Both gubernatorial elections in Korea and

provincial legislative elections in Canada represent different local partisan characteristics

76

across the country and inﬂuence the degree of electoral competition in legislative
elections across “ridings”l in any given region.

This chapter ﬁrst provides a preliminary analysis of party system developments at
both national and sub-national levels. Then, it proceeds to the analysis of the
determinants of the degree of electoral partisan fragmentation at the district level with an
emphasis on the local/district partisan characteristics, while controlling for other district-

level confounding factors.

Degree of Electoral Multipartism at National and Sub-national Levels

Table 3-1 Division of Seats and Vote Shares at the National Level in Canada

 

 

Elections/ 1997 Election 2000 Election

Parties Seats Votes Seats Votes
Bloc Quebecois 44 10.7% 38 10.7%
Reform Party/ 60 19.4% 66 25.5%
Canadian Alliance*

Liberals 155 38.5% 172 40.8%
New Democratic Party 21 11.0% 13 8.5%
Progressive Conservative 20 18.8% 12 12.2%
Total 300 98.4% 301 97.7%

 

Note: * the Canadian Alliance was “Reform Party” in 1997; the total number of seats is 301 for both
elections.

The Canadian case has been considered to be an exception to the conventional

understanding of the relationship between electoral and party system development

 

l“Riding” is the Canadian term used for district or constituency. The term will be used in this chapter
unless otherwise indicated.
2 N = (X p,- 2)’l (When P is the proportion of popular votes for party (or candidate) i)

77

(Gaines 1999; Palfrey 1989; Taagepara and Shugart 1985, 1989). Contrary to traditional,
institutionalist expectations, the degree of electoral multipartism at the national level has
resulted in a multiparty instead of a two-party system. The effective numbers of electoral
parties at the national level are 4.09 for the 1997 election and 3.77 for the 2000 election,
following the Laakso-Taagepara Index.2

Table 3-1 shows the division of seats and vote shares in recent parliamentary

__'

elections at the national level in Canada, indicating that there have been ﬁve major
parties involved. These ﬁve parties occupy all the legislative seats and more than 95% of

vote shares at the national level. These two elections are no exception of Canadian norms.

 

The average effective number of electoral parties at the national level in Canada from
1935 to 1993 was 3.1 (Gaines 1999), suggesting that there have been more than three

parties in the Canadian national political landscape for some time.

Table 3-2 Comparison of Electoral Multipartism at
National and Riding Levels

1997 Election 2000 Election
National Riding National Riding

 

 

4.09 2.90 3.77 2.68

 

Note: The riding level degree of electoral multipartism is an average of all
the local tidings following the formula: DN = (Z N, )/ d (When N is the
effective number of parties at district i, dis the total number of local tidings)
However, one ﬁnds a different picture when we introduce the average degree of
electoral competition across ridings, as shown in Table 3-2 above. The average levels of
electoral partisan fragmentation at the riding level in the 1997 and 2000 elections are

relatively closer to two-party electoral competitions across ridings than those at the

national level. These two elections are also no exception to Canadian political history.

78

The average degree of electoral competition at the riding levels in Canada has been
relatively closer to two-party systems.3 Therefore, it appears safe to say that the
institutional argument also works in the Canadian case at the riding, although not the
national level

The simultaneous comparison of the degree of electoral partisan fragmentation at

national and riding levels clearly shows that the appropriate unit of analysis is the local

 

ridings, in order to better understand the degree of electoral competition in the Canadian 1'
l

legislative elections under SMSP electoral systems. The level of electoral competition at F

the riding level is clearly different from that of the national level. While a relatively

multiparty system emerges at the national level, the degree of electoral competition at the
Ii

riding level seems to be relatively closer to a two-party competition.

As suggested above, the shape of party systems at the national level is an
aggregation of local-level party systems (Chibber and Kollman 1998). In this sense, the
relative two party electoral competitions across ridings in Canada appear to lead to
multiparty systems at the national level. Also, the patterns of electoral partisan
competition for legislative seats across local ridings, differs from riding to riding, and,
subsequently, province to province (as an aggregation of tidings), contributing in a
variety of ways to the formation of the party system at the national level.

The degrees of electoral multipartism of some tidings, or a region, are higher than
that of the riding average, and relatively closer to that of the national level. In these cases,
one sees multiparty electoral competition (high electoral competition). The levels of
electoral partisan fragmentation in other ridings or regions, on the other hand, are lower

than that of the national-level, and closer to the average competition at the riding level.

 

3 The average effective number of electoral parties at the riding levels from 18 legislative elections (1935 to
1993) is 2.54 (Gaines 1999).

79

We can see relative two party electoral competitions in those ridings (low electoral
competition). Thus, when we focus on the degree of electoral competition at sub-national
levels, we ﬁnd more dynamics in the levels of electoral multipartism across a nation (Cox
1997)

A clue emerges here for understanding the “Canadian exception” in terms of
electoral and party system developments. As mentioned earlier, the interaction between
federalism and party system development have generated a wide range of important
questions for comparative political research. Numerous research questions have been
raised concerning the impact that a federalist polity structure has in providing multiple
arenas for electoral competitions for political ofﬁces within a given country.

Since there are a variety of institutional contexts and electoral systems where
electoral partisan competition occurs, the possibility arises of differing patterns or
degrees of electoral competition in the various constituent elements of the federation. The
Canadian federalist system has historically seen local provincial governments maintain or
increase their autonomy in many areas of public policy (Shannan 1994).

In this sense, such scholars as Palfrey, Taagepara, and Groﬁnan have maintained
that Canada has bipartisan provincial politics, while it is a kind of multiparty system at
the national level. Palfrey (1989) represents the conventional view concerning the
Canadian exception by arguing that “The most well-known exception is Canada, where
provincial party systems are bipartisan, but not all provinces have the same two dominant
parties. This produces a patchwork national party system that is a conglomeration of

strong regional parties” (p. 69).

80

Table 3-3 Provincial-Level Comparison of Electoral Multipartism
in Canadian Legislative Elections

 

 

Province 1997 Election 2000 Election Average 1935-93*
Alberta 2.63 2.43 3.1
British Columbia 3.26 2.95 3.4
Manitoba 3.85 3.80 ~ 3.1
New Brunswick 3.55 3.28 2.4
New Found Land 3.05 2.94 2.2
Northwest Territories 3.61 3.12 **
Nova Scotia 3.60 3.51 2.4
Nunavut 1.93 **
Ontario 3.04 2.87 2.7
Prince Edward Island 2.70 2.64 2.1
Quebec 3.05 2.76 2.7
Saskatchewan 3.42 2.93 3.0
Yukon 4.44 3.45 **

 

Note: * provided by Gaines (1999); ** Not included in Gaines’ analysis. The provincial-level degree of
electoral competition is based on parties’ vote share at the province level.

Taagepara and Grofman (1985) agree with Palfrey, saying that “Canada, which

might appear another exception, with three major parties, has at the local level two-party

politics - the parties are simply not always the same two throughout the country” (p. 342).

Furthermore, Rae (1971) identiﬁes a possibility of regional parties surviving locally and

contributing to a national-level multiparty system. He indicates that: “Intense hostility

81

between overlapping regional, cultural, and linguistic groups produced a strong basis of
support for locally strong minority parties” (p. 94).

Therefore, with different (mostly) two parties competing, Canadian provinces
have two party electoral competitions, producing multipartism at the national level. The
very reason why Canada has a multiparty system at the national level is the provincial-
level two party competitions for legislative seats with alternating regional combinations
of the two competing parties. Conventional arguments simply do not stand up to
empirical scrutiny. For example, Gaines argue that “This claim about Canadian federal
elections, as a “stylized fact,” emphasizes the adjective at the expense of the noun—it is
scarcely a fact at all” (Gaines 1999: 842). With the national legislative elections from
1935 to 1993 in Canada, the degree of electoral competition, using the data on
‘provincial-level election returns’, clearly indicate that provincial-level electoral
multipartism is consistently above 2 and almost always above 3. With exactly two parties
in an electoral competition, the degree of electoral partisan fragmentation cannot exceed
2. However: “There are only 9 cases out of 177 (roughly 5%) of a province having an N
of 2 or less” (p. 842).

The elections of 1997 and 2000 were no exception to the trend of multiparty
electoral competition at the province level in Canada. The average degree of electoral
competition at the provincial level for the 1997 election was 3.35 and 2.97 for the 2000
election. This suggests that more than two parties competed at the provincial level for
seats in Canada’s national legislature.

We can ﬁnd dynamic electoral competition across provinces as shown in Table

3-3. Some provinces show relatively higher degrees of electoral partisan fragmentation

82

than the riding average, closer to national-level multipartism. A relatively lower level of
electoral multipartism emerges than with the national-level party system and closer to the

riding-level average in electoral competition in some provinces.

Table 3-4 Riding-Level Comparison of Electoral Multipartism
in Canadian Legislative Elections

 

 

Province 1997 Election 2000 Election
Alberta 2.54 (26) 2.31 (26)
British Columbia 3.06 (34) 2.78 (34)
Manitoba 3.24 (14) 3.05 (14)
New Brunswick 3.01 (10) 2.85 (10)
New Found Land 2.42 (7) 2.55 (7)
Northwest Territories 3.43 (2) 3.12 (1)
Nova Scotia 3.36 (11) 3.l4(11)
Nunavut * 1.93 (1)
Ontario 2.88 (103) 2.72 (103)
Prince Edward Island 2.6] (4) 2.59 (4)
Quebec 2.75 (75) 2.55 (75)
Saskatchewan 3.25 (14) 2.74 (14)
Yukon 4.48 (l) 3.45 (1)
National Average 2.90 (301) 2.68 (301)

 

Note: * the province of Nunavut was included in the Northwest Territories in the 1997
election; () are the number of tidings allocated in the province; The riding-level degree of
electoral competition is an average degree of electoral multipartism at the riding level in
each province using formula: Dp = (Z N, )/ d (When N is the effective number of parties
at district i, dis the total number of local ridings in a province)

83

The provincial-level degree of electoral multipartism is an aggregation of riding-
level electoral competition across a province. A riding is a basic unit of electoral
competition from which a national-level party system is formed under the SMSP electoral
system. Canadian ridings have historically had more than two parties competing for
legislative seats. According to Gaines (1999), “Most of these elections do not seem to ﬁt
the generalization that Canada has bipartisan electoral districts” (pp. 843-855).
Speciﬁcally, at the riding level where “the bite of single member simple plurality
electoral systems is felt,” the degree of electoral competition spans the two-to-four-party
continuum and, furthermore, it hasn’t changed over time until 1993.

Table 3-4 shows the average degree of electoral competition at the riding level
across a province. The table clearly indicates the dynamics of electoral partisan
fragmentation at the riding level. The ridings of some provinces experience relatively
multiparty types of electoral competition like that of the national level, while relatively
two-party competition also emerges as reﬂected in the national riding average. There are
around three or four parties competing with each other for legislative seats in some
provinces. At the same time, two or three major parties emerge in other provinces. There
are 10 out of 25 provinces where the degree of electoral competition is closer to the
national riding average, but quite distant from the national-level partisan fragmentation.
The remaining 15 provincial cases are closer to the national level party system and distant
from that of the national riding average.

We ﬁnd evidence here that is contrary to conventional wisdom. Canadian party
systems have not been two-party competitions at the provincial level throughout history.

There have also been more than two parties competing ‘at the riding level’ in Canadian

84

 

 

parliamentary elections. Divergent degrees of electoral competition across tidings and
provinces following regional lines have emerged. Some ridings and provinces are
relatively closer to two-party systems (national riding average) whereas others are closer
to the national-level degree of electoral competition (multiparty systems). Thus, the
national-level party system in Canada is an aggregation (or mixture) of local-level (riding
and provincial) electoral partisan fragmentations.

Furthermore, when one reviews electoral returns at the riding and provincial
levels, it seems that there have been different combinations of parties competing at the
provincial level, indicating that there have existed ‘regionally strong but nationally weak
parties’ in Canada. There have been a variety of political parties in Canada. But, the
liberals (Liberal Party of Canada) and conservatives (Progressive Conservative Party of
Canada) are the only two ever to have formed the national government. Overall,
Canadian parties in Canada have been strong in certain regions and weak in others
(Arseneau 1994).

The “minor” party with the greatest longevity has been the NDP (New
Democratic Party). Its national election performance has always been best in the west, in
which it has formed several provincial governments, having only moderate strength in
Ontario, dismal in Quebec, and generally bad on the Atlantic Canada.4 According to
Morley (1994), the transition from a party with a highly centrist ideology (majoritarian
approach to government) to a party with a commitment to protect the inﬂuence of

provincial governments, explains the relative lack of success of the NDP at the national

 

4 Provinces of Newfoundland, Prince Edward Island, New Brunswick and Nova Scotia

85

level when compared with its success in forming provincial governments in some

provinces.

Table 3-5 Regional Bases of Canadian Parties (Seats)
Parties Atlantic (32) Westem* (88) Ontario (103) Quebec (75) Party’s

 

 

Total Seats
BQ - - - 38 38
CA - 64 2 - 66
Liberal 19 14 100 36 169
NDP 4 8 1 - 13

PC 9 2 - l 12

 

Note: * Provinces of Alberta, British Columbia, Manitoba and Saskatchewan.

Table 3-6 Regional Bases of Canadian Parties (Vote Shares)

 

 

Parties Atlantic Western Ontario Quebec
BQ - - - 39.9%
CA - 49.9% 23.6% -
Liberal 40.7% 25.3% 51.5% 44.2%
NDP 16.6% 12.3% 14.4% -

PC 31.3% 9.97% - 5.6%

 

First-past-the-post electoral systems generally reward the winning party and
penalize the second and especially third place parties. Yet, the same electoral systems can
work to a third party’s advantage, if the party’s political support is regionally (or

geographically) concentrated. Pluralistic electoral systems can also over represent

86

regional parties. This is true in the Korean as well as the Canadian case.5 The CA
Canadian Reform Conservative Alliance) has thrived in the west, the base of 64 out of a
total of 66 seats (50% of vote shares) in the 2000 general elections.

Table 3-5 and 3-6 show the division of seats and vote shares across Canada.
While the CA is dominant in the Western provinces, it competes with the Liberals and the
NDP for legislative seats in the region. Even though the CA is not on the ballot in
Atlantic Canada, it is the main contender with the NDP against the Liberals in the
province of Ontario. There emerge three major parties competing in Atlantic Canada
(Liberals, Conservatives and NDP). The province of Quebec has two major parties
competing—Bloc Quebecois and Liberals. The BQ is the classical regional party whose
seats are totally from one province, yet is the third largest party in the Canadian House of
Commons.

It appears to be clear, therefore, why we should focus on riding-level electoral
competition in Canadian legislative elections. The party system at the national level
shows a mixed picture since it includes different local (ridings and regions) electoral
partisan fragmentation across the country. Furthermore, the combination of parties
competing for legislative seats changes from province to province in Canada. Thus, a
relative two party system in some ridings (provinces), where a certain party (or two

parties) has relatively strong electoral support, blends with a multiparty system in other

 

5 According to Arseneau (1994), the rise of the Reform Party (until the 1997 general elections, changing to
CA before the 2000 general elections) as a new party of the right represents a case that makes use of the
federal system. First of all, it came from the province of Alberta, a prairie province with a history of
producing new political parties on the right of the political spectrum. The regional unresponsiveness of the
(5 cont’d)traditional parties led to the formation of a new regionally sensitive political party to run
provincially. Then, the party used regional dissatisfaction to reinforce its partisan appeal for radical change
of the political agenda at the national level, eschewing provincial politics. Therefore, Arseneau describes
the CA (former Reform Party) as a strange hybrid of regional and national components, drawing political
support from one element of the federalist system, but operating in another.

87

ridings (provinces), where three or four major parties are competing. Recent Canadian
parliamentary elections saw different levels of partisan strength and, subsequently,

different degrees of electoral competition along regional lines.

Determinants of Riding-Level Electoral Multipartism in Legislative Elections
Data and Method

Under a federalist institutional context political powers exist at multiple levels
and political/social diversity emerges along regional lines. As shown in the preliminary
analysis section of the chapter, different patterns of electoral partisan fragmentations
follow regional lines in Canada, and, therefore, it is highly likely that the local/riding
partisan characteristics affect the degree of electoral competition when electing the
members of the Canadian House of Commons across local tidings.

The dependent variable is the effective number of electoral parties at the riding
level in the 2000 general elections, based on the Laakso-Taagepara Index. The primary
independent variable under review is the local/riding partisan characteristics. The
variable is reﬂected in different levels of partisan strength and electoral partisan
fragmentation in a given riding. It is measured through the degree of electoral partisan
fragmentation at the riding level in provincial legislative elections.

Canadians elect provincial legislative members so that the majority party of the
provincial legislative body forms the provincial local government. While Koreans elect
their governor directly, Canadians elect their provincial prime minister (governor)
through provincial legislative members. Provincial elections are used to measure the

local/riding partisan characteristics in a given riding following the same formula used for

88

the dependent variable. The most recent provincial elections for the 2000 legislative
election are used for the local/riding partisan characteristics.

Table 3-7 Provincial Elections in Canada

 

Year of Provincial Electoral

 

Province Election Division Riding (301)
Alberta 1997 83 26
Chilﬁilibia ‘996 75 3"
Manitoba 1999 57 14
New Brunswick 1999 55 10
Newfoundland 1 999 48 7
Northwest Territories 1999 19 1
Nova Scotia 1999 52 11
Nunavut 1999 19 1
Ontario 1999 103 103
Prince Edward Island 1996 27 4
Quebec 1998 124 75
Saskatchewan 1 999 5 8 14
Yukon 2000 17 1

 

Table 3-7 summarizes the information about provincial legislative elections in
Canada. The electoral returns for electoral divisions in provincial elections in a province

are aggregated to match each riding allocation in the province. With the exception of the

 

6 According to Shugart and Carey (1992), legislative elections in parliamentary democracies (like United
Kingdom or Canada) are more party —— centered since they are to form a national government unlike
presidential systems. Thus, the aggregation of electoral returns of electoral divisions in a given riding
seems not to distort the analysis. However, a separate analysis of the province of Ontario will be performed
because the ridings in the province are the electoral divisions for provincial legislative elections.

89

province of Ontario, each riding consists of several electoral divisions across a province.
Therefore, the electoral returns of electoral divisions in each riding are aggregated
following partisan afﬁliations in each riding in accordance with the local/riding partisan
characteristics.7

Since the presence of control variables improves the ﬁt of the model, this analysis
introduces control variables. First, the presence (or absence) of incumbent legislative
members running for re-election in a given riding controls for the impact of a candidate
running for re-election on the level of electoral competition in the riding. Ridings where
incumbents ran were coded zero while open seats were coded one. The absence of
incumbents is expected to exert an increasing effect on the level of electoral competition
across tidings.

While the Korean case provides measures for challenger candidate quality8 and
the nature of districts, the Canadian case supplies more in-depth information about the
nature of ridings, called “district diversity.” Students of congressional elections have
offered theoretical reasons to expect heterogeneous districts to be more competitive than
homogeneous districts (Fiorina 1974). District diversity — the mix of social demographic
characteristics in a given congressional district — is a central trait that underlies the
degree of electoral competition for legislative seats in elections to the US. House.

There are clear theoretical reasons to believe that district diversity should have

discernible ramiﬁcations for electoral politics. Certain types of social, economic, and

 

7 According to Shugart and Carey (1992), legislative elections in parliamentary democracies (like United
Kingdom or Canada) are more party — centered since they are to form a national government unlike
presidential systems. Thus, the aggregation of electoral returns of electoral divisions in a given riding
seems not to distort the analysis. However, a separate analysis of the province of Ontario will be performed
because the tidings in the province are the electoral divisions for provincial legislative elections.

8 Given the party-centered electoral politics of parliamentary systems and the regionally different
combinations of parties competing for legislative seats in Canada, the presence ( or absence) of a quality
challenger seems not to exert an important impact on the degree of electoral competition.

90

demographic characteristics are positively (or negatively) related to support for parties of
the left or right. The set of traits associated with one or another political party seems to
have implications for the distribution of political opinion within a district. In turn, the
level of district diversity will inﬂuence the degree of electoral partisan fragmentation in
the district (Koetzle 1998).

For example, Fiorina’s (1974) examination of US. Senate elections demonstrates
in general that the diversity of a state does not affect the levels of electoral competition.
Nevertheless, state-level diversity, in particular, increased the competitiveness of senate
races in the South. Bond (1983), however, using the same measure for district diversity,
found no signiﬁcant relationship between constituency diversity and electoral
competition for House elections in the 1970’s. On the other hand, Hermson and Gimpel
(1995) found evidence that district diversity did have electoral consequences, but only in
Democratic House primaries. Therefore, empirical research so far has failed to ﬁnd
systematic evidence that the population diversity of congressional districts has a direct
impact on the level of electoral competition in US. congressional elections.9

The typical measure for riding diversity is the Sullivan index. Sullivan (1973)

offered a straightforward index of district diversity, which allows for the inclusion of key

 

9 Given these unexpected empirical ﬁndings, there have been a variety of efforts to revise district diversity
measures. For example, Koetzle (1998) suggests the use of “political diversity” that takes into account the
differential partisan impact of district diversity. With an assumption that American party support is closely
tied to certain demographic groups, he developed a measure of political diversity that is informed by the
relationship between certain demographic variables and electoral outcomes, and ﬁnds that diverse districts
experienced signiﬁcantly more electoral competition in the US. House elections held between 1962 and
1996. On the other hand, Bond and his colleagues (2001) argue that we should view constituency diversity
as a context in which electoral competition occurs rather than a direct cause of it. Since the population
characteristics that make a district homogeneous or heterogeneous are “prepolitical,” these prepolitical
characteristics are unlikely to have political consequences unless there are conditions present that make the
incumbent candidate vulnerable (unfavorable partisan forces and contrary voting records) and there is a
challenger with the skills and resources (campaign spending and prior elective experience) necessary to
exploit the incumbent’s vulnerabilities in a vigorous campaign (Bond 1983).

91

social demographic characteristics. The index is an expression of the probability of the
average proportion of unshared characteristics between any two members of a population.
Therefore, the closer to one the index, the more diverse the district is. The diversity index
developed and used in this chapter is based on ﬁve variables that are considered to tap
ethnic, educational, generational and economic dimensions of district characteristics
(Table 5-8 summarizes the variables used for the index in this study). The formula for
constructing a district diversity index is:
Aw=1—{Z Yz/V},
Where:
Aw = the index of diversity within a population

Y2 = the proportion of the population within a given category of each variable
used in the index

2 1 through P, where P is the total number of categories within all variables
V = the number of variables used in the index

Table 3-8 Construction of Riding Diversity Index

 

 

Variables Categories
Age Seniors (Older Than 65); Non Seniors
Education College Education; Less than College Education

Home Ownership Home Owner; Rent
Employment Employed; Not Employed

Language English; French; Other Languages

 

Note: All the statistics used for the construction of riding diversity index are based on the 1996
Canadian Census data.

The control variables should have a positive inﬂuence on the degree of electoral
competition across ridings in the 2000 election. Accordingly, the level of electoral

competition at the riding level is expressed in the following equation:

92

Y = O F(P, D, I),
Where Y is the effective number of electoral parties in a given riding in the 2000 election,
P is the local/riding partisan characteristics, D is the riding diversity combining the
variables of age, education, home ownership, employment and language. I is whether
there is an incumbent(s) and (9 is a stochastic factor. Function F is assumed to be linear

and the above function can be transformed into a regression line as follows:

Y =3] + 32X1+ B3X2 + B4X3 + B5X4+ u,

in which B's are unknown but ﬁxed parameters, X's 1 through 4 represent chosen
independent variables (local/riding partisan characteristics, riding diversity and open seat/or

not) and u represents the stochastic factor, O.

Analysis and Discussion

Table 3-9 Local Partisan Characteristics and Legislative Electoral Multipartism
Across Ridings in the 2000 Canadian Elections

 

 

Variables Unstandarized Coefﬁcients Robust Standard Error
Riding Diversity -.629 .415
Open Seat or Not 143 * .087
Local Partisan Characteristics . 180* * .074
Constant 2.552*** .236

 

RI: .039, N=299; * Signiﬁcant < .10, ** Signiﬁcant < .05, *** Signiﬁcant < .001

Table 3-9 shows how local partisan characteristics inﬂuence the level of electoral

competition across ridings in the 2000 Canadian parliamentary elections, while

93

controlling for potentially confounding factors at the riding level. It also shows the
relevancy of the local partisan characteristics from sub-national elections for
understanding the degree of electoral competition at the riding level in the Canadian case.
The level of electoral competition from the closest provincial election in a given riding is
signiﬁcantly inﬂuential in deciding the effective number of electoral parties in the riding
in the legislative election. Given the basic unit of electoral competition in legislative
election under the SMSP electoral system, it appears to be the case that the local/riding
district partisan characteristics decide the degree of electoral competition across ridings
in legislative elections.

Where the degree of local partisan characteristics is relatively high (more parties
in the last local elections) in a given riding, for example, the degree of electoral
competition in the riding tends to be higher (more parties) in the following legislative
election. In contrast, a riding in which the level of electoral partisan fragmentation is
relatively low (fewer parties), tends to have fewer effective electoral parties in the
subsequent legislative election. The estimated regression coefﬁcient for the variable is
signiﬁcant at the .05 level for a two- tailed t test. The estimated coefﬁcient of .180
indicates that, in an average riding, an increase in riding partisan characteristics from 1 to
2 would result in an increase of approximately .180 of an effective number of electoral
parties at the riding level, other factors remaining constant.

As mentioned earlier, previous empirical research of US congressional elections
has failed to ﬁnd systematic evidence that the population diversity of a congressional
district has a direct inﬂuence on the level of electoral competition (Bond 1983; Bond et al

2001). The sign on the regression coefﬁcient is not in the predicted direction, and the

94

models that have been developed have little explanatory power (Koetzle 1998). One can
also ﬁnd the same pattern from the Canadian case. The variable of riding diversity is not
statistically signiﬁcant either, and the relationship runs in the opposite direction from
what was expected. The estimated coefﬁcient indicates that the more diverse ridings are
likely to have a lower degree of electoral competition.

However, the presence of an incumbent(s) running for re-election is statistically
signiﬁcant at the level of 90% for a two-tailed t test and indicates that the absence of
incumbent candidates in a given riding tends to increase the level of electoral partisan
fragmentation in the riding. For instance, a legislative contest without incumbents in a
given riding results in a .143 increase in the effective number of electoral parties in the
riding.

Table 3-10 Riding Partisan Characteristics and Legislative Electoral Multipartism
Across Districts in the 2000 Canadian Elections (Ontario Only)

 

 

Variables ENPV Electoral Margin
Riding Diversity -3.572*** (.604) 95.408*** (24.708)
Open Seat or Not .159 (.260) 2.192 (10.623)
Riding Partisan .334** (.117) -8.494* (4.768)
Characteristics

Constant 3.203*** (.394) 15.142 (16.127)

R2 .282 .139

N 103 103

 

* Signiﬁcant < .10, ** Signiﬁcant < .05, *** Signiﬁcant < .001; () are robust standard error of the estimate.
As Korea’s mayoral elections are held at the same district boundary as the district

for legislative elections, this provides a good monitor of district-level partisan

95

characteristics; the elections of provincial legislative members in Ontario are held at ‘the
same riding boundary’ as the federal parliamentary elections. Therefore, this could be
said to reﬂect the riding-level dynamics of electoral competition or partisan
fragmentation in inﬂuencing legislative elections in a given riding.

The variable of riding partisan characteristics is statistically important, indicating
that a riding with relatively high level of electoral partisan fragmentation in local
elections tends to have a higher level of electoral competition in legislative elections,
while controlling for other factors. Speciﬁcally, the change from 1 to 2 in the effective
number of electoral parties across ridings in local elections results in a .334 increase in
the level of electoral multipartism in a legislative election.

While the presence of incumbent legislators has no impact on the degree of
electoral competition, riding population diversity becomes statistically signiﬁcant for
both models. Table 3-10 introduces two different dependent variables to measure the
same concept-the degree of electoral competition at the riding level. They are the
effective number of electoral parties and the electoral margin between the vote shares of
the top two ﬁnishers. Even though riding diversity is statistically important at 99% for
both models, they are different in terms of substantive implications.

The second column of the table indicates that a riding with relatively higher
population diversity tends to have a lower degree of electoral competition. This runs
directly counter to the theoretical expectation that district diversity leads to a higher
degree of electoral competition. However, if we use the effective number of electoral
parties as our dependent variable, a relatively high number of electoral parties (or

candidates) indicate a higher degree of electoral competition.

96

There have been a variety of indicators used to measure the degree of electoral
competition at the district level.'0 Some scholars used the electoral margin, while others
incumbents’ vote share. Also, the number of candidates in a given district is utilized as a
dependent variable. Under the SMSP electoral system, the nature of duality prevails in
voters’ choice and would-be third-party candidates could anticipate their candidacies
being ruined by strategic voting (Cox 1997; Duverger 1954). This indicates the
emergence of two main contenders in most electoral districts, “M+1” when M (district
magnitude) equals one.

According to Gaines (1999), “We should make use of appropriate measure of the
extent of party competition, according to whether their theories apply to constituency
competition or to national competition (p. 56). If our purpose is to know the degree of
electoral competition under the electoral system in which two candidates “by nature”
compete for a legislative seat, the answer appears to be the electoral margin between the
top two ﬁnishers (winner and runner-up). However, we ﬁnd the same patterns when we
use the electoral margin as our dependent variable. The variable of riding diversity runs

directly counter to our theoretical expectations.ll

 

Table 3-1 1 Different Measures of Degree of Electoral Competition

 

 

Research Measure Used for Electoral Competition
F iorina (1974) Electoral Marginality
Bond (1983) Pre-election Forecasts (Safe/Unsafe Seats)
Hermson/Gimpel (1995) Number of Candidates
Koetzle (1998) Electoral Marginality
Bond et al (2001) Incumbent’s Vote Share

 

” In this sense, we are supposed to develop and use different measures for the degree of electoral
competition depending on cases and purpose of research. Given the nature of multiparty competition at the
national level and regionally different combinations of parties competing at the province level, the total
number of candidates at the riding level appears to be a relevant measure in the Canadian context. Separate
analysis shows that population diversity tends to increase the number candidates across ridings controlling
for other factors.

97

Nonetheless, riding partisan characteristics show statistical and substantive
importance in explaining the degree of electoral competition in the form of electoral
margins. This point indicates the robustness of the riding partisan characteristics in
explaining the degree of electoral competition across ridings. A riding in which there is a
relatively high degree of electoral partisan fragmentation tends to have smaller electoral
margins indicating relatively high degrees of electoral competition in legislative election.
For example, the change from 1 to 2 in the effective number of electoral parties in a
provincial legislative election leads to an 8.5 percentage point decrease in the electoral
margin between the top two ﬁnishers.

This analysis seeks to apply the ﬁndings from the Korean case to the Canadian
case - the same electoral system but different institutional contexts. This chapter has
suggested that the local /riding partisan characteristics, to some extent, the degree of
electoral competition across ridings in the recent Canadian parliamentary elections.
Macro analyses of Canadian and Korean cases clearly indicate that we should focus on
the local/district-level partisan characteristics if we want to understand the degree of
electoral competition for legislative seats under the SMSP electoral system at the district
level. Local elections appear to be accurate reﬂections of the local or district-level

partisan characteristics.

98

CHAPTER FOUR

MICRO ANALYSIS OF THE KOREAN CASE

Introduction

The preceding macro-analysis of the Korean case sought to better understand the
sources of partisan ﬁagmentation at the district level in legislative elections with various
data at the district level. In fact, I have argued that the degree of electoral competition in
legislative elections must be understood at the local district level since the local/district
partisan characteristics of prominent local or district-level elections are of crucial
importance for determining the level of electoral partisan fragmentation across districts in
the recent Korean legislative elections under the SMSP electoral systems while
controlling for other district — level factors.

In addition, the macro analysis of the Korean case has found that the local/district
partisan characteristics across districts are variable following regional lines. Some
regions have relatively higher degree of electoral competitions, while other show
relatively low level of electoral partisan fragmentation. These kinds of variations in the
degree of electoral competition patterns across districts ultimately depend on the
distribution of electoral supports at the individual level. Therefore, the districts that have
relatively low degree of electoral competition pattern are those in which the electoral
supports at the individual level are concentrated to a relatively small number of parties.
However, those districts that have relatively higher levels of electoral competition

patterns tend to have more dispersed electoral supports at the individual level.

99

 

This chapter is speciﬁcally to identify whether there exists the concentration of
electoral supports at the individual level following regional lines across districts, making
the degree of electoral competition acrOss districts to be high or low. When the shape of a
party system within a democracy is built on the distribution of votes cast by individual
voters in a given election, it is important to seek as full an understanding as possible
concerning how individual voters decide whom to vote for in elections, and according to
which key factors. This will also contribute to our understanding of sub-national-level
analyses of the degree of electoral competition in legislative elections.

This chapter, therefore, represents a micro-level analysis of the Korean case
utilizing individual-level survey data related to the legislative elections under
consideration, in search of a more reﬁned understanding of the dynamics surrounding
recent parliamentary elections and sources of partisan fragmentation at the individual
level. This chapter also seeks to identify the existing political cleavages upon which the
shape of the Korean political, electoral, and party systems is based. In what follows,
therefore, necessary information is presented concerning data, measurement and
methodology. Next, we proceed to an analysis of the determinants of partisan choices in

Korean legislative elections.

Data and Measurement
This analysis utilizes the Korean Election Survey (KES) data for the three most

recent legislative elections (1992, 1996 and 2000).1 This includes individual post-election

 

’ There was no publicly accessible public opinion survey data for the 1988 legislative elections, which was
the ﬁrst parliamentary election after the restoration of procedural democratic requirements in 1987.

100

surveys of around 1200 Korean voters for each election, which help us to better

understand how Koreans make their political decisions.

Table 4-1 2000 Legislative Election

 

 

Partisan Choices KES Ofﬁcral Difference
Result

Grand National Party 43.4 38.96 4.44

New Millennium Democratic Party 32.8 35.87 -3.07

Coalition for Liberty and Democracy 6.1 9.84 -3.74

Democratic National Party 1.7 3.68

Minor Parties .6

Independents 7.5 9.39

No Answer 7.8

Total 100.0 97.74

 

Table 4-2 1996 Legislative Election

 

 

 

 

 

Partisan Choices KES Ofﬁcral Difference
Result
New Korea Party 42.8 34.52 8.28
National Congress for New Politics 24.9 25.30 -.04
Democratic Party 8.3 11.23
Coalition for Liberty and Democracy 14.8 16.17 -1.37
Minor Parties .2
Independents 8.9 1 1.85
Total 100.0 99.07
Table 4-3 1992 Legislative Election
Partisan Choices KES Officral Difference
Result
Democratic Liberal Party 39.0 38.49 .51
Democratic Party 33.6 29.17 4.43
Uniﬁcation National Party 14.8 17.37 -2.57
Party for New Politics .6 1.79
People’s Party 2.3 1.55
Independents 9.3 1 1.52
Minor Parties .4
Total 100.0 99. 89

 

101

Before proceeding to the analysis of individual voting behavior, however, it is
important to check the validity of the data by comparing the differences between
aggregate and individual levels. Tables 4-1 through 4-3 compare the proportion of vote
choices at both levels to illustrate how they diverge from each other. The tables show that
a 3.2 percentage point difference exists, on average, between aggregate and individual
levels. Among the elections under review, the 2000 election shows the largest average
difference (3.75 percentage points), as opposed to only 2.5 percentage points, on average,
for the 1992 election. This is especially important insofar as non-negligible differences
between aggregate and individual levels may cause serious problem in terms of validity.

Apparently, our data is not ﬂawed in this way. The 3.2 percentage point average
difference is not great.2 Thus, using this survey data is not likely to endanger the validity
of the research since the distribution of partisan preference at the individual level in each
legislative election is highly reﬂective of that of the aggregate level. However, those who
do not express a vote intention or those who indicate a preference for small parties or
independent candidates are not included in the following analyses.

There is probably no subject that has been as thoroughly studied by political
scientists as elections, in general, and voting behavior in particular. To identify the
sources of partisan preference at the individual level, we examine various sets of
variables. Speciﬁcally, these variables include demographic characteristics, party
identiﬁcation, political ideology, political leadership evaluation, policy issues, evaluation

of the economic situation, and certain country-speciﬁc factors. Table 4-4 shows the list of

 

2 For example, the difference between ofﬁcial electoral returns and American National Election Studies for
the presidential elections in 1996 is 7.64 percentage points on average. See the Canadian micro — level
analysis chapter for the comparison of KES, ANES and CBS in terms of the survey data validity.

102

variables across elections. Data for some variables are available for certain elections but

 

 

not others.

Table 4-4 Categorization of Variables across Elections in KES
Category Variable 1992 1996 2000
Demographics

Gender X X X
Age x x x
Education X X X
Income X X X
Size of x x
Residence
Regionalism
Place of Birth X X X
Regional _ _ x
Antipathy
Party Identiﬁcation
Partisanship X ' '
Partisan .. x x
Tendency I/O
Presidential Job Approval X X X
Political Ideology
R/L Self _ _ x
Placement
Social Order vs.
Individual - x -
Freedom
Retrospective Economic Voting
Personal X X '
National " X '
3 Party Merge (1990) X ' ’

 

Note: * Incumbent or Opposition parties.

103

Among the demographic variables, residence size (urban vs. rural) is especially
relevant to the Korean case. Korea has undergone rapid economic growth over the last 40
years. Unlike rural areas, urban areas are mixed with many people from all over the
country, resulting in weaker regional identities than in rural areas. There is also more
mobility among people who live in urban areas, and they tend to be younger and more
highly educated than their rural counterparts. Before the emergence of regionalist party
politics with the 1987 presidential election, urban areas had tended to vote for opposition
parties and rural areas for the incumbent party. This urban-rural distinction in terms of
electoral support for certain parties has long been an established voting pattern in Korea
(Cho 1998; Kim and Browne 2001; Kim and Khil 1988).

While urban residents tend to support opposition parties, they are not as likely to
vote as rural residents. According to Brady and his colleagues (1992), there is a greater
voter turnout in rural as opposed to urban areas. People from rural areas tend to be more
participatory because of the way in which rural residents are mobilized to participate
(Kim and Khil 1988). The relatively low voter turnout in urban areas reﬂects the
dissatisfaction of urban residents with the undemocratic politics of the pre-
democratization period,3 and, therefore, urban residents tend to support opposition parties
if they participate at all.

It is important to test to what extent urban residents are still likely to support an
opposition rather than an incumbent party in the legislative elections that have occurred
since democratization. There have been more than two opposition parties against the
incumbent parties in the recent Korean parliamentary elections. In this sense, given the
asymmetry in electoral supports for opposition and incumbent parties between urban and
rural areas, there tends to compete relatively more parties in urban areas than rural areas.

It is to know whether the size of residence makes difference in terms of the degree of

 

3 It is called as an “alienation hypothesis” that is directly opposite to the “mobilization hypothesis” (Brady,
Mo and R0 1992).

104

electoral competition across districts at the individual level. That is, the urban areas tend
to have more competing parties, whereas the rural areas are likely to have fewer parties.

Another important, Korea-speciﬁc variable is regionalism" or provincial
favoritism. Regions are of special political signiﬁcance since they represent convenient
units to which parties or politicians can appeal for electoral success. As the macro-
analysis of the Korean case suggests, the SMSP electoral systems render the geography
of electoral support especially consequential. Along with class, religion, and ethnicity,
regionalism has proven to be one of the most consistent inﬂuences on individual political
behavior (McAIlister and Studlar 1992).

As stated above, Korea has become well known for its stridently “region-based”
party systems and “regionalist politics” following democratization (Clark 1987; Kang and
Jaung 1999; Lee and Brunn 1996; Morris 1996). Its electoral politics after
democratization have resulted in struggles among regionally-based political factions or
parties. Regional concentrations of electoral support along regional lines within the
country were particularly pronounced in the 1987 and 1992 presidential elections and the
elections to the National Assembly in 1988 and 1992.

Since the 1987 presidential elections, most parties have had their own strong
regional bases. Generally speaking, a political party based in a speciﬁc region gets almost
all the seats allocated to the region in legislative elections. This is particularly evident in
Yungnarn and Honam regions. In fact, regionalism has become the most dominant

political cleavage line for framing political appeals and voter decisions in Korea (Kang

 

4 Why regionalism appears is beyond the scope of this research since we focus on the political
consequences (partisan preference) of regionalism. In explaining the rise (cause) of regionalism, analysts
have focused on excessive underdevelopment between regions, the biased recruitment of elites by military
regimes and the manipulation of regional sentiment by political elites. David Kang, “The Institutional
Foundations of Korean Politics,” in Understanding Korean Politics: An Introduction. Eds. Soong Hoom
Khil and Chung-in Moon (State University of New York Press, 2001).

105

and Jaung 1999). People from the Yungnam region have traditionally been very
supportive of local parties and people from the Honam region have been very supportive
of a regional party. Frequently, divisions between two particular regions have been the
main factors determining individual vote choices (Cho 1998).

While Regionalism has typically been understood as the geographic concentration
of electoral supports for a party at the aggregate level in terms of the electoral politics,
there have been some attempts to gauge its importance at the individual level. Surveys
have been taken, for example, asking respondents to identify their place, so as to better
understand how this makes a difference in terms of partisan political support. Place of
birth was the principle variable studied for the 1992 and 1996 elections. Since the politics
of regionalism generally involves at least two regions, however, it should be coded as two
different variables. Here, whether a respondent is from the Honam (Regionalism 2) or
Yungnam regions (Regionalism 1) will be used for our analysis of the 1992 and 1996
legislative elections. If someone is from one of these regions, the variable is coded one,
otherwise zero.

The survey in 2000, however, rather than asking respondents their places of birth,
asked them whether they had a certain antipathy towards people from certain regions.
Relying on development in the ﬁeld of social psychology, this survey focused on the
individual psychological sentiments, either attachment to or antipathy towards certain
regions, namely Yungnam and Honam, and the impact that this had on voting decisions.
The concentration of electoral supports along regional lines is based on the political
support of people from certain regions. The political support of those who are not from a
given region is more difﬁcult to explain.

Thus, it seems most appropriate to measure the impact of regionalism according

to individual regional sentiment or antipathy towards people from certain regions. Either

106

way, these variables had a profound inﬂuence on individual voting choices in recent
legislative elections. Our survey asked respondents how they would like to have someone
from the Yungnam (Regional Antipathy 2) or Honam regions (Regional Antipathy l) as
their business partner. If one would prefer not to have someone from a certain region as
one’s business partner, we can classify him or her as having a certain antipathy towards a
certain region. Clearly, the presence (or strength) of antipathy towards people from a
certain region has a profound inﬂuence on individual, partisan voting decisions.

Along with the variable of residence size, the regionalism — related variables are
to know the sources of political cleavage in general and of electoral partisan
fragmentation in particular at the individual level. It speciﬁcally tests the impact that the
nature of a district (urban vs. rural; concentration of electoral supports along regional
lines) has on individual partisan preference at the individual level. If there exists a
concentration of electoral supports at the individual level along regional lines, we can
understand why there exist different degrees of electoral competition following regional
lines in the Korean case. It is the extent to which the concentration of electoral supports
at the individual level along regional lines that determines the degree of electoral
competition at the district level in legislative elections. This point will clearly conﬁrm the
existence of the geography of elections in Korea at the individual level.

One of the factors that have traditionally inﬂuenced the choices of individual
voters choices is identiﬁcation with a particular political party. Partisan identiﬁcation is
an enduring and stable component of mass political behavior in presidential and
congressional elections in the US, for example. The relative strength of Korean political
parties, however, since the restoration of procedural democratic requirements in 1987 has

been much less stable than in the US. Parties tend to emerge, merge, and disappear from

107

one election to the next. There has been little consistency in terms of parties’ names and
histories in Korean politics.

Thus, it is difﬁcult to identify individual partisan identiﬁcation with certain
parties over time. While the 1992 survey did include partisan identiﬁcation with speciﬁc
party names that existed at the time of survey,5 here, we have used an alternative measure
for individual partisanship, asking respondents about their partisan tendency (or leaning)
towards incumbent vs. opposition parties. Rather than providing respondents with
speciﬁc party names (like Democrat or Republican in the US), we asked respondents to
identify whether their partisan views or leaning were relatively closer to either the current
incumbent or opposition party.6 Given the brief history of democratic politics and high
ﬂuidity of parties, this measure of partisan tendency seems appropriate for Korean
electoral politics (Cho 1988). Thus, it is expected that people who aligned themselves
closer to the incumbent or opposition party were likely to support that party in the
legislative elections.

Political ideology is an important factor in deciding individual voting choices. It
is used as an information shortcut (or heuristic device) to make sense of politics and
political decisions (Lau and Redlawsk 2001). Political ideology is measured with
different variables in this analysis. While the 1996 survey asked respondents to locate
themselves on a left-right continuum, the 2000 survey used a proxy measure to identify .
individual political ideology on a continuum of maintaining social order on one extreme

and individual freedom on the other. When one emphasizes maintaining social order over

 

5 According to the list of party names available at that time, this survey asked respondents to pick a party
that they liked. The variable was recoded into three dummy variables; where someone identiﬁed his/herself
as belonging to a certain party, the variable was is coded 1, otherwise 0.

6 There have been more than two opposition parties in Korean legislative body. Thus, the reference of
“opposition party” in the survey refers to all the parties when they are not the incumbent party at the time
of survey.

108

individual freedom he or she can be categorized as conservative, or leaning to the
political right of the spectrum. Giving priority to individual freedom, on the other hand, is
closer to the liberal ideology of the political left. This variable was chosen because of the
way in which it serves to distinguish ideological divisions among the electorate in terms
of partisan choices. It is expected that there exists a possibility of ideological divisions in
the electorate in terms of its partisan preference given the severe inﬂuence of regionalism
in legislative elections.

While the variables of party identiﬁcation and political ideology are
representative of long-term individual psychological preferences or attachments that
inﬂuence one’s political decision - makings, retrospective economic evaluations are
based on issues and relatively short-term factors, especially the economic situation at the
aggregate and individual levels. It also represents the most straightforward statement of
the “revisionist” position that regards partisanship as merely a cumulative tally of issues,
based on retrospective evaluations (Witheley 1988). Two different kinds of voting
behaviors have been identiﬁed that are based on economic considerations: “pocket book
voting” and “sociotropic voting.” Between these two types of economic voting, it is
collective national economic judgments rather than one’s personal economic situation
evaluation that mainly inﬂuence individual voting choices (Lewis — Beck 1985; Weisberg
1991).7

Lastly, among Korea-speciﬁc variables the “three party merge” was speciﬁc to
the 1992 legislative election alone. The 1992 election was held after the three main

parties that participated in the 1988 parliamentary elections merged into one party in

 

7 There is another type of economic voting, prospective voting. that is based on an individual’s prospective
evaluation of which party or candidate’s economic program is seen as more relevant to future economic
prosperity. However, the KES data doesn’t include items that measure prospective economic voting.

109

1990. Very heated political debates took place over whether the merger was politically
right or not, and this was seen as one of the main factors inﬂuencing the electoral
fortunes of the parties concerned. It is expected that individuals who supported the party
merger would be likely to support the “big” incumbent party. In contrast, voters who
criticized the merger were more likely to support the main opposition party in the 1992
legislative election.

Along with the substantive factors “mentioned above, this study also introduces
certain demographic control variables such as age, gender, level of education and income.
Level of education is categorized into elementary, middle, high school and college
educations. Income is respondents’ monthly income. These variables are relevant to the
traditional cleavage structures upon which a party system can be based. Depending on
age, gender, and social class (determined by both education and income level),
individuals are likely to vote for a certain party.

For example, older people and the poor tend to support the incumbent party, since
they tend to see the incumbent party’s legislative majority in the National Assembly as a
necessary condition for political stability and economic prosperity in Korea (Cho 1998).
In contrast, the well-educated and younger people are more likely to support opposition
parties rather than endorsing the incumbent party because they tend to be critical of the
sitting president and incumbent party. These demographic variables provide possible
general political cleavages upon which a party system may be based and sources of
partisan fragmentation at the individual level.

Finally, the dependent variable for this analysis is the individual choice of a

certain party in the legislative elections under consideration. The major goal of this

110

 

chapter is to better understand how an individual voter makes his or her partisan choice.
Thus, a logistic regression analysis using a dichotomous dependent variable (1=choice of
a certain party; 0: no choice of the party) is employed to determine the inﬂuence of the
set of independent variables. Separate logistic regressions will be performed for each
major party participating in an election (see Table 4-5).

In the case of a binary dependent variable, the effects of any single independent
variable are best evaluated with respect to a particular point on the probability
distribution for a certain voting choice. I generally examine the magnitude of change in
the predicted probability of choosing a particular party across the range of one variable,
with an estimation of maximum likelihood, other variables being held constant at typical
values (usually the mean). The associated change in terms of the predicted probabilities is
calculated from a baseline model using the Tomz, Wittenberg, and King’s (1999)
CLARIF Y program.

Speciﬁcally, the interval and ordinal-level variables in the model are held at the
mean and the binary dichotomous variables are held at the value of 0 (the absence of
certain characteristics). For the interval (ordinal)-level variables, the change reﬂects an
increase of one standard deviation from the mean value of the variable. The associated
changes in terms of the predicted probabilities range from one standard deviation below
the mean to one standard deviation above the mean. However, for the dichotomous
variables, the values reported reﬂect the associated changes in its predicted probabilities
from one category (absence of certain characteristics) to another category (presence of

certain characteristics)8 (King, Tomz and Wittenberg 2000).

 

8 From female (0) to male (1), for instance.

11]

 

The logistics regression coefﬁcients provide direction for the relationship as well
as the statistical signiﬁcance of the relationship. But they are not directly interpretable,
making it necessary to assess how each variable changes the probability of a partisan
choice, while holding all other factors constant at the their mean values. It is necessary to
have a procedure that enables us to understand the relative impact of the variety of
independent variables. Thus, I will take into account changes in the predicted probability
(either positive or negative to the dependent variable) when there is a corresponding
change in the selected independent variable that is statistically signiﬁcant. Therefore, the

individual choice of a certain party can be expressed in the following equation:

Y =31 + Bzxr + B3X2 + B4X3 + B5X4 + B6X5 + B7X6,

in which Y represents the individual choice of a certain party or not and B's are unknown
but with ﬁxed parameters. X's 1 through 6 represent the chosen set of independent
variables: demographic characteristics, regionalist attitudes, party identiﬁcation,
president job approval, political ideology, and retrospective evaluation based on
economic issues.

At least three major parties have been key players in Korea’s recent legislative
elections. This indicates that voters have been faced with multiple choices, not binary
choices. When there are only two parties, however, for the electorate to choose from,
“The electoral space can be represented as a single dimension since the probability of
voting for one party is directly related to the probability of voting for the other party”

(Whitten and Palmer 1996: pp. 232-233). However, if more than two parties, exist, as in

112

multiparty systems, electoral space becomes more complex because it is multi-
dimensional.

Therefore, statistical models need to capture the multifaceted process of choosing
one party among multiple available parties. A statistical model should allow us to make
inferences about the electoral dynamics of a multiparty system. The empirical results
should help us to assess what factors in the model are most inﬂuential for determining the
distribution of electoral support between particular parties at the individual level. In this
sense, a less restrictive model, multinomial logistic regressions, will incorporate the
variables that affect voters’ probability of choosing among the parties competing in any
given election (Aldrich, Magaloni and Zechmeister 2001; Seligson and Carrion 2002;
Whitten and Palmer 1996). Thus, I will perform a logistic regression analysis ﬁrst and

then use a multinomial logistic regression to arrive at more in-depth explanations.

Analysis and Discussion

Table 4-5 indicates how individual voters arrive at their partisan preferences in
the 1992 parliamentary elections. Generally speaking, Korean voters follow their partisan
loyalty to a certain party. For example, the associated change in predicted probabilities
is .535 for those who identify themselves closer to the incumbent party, indicating that
the probability of choosing the incumbent party is more than ﬁfty percentage points
higher when an individual identiﬁes him or herself as a supporter of the incumbent party,
while other factors remaining constant. This becomes especially evident when we look at
the variable of partisan tendency (the relative closeness on a continuum to either the

incumbent or opposition parties at both extremes).

113

 

Table 4-5 Determinants of Partisan Choice in 1992 Legislative Election

 

 

Election/ Incumbent Party 1St Opposition Party 2"d Opposition Party
Variable Probability Change Probability Change Probability Change
Age .10 (.17) - -.03* (.02) -.15 .02 (.02) -
Gender .01 (.01) - .09 (.34) - .03 (.40) -
Education -.42 (.32) - -.13 (.21) - .33*(.19) .15
Income -.24***(.09) -.132 .ll (.11) - .09 (.13) -
Big Cities .87**(.36) .174 1.27 ***(.35) .28 -l.15***(.40) -. 19
Counties 1.07**(.47) .223 -.20 (.46) -.95*(.49) -.l6
Regionalism 1" .44 (.35) - -l .29***(.39) -.15 -.99**(.39) -.17
Regionalism 2b .08 (.43) - 1.65***(.37) .37 -2.06***(.56) -.26
Incumbent 2.41 ***(.57) .535 -l .94***(.62) -. l 8 -.32 (.73) -
1” Opposition -.57 (.61) - 2.10***(.47) .48 -l .73***(.57) -.24
2nd Opposition .14 (.63) - -l .07*(.56) -.13 l.84***(.57) .43
Partisan
Tendency -.76***(.28) -.183 .15 (.26) - .73**(.34) .23
President
Approval -.47*(.28) -.10 .35 (.25) - .40 (.30) -
Personal .10 (.17) - -.5***(.17) -.15 .03 (.20) -
Economy
3 Party Merge -.24 (.35) - -.34 (.48) - .05 (.48) -
-5.53
Constant 1.94 (1.48) .18 (1.66)
***(1.97)
Pseudo R2 .46 .51 .37
Wald c1112 170.22*** 145.25*** 112.08***
Log Likelihood -l71.10 -162.08 -13 1.67
% Correction 86.1% 85.7% 84.9%
Base Probability .307 .203 .062
N 490 490 490

 

Note: * signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01;Robust standard error of the
estimate is in the parentheses; a from Yungnam Provinces; b from Honam provinces.

People who are closer to the incumbent party are more likely to vote for that party,

whereas those who consider themselves as to be closer to the opposition carnp tend to

114

 

support the opposition parties. The associated change in the predicted probabilities is -.18,
suggesting the odds of voting for the incumbent party is almost twenty percentage points
higher when some feels closer to the party, while other factors are held constant.

As expected, the variable of regionalism appears to be inﬂuential in determining
the electoral fortunes of a party at the individual level. The most popular opposition party
in the Honam region and least preferred party in the Yungnam region serves to illustrate
the stark divisions of regionalist politics at the individual level. Speciﬁcally, when the
associated change in predicted probabilities is -.15, this suggests that the probability of
voting for the I” opposition party is about ﬁfteen percentage points lower when an
individual is from the Yungnam region. However, it is almost forty percentage points
higher when one is from the Honam region while holding other factors at their means.

The 2nd most popular opposition party, with no strong regional base like other
parties, was not preferred by those who are from either the Yungnam or the Honam
regions. When one identiﬁes oneself as from either of these two regions, the probability
of choosing the 2nd most popular opposition party declines by seventeen or twenty-six
percentage points while controlling for other factors. Thus, the 2“‘1 opposition party gets
its electoral support from non-Yungnam and non-Honam regions. However, the 2Ind
opposition party does seem to do well in the mid—sized cities. The table shows that the
associated change in the predicted probability of voting for that party is -.19 and -.16,
indicating that the probability of choosing the 2"d opposition party is about twenty
percentage points lower when an individual voter resides in either a big cities or in the

countryside.

115

 

Other demographic variables such as income and age are also important in the
decision of which political party to support. The main opposition party, for example, has
the most electoral support from young people (Cho 1998). Table 4-5 conﬁrms this point.
More speciﬁcally, for example, when the associated change in predicted probabilities is -
.15, this suggests that the probability of choosing the 1"t opposition party is ﬁfteen
percentage points lower when an individual gets old. This indicates very clearly that
those who are relatively younger are more likely to support the lSt opposition party. This
kind of generational division in partisan supports seems to continue since the 1992
election.

In addition to the generational divisions, there are also divisions between the
wealthy and the poor. For example, when the associated change in predicted probabilities
is -.132, this indicates that the probability of choosing the incumbent party is about
thirteen percentage points lower when an individual voter has a low income, which
means that wealthy people are more likely to support the incumbent party. The division
of political support between generations and income levels seems to have continued in
subsequent elections. While the older and the wealthy are likely to support the incumbent
party, the younger and poor tend to support the opposition parties in Korean politics.

We have seen how the electorate chooses its partisan preference in legislative
elections according to certain factors. But we do not know the relative impact of a series
of independent variables between particular parties. Logistic regression appears to be
inadequate for describing the effects of independent variables in changing individual

partisan supports from the incumbent party to the 1‘” or 2nd opposition parties.

116

Table 4-6 Multinomial Logistic Estimation of 1992 Election

 

 

Election/ ls‘Opposition/ 2"d Opposition/ lSt Opposition/
Variable Incumbent Incumbent 2"d Opposition
Age .02 (.018) .02 (.019) -.O38* (.02)
Gender -.68 (.442) -.57 (.48) -.10 (.49)
Education .057 (.26) .36 (.26) -.30 (.30)
Income .21 (.13) .26* (.14) -.054 (.14)
Big Cities -.68 (.50) -1.26** (.49) 194*" (.52)
Counties .36 (.57) -1.51** (.61) 1.15“ (.65)
Regionalismla -l .40*** (.49) -l.ll ** (.49) .29 (.55)
Regionalism 2b -75 (.48) -193 *** (.72) -2.67*** (.68)
Incumbent -.036 (.52) -l .82“ (.80) .095 (.62)
agggym 1.19 *** (.70) 1.25 (.83) 2.44 *** (.66)
figfgé’rssmm 1.01 (.77) - 1.02 (.78) -204 *** (.65)
Partisan Tendency .65 ** (.33) 1.15*** (.34) -.49 (.37)
Pres’dem Approval .62* (.33) .63* (.36) .015 (.37)
Personal Economy -.47** (.23) -.085 (.24) .091 (.62)

3 Party Merge .21** (.08) .33 (.30) .32 (.28)
Constant -4.07* (2.33) -l2.6l*** (2.68) 8.54***
Pseudo R2 .67

Chi Square 478.33***

Log Likelihood 413.64

N 433

 

Note: * signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Standard error of the estimate
is in the parentheses; a from Yungnam Provinces; b from Honam Provinces

117

 

Table 4-6 speciﬁcally shows which factor becomes more relevant to the
individual choice of a certain party amongst competing parties in the 1992 legislative
elections. It provides us with a more in-depth explanation of the impact of independent
variables on individual partisan preferences. Thus, the multinomial model allows us to
evaluate the different calculi made by individual voters supporting different parties.

The respondents who support the incumbent party comprise the baseline category
of the dependent variable except for the last column where the category of choosing the
2nd opposition party is used as a reference group for comparing individual partisan
choices between the two main opposition parties. The task here is to identify what set of
predictors make it more likely for a respondent to support the lSt or “2nd opposition party
rather than the incumbent party. We have two sets of independent variables:
ordinal/interval and nominal-level measures. In the case of ordinal/interval variables, the
interpretation of coefﬁcients is straightforward, representing the relative change in the
odds of a person falling into a certain category of the dependent variable relative to the
baseline category for each incremental unit of the independent variable. However, in the
case of nominal variables, the multinomial logistic regression coefﬁcient indicates the
change in the odds of falling into a given category relative to a reference category of the
dependent variable (Seligson and Carrion 2002).

The ﬁndings from a multinomial logistic analysis, in general, conﬁrm those from
a logistic regression analysis. This indicates how regionalism changes the combination of
parties competing with each other along regional lines. For people from Yungnam region,
for example, the critical issue seems to be electoral competition between the incumbent

party and the 2nd opposition party. The main opposition party is the least preferred party

118

 

for people from that region. In this situation, the newly formed 2"d opposition party
becomes the main contender against the incumbent party. Speciﬁcally, where the
Democratic Party (DP) was the least preferred and the incumbent Democratic Liberal
Party (DLP) was also less popular than in other regions, the newly-created Uniﬁcation
National Party (UNP) became a viable contender against the DLP, indicating that the rise
of a third party was also associated with political cleavage along regional lines (Kang
1998). r
As it has been identiﬁed in the macro-level analysis of the Korean case, different I

degrees of electoral competition have existed along regional lines. While some regions

 

have a relatively high level of partisan fragmentation, other regions show a relatively low
degree of partisan fragmentation. The macro — level ﬁndings are conﬁrmed by the micro
— level analysis. Speciﬁcally, where a given party is relatively strong, as in the Honam
and Yungnam regions, there appears to be a relatively low level of electoral competition
since people of either region tend to be very supportive of the one party that is based in
that region. The micro-level analysis makes it clear that the degree of electoral
competition across districts in legislative elections is based on the distribution of electoral
supports along regional lines at the individual level.

The generational division also becomes clearer in this light. The multinomial
analysis conﬁrms that young people are generally more likely to support the opposition
and they are generally more supportive of the 1St rather than the 2nd opposition party. This
may be due in part to the negative image of the 2nd opposition party (UNP) among young
people, since it was founded by one of the country’s largest conglomerates, Chaebol.

However, the 2nd opposition party does receive a great deal of political support from the

119

wealthy who are generally considered to focus more on the national economy than any
other group. The binary logistic regression shows that the wealthy tend to vote for the
incumbent party rather than the opposition parties (see Table 4-5). However, when faced
with a choice between the incumbent and the 2"GI opposition party, many wealthy people
do endorse the 2nd opposition party.

While the variable of “three party merge” is not statistically relevant to the
logistic analysis, it does become consequential in the multinomial analysis. Those who
are critical of the three party merge are likely to support the 1""1 opposition rather than the
incumbent party. This variable beneﬁts the main opposition party at the expense of the
incumbent party. This seems reasonable since the three party merge has been intended to
exclude the lSt opposition party alone, combining the remaining three parties into one.
When voters face a partisan choice between the incumbent and the lSt opposition parties,
positive or negative evaluations of political events become critically important in
deciding the electoral fortunes of a party. Also, the fact that the younger tend to support
the lSt opposition party rather than the 2"d opposition party is important here.

Table 4-7 shows how individuals tended to make political decisions in the 1996
legislative elections. Along with the strong inﬂuence of regionalism on individual
partisan choices, the 1996 election saw the revival of old-fashioned voting behavior in
Korea (Cho 1988; Browne and Kim 2001). Speciﬁcally, the change in the probability of
choosing the incumbent party is .20, indicating that the odds of voting for the incumbent
party are twenty percentage points higher when an individual voter is older, other factors
remaining constant. This results in a generational gap in terms of partisan support: the

young for the main opposition party, the old for the incumbent party.

120

Table 4-7 Determinants of Partisan Choice in 1996 Legislative Election

 

 

Election/ Incumbent Party 1” Opposition Party 2nd Opposition Party
Variable Probability Change Probability Change Probability Change
Age .34*** (.09) .20 -.12 (.12) - -.06 (.l l) -
Gender -.065 (.18) - -.21 (.29) - .30 (.22) -
Education .065 (.1 1) - -.05 (.15) - -.37***(.l4) -.06
Income .081 * (.42) .09 -.10* (.05) -.06 .02 (.06) -
Big Cities -.09 (.19) - 1.33***(.27) .26 -.63** (.25) -.04
Counties .91 * (.34) .22 -.38 (.37) - -.68 (.42) -
Regionalismla -.ll (.22) - -2.19***(.45) -.15 .Ol (.28) -
Regionalism 2b -l.03*** (.28) -.235 2.1 l***(.29) .46 -2.28*** (.51) -.08
Regionalism 3c —.69*** (.25) -.16 -.61* (.32) -.07 132*” (.29) .18
Partisan Tendency -.93*** (.09) -50 .89***(.11) .30 .44***(.10) .09
President

Approval -.67*** (.16) -.20 .13 (.20) - .64*** (.20) .06
Personal .07 (.13) - -.04 (.18) - -.19 (.17) -
Economy

National .48*** (.12) .22 -.32* (.1'6) -.08 -.14 (.15) '
Economy

Political Ideology --02 (.20) - «42 (.27) - .38 (.26) -
Constant 1.15 (.77) -2.51* (1.04) -2.32*** (1.01)
Pseudo R2 .26 .43 .18

Wald Chi2 174.62*** 189.24*** 107.32
LogLikelyhood -410.1 1 -260.53 -282.08

% Correction 76.8% 87.7% 85.2%

Base Probability 383 '119 '10]

N 818 818 818

 

Note: * signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Robust standard error of the
estimate is in the parentheses; a from Yungnam Provinces; b from Honam Provinces; ° from
Chungchong Provinces.

Along with this generational division in partisan support, the concentrations of

electoral support in metropolitan areas and the countryside (urban areas for the opposition

121

 

and rural areas for the incumbent party) reappears in the 1996 election. For example,
when the associated change in the predicted probability of voting for the incumbent party
is .22, this means that the probability of choosing the incumbent party is around twenty
percentage points higher when one identiﬁes oneself as a resident of a rural area, other
factors remaining constant.

In contrast, the main opposition party seems to garner more political supports than
the incumbent party in urban areas. The probability of choosing the 1"t opposition party
is .26, indicating that the odds of voting for the main opposition party are around one-
quarter percentage points higher if a respondent resides in an urban area, other factors
remaining constant.

The impact of regionalism, therefore, becomes very clear. For example, the
probability of choosing the incumbent party declines by around twenty-ﬁve percentage
points when an individual identiﬁes oneself as from the Honam as opposed to the
Yungnam region. However, the odds of voting for the 1St opposition party increase by
about ﬁfty percentage points under the same situation. This indicates that regionalism is
the single most important factor inﬂuencing individual partisan choice in the election.

Furthermore, it becomes clear why the degree of electoral competition between
parties is low in these regions. Therefore, the degree of electoral competition in those
regions where a certain party is relatively strong tends to be lower than that of the
national level and closer (or even lower) than that of the average district-level electoral
competition. We can ﬁnd the same phenomena with the individual survey data. These

points clearly show the existence of the geography of politics at the individual level.

122

Economic voting was also quite visible in the 1996 legislative elections. Between
pocketbook and sociotropic voting, the ﬁndings show that individual evaluations of the
national economic situation are indeed quite consequential. The probability of voting for
an opposition rather than the incumbent party is around twenty percentage point higher
when one is quite dissatisﬁed with the national economic situation, other factors
remaining constant. The electorate tends to attribute the weak national economy to the
policies of the incumbent party. The same patterns appear when we look at the impact of
presidential job approval on individual partisan choices. Those who are not satisﬁed with
the President’s job are more likely to support one of the opposition parties.

Choices between the incumbent and main opposition party seems to be based a
little on one’s income level. It is likely that people who have relatively high incomes tend
to support the incumbent party, whereas people from the lower classes are more likely to
endorse the main opposition party. For example, the associated change in the predicted
probabilities of choosing the incumbent party is around .09, suggesting that the
probability of voting for the incumbent party is almost ten percentage points higher when
one has a relatively high income (the odds are around six percentage points higher for
choosing the 1S” opposition party when a respondent has a relatively low income) other
factors remaining constant.

All of this becomes clearer when we review the ﬁndings of the multinomial
analysis. A relatively low income level beneﬁts the main opposition party at the expense
of the incumbent party. While the main opposition party gets more support from lower

class people, this party is also considered as to be more liberal than the 2Ind opposition

123

party. Conservatives, therefore, tend to support the 2nd opposition party at the expense of

the 1S" opposition party.

Table 4-8 Multinomial Logistic Estimation of 1996 Election

 

 

Election/ IS! Opposition/ 2nd Opposition/ lSt Opposition/
Variable Incumbent Party Incumbent Party 2nd Opposition
Age -.37** (.14) -.31* (.13) .065 (.16)
Gender .23 (.27) -.25 (.26) .48 (.31)
Education .14 (.17) -.40* (.17) .26 (.20)
Income -.15** (.07) .028 (.06) -.12 (.08)

Big Cities 18*“ (.33) .30 (.28) 2.1*** (.36)
Counties .76 (.48) -l .28*** (.45) -.46 (.55)
Regionalisml" - 2.25*** (.46) .007 (.34) - 2.24*** (.49)
Regionalism 2b 2.1*** (.37) -.92* (.54) 3.02*** (.53)
Regionalism 3c .019 (.39) .147*** (.33) - l.49*** (.40)
Partisan Tendency l.36*** (.14) .93*** (.13) .43*** (.15)
President Approval .30 (.25) .83*** (.23) -.54** (.28)
Personal Economy “16 (20) T16 ('20) '11 ('23)
National -.59***(.l 8) -.59*** (.18) -.22 (.21)
Economy

Political Ideology .41 (.31) .41 (.31) - .70** (.35)
Constant -1.5 (.1 .33) -2.73* (1.29)

Pseudo R2 .56

Chi Square $51.92*"

Log Likelihood 813.8]

N 675

 

Note: 8 signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Standard error of the estimate
is in the parentheses; a from Yungnam Provinces; b from Honam Provinces; c from Chungchong

Provinces.

124

 

In contrast, people who consider themselves to be liberal are more likely to
endorse the main opposition party. Similarly, young people are more likely to support
opposition parties at the expense of the incumbent party. Therefore, in the face of a
decline in the inﬂuence of regionalism, the possibility emerges of a party system based
on ideological rather than regional divisions.

Urban areas appear to beneﬁt the main opposition party at the expense of the
incumbent party, as shown in the logistic analysis. The 1St opposition party gets more r—
electoral support than the incumbent party and the 2nd opposition party in urban areas.
The national economic situation also becomes important in generating political support

for one party over another. Those who are critical of the current national economy

 

situation tend to support the opposition parties at the expense of the incumbent party
since they believe that a poor economy is the fault of the ruling party and president.

The legislative elections of 1992 and 1996 had three major parties at the national
level. The third party received more than 15% of the national vote and comfortably
maintained a legislative bloc in the Korean National Assembly in those two
parliamentary elections. The legislative election of 2000, however, shows a different
picture. The third party (CLD) received less than 10% of the national vote and failed to
become a legislative bloc in the national legislative body. The two major parties together
garnered more than 75% of the national vote at the aggregate and individual levels. Thus,
the third party can be removed from the analysis of the election for purposes of logistic

. . 9
regressron analysrs.

 

9 However, a separate multinomial analysis renders results that are almost identical to the logistic
regressions of the two major parties.

125

Table 4-9 Determinants of Partisan Choice in 2000 Legislative Election

 

 

Election! Opposition Party Incumbent Party
Varlable Probability Change Probability Change
Age -.013(.01) - .013(01) -
Gender .18 (.22) - .05 (.22) -
Education -.14 (.18) - -.06 (.15) -
Income .11* (.06) .10 -.02 (.06) -
Big Cities .20 (.25) - .01 (.25) -
Counties -.33 (.31) -- .02 (.31) -
Regionalisml" 18*“ (.28) .42 -l.38*** (.31) -.19

Regionalism 2b -1.02***(.33) .18 .22(.27) -

Regional Sentiment IC 30*” (.20) 25 -.83*** (.22) _.23

Regional Sentiment 2d -.47* (.26) —.12 .32 (.25) -

Partisan Tendency '54,”... (.12) .22 -.83*** (.12) .30

President Approval

fl

 

.46** (.20) .12 -.37** (.20) -.09
Left Ideology -.38 (.26) - .15 (.25)
Right Ideology .40 (.26) - -.90*** (.28) -.14
Constant -.358*** (1.1) 3.36*** (1.11)
Pseudo R2 .31 .27
Wald Chi2 l41.75*** 141.75***
Log Likelihood -275.91 —274.99
% Correction 76.2% 78.3%
Base Probability .508 .266
N 580 580

 

Note: signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01;Robust standard error of the
estimate is in the parentheses; " from Yungnam Provinces; b from Honam Provinces; ° Antipathy
toward People from the Honam Provinces; " Antipathy toward People from the Yungnam
Provinces

126

Table 4-9 above shows how individual voters decided their partisan preferences in
the 2000 legislative elections. The ﬁndings indicate the continuity of the importance of
the following factors: regionalism, residence size, income level, and individual partisan
tendency toward the incumbent or opposition party. As suggested earlier, in addition to
the relatively strong inﬂuence of regionalism, we also ﬁnd that the level of income
becomes an important factor.

For example, the associated change in the predicted probability is .10, indicating
that the probability of voting for the opposition party is more than ten percentage points
higher when a respondent has a relatively high income, while other factors are being held
constant. The electoral division following income level has been consistent in the Korean
parliamentary elections.

The main opposition party in the 2000 legislative election was the incumbent
party in the 1996 legislative elections. Older and more wealthy people were more likely
to support the old incumbent party (the current main opposition party). If we assume that,
generally speaking, the old and the wealthy are more likely to be conservative, then the
supporters for the main opposition party were more likely to be conservative than the
supporters of the incumbent party in the 2000 elections. This point becomes clearer when
we see the associated change in the predicted probabilities of voting for the incumbent
party. Speciﬁcally, the probability of choosing the incumbent party is almost ﬁfteen
percentage points lower when we identify someone as having a right wing or
conservative political ideology. This suggests that conservative people are more likely to

vote for the main opposition party and liberals to support the incumbent party.

127

However, the impact of regionalism is still strong and seems dominant. The
associated change in the predicted probability of choosing the main opposition party
is .42. This point suggests that the probability of voting for the opposition party is more
than forty percentage points higher when someone is from the region where the
opposition party is based, while other variables remaining constant.

Unlike previous surveys, the 2000 survey introduced a variable designed to
measure individual antipathy towards people from either of the two typical regions.
Those who have relatively high levels of antipathy towards the region where the 1
incumbent party is based are more likely to support the opposition party. For instance, the

probability of voting for the opposition party is one-quarter percentage point higher if a

 

respondent has strong antipathy towards the people from the incumbent party’s regional
strongholds.

In contrast, those who feel less antipathy towards people from the incumbent
party’s regional bases are more likely to endorse the incumbent party. The associated
change in the predicted probabilities is .28, suggesting that the probability of voting for
the incumbent party is almost one-third percentage points higher when someone feels
antipathy towards the people from the region(s) in which the opposition party is
traditionally based while we held other factors constant at their means.

In overall, we ﬁnd a slim ideological division among the Korean electorate based
on income, age, and political ideology. The young, liberal, and those with low incomes
tend to support the party that is regionally based in the Honam region. Whereas, older,
more conservative people with higher incomes are more likely to support the party that is

regionally based in the Yungnam region. For this reason, therefore, a party system based

128

on ideological division may appear and it may contribute to the institutionalization of the
party system on its road to a fuller democratic participation in Korea.

However, the three legislative elections under discussion serve to illustrate the
dominant inﬂuence of a certain nature of region and regionalism on the distribution of
electoral supports across the nation at the individual level. The ﬁndings of the micro-level
analysis speciﬁcally show a clear concentration of electoral support along regional lines
at the individual level. Also the nature of one’s residence (urban or rural) has a direct
implication in deciding individual partisan preferences. These points suggest that certain
characteristics within sub — national units in the country (i.e., urban vs. rural; degree of
concentration of electoral supports along regional lines) are the major source of political
cleavages and partisan fragmentation at the individual level in the Korean electoral
politics.

The local/district partisan characteristics that determine the degree of electoral
competition across districts in legislative elections at the aggregate level appear to have
the ﬁrm roots in the distribution of electoral supports at the individual level. In turn, the
concentration of electoral supports along regional lines at the individual level determines
the level of electoral competition across districts in the Korean legislative elections.
Furthermore, the micro — level analysis indicates that the local/district partisan
characteristics are variable following regional lines in determining the degree of electoral
partisan fragmentation in most elections. For example, those districts in which there
exists relatively strong support for a certain party at the individual level tend to have

relatively low degree of electoral partisan fragmentation, or vice versa.

129

The micro — level analysis of the determinants of partisan choice in the recent
Korean legislative elections also indicates the necessity of district — level analysis of the
degree of electoral partisan fragmentation in legislative elections under the SMSP
electoral systems. We found the dynamics of the local/district partisan characteristic at
the individual level. Therefore, an understanding of the degree of electoral competition at
the district as opposed to the national level is clearly called for with respect to legislative
elections. The macro and micro-level analysis of the Korean case clearly supports the

district-level analysis of the degree of electoral competition in legislative elections.

130

CHAPTER FIVE

MICRO ANALYSIS OF THE CANADIAN CASE

Introduction

In addition to comparing cases under the same electoral system (SMSP) but under
different institutional contexts (decentralized unitary presidential vs. federalist
parliamentary systems), it is useful to compare the micro-level analyses—here, the
Korean and Canadian cases. It is also important to analyze how the variations in electoral
support and the different combinations of parties competing along regional lines at the
macro level in Canada has an impact on the distribution of partisan choices at the
individual-level. Furthermore, this chapter is to know whether the analysis of the sources
of political cleavage and electoral partisan fragmentation at the individual level in the
Canadian case support the necessity of sub-national (district and regional level) analyses
of the degree of electoral partisan fragmentation in legislative elections as in the Korean
case.

This chapter, therefore, represents a micro-level analysis of the Canadian case by
utilizing the individual-level survey data related to the most recent legislative elections
under consideration. We seek a more reﬁned understanding of the dynamics surrounding
the most recent Canadian parliamentary elections and the sources of partisan
fragmentation that occurred at the individual level in this country. This chapter also seeks
to identify the existing political cleavages upon which the current shape of the Canadian
political, electoral, and party system is based. In what follows, therefore, relevant

information is presented concerning data, measurement, and methodology. Next, I

131

 

proceed to an analysis of the determinants of partisan choices in Canadian legislative

elections.

Data and Measurement

This chapter utilizes the 2000 Canadian Election Studies (CES) data that surveyed
around 3000 Canadians to verify how they made their political decisions in the 2000
parliamentary elections.

Table 5-1 Comparison of Partisan Choices at Individual and Aggregate Levels

 

 

Parties CES Ofﬁcial Difference
Result

Other 2.12%

Liberal 43.32% 40.82 2.48

Alliance 24.69% 25.46 -.77

Conservative 8.13% 12.21 -4.08

NDP 7.84% 8.51 .67

Bloc Quebecois 13.90% 10.72 3.18

Total 100% 97.72%

 

Table 5-1 represents the difference between individual (reported) and aggregate
(ofﬁcial) electoral returns in terms of partisan preferences. There is a 2.24 percentage
point difference, on average, for ﬁve partisan choices, between individual and aggregate
levels. The largest difference, for the New Democratic Party (N DP), is more than four
percentage points, whereas the smallest difference, for the Canadian Alliance (CA), is
less than one percentage point.

While the average difference between individual and aggregate levels is 3.2
Percentage points in the Korean case, the deviation of 2.24 in the Canadian case, on

average, between reported and ofﬁcial electoral returns, indicates that the data is not

132

ﬂawed for our analysis. This point becomes clearer when we compare the difference
between individual reported partisan preferences and aggregate ofﬁcial electoral results
from comparative perspective.

Table 5-2 Comparisons of Survey and Ofﬁcial Data in
Comparative Perspective

Surveys Average Difference
CES2000 2.24%
KES* 3.20%
NES 1996 7.64%
NES 2000 2.36%
*Average difference of three election surveys (1992, 1996 and 2000)

 

 

 

Table 5-2 compares the average difference between individual and aggregate level
electoral partisan preferences in comparative perspective. There existed ﬁve — percentage
point difference, on average, in the American National Election Studies (1996 and 2000
combined). However, we found the deviations of 2.24 and 3.2 percentage points from the
Canadian Election Studies and Korean Election Studies, respectively. The degrees of
deviation of individual reported partisan preference from the ofﬁcial electoral returns in
Canada and Korea appear to be negligible.

Thus, using the CES 2000 survey data in this analysis is not likely to endanger the
validity of the research since the distribution of partisan preference at the individual level,
in this parliamentary election, is highly reﬂective of what is shown by the ofﬁcial
election returns at the aggregate level. However, those who do not express a vote
intention or those who indicate a preference for small parties or independent candidates
are not included in the following analyses. In this sense, the Bloc Quebec (BQ) is also
excluded from the analysis since it is on the ballot for the province of Quebec alone. The

Bloc Quebec is a typical example of regional party, with all of its 38 legislative seats

133

 

coming from the province of Quebec (75 seats total for the province); it does not ﬁeld a
candidate outside of the province. Nevertheless, the Bloc Quebec has become the third
largest party in the Canadian national legislative body. Here, however, our focus is on
partisan choices among the signiﬁcant ‘national’ political parties that are viable on a
national level.

The dependent variable in this analysis is the individual vote choice of a certain
party in the 2000 legislative elections. Since our major goal is to better understand how
an individual voter makes his or her partisan choice, a logistic regression analysis using a
dichotomous dependent variable (1=choice of a certain party; 0=no choice of the party) is
employed to determine the inﬂuence of the set of independent variables. Separate logistic
regressions are performed for each major party participating in an election (Liberal,
Canadian Alliance, Progressive Conservative, and New Democratic Party).

As with the micro-level Korean case study, this chapter provides the associated
changes in the predicted probabilities for selected variables while controlling for other
variables; logistic regression analysis will show the associated changes for the variables
when they change from one standard deviation below the mean to one standard deviation
above the mean for the interval/ordinal variables, and from the absence to the presence of
certain characteristics while holding other variables at their mean values.l

As suggested, the Canadian voters were faced with a multiple choices of parties.
The macro —— level analysis indicates that there have existed more than four parties at the
national level. Therefore, it seems necessary to take into account the nature of multiparty
system in Canada. Binary logistic regressions will be supplemented by a multinomial

logistic analysis in order to assess what factors in the model are most inﬂuential for

 

' CLARIFY program is used for the calculation of the associated changes in the predicted probabilities.

134

determining the distribution of electoral support between particular parties at the
individual level in the 2000 parliamentary election.

Understanding the individual partisan choice in the most recent Canadian
legislative elections also involves a variety of variables. First, we expect that a substantial
portion of the Canadian electorate has developed a long-terrn attachment to a particular
party and this type of psychological attachment almost always determines individual vote
choices. While evaluations of personal and national economic situations tend to revolve
around short- term issues that inﬂuence individual voting choices, partisan identiﬁcation,
in the words of Miller and Shanks, is “the most enduring of political attitudes, responsible
for shaping a wide variety of values and perceptions, and therefore, an appropriate
starting point for an analysis of partisan political preference, such as the choice of
presidential candidates” (1996: 1 17).

However, the long history of multiparty system in Canada prevents the use of
traditional measure for partisan identiﬁcation since we have more than 2 major parties.
Rather than using the traditional party identiﬁcation measure, this analysis uses the
feeling thermometer for each major party. If two major parties exist, we can put those
two parties at the two extremes and identify where each is located on the continuum.
However, Canada has a multiparty system consisting of four or ﬁve parties. Therefore,
the degree of individual attachment to a particular party expressed in the form of a feeling
thermometer appears to be more relevant for measuring individual partisan attachment.
We asked respondents how much they like or dislike a particular party on a scale where
zero indicates that the person dislikes the party and 100 means that he or she likes this

party the best.

135

While the partisan feeling thermometer is relatively long—terrn factor to inﬂuence
individual vote choice, I focus on feelings about the economy as a short-terrn factor:
personal vs. national and retrospective vs. prospective. The conventional wisdom found
in the empirical literature suggests that the political fate of the incumbent government
depends to a large extent upon economic conjunctures at the time the election is held
(Lewis-Beck 1985). When the economy is in good shape, the incumbent party is likely to
be reelected; however, in times of economic downturn, the party in power may be in
political trouble (Lewis-Beck and Paldam 2000).

I introduce two different types of variables to tap into the Canadian perceptions of
the economy: personal and national economic situation. The ﬁrst two variables are
concerned with the individual or personal economic situation: one retrospective and the
other a prospective evaluation. The other two variables deal with individual perceptions
of the Canadian economy as a whole, retrospective and prospective. It is expected that
those who have positive attitudes about their personal and the national economic situation
are likely to support the incumbent party in the 2000 parliamentary election. Also, people
who have positive prospects about their personal and national economic situations tend to
support the incumbent Liberal party in the election.

Given the nature of its parliamentary democratic institutions, the Canadian
electorate chooses the party that will form its national government through the election of
local legislative members. Therefore, individual evaluations of party leadership, who may
become prime minister, appear to be consequential in determining individual partisan
choice in the election. There is a vast amount of literature that suggests that individual

voters make up their political minds about a party on the basis of how they feel about its

136

leader (McAIlister and Studlar 1996). Furthermore, there is evidence that suggests that
how the electorate feels about party leaders is particularly important in Canada (Gidengil,
Blais, Nadeau and Nevitte 2000).

As there exist multiple parties, multiple party leaders appear. Like the partisan
thermometer, we asked our respondents how much they liked or disliked a particular
party leader on a scale where zero indicates that the person dislikes the leader and 100
means that he or she sees this leader as the best candidate. I included four party leaders
for the four major parties under consideration: Chretien for the Liberals, Day for the
Canadian Alliance, Clark for the Progressive Conservatives and McDonough for the New
Democratic Party.

As the macro-level analysis of the Canadian case shows, it is a well-established
fact that political parties do not get the same level of electoral supports from voters from
different regions in Canada. The role of regions is actually more important than it was
thirty years ago (Blais et al 2001) and this depth of regional division is clearly evident in
election returns at the aggregate level. The Liberals obtained 52% of the vote in the
province of Ontario, for example, and only 25% in the provinces of the Canadian West.
The Canadian Alliance, on the other hand, got 50%of the vote in the western provinces
but only 10% in the Atlantic Canada. The Progressive Conservatives garnered 31% of the
vote in Atlantic Canada and 10% in the West. Finally, the New Democratic Party
obtained 17% on the Atlantic coast and 8% in the province of Ontario.

The primary variable that we need to focus on in this micro-level analysis is the
regional variation of electoral support for a particular party at the individual level: the

concentration of electoral supports at the individual level following regional liens in

137

 

Canada. The macro-level analysis found that the local/riding-level partisan characteristics
that can be measured through local or riding-level electoral competition patterns are the
main determinants of the degree of electoral partisan fragmentation in the Canadian
national legislative elections at the riding level. This suggests that the sources of electoral
partisan fragmentation in particular and political cleavages in general at the individual
level are certain sub-national units (provinces and ridings). Furthermore, this point
indicates that a concentration of electoral supports exists along regional lines at the
individual level in the country.

Therefore, I introduce the region dummy variables to test for the presence of

 

geographical factors inﬂuencing electoral choices at the individual level. For this purpose,
I categorize the 13 provinces of Canada into several regions and then introduce the region
dummy variables of Western and Atlantic provinces along with the province of Ontario
following the general divisions of provinces in Canada. These variables serve to test the
extent to which sources of partisan fragmentation in Canada are also represented by
certain sub-national units at the individual level.

I also introduce the variable of political ideology. As in the Korean case, this is a
dummy variable: respondents were asked to identify themselves as belonging to the right
or the left. Along with the substantive factors mentioned so far, this analysis includes
such demographic control variables as gender, education, income, religiosity (protestant
or catholic) and union membership. While Korea is a relatively homogeneous society,
Canada appears to be more diverse in terms of ethnicity, language, religion and cultures.
Therefore, it is expected that certain demographic characteristics be related to the

electoral support to a certain party. Table 5-3 summarizes the factors that were mentioned

138

so far to inﬂuence the individual vote choice in the 2000 Canadian parliamentary
elections.

Table 5-3 Categorization of Variables in 2000 CES

 

Category Variable

 

Demographics
Gender
Age
Level of Education

Monthly Income
Religiosity
(Protestant/Catholic)

Union Membership
Regional Dummy

Western

Atlantic

Ontario

Partisan Feeling Thermometer
Leadership Thermometer
Political Ideology

Prospective Economic Voting

Retrospective Economic Voting

F our Parties
Four Parties
Right

Personal

National

Personal

National

 

Finally, I will take into account the associated changes in the predicted
probabilities (either positive or negative with respect to the dependent variable) when

there are corresponding changes in the selected independent variable that are statistically

139

signiﬁcant. Therefore, the individual choice of a certain party can be expressed in the

following equation:

Y =31 + Ble + B3X2 + B4X3 + B5X4 + Boxs + B7X6,

in which Y represents the individual choice of a certain party or not and B's are unknown
but with ﬁxed parameters. X's 1 through 6 represent the chosen set of independent
variables: demographic characteristics, partisan feeling thermometers, party leadership
feeling thermometers, political ideology, personal economic evaluation, national

economic evaluation, and regional dummies.

Analysis and Discussion

Tables 5-2 and 5-3 Show how Canadians made their political decisions in the most
recent legislative elections. The ﬁrst table is concerned with two major parties (Liberal
and Alliance), which obtained 238 out of 301 total seats (80%). The second table deals
with relatively small parties, the Progressive Conservative and New Democratic parties,
which together had only 25 seats in the Canadian House of Commons. Both tables
indicate that the variables chosen explain the variation of individual partisan preference
in the most recent Canadian parliamentary elections. For example, the pseudo R2 ranges
from .313 to .558, indicating that the model can explain more than ﬁfty percent of the

variation in the dependent variable at the maximum.

140

 

Table 5-4 Determinants of Partisan Choice (Two Major Parties)

 

 

P arti e s / Liberal Canadian Alliance
Variables Probability Change Probability Change
Gender -.213 (.149) - .855***(.226) .03
Education -.010 (.027) - -.028**(.015) -.003
Income .00002 (.0001) - .0003 (.0002) -
Leadership .018***(.005) .21 .027***(.010) .04
Partisanship .066***(.007) .65 .068***(.010) .12
;gtrrzzrgilcﬁsgnomy -.008 (.041) - .021 (.060) -
Personal Economy Prospective .023 (.0370 - .006 (.050) -
National Economy -.019 (.040) - —.005 (.053) -
Retrospective

National Economy Prospective -.l l4***(.037) -.ll .126**(.054) .012
Atlantic Regions -.088 (.288) - .162 (.508) -
Western Regions -.086 (.227) - 2.07***(.4l4) .14
Ontario .46**(.217) .12 l.13***(.403) .05
Protestant -.631***(.205) -.15 .075 (.266) -
Catholic .197 (.215) - -.595**(.322) -.01
Union Membership -.314**(.152) -.08 .032 (.221) -
Right Ideology -.558***(.172) -.13 .928***(.207) .04
Constant -4.28***(.584) -8.54***(.857)

Pseudo R2 .362 .558

Wald Chi2 252.99*** 223.14***

Log Likelihood -550.48 -307.12

% Correction 81% 89.6%

Base Probability .467 .026

N 1246 1079

 

Note: signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Robust standard error of the

estimate is in the parentheses

141

Table 5-5 Determinants of Partisan Choice (Two Minor Parties)

 

 

Parties/ PC NDP

Variables Probability Change Probability Change
Gender -224 (.228) — .021 (.236) -
Education .014 (.018) - .019 (.032) -
Income .0001 (.0002) - -.0003 (.0003) -
Leadership .026***(.008) .02 .021**(.010) .04
Partisanship .057***(.009) .05 .055***(.010) .10
giggliggnomy .008 (.068) - .106 (.064) -
Personal Economy Prospective .056 (.059) - .034 (.058) -
National Economy -.029 (.061) - .110**(.062) .11
Retrospective

National Economy Prospective .107** (.057) .008 .033 (.061) -
Atlantic Regions 1.76***(.410) .08 1.26***(.474) .03
Western Regions .020 (.434) - .914**(.432) .08
Ontario .538 (.384) - .770 (.433) -
Protestant .392 (.321) - -1.12***(.304) j03
Catholic .447 (.360) - -1.21***(.309) :03
Union Membership -.165 (.263) - .439**(.252) :03
Right Ideology .055 (.260) - -1.85***(.569) j03
Constant -8.55***(.825) -7.69***(858)
Pseudo R2 .313 .369

Wald Chi2 144.87*** 159.50***

Log Likelihood -266.82 -231.04

% Correction 92.2% 92.2%

N 1249 1096

 

Note: signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Robust standard error of the

estimate is in the parentheses

142

As expected, the individual evaluative perceptions of the party leadership and
individual attachment to a particular party play a signiﬁcant role in determining the
electoral fortunes of each party. For example, the relevant associated change in the
predicted probabilities is .21 for those who evaluate Liberal party leadership positively,
indicating that the probability of choosing the incumbent Liberal party in the 2000
election is more than twenty percentage points higher when a person likes the party’s
leader (Chretien), with the other variables set at their mean values.

However, the party leadership variable appears to be most the important for the
Liberals among parties. The individual evaluative perceptions of the Alliance party leader
(Day) appear to be less important for the individual choice of the party even though it is
statistically important at the level of 99% for a two-tailed test. For instance, the
associated change in the predicted probabilities is .04, suggesting that the odds of
choosing the Canadian Alliance is four percentage points higher when someone evaluates
the party leadership positively while other factors are held constant.

The electorate’s perception of the party as gauged by the partisan feeling
thermometers is also statistically signiﬁcant. The associated change in the predicted
probabilities is .65 for those who feel close to the Liberal party, suggesting that the odds
of voting for the Liberal party are almost seventy percentage points higher when the
person likes the party, other factors remaining constant. The same patterns appear for the
Canadian Alliance. The associated change in the predicted probabilities for those who
like the party is .114. This means that the probability of choosing the Canadian Alliance
is more than ten percentage points higher when someone likes the party, while other

factors held at their means.

143

We can identify ideological division among the electorate based on partisan
preference. Those who identify themselves as belonging to right side of the spectrum tend
to support the Canadian Alliance rather than the Liberal or New Democratic Party. In
contrast, people who classify themselves as liberal are likely to support the Liberal party
candidate. The associated change in predicted probabilities for those who are liberal is
.13, suggesting that the probability of choosing the Liberal party is around thirteen
percentage points higher when one identiﬁes oneself as liberal, while controlling for
other factors.

Similarly, we can also ﬁnd the impact of economic concerns in the Canadian
voting pattern. Rather than pocketbook voting, it is sociotropic economic voting.
However, it appears statistically signiﬁcant only for the minor opposition party. For
example, those who are not satisﬁed with the retrospective national economy are likely to
support the New Democratic rather than any other party. The associated change in the
predicted probabilities for those who evaluate the national economy negatively is .11,
indicating that the probability of voting for the NDP is more than ten percentage points
higher when someone is dissatisﬁed with the national economic situation, while
controlling for other factors.

Unlike previous studies on prospective economic voting in the advanced
democracies (Lewis-Beck 1985; Lewis-Beck and Paldam 2000), we ﬁnd the importance
of prospective national economic evaluation from the Canadian case. Those who are
positive about the future of the Canadian national economy are likely to support the
incumbent Liberal party. People who have negative attitudes or evaluations about the

national economy, on the other hand, tend to vote for either the Canadian Alliance or the

144

 

Progressive Conservative parties. For example, the associated change in the predicted
probabilities for those who have positive feelings about the national economy is .10,
indicating that the probability of choosing the Liberal party in the election is ten
percentage points higher when people anticipate a positive national economy, while other
factors are held constant.

However, we ﬁnd the most important impact of regional dummy variables in the
2000 Canadian parliamentary elections. Most parties seem to have their regional
stronghold and to become a major party at the national level with regionally concentrated
electoral supports. This point especially becomes true for the Canadian Alliance party.
For instance, the largest associated change in terms of the predicted probabilities for
those who chose the Canadian Alliance involves one of the regional dummy variables,
especially the variable of Western provinces. The macro-level analysis already indicated
that the Canadian Alliance party obtained 64 out of its total of 66 seats from the Western
provinces of Alberta, Manitoba, Saskatchewan and British Columbia.2

The micro — level analysis also shows the existence of the concentration of
electoral supports at the individual level following regional lines in Canada. The
associated change in the predicted probabilities for those who reside in the provinces of
Western Canada is .14, indicating that the probability of voting for the Canadian Alliance
was almost ﬁfteen percentage points higher when one resided in the Western provinces,
controlling for other factors constant. The Canadian Alliance became the main opposition
party (and second largest party) in the Canadian House of Commons with strong regional

support from the Canadian Westerns.

 

2 The total seats allocated to the Western provinces are 88.

145

f

The Liberal party is no exception with respect to the concentration of electoral
support along regional lines at the individual level in Canada. The Liberal party appears
to be stronger in the province of Ontario. The associated change in the predicted
probabilities for those who reside in the province of Ontario is .11, suggesting that the
odds of choosing the Liberal party in the 2000 parliamentary elections were more than
ten percentage points higher when we identiﬁed someone as residing in Ontario rather
than another region, while other factors are held at their means.

These factors become even more salient as we review the electoral returns from
the Western provinces and the province of Ontario at the aggregate levels. While the
Canadian Alliance obtained more than 70% of the total seats allocated to the entire region
of Westerns, the Liberal party got 100 seats out of 103 seats from the province of
Ontario. The micro — level analysis of Canadian case conﬁrms that the geography of
elections clearly exists at the individual level.

The concentration of electoral supports at the individual level along regional lines
is not common to the major parties alone. The other two minor parties (Progressive
Conservative and New Democratic Party) also have their regional strongholds. The New
Democratic Party is relatively strong in Western Canada, whereas the Progressive
Conservative party is strongest in the Atlantic Canadian provinces. For example, the
associated change in the predicted probabilities for those who reside in both regions is
around .8, indicating that the probability of choosing either the Conservative or the NDP
is almost eight percentage points higher when someone resides in either the Western or

the Atlantic provinces of Canada, while controlling for other factors.

146

Table 5-6 Multinomial Logistic Estimation of Partisan Choice

 

Parties/

 

. Alliance/ Conservative/ NDP/
Var lables Liberal Liberal Liberal
Gender -.741 (.485) -.048 (.420) -.011 (.476)
Education -.249**(.098) -.0008 (.037) .054 (.046)
Income .0005 (.001) -.0002 (.001) .0004 (.001)
Liberal Leadership -.044**(.013) -.04*** (.013) -.02 (.015)
Alliance Leadership .060***(.016) .006 (.014) .014 (.017)
Conservative Leadership -.016 (.014) .03**(.016) -.03 l *(.018)
NDP Leadership .009 (.014) .006 (.015) .059**(.019)
Liberal Partisanship -.130***(.017) -.108 (.017) -.123***(.019)
Alliance Partisanship .125***(.020) -.013 (.016) -.013 (.019)
Conservative Partisanship .028 (.017) .l27***(.020) .006 (.018)
NDP Partisanship -.050**(.017) -.02 (.016) .098***(.020)
Personal Economy Retrospective .084 (.1 13) .079 (.l 15) .183 (.138)
Personal Economy Prospective -.125 (.105) .147 (.107) -.077 (.125)
National Economy Retrospective -.l 17 (.l 15) -.159 (.109) .120 (.131)
National Economy Prospective .1 16 (.103) .065 (.098) .056 (.123)
Atlantic Regions .220 (.872) -.610 (.700) —.845 (.982)
Western Regions l.32**(.742) .276 (.703) -l.204 (.934)
Ontario -.349 (.733) .292 (.642) -l .08 (.919)
Protestant 1.22**(.537) 1.27" (.591) -.107 (.577)
Catholic .206 (.555) .996* (.592) .185 (.598)
Union Membership .1 19 (.445) .138 (.441) 1.285**(.470)
Right Ideology -.475 (.474) -.244 (.516) 16.8 (.176)
Constant 363* (2.14) .268 (2.04) 15.92***(17.3)
Pseudo R2 .857
Chi Square 1489.50***
Log Likelihood 598.696
N 766

 

Note: signiﬁcant < .10, ** signiﬁcant < .05, *** signiﬁcant < .01; Standard error of the estimate
is in the parentheses

147

Table 5-6 presents the ﬁndings from the multinomial logistic analysis that takes
into account the relative impact of variables on individual partisan preference between
particular parties competing in the 2000 parliamentary elections (CA vs. Liberals, NDP
vs. Liberals and PC vs. Liberals). The multinomial logistic analysis conﬁrms the
existence of the geography to inﬂuence on electoral decision - making at the individual
level. For example, the Western provinces of the country beneﬁt the Canadian Alliance
party at the expense of the Liberals. By contrast, the province of Ontario beneﬁts the
Liberals at the expense of the CA, even though this difference is not statistically
signiﬁcant.3

Overall, the micro-level analysis makes clear that the concentration of electoral
support along regional lines is a source of political cleavage and partisan fragmentation at
the individual level in Canada. This ﬁnding is in line with those from the macro-level
analysis that the local/riding electoral partisan characteristics are the main determinant of
the degree of electoral partisan fragmentation at the riding level in legislative elections.
As suggested, the Local/district partisan characteristics refer to the existing patterns or
degrees of electoral competition in a given riding or locality, and they can best be
measured on the basis of sub-national elections.

Given the concentration of electoral support and the different combinations of
political parties competing along regional lines in Canada at the individual level, the
local/riding partisan characteristics are the primary determinant of the degree of electoral
partisan fragmentation at the riding levels in the most recent Canadian parliamentary

elections. Furthermore, the local/riding partisan characteristics seem to represent a

 

3 The statistical insiginiﬁcance of the region dummy variables for the New Democratic Party and the
Progressive Conservatives can be attributed to the relatively low levels of electoral supports for these
parties (12% and 8%, respectively), compared with those of the Liberals and the Canadian Alliance.

148

particular type or pattern of electoral competition in a given sub -— national unit in most
elections since it is localized pattern of electoral competition in a given locality.

It is also highly likely that local and/or riding-level patterns of electoral
competition differ from district to district and from locality to locality (as an aggregation
of local legislative districts) because of the geography of elections at the individual level.
Therefore, there emerges relatively high degree of electoral partisan fragmentation in the
ridings (regions) where the concentration of electoral supports at the individual level is
relatively low. In contrast, those ridings that show relatively high concentration of
electoral supports at the individual level are likely to have relatively low levels of
electoral competition.

The micro-level analysis of the Canadian case also shows that different
combinations of parties compete against each other in the Canadian provinces. For
example, even though the Canadian Alliance as the dominant party in the West, the
Liberals and the New Democratic Party are also major competing parties in this region.
However, the Liberals, as the predominant party, and the Progressive Conservatives are
also major competitors in Atlantic Canada.

While the Liberals are predominant in the province of Ontario, it competes mainly
with the Bloc Quebecois for national legislative seats in the province of Quebec. As also
suggested above, the probability of choosing the New Democratic Party increases when
someone resides in the provinces of Western Canada, other factors held constant.
Furthermore, the odds of voting for the Progressive Conservative increase when someone

resides in Atlantic Canada, other factors constant.

149

Combined with the concentration of electoral supports following regional lines at
the individual level, the different combinations of political parties competing each other
along regional lines shed light on the importance of the local/riding partisan
characteristics. For instance, the combination of the Canadian Alliance, Liberals, New
Democratic Party and Progressive Conservatives in the Western provinces of Canada
results in four-party electoral competition. However, we ﬁnd relative two-party
competition in the province of Ontario (Liberals and Canadian Alliance) and Quebec
(Liberals and Bloc Quebec). The Atlantic Canada shows relative three-party electoral
competition in most elections (Liberals, New Democratic Party and Progressive
Conservatives).

These points, furthermore, support the necessity of a sub-national level (riding
and province) analysis of the degree of electoral partisan fragmentation in legislative
elections. Given the dynamics of the riding/local partisan characteristics within the
country, the level of electoral competition at the riding level in legislative election under
the SMSP electoral systems appear to be different from riding to riding and from locality
to locality. The macro and micro-level analysis of the Canadian case conﬁrms the riding-
level analysis of the degree of electoral partisan fragmentation in national legislative

elections.

150

CHAPTER SIX

CONCLUSION

The current understanding of party system development within the tradition of the
diversity of political institutions has focused on data from the national level and if any
provincial level. However, the basic unit of electoral competition for legislative seats is
the local electoral district. While most research has focused on the impact of the national-
level election on the degree of electoral multipartism in legislative elections, there are
two studies to date that put an emphasis on local elections in understanding the level of
electoral competition in legislative elections under the federalist presidential
democracies.

Extending the current research on the levels of electoral competition in legislative
elections, this study has analyzed the degree of electoral partisan fragmentation at the
district level in the Single Member Simple Plurality (SMSP) electoral systems in both a
presidential decentralized unitary (Korea) and a parliamentary federalist system
(Canada). This ﬁnal chapter ﬁrst summarizes the major empirical ﬁndings from previous
macro and micro-level analyses and then looks toward avenues for future research.

Relying on the legislative district-level electoral data from national, provincial,
and local elections in these two cases, we ﬁrst examined the sources of electoral partisan
fragmentation at the district and national levels in Canadian and Korean legislative
elections. The main arguments of the study were then supplemented by micro-level
analysis of the sources of political cleavage and partisan fractionalization at the

individual level, through the utilization of micro-level survey data.

151

What theoretical implication can be added from this study to the scholarly
literature on the degree of electoral partisan fragmentation in legislative elections? The
starting point for this study is to acknowledge the importance of other political
institutions in understanding the degree of electoral competition in legislative elections
either at the national or at the district level. Unlike much research that has focused on the
inﬂuence of such national-level elections as that for President on the party system
development in legislative elections, this study puts an emphasis on sub-national and ﬂ:
sub-provincial elections.

Most legislative elections are held at the local level and this is especially evident

 

under the SMSP electoral system. Given the local district as a basic unit of electoral

yr

competition in legislative elections, it is of the utmost consequence to incorporate the
local pattern of electoral competition to understand the degree of electoral partisan
fragmentation in legislative elections in a given locality. The local voting patterns for
other levels of government such as gubernatorial, municipal, or provincial legislative
elections appear to be reﬂective of the local (or district)-level pattern of electoral
competition.

Overall, this study found signiﬁcant correlations between local elections and the
degree of partisan fragmentation across districts in legislative elections. The local/district
partisan characteristics based on the closest local elections in a given district or province
is reﬂective of the local partisan strength and subsequent patterns of electoral competition
in a given locality. The current pattern or level of electoral partisan competition in a
given legislative district from the closest provincial or district-level local elections is the

primary determinant of the degree of electoral partisan fragmentation at the district level

152

in legislative elections under the SMSP electoral systems. The existence of dynamic
local/district partisan characteristics across localities in both Canada and Korea was
substantiated by the district and individual-level analysis of the sources of partisan
fragmentation.

Speciﬁcally, the local elections appear to provide different incentives or motives
for voters and rational politicians to associate with any given party for electoral success
in legislative elections across a locality. Among the incentives or motives are the local
strength of a party and subsequent pattern of electoral competition in a given locality.
Thus, such prominent local elections as gubernatorial, municipal and provincial

legislative elections are very reﬂective of local-level partisan strength and subsequent

 

partisan fragmentation. Analyses of the degree of electoral partisan fragmentation at the
district level in legislative elections can beneﬁt from the consideration of local elections
within the diversity of political institutions.‘

The major ﬁndings are as follows: First, this study conﬁrms the necessity of sub-
national and district-level analysis of the degree of electoral partisan competition in
legislative elections, especially under the SMSP electoral systems. In our macro-level
study, we found that the party system at the national level is an aggregation of the
dynamics of electoral partisan fragmentation across districts and localities within a
country. In other words, the party system at the national level of a polity is a product of

electoral partisan fragmentations across the local district.

 

' In addition, the importance of local elections in understanding the level of electoral competition across
local districts in legislative elections became more evident when we considered a particular type of
institutional context that may foster political and social diversity following regional lines within a country.
We found, however, that other institutional systems like decentralized unitary admit the existence of
diverse local party systems within a polity.

153

More speciﬁcally, the comparative analysis of the levels of electoral partisan
competition for legislative seats indicates that different patterns or degrees of electoral
partisan fragmentation along regional lines have emerged in the Korean legislative
elections. Furthermore, the dynamic patterns of electoral competition following regional
lines in the Korean case suggests that these phenomena would happen for other types of
elections. We found how the degree of electoral partisan fragmentation varies depending
on localities in both presidential and local elections since 1987.

In most types of elections in Korea, for example, the effective number of electoral
parties across a province (and local districts) is closer either to a relative two party system
or a relative multiparty system depending on locality. Furthermore, local elections are
more reﬂective of the local strength of a party since the basic unit of electoral
competition under the SMSP electoral systems is the local district. Given the regional
dynamics of electoral competitiveness in most types of elections, the national-level party
system in the Korean legislative elections is a mixture of diverse local (province and
district) party systems.

As suggested, the particular structure of a country (i.e., federalism) may foster
sub-national political and social diversity, and lead to a variety of electoral party systems
across a nation. As in the case of Korea, the degree of electoral partisan competition at
the riding level in Canada has been relatively closer to two-party systems, while at least
more than four maj or parties have competed at the national level. Thus, the national-level
party system incorporates the dynamics of local-level electoral partisan fragmentations

across ridings and provinces.

154

 

r1-—

Furthermore, we found that the very reason Canada has a relative multiparty
system at the national level is the provincial-level two-party competitions for legislative
seats with alternating regional combinations of the competing parties. Given the
multiparty electoral competition across provinces, the major competing parties vary
following regional lines. Speciﬁcally, some Canadian provinces show relatively higher
degree of electoral partisan fragmentation than the riding average (relative two party
systems), approaching the national-level of multipartism. In contrast, other provinces
display relatively lower levels of electoral competition than the national-level party
system and are closer to the riding-level average.

Therefore, it is evident why we should focus on the district-level degree of
electoral competition in national legislative elections under the SMSP electoral systems.
The party system at the national level is a mixed picture that includes dynamic local
(districts and/or provinces)-level electoral partisan fragmentations across the country.
While the Korean case shows differing partisan strength along regional lines, the
Canadian case indicates that the different combinations of parties competing for
legislative seats follows provincial lines. However, both cases clearly suggest that there
have existed dynamic degrees/patterns of electoral partisan fragmentation across districts
in legislative elections. Thus, the national-level party system is an aggregation of local-
level electoral partisan fragmentations that are diverse along regional lines.

Given the dynamic variation of electoral competition following regional lines
(mostly provincial level) in national and sub-national elections, the local/district partisan
characteristics are the key factor for understanding the degree of partisan fragmentation at

the district level in legislative elections under the SMSP electoral systems. Since the

155

basic unit of electoral competition for legislative seats is the local district, across a polity,
the currently existing pattern of electoral competition in a given district or locality,
local/district partisan characteristics, can be best measured through district and/or local-
level elections, which is critical for explaining the levels of electoral partisan
fragmentation across the districts in legislative elections.

The Korean case study speciﬁcally shows that it is the degree/pattem of electoral
partisan competition in prominent local elections (gubernatorial and municipal) that
determine the degree of partisan fragmentation at the district level in legislative elections.
While the provincial-level gubernatorial elections represent local (provincial)-level
partisan characteristics, Korean municipal elections tend to turn on district-level partisan
characteristics. These two measures are highly complementarily for explaining the degree
of electoral competition across districts in national legislative elections.

This study, therefore, extends the ﬁndings from the Korean case to the Canadian
case, where there is a long history of federalism characterized by strong regionalist party
systems at the national level. The importance of local elections for explaining the degree
of electoral competition in legislative elections is highlighted by the presence of political
powers at multiple stages and subsequent political/social diversity along regional lines in
a federalist system. The Canadian case also illustrates the variations of local/riding
partisan characteristics along regional lines, which has a signiﬁcant inﬂuence on the level
of electoral partisan fragmentation at the riding level in the most recent Canadian
legislative elections.

The national-level party system, therefore, can be seen as a “patchwork” of

localized (or regionalized) party systems in both countries, because a variety of types of

156

electoral partisan characteristics follow regional lines. In Korea, there is a relatively low
degree of electoral partisan competition in most elections across the locality where a
certain party is especially strong. Relatively high levels of electoral partisan
fragmentation emerge, however, across the locality in regions where no particular party is
especially strong and several regional parties compete for electoral supports in most
elections. As a result of the concentration of electoral support along provincial lines in
the Canadian case, we found different combinations of parties competing within sub-
national units.

We can identify the dynamics of local/district partisan characteristics following
regional lines, therefore, in most types of elections in both Canada and Korea. Local and
district-level elections are representative of the existing pattern of local and/or district
level electoral competition in a given sub-national unit. These local/district partisan
characteristics play a critical role in determining the degree of electoral partisan
fragmentation at the district level in national legislative elections. Furthermore, given the
different types of party systems or patterns of electoral competition across provinces, we
can suggest that different types and patterns of electoral partisan competition have existed
along regional lines in Canada and Korea in most local and national elections.

Second, utilizing survey data, this study has extended the district-level analysis to
individual-level analysis of the sources of partisan fragmentation in particular and
political cleavage generally in these two cases. Since the shape of a party system depends
on the distribution of votes cast by individual voters, this serves to conﬁrm the major
ﬁndings of the macro-level analysis. In other words, the ﬁnding that local/district partisan

characteristics are the primary source of electoral partisan fragmentation in legislative

157

 

elections are conﬁrmed by the micro-level analyses. Furthermore, we found that certain
sub-national units (district, province and region) represent the major source of political
cleavage and electoral partisan fragmentation in Canada and Korea.

Concentrations of electoral support for parties following regional lines at the
individual level in Canada and Korea have a substantial history. The degree of partisan
competition across districts in legislative elections appears to change depending on the
degree of concentration of electoral partisan support along regional lines. Levels of
electoral competition are low when a relatively high degree of electoral support
concentration exists, and vice versa. Also, we established the existence of different
combinations of parties competing with each other along regional lines in the Canadian
case.

All of these empirical ﬁndings suggest that there have existed certain types or
patterns of partisan competition at certain sub-national units in local and national
elections. The micro-level analysis conﬁrms the existence of different degrees of
electoral competition and local/district partisan characteristics at the individual level
following regional lines in both Canada and Korea. The individual-level analysis of the
sources of partisan fragmentation also makes clear the necessity of studying the degree of
electoral competition in legislative elections at the sub-national level, since the national-
level analysis of the degree of partisan fragmentation in legislative elections does not
serve to identify the dynamics of electoral competition within sub-national units. Both
macro and micro-level analyses of the Canadian and Korean cases contribute to our
understanding of the degree of electoral partisan fragmentation in national legislative

elections.

158

This research is intended to serve as a starting point for future research. First,
there should be more study of the dynamics of levels of electoral competition across
legislative districts in legislative elections under the SMSP electoral systems. The degree
of electoral competition in legislative elections can be better understood at the district
level by data collected at that level. If we want to understand better the interrelationships
between different types of elections, the basic unit of analysis for the degree of electoral
partisan fragmentation in legislative elections should be the local legislative electoral
district.

This study has focused on cases under the institutional context of decentralized
unitary presidential democracy (Korea) and federalist parliamentary system (Canada). It
dose not directly speak to the degree of electoral competition in legislative elections in
federalist presidential institutions and unitary parliamentary democracies. It seems
necessary to analyze the impact of different institutional contexts on the degree of
electoral competition across districts in legislative elections. By analyzing the social and
cultural diversity following regional lines in a country, we may better understand the
existence of dynamics of local-level electoral partisan competitions.

Related to the above point, it is also required to develop more relevant measures
for the local/district partisan characteristics that are presumed to inﬂuence the degree of
electoral partisan fragmentation across districts in legislative elections. This study used
the closest local elections (gubematorial/municipal for Korea and provincial legislative
elections for Canada) to measure the concept. The local/district partisan characteristics
refer to the existing pattern or degree of electoral partisan competition in a given locality

or district. While the closest local elections are reﬂective of the current pattern of

159

 

electoral competition in a given locality, there also exists a possibility of a regionalized
type of electoral competition. For example, we found two-party competition in the
province of Quebec (Liberals and Bloc Quebecois). The same pattern appears in most
types of elections in the provinces of Canada. Therefore, combining two or more local
elections might be relevant to accurately measure the local/district partisan characteristics
in a given locality.

Further research is also called for with respect to how a party system at the
national level is formed on the basis of diverse local party systems, through
“aggregation” or the “linkage process of party systems.” If the national-level party
system represents an aggregation of local-level electoral partisan fragmentation, which
differs from locality to locality, then it is necessary to provide adequate explanation of
how the national party systems of Canada and Korea have formed. While the existence of
federalism is the main reason for diverse local party systems within sub-national units in
Canada (Gaines 1999) and possibly Argentina (Jones 1997) and Brazil (Samuels
20003/b) as well, we can also identify the impact of certain political/social cleavage
structures on the development of diverse local party systems. Given the importance of
political institutions to party system development, further research is also called for with
respect to the cleavage structure of a country, so as to understand better the process of
party system linkage between local and national levels.

We have identiﬁed several measures of the degree of electoral competition at the
district level in legislative elections. Depending on the substantive purpose of the
research in question, it is important to choose the appropriate measures for the variables

under consideration (Gaines 1997). Studies of the inﬂuence of district population

160

a 0

diversity on levels of electoral competition in legislative elections have found that the
impact of district diversity is conditioned by candidate-related characteristics (Bond
1983; Bond et al 2001). While the Canadian macro-level analysis found no direct
inﬂuence of district diversity on legislative electoral competition patterns in this study, it
is still quite possible that district diversity becomes consequential when we use different
types of dependent variables to assess the degree of electoral partisan competition across
districts in legislative elections. i-
In choosing which measure to use for the dependent variable, one must take into

account the nature of party systems in any given country. Given the two-party

 

competition that exist in most cases of SMSP electoral systems, as in the US, the

electoral margin or the incumbent vote share are the appropriate measures. However, we
ﬁnd different types of party systems coexisting within the sub-national units in Canada.
Given the different combinations of parties competing along regional lines, therefore, the
number of candidates in each riding, rather than the electoral margin or incumbent vote
share, may be more appropriate for measuring the degree of electoral competition in
Canada. In light of the need for district-level analyses of legislative elections from a
comparative perspective, additional analysis of riding population diversity in Canada is
called for.

Finally, our micro-level analysis found that the concentration of electoral support
along regional lines is the main determinant of electoral fortunes for parties in both
Canada and Korea. However, we do not yet know to what extent sub-national units in
these two cases act as surrogates for a wide variety of other factors related to partisan

preference. There may be regional concentrations of certain demographic characteristics

161

or different distributions of political attitudes and orientations that help to determine
individual partisan preference along regional lines.2 It seems important, however, to
uncover the political implications of regional variations of electoral support for parties in
these two countries. Also, the possibility of party system institutionalization should be
probed further, especially in light of the regionalist party system in the case of Korea. In
fact, party system institutionalization may occur through the uniﬁcation of regionalist
party systems with respect to ideological divisions.

These are the primary questions that need to be answered to achieve a more
systematic understanding of party system development in particular, and of political
institutions in general. This study of the sources of electoral partisan fragmentation in
legislative elections at the macro- and micro-levels is intended to contribute to our
understanding of the degree of electoral competition in legislative elections under the
SMSP electoral systems. Our understanding of party system deve10pment in legislative
elections can be greatly strengthened by district-level analyses of party system

development wherever adequate district-level data are available.

 

2 For example, Blais and his colleagues (2001) and McAllister and Studlar (1996) address whether the
regional gap in electoral supports reﬂects the difference in political attitudes/orientations or concentration
of demographic characteristics in Canada and United Kingdom, respectively .

162

 

BIBLIOGRAPHY

Achen, C. (2000). Why Lagged Dependent Variable Can Suppress the
Explanatory Power of Other Independent Variables. Paper presented at the
American Political Science Association Annual Meeting.

Aldrich, J., Magaloni, B., and Zechmeister, E. (2001). Electoral Considerations and
Party Change in Mexico and Taiwan. Presented at the American Political
Science Association annual meeting.

Aldrich, John, H. (1995). Why Parties? Chicago: University of Chicago Press.

Anckar, C. (2000). Size and Party System Fragmentation. Party Politics 6 (3): 305-
328.

Arseneau, T. (1994). Federalism and ‘New’ Parties of the Right in Alberta: From the
Progressive to Reform. In Parties and Federalism in Australia and Canada. Ed.
Campbell Sharman. Australian National University.

Bawn, K. (1993). The Logic of Institutional Preference: German Electoral Law as a
Social Choice Outcome. American Journal of Political Science 37 (4): 965-989.

Beck, P. (1997). Party Politics in America. NY: Longman.

Bianco, W. (1984). Strategic Decisions on Candidacy in US Congressional Districts.
Legislative Studies Quarterly 2: 351-364.

Blais, A & Carty, K. (1991). The Psychological Impact of Electoral Laws: Measuring
Duverger’s Elusive Factor. British Journal of Political Science 21 ( 1): 79-83.

Blais, A. & Massicotte, L. (1996). Electoral Systems. In LeDuc, L., Niemi. R., Norris, P.
(Eds), Comparing Democracies: Elections and Voting in Global Perspective.
Thousand Oaks: Sage Publications Inc.. 49-81.

Blais, A., Massicotte, L. & Dobrzynska, A. (1997). Direct Presidential Elections: A
World Survey. Electoral Studies 16: 441-455.

Blais, A. Gindegil, E. , Nevitte, N., & Nadeau, R. (2001 ). Making Sense of Regional
Voting in the 1997 Federal Election: Liberal and Reform Support Outside
Quebec. Canadian Journal of Political Science 32: 247-272.

Bond, J. (1983). The Inﬂuence of Constituency Diversity on Electoral Competition in
Voting for Congress, 1974-197. Legislative Studies Quarterly 8 (May): 201-217.

163

 

Bond, J ., F leisher, R. & Talbert, J. (1997). Partisan Differences in Candidate Quality in
Open Seat House Races, 1976-1994. Political Research Quarterly 50 (2): 281-
299.

Bond, J ., Campbell, K., & Cottrill, J. (2001). The Puzzle of Constituency Diversity
Revisited: Conditional Effects of District Diversity on Competition in
Congressional Elections. Paper presented at the Southern Political Science
Association annual meeting.

Brady, D. & Mo, J. (1992). Electoral Systems and Institutional Choice: A Case Study of
the 1988 Korean Elections. Comparative Political Studies 24 (4): 405-429.

Brown, R. & Woods, J. (1991). Toward a Model of Congressional Elections. Journal of
Politics 53 (2): 454-473.

Browne, E. & Kim, S. (2001). Regionalism in South Korean National Assembly
Elections: A Vote Component Analysis of Electoral Change. Paper presented at
the 2001 American Political Science Association Annual meeting.

Canon, D. (1990). Actors, Athletes, and Astronauts: Political Amateurs in the United
States Congress. Chicago: University of Chicago Press.

Carey, J ., Niemi, R., & Powell, L. (2000). Incumbency and the Probability of Reelection
in State Legislative Elections. Journal of Politics 62 (3): 671-700.

Carson, J ., Jenkins, J ., Rohde, D., & Souva, M. (2001). The Impact of National Tides and
District — Level Effects on Electoral Outcomes: The US Congressional Elections
of 1862-63. American Journal of Political Science 45 (4): 887-898.

Carson, J ., Jenkins, J., & Schickler, E. (2001). Investigating the Electoral Roots of the
Conditional Party Government. Paper presented at the 2001 American Political
Science Association annual meeting.

Chhibber, P. & Kollman, K. (1998). Party Aggregation and the Number of Parties in
India and the United States. American Political Science Review 92 (2): 329-342.

Cho, K. (1998). Regionalism in Korean Elections and Democratization: An Empirical
Analysis. Asian Perspectives 22 (1): 135-156.

Clark, D. (1987). The Kwangju Uprising: Shadows Over the Regime in South Korea.
Westview.

Coppedge, M.(1997). District Magnitude, Economic Performance and Party System
Fragmentation in Five Latin American Countries. Comparative Political Studies
30(2): 156-185.

164

Coppedge, M. (2000). Presidential Runoffs Do Not Fragment Legislative Party Systems.
Paper presented at the 2000 American Political Science Association Annual
meeting, Washington, D. C.

Cox, G. & Munger, M. (1989). Closeness, Expenditures, and Turnout in the 1982 US
House Elections. American Political Science Review 83: 217-231.

Cox, G. (1990). Centripetal and Centrifugal Forces in Electoral Systems. American
Journal of Political Science 34 (4), 903-935.

Cox, G. & Katz, J. (1996). Why Did Incumbency Advantage in US House Elections
Grow? American Journal of Political Science 40 (3): 478-497.

Cox, G.. 1997. Making Votes Count. Cambridge University Press.

Dalton, R., Flanagan, S. & Beck, P. (eds.). (1984). Electoral Change in Advanced
Industrial Democracies. Princeton University Press.

Down, A. (1957). An Economic Theory of Democracy.

Duverger, M. (1954). Political Parties: their Origins and Activity in the Modern State.
John Wiley & Sons Inc.

Duverger, M.. (1986). Duverger’s Law: Forty Years Later.” In: Grofman, B., Lijphart,
A. (Eds). Electoral Laws and Their Political Consequences. Agathon Press,
69-84.

Evans, G. & Whiteﬁeld, S. (1993). Identifying the Bases of Party Competition in Eastern
Europe. British Journal of Political Science 23 (4): 521-548.

Farrell, D. (1997). Comparing Electoral Systems. London: Prentice Hall.

Fey, M. (1997). Stability and Coordination in Duverger’s Law: A Formal Model of
Preelection Polls and Strategic Voting. American Political Science Review 91 (1):
135-147.

Fish, S. (1998). Reversal and Erosion of Democratization in the Postcommunist World.
Paper presented at the American Political Science Association Conference.

Gaines, B. & Nokken, T. (2001). The Presidential Shadow on Midterm House Elections:
Presidential Support, Presidential Agenda, Vote and Seat Loss. Paper presented at
the Midwest Political Science Association Annual Meeting.

Gaines, B. (1999). Duverger’s Law and the Meaning of Canadian Exceptionalism.
Comparative Political Studies 32 (7): 835-961.

165

 

Gaines, B. (1997). Where to Count Parties. Electoral Studies 16 (1): 49-58.

Geddes, B. (1991). Paradigms and Sand Castles in Comparative Politics of Developing
Areas. In William Crotty (Ed.). Political Science: Looking to the Future.
Northwestern University Press.

Gidengil, E. Blais, A., Nadeau, R. & Nevitte, N. (2000). Are Leaders Becoming More
Important to Vote Choice in Canada. Paper ppresented at the American Political
Science Association annual meeting.

Gierzynski, A. & Breaux, D. (1991). Money and Votes in State Legislative Elections.
Legislative Studies Quarterly 16: 203-217.

Grofman, B., Lee, S-C., Winckler, E., & Woodall, B. (1999). Elections in Japan, Korea
and Taiwan under the Single Non — Transferable Vote: The Comparative Stuay of
Embedded Institution. Ann Arbor: The University of Michigan Press.

Haggard, S. & McCubbins, M. (2001). Presidents, Parliaments and Policy. Londong:
Cambridge University Press.

Hermson, P. & Gimpel, J. (1995). District Conditions and Primary Divisiveness in
Congressional Elections. Political Research Quarterly 48 (March): 117-134.

Hickman, J. (1992). The Effect of Open Seats On Challenger Strength In Japanses Lower
House Elections. Legislative Studies Quarterly 17 (4), 573-584.

Hogan, R. (2001). Gubernatorial Coattail Effects in State Legislative Elections. Paper
presented at the Midwest Political Science Association Annual Meeting.

Holbrook, T. & Van Dunk, E. (1993). Electoral Competition in the American States.
American Political Science Review 87 (4): 955-962.

Jacobson, G. (1990). The Electoral Origins of Divided Government: Competitions in US
House Elections, 1946-1988. Boulder, CO: Westview Press.

Johnston, R. & Ballantyne, J. (1979). Geography and the Electoral System. Canadian
Journal of Political Science 77: 857-867.

Jones, M. (1995). Electoral Laws and the Survival of Presidential Democracies.
Notre Dame: University of Notre Dame Press.

Jones, M. (1997a). Federalism and the Number of Parties in Argentine Congressional
Elections. Journal of Politics 59 (2): 538-549.

166

Jones, M. (1997b). Racial Heterogeneity and the Effective Number of Candidates in

Majority Runoff Election: Evidence from Louisiana. Electoral Studies 16: 349-
358.

Jones, M. (1999). Electoral Laws and the Effective Number of Candidates in Presidential
Elections. Journal of Politics 61 (1): 171-184.

Jones, M. (2001). Electoral Institutions, Ethnic and Ideological Cleavages, and Candidate
Competition in Presidential Elections. Paper presented at the annual Western
Political Science Association Meeting.

Kang, W-T. (1998). The Rise of a Third Party in South Korea: the Uniﬁcation National
Party in the 1992 National Assembly Election. Electoral Studies 17 (1), 95-110.

Kang, W-T & Hun, J. (1999). The 1997 Presidential Election in South Korea. Electoral
Studies 18 (4), 599-608.

Katz, R. (1997). Democracy and Elections. London: Oxford University Press.

Kemell, S. & Jacobson, G. (1981). Strategy and Choice in Congressional Elections. New
Haven: Yale University Press.

Kim, H-M. (1997). Rational Choice Theory and Third World Politics: The 1990 Party
Merger in Korea. Comparative Politics 30 (1):83-100.

Kim, I. & Khil, Y. (1988). Political Change in Korea. Paragon.

Kim, J. & Ohn, M. (1992). A Theory of Minor-Party Persistence: Election Rules, Social
Cleavages and the Number of Parties. Social Forces. 70 (3), 575-599.

Kim, Y-H. (1998). Korea. In W. Sachsenroder and F rings (Eds.), Political Party Systems
and Democratic Development in East and Southeast Asia. Ashgate

King, G., Tomz, M. & Wittenberg, J. (2000). Making the Most of Statistical Analyses:
Improving Interpretations and Presentation. American Journal of Political Science
44 (2): 347-361.

Koetzle, W. (1998). The Impact of Constituency Diversity upon the Competitiveness of
US. House Elections, 1962-96. Legislative Studies Quarterly 33 (4): 561-573.

Laakso, M. & Taagepera, R. (1979). Effective Number of Parties: A Measure With
Application to West Europe. Comparative Political Studies 12 (1): 3-27.

LaPalombara, J. & Weiner, M. (1966). Political Parties and Political Development.
Princeton University Press.

167

 

Lau, R. & Redlawsk, D. (2001). Advantages and Disadvantages of Cognitive Heuristics
in Political Decision Making. American Journal of Political Science 45 (4): 951-
971.

Leamer, E. (1990). Let’s Take the Con Out of Econometrics. in C. W. Granger, (ed.),
Modelling Economic Series: Readings in Econometric Methodology. Oxford:
Clarendon Press. (29-49)

Lee, D-O. & Brunn, S. (1996). Politics and Regions in Korea: An Analysis of the Recent
Presidential Elections. Political Geography 15: 99-119.

Lee, K-Y. (1997). Hankookeui Sunkuwa Jiyulg'ueui (Korean Elections and Regionalism).
Seoul: Oreum Press.

Lewis-Beck, M. (1985). Economic Conditions and Electoral Outcomes. New York:
Agathon Press.

Lewis-Beck, M. & Paldam, M. (2000). Economic Voting: An Introduction. Electoral
Studies 19: 113-121.

Lijphart, A. & Grofman, B. (1984). Choosing an Electoral System: Issues and
Alternative. Preaeger.

Lij phart, A.. (1984). Democracies: Patterns of Majoritarian and Consensus Government
in Twenty — One Countries. New Haven: Yale University Press.

Lijphart, A. (1990). The Political Consequences of Electoral Laws, 1945-85. American
Political Science Review. 84 (2), 481-496.

Lij phart, A. (1994). Electoral Systems and Party Systems: A Study of Twenty-Seven
Democracies, 1945-1990. Oxford University Press.

Lihphart, A. (1997). The Difﬁcult Science of Electoral Systems: A Commentary on the
Critique by Alberto Penades. Electoral Studies 16: 73-77.

Linz, J. & Stepan, A. (1992). Political Identities and Electoral Consequences: Spain,
Soviet Union, and Yugoslavia. Deadalus 121 (2): 510-522.

Linz, J. (1994). Presidential or Parliamentary Democracy: Does It Make a DifferenCe? in
Juan J. Linz and Arturo Valenzuela (Eds), The Failure of Presidential Democracy.
Baltimore: Johns Hopkins University Press.

Lipset, S. & Rokkan, S. (1967). Party Systems and Voter Alignments: Cross-National
Perspective. Free Press.

Macdonald, D. (1990). The Korea: Contemporary Politics and Society. Westview.

168

Mainwaring, S. (1999). Rethinking Party Systems in the Third Wave of Democratization:
The Case of Brazil. Stanford University Press.

Mainwaring, S. & Shugart, M. (1997). Conclusion: Presidentialism and Party System.
In Scott P. Mainwaring and Matthew S. Shugart (eds). Presidentialism and
Democracy in Latin America. Cambridge University Press.

Mainwaring, S. & Scully, T. (1995). Building Democratic Institutions: Party Systems in
Latin America. Stanford University Press.

Mainwaring, S. (1993). Presidentialism, Multiparty Systems, and Democracy: the
Difficult Equation. Comparative Political Studies 26: 198-228.

Mainwaring, S. (1991). Politicians, Parties, and Electoral Systems: Brazil in Comparative
Perspective. Comparative Politics 25: 21-43.

McAllister, I. & Studlar, D. (1996). Region and Voting in Britain, 1979-1987: Territorial
Polarization or Artifact? American Journal of Political Science 36 (1): 168-199.

Miller, W. & Shanks, M. (1996). The New American Voter. Harvard University Press.

Mo, J ., Brady, D., & Jaehun, R. (1991). Urbanization and Voter Turnout in Korea: An
Update. Political Behavior 13 (1): 21-32.

Moe, T. & Caldwell, M. (1994). The Institutional Foundations of Democratic
Government: A Comparison of Presidential and Parliamentary Systems. Journal

of Institutional and Theoretical Economics 1 50: l 71 -1 95.

Morley, T. (1994). Federalism and the New Democratic Party. in Parties and Federalism
in Australia and Canada. Ed. Campbell Sharman. Australian National University.

Morris, P. (1996). Electoral Politics in South Korea. Electoral Studies 15 (4): 550-562.

Neto, O. & Cox, G. (1997). Electoral Institutions, Cleavage Structures, and the Number
of Parties. American Journal of Political Science 41 (1), 149-174.

North, D. (1990). Institutions, Institutional Change and Economic Performance.
Cambridge University Press.

Ordeshook, P. & Shevetsova, O. (1994). Ethnic Heterogeneity, District Magnitude, and
the Number of Parties. American Journal of Political Science 38 (1), 100-123.

Ordeshook, P. (1996). Russia’s Party System: Is Russian Federalism Viable? Post —
Soviet Affairs 12: 195-217.

Palfrey, T. (1989). A Mathematical Proof of Duverger’s Law. In Models of Strategic

169

Choice in Politics. (Eds), Peter Ordeshook. Ann Arbor: University of Michigan
Press.

Park, M (Forthcoming). Constituency Level Electoral Multipartism in Single Member
Simple Plurality Legislative Elections: Evidence from Korea. Party Politics.

Park, M (Forthcoming). Political Cleavage and Linkage of Party Systems between Local
National Levels: Evidence from Recent Korean Legislative Elections. Asian
Survey. .

Rae, D. (1967, 1971). The Political Consequences of Electoral Laws. New Haven: Yale
University Press.

Reed, S. (2001). Duverger’s Law is Working in Italy. Comparative Political Studies 34
(3): 312-327.

Remington, T. & Smith, S. (1996). Political Goals, Institutional Context and the Choice
of an Electoral System: the Russian Parliamentary Electoral Law. American
Journal of Political Science 40 (4): 1253-1279.

Reynolds, A. (2000). The Comparative Study of Electoral Systems: Is It Time for a Middle
— Life Crisis? Ppaper presented at the American Political Science Association
Annual meeting. '

Riker, .W. (1976). The Number of Political Parties. Comparative Political Studies 9 (1):
93-106.

Riker, W. (1982). The Two-Party System and Duverger's Law: An Essay on the History of
Political Science. American Political Science Review 76 (4): 753-766.

Riker, W. (1986). Duverger’s Law Revisited. In Bernard Grofman and Arend Lijphart
(Eds), Electoral Laws and Their Political Consequences. Agathon Press, Inc., pp.
19-42.

Rokkan, S. (1970). Citizens, Elections, and Parties. Mackay.

Samuels, D. (2000a). The Gubernatorial Coattail Effects: Federalism and Congressional
Elections in Brazil. Journal of Politics 62 (1): 240-253.

Samuels, D. (2000b). Concurrent elections, discordant results — Presidentialism,
Federalism, and Governance in Brazil. Comparative Politics 33 (1): 1-20.

Sartori, G. (1976). Parties and Party Systems. Cambridge University Press.

Sartori, G. (1986). The Inﬂuence of Electoral Systems: Faulty Laws or Faulty Method?
In Groﬁnan, B., and Lijphart, A., (Eds), Electoral Laws and Their Political

170

 

 

Consequences. Agathon Press, Inc., pp. 43-68.

Schlesinger, J. (1991). Political Parties and the Winning of Oﬂice. University of
Michigan Press.

Seligson, M. & Carrion, J. (2002). Political Support, Political Skepticism and Political
Stability in New Democracies: An Empirical Examination of Mass Support for
Coups d’etat in Peru. Comparative Political Studies 35 (1): 58-82.

Sharman, C. (1994). Parties and Federalism in Australia and Canada. Australian
National University.

Shepsle, K. & Bonchek, M. (1997). Analyzing Politics. Norton & Company.

Shugart, M. (1995). The Electoral Cycle and Institutional Sources of Divided Presidential
Government. American Political Science Review 89 (2): 327-343.

Shugart, M. & Carey, J. (1992). Presidents and Assemblies: Constitutional Design and
Electoral Dynamics. Cambridge University Press.

Shugart, M. & Taagepara, R. (1994). Plurality versus Majority Election of President: A
Proposal for a ‘Double Complement Rule. Comparative Political Studies 27: 323-
348.

Stimson, J. (1985). Regression in Space and Time: A Statistical Essay. American Journal
of Political Science 29: 914-947.

Siavelis, P. (1997). Continuity and Change in Chilean Party System: on the
Transformational Effects of Electoral Reform. Comparative Political Studies 30

(6): 651-674.

Sullivan, J. (1973). Political Correlates of Social, Economic and Religious Diversity in
the American States. Journal of Politics 35 (February): 70-84.

Taagepara, R. & Grofman, B. (1985). Rethinking Duverger’s Law: Predicting the
Effective Number of Parties in Plurality and RP Systems-Parties minus Issues
Equals One. European Journal of Political Research 13 (4): 341-3 52

Taagepera, R. & Shugart, M. (1989). Seats and Votes: The Seats and Determinants of
Electoral Systems. Yale University Press.

Taagepara, R. (1999). The Number of Parties as a Function of Heterogeneity and
Electoral System. Comparative Political Studies 32 (5): 531-548.

Taagepara, R. (2000). Complete Institutional Theory of Simple Electoral and Party

171

 

Systems. Paper ppresented at the Annual American Political Science Association
Meeting.

Tomz, M., Wittenberg, J. & King, G. (1999). CLARIFY: Software for Interpreting and
Presenting Statistical Results. Version 1.21. Harvard University, March 21.
http://Gking.harvard.edu

Van Dunk, E. & Weber, R. (1997). Constituency — Level Competition in the US States,
1968-1988: A Pooled Analysis. Legislative Studies Quarterly 32 (2): 141-159.

Wattenberg, M. (1990). The Decline of American Political Parties 1952-1988. Harvard
University Press.

Weisberg, H. (1991). The Inﬂuence of Economy on Party Identiﬁcation in the Reagan
Years. Journal of Politics 53 (4): 88-121.

Whitten, G. & Palmer, H. (1996). Heightening Comparativsts’ Concern for Model
Choice: Voting Behavior in Great Britain and the Netherlands. American Journal
of Political Science 40 (1): 231-260.

Wildavsky, A. (1959). A Methodological Critique of Duverger’s Political Parties.
Journal of Politics 21: 303-318.

Witheley, P. (1988). The Causal Relationship between Issues, Candidate Evaluations,
Party Identiﬁcation and Vote Choice — the View from Rolling Thunder. Journal
of Politics 50 (4): 961-984.

Wright, S. & Riker, W. (1989). Plurality Rule and Runoff Systems and Number of
Candidates. Public Choice 60: 155-175.

172