.é E I. . an.
u . ._ :0 clinic-:3, ¢ 5 9")!
41...!”

«c w .

i . .
k 1.3...

.6! .

r ” , .

.mmfivrdufir‘ as: 12.3 .

< 1. .s.u...n.
dmwfq ... é...)

 

(fit .
'HHJ.'.N

 

.ms .

v 39.»:

THESIS
2

J
"1

(303

This is to certify that the

dissertation entitled

-The Electoral Consequences Of Mixed Electoral Systems:
Contamination Effects And Strategic Behavior In Japan

presented by
Misa Nishikawa
has been accepted towards fulfillment

of the requirements for

Ph.D. degreein Political Science

 

 

\3 Major professor

Date June 21, 2002

MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771

 

 

 

 

LIBRARY
Michigan State
University

 

 

 

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

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJClRC/DateDuepss-p. 1 5

THE ELECTORAL CONSEQUENCES OF MIXED ELECTORAL SYSTEMS:
CONTAMINATION EFFECTS AND STRATEGIC BEHAVIOR IN JAPAN

By

Misa Nishikawa

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

THE ELECTORAL CONSEQUENCES OF MIXED ELECTORAL SYSTEMS:
CONTAMINATION EFFECTS AND STRATEGIC BEHAVIOR IN JAPAN
BY
Misa Nishikawa

Two types of mixed electoral systems have been used in Japan’s post-war
history. These mixed systems combine a proportional representation rule with
either a single member district system or a medium size district system to
decide seat allocations. The use of these mixed electoral systems has
produced a number of political parties that violates our expectations as stated
in Duverger’s Law and the M+1 Rule. In this dissertation, I explain why our
expectations do not hold up and replace them with different expectations that
are more sensitive to the institutional features of mixed electoral systems. The
focus of my argument is how the use of a proportional representation rule
contaminates the effects that are normally associated with plurality rules,
especially how voters contribute to violating Duverger’s Law and the M+1

Rule.

Copyright by
Misa Nishikawa
2002

For My Parents

and Brother

ACKNOWLEDGEMENTS

Many people have helped me to complete my Ph.D. program. However, I
owe special thanks to my advisor, Dennis Patterson, for being an excellent
mentor. Dennis has spent years helping me develop my intellectual ability.
When I first met Dennis, I was an undergraduate student in the Department of
Political Science at the University of California, Los Angeles. Since that time,
Dennis has shaped my academic career, and it was because of Dennis that I
was able to enter and finish the Ph.D. program in the Department of Political
Science at Michigan State University.

I also owe a great deal of thanks to Scott Gates, Mark Jones, and
Brian Silver, the members of my dissertation committee. My committee
members were always willing to help me to succeed in my work. They
supplied me with many constructive comments on my dissertation as well as
other work. Additional help and advice were given by Eric Browne, Mark
Elder, Timothy Frye, Jim Granato, Masaru Kohno, Tse-Min Lin, Charles
Ostrom, Steven Reed, Junichiro Wada, and Kenneth Williams.

My thanks also goes to the members of the Japan Election and
Democracy Study project, including Ken’ichi Ikeda, Susan Pharr, Bradley
Richardson, Aiji Tanaka, and my advisor, Dennis Patterson, who were
generous to make the JEDSQS survey data accessible before it became

available at the ICPSR data archive. I also thank Motoshi Suzuki for getting

me a permission to use the library at Kwansei Gakuin University in Japan. I
was also fortunate to have access to the library at Kobe Gakuin University.
Toshio Nagahisa and Reiko Niwa helped me to set up interview with Diet
members in Japan.

I also would like to express my gratitude to my friends in the
Department of Political Science at Michigan State University. Erik Herron and
I started to write a paper together on mixed electoral systems. Writing a paper
with Erik was not only great fun but also was a success in publishing and
winning an award.

Many of my friends gave me advice while I was in the process of
carrying out my research. These friends include Chris Bonneau, Chris Butler,
Jamie Carson, Justin Fox, Kevin Joldersma, Ko Maeda, and Becky Tothero.
Wonjae Hwang provided me with election data which saved my time
tremendously. I would also like to thank Sara Benesh, Michelle Kuenzi, and
Gina Lambright, my roommates who are from the Department of Political
Science at Michigan State University. I especially would like to thank Michelle
for spending many nights listening to me describe my research trials and
tribulations. Despite all of the efforts and advise by my friends, all errors and
omissions in my dissertation are my own.

Finally, I am deeply indebted to my parents and brother for their
untiring support. My parents not only encouraged me to achieve my
academic goals in the United States, but also provided me with both the
financial and psychological support I needed to complete my degree.

vi

TABLE OF CONTENTS

LIST OF TABLES

LIST OF FIGURES

Part I. Introduction

Chapter 1. Introduction

Chapter 2. Mixed Electoral Systems in Japan
History of Mixed Electoral Systems in Japan
Description of Japan’s Mixed Electoral Systems
Election Results

Part II. Explaining Japanese Party System

Chapter 3. Strategic Behavior of Political Parties

Duverger’s Law
Duverger’s Law and Mixed Superposition Systems

Contamination Effects and Candidate Nomination Process

Chapter 4. Strategic Behavior of Voters

Strategic Voting and Contamination Effects
Abstention and Contamination Effects
Impact of Sincere Voting on Number of Parties

Part III. Empirical Assessment of Strategic Behavior in Mixed
Superposition Electoral Systems

Chapter 5. Testing Parties’ Strategic Entries

Empirical Patterns of Candidate Placement
Hypotheses for Strategic Entries

Testing Hypotheses for Mixed Electoral Systems with SMD and PR

18

47

48

67

84
85

Testing Hypotheses for Mixed Electoral Systems with SNTV and PR

Summary

vii

Chapter 6. Voting Environment and Occurrence of Strategic Voting 106

Contamination Effects and Voting Conditions
Classification of Voters

Occurrence of Strategic Voting

Conclusion

Chapter 7. Impact of Sincere Voting 128
Significance of Opportunity Factor

SF Ratio Test
Conclusion
Implication
Part IV. Conclusion 153

Chapter 8. Conclusion 154

Impact of Superposition Systems on Democratic Performance
Generalizability of the Japanese Findings

Bibliography 172

Appendix 181

viii

LIST OF TABLES

2-1 Electoral Rules for the 1996 and 2000 Lower House Elections

2-2a The 1996 Lower House Election Results

2-2b The 2000 Lower House Election Results

2-3a Number of Seats by Political Parties in the 1983 Upper House Election
Results

2-3b Number of Seats by Political Parties in the 1983 Upper House Election
Results

2-4 Mean Effective Number of Legislative Parties for Different Electoral
Systems

2—5 Seat Distribution and Effective Number of Legislative Parties in the US.
2000 Congressional Election

2-6a Effective Number of Electoral Parties in the SMD Plurality Section of
Japan’s Lower House and US. Congress

2-6b Effective Number of Electoral Parties in the Plurality Section of Japan's
Upper House

2-6c Effective Number of Electoral Parties in the Single Nontransferable Vote
Systems-the Japanese Lower House Between 1983 and 1996

5-1 Patterns of Candidate Placement in the Lower House Elections-1996 and
2000

5-2a District Magnitude in the Plurality Section of the Upper House Elections
Between 1983 and 2001

5-2b Patterns of Candidate Placement-the Upper House Elections Between
1983 and 2001

5-2c District Magnitude and Patterns of Candidate Placement-the Upper
House Elections Between 1983 and 2001

5-3 OLS Results for the 1996 Lower House Election with INCUMBENT and
POPULARITY Variables

5-4 OLS Results for the 1996 Lower House without POPULARITY Variable
5-5 OLS Results for the 1995 Upper House Election with INCUMBENT and
POPULARITY Variables

5-6 OLS Results for the 1995 Upper House Election without INCUMBENT
Variable

6-1a Reduction in the Size of Potential Strategic Voters by Opportunity and
Marginality Factors - Lower House

6-1b Reduction in the Size of Potential Strategic Voters by Opportunity and
Marginality Factors - Upper House

6-2 Reduction in the Size of Potential Strategic Voters by Candidates'
Probability of Vlfinning and Utility Differences

6-3 Occurrence of Strategic Voting

7-1 Impacts of Presence of Most Preferred Competitor

LIST OF FIGURES

7-1a Bimodal Distribution of SF Ratios

7-1b Non-bimodal Distribution of SF Ratios

7-2a SF Ratios in the 1996 Lower House Election- 2-3 Candidates

7-2b SF Ratios in the 1996 Lower House Election- 4 Candidates

7-2c SF Ratios in the 1996 Lower House Election- 5 Candidates

7-2d SF Ratios in the 1996 Lower House Election- 6 Candidates

7-3a SF Ratios in the 1996 Upper House Election- 0 to 1 Excess Candidates
7-3b SF Ratios in the 1996 Upper House Election- 2 Excess Candidates
7—3c SF Ratios in the 1996 Upper House Election- 3 or More Excess
Candidates

Part I.

Introduction

Chapter 1

Introduction

In the last two decades, more and more democracies have adopted mixed
systems under which to conduct elections to their national legislatures.1
These mixed systems are composed of two distinct sets of electoral rules that
include district systems with a plurality rule and a proportional representation
(PR) rule.2 These more recent mixed systems are different from the more well
known form of mixed electoral systems, such as that found in Germany,
because in the German mixed system, the PR component is used to
compensate for the disproportionality that is caused by the single member
district (SMD) component of the system.3 Newer mixed election systems are
true parallel systems that were intended to operate independently of each
other.

Electoral institutions are usually changed with some political purpose in
mind. Such purposes typically involve attempts to improve the quality of a
nation’s electoral politics in some way, but changes in electoral rules are
tricky because they always touch directly on the electoral fortunes of elected
officials and the parties to which they belong. We know much about the

impacts that certain electoral institutions carry for political actors, but we know

 

‘ For example, these countries include Russia, Ukraine, Lithuania, Armenia, Azerbaijan and
Georgia.
2 Seats are distributed to political parties based on their vote shares and are allocated
proportionally to political parties.

With an SMD plurality rule, a candidate who gains the largest number of votes in a district
wins a seat.

little about the partisan consequences of mixed systems especially those
mixed electoral systems that have been adopted in the more recent past. The
focus of dissertation then is to demonstrate how these mixed systems
influence party systems.

The effect of mixed electoral systems is unclear so far because such
systems have not influenced party systems in the way that some scholars
expected. The cases of Japan (1993) and Russia (1994) adopting mixed
systems for elections to their Lower Houses are illustrative. Until 1993,
Japan’s Lower House held its postwar elections under a medium-sized district
system with a single nontransferable vote (SNTV) rule.4 In the late 1980s and
the early 19908, the Japanese government was under constant pressure from
the general public to carry out political reform because of repeated incidences
of corruption. One of the political reforms that the government chose to carry
out was a change in its electoral rules. It chose to do so since some scholars
suggested that its medium size district system promotes corruption. The
reason for this is related to the fact that major political parties have to place
more than one candidate in each district to win a majority in the Diet. The
resultant intra-party competition was said to promote corruption.

The Liberal Democratic Party (LDP), the largest party in Japan, tried to
adopt a mixed system that would reflect the characteristics of the regular

plurality SMD rule. LDP leaders advocated this system because they

 

‘ Under a medium size district system, multiple representatives are elected. For this reason,
the system should be called a multimember district system. However, I will use the name, the
medium size district system, in order to distinguish it from proportional representation rules.

expected such a system to give them a substantial seat bonus. LDP leaders
also expected the system to produce a two party system in which the LDP
could maintain its majority party status for at least the foreseeable future.
Although the regular SMD system was more attractive to the LDP members, a
mixed system with an SMD component was proposed instead. This is
because the mixed system was expected to have a greater likelihood of
passing in both houses. Having lost its majority in the Upper House of the
Diet in 1989, the LDP needed the support of other parties to pass an electoral
reform bill. The mixed system that was ultimately advanced utilizes a PR rule
that would distribute seats more proportionately with parties’ vote shares
compared to the regular SMD system. With a PR rule, small parties could
expect to survive and, as a result, were expected to support the mixed
system.

The reason why LDP leaders expected their party to benefit from the
use of the regular SMD system is found in scholarly writing. Many researchers
expected that the mixed system that Japan was to adopt would reduce the
number of parties. They came to this conclusion because they expected the
SMD portion of the system to be dominant. The reasoning behind this view is
based on Maurice Duverger’s idea that a plurality SMD system leads to two
party competition at the district level (Reed and Thies 2001). For example,
Murakami (1991) suggests that the mixed system would create a two-party

system in Japan because of the use of SMDs. Murakami also suggests that

 

The medium size district system uses a plurality rule, whereas the multimember district
systems usually use proportional representation rules.

the use of eleven separate electoral districts in the plurality section of the
system, instead of the use of a single national district, would further enhance
the system’s ability to reduce the number of competitive parties (Murakami,
1991). Miyake (1989) suggested that the mixed system’s ability to reduce the
number of parties will be stronger, compared to Japan’s old medium size
district system with SNTV.5 Budge et al. (1997) indicates that the mixed
system would work more like the plurality SMD rule except that the impact on
electoral outcome would be “softened.“5

The case of Russia also illustrates the point that political actors have
attempted to adopt an electoral system for some political purpose. When
democracy was established, reformers were divided over what kind of
electoral system would maximize their benefits, i.e. the number of seats.
Some reformers expected that they would do well in the next election if they
adopted the regular SMD system since the latest referendum at the time
showed that the Russian electorate was more pro-reform than anti-reform.
What this means is that, based on Duverger’s Law, these reformers expected
the system to produce a two party system in which reformers would dominate
(Remington and Smith 1996).

Although the regular SMD rule was attractive to these reformers, other

reformers in Russia were not sure about the benefits of the regular SMD

 

5 Miyake (1989) states that the impacts of the mixed system are closer to that of the plurality
SMD than that of the SNTV medium size district system when the constituency is divided into
districts by prefectures. Miyake also states that in the mixed system with compensatory seats,
the impacts come more from the PR portion of the system.

° Sato (1997) also reports that the electoral reform was carried out because the practitioners
expected a superposition system to produce a two party system.

system. Refon'ners understood that they would have difficulties in placing
candidates in some of the rural districts where communists were still
dominant, even though they knew that they would do well in large cities.
Moreover, reformers themselves were composed of a number of small parties
that may not do well under the regular SMD rule. Under such conditions, the
use of a mixed system with a PR rule appeared to be the best solution.
Russian reformers expected that they could benefit from the SMD rule in
Russia’s cities, while the plurality section of the system would reduce the
fragmentation of parties (Remington and Smith 1996, White, Wyman and
Kryshtanovskaya 1995).

Unfortunately for those involved in the creation of these two mixed
systems, election results have proven to be disappointing. The problem is that
the Japanese and Russian mixed systems have not reduced the number of
parties to the level that Duverger’s Law predicted. Briefly, the mixed system of
Japan’s Lower House produced a 4.1 effective number of political parties in
1996, while the SMD portion of the system alone produced a 2.9 effective
number of parties.7 This means that the SMD part of the system produced

one extra party than Duverger’s Law predicts.8 Russia's mixed system as a

 

7 The description on the number of parties will be seen in more detail in Chapter 2.

° There is a .9 difference between the number that Duverger’s Law predicts and the number
that the Japanese Lower House election produced. The effective number of 2.9 is
substantially larger compared to the numbers that other countries with plurality SMD have
produced. For example, the values for the US. House and Turkey are 2.1. The values for the
US. Senate and Norway are 2.2. The values for South Africa and Sweden are 2.3. New
Zealand has produced 2.4, and United Kingdom has produced 2.5. Substantively, the
difference of .9 means that the mixed system produced one extra party, compared to the
regular SMD system.

whole produced 5.6 in 1995, while the SMD part alone produced 4.7, that is,
at least two more parties than the law predicts.

One explanation for these unexpected results can be attributed to time,
that is, the data that are available now are insufficient to judge the long-term
effects of these new mixed electoral systems.9 The idea that electoral rules
take some time to shape party systems is also based on Duverger’s Law.
More specifically, Duverger’s Law is formulated on the idea that voters are
strategic in that they try not to waste their votes by supporting a competitor
who has little chance of winning. Moreover, political contributors will not waste
their resources by investing in competitors who are not likely to win. The point
is that such decisions may take some time because they require that voters
and contributors identify which candidates and parties will never be electorally
competitive.

Even though there may be some truth to the argument that electoral
institutions require more than one election to shape party systems, time may
not be the principal factor explaining why the Japanese and Russian Lower
Houses have not produced two party systems. This is because there is one
other case in which a mixed electoral system has not produced the expected

outcome, even though elections have been held a number of times. The

 

9 It is usually expected that electoral systems take some time to shape a party system.
According to Horie (1989), a simultaneous change of a party system was not seen in 1924
when a medium size district system replaced a single member district system in Japan. The
change from a two party system to a multiparty system was observed around 1931.

reference here is to Japanese Upper House elections which have produced
results that do not follow Duvergerian expectations.1o

Japan’s Upper House is composed of a medium size district system
with an SNTV rule and a PR rule.11 A large number of researchers accept the
idea that the number of parties should approach M+1 under the SNTV rule
where M is the district magnitude (lshikawa 1981; Miyake 1989; Reed 1990;
Cox 1997). The relationship between Duverger’s Law and the M+1 Rule is
that the M+1 Rule is considered to be a more generalized version of
Duverger's Law. This is because both the SMD system and the medium size
district system utilize a plurality rule. The difference is that a single winner is
produced in the SMD system and multiple winners are elected in the medium
sized district system.

What we see in the election results under the Upper House version of
the Japanese election system is that the numbers of competitors have long
been above the expected equilibrium. The effective number of parties for the
medium size district portion of the system in 1983 is 5.15 where the district
magnitude is 3.12 The numbers for the 1986 and 1992 were 4.9 in both years.
Overall, the mixed system for Japan’s Upper House has produced two more

effective parties than the M+1 Rule predicts.13 What these results suggest is

 

'° The Upper House has carried out elections by using a mixed system for six times.
" With the SMD rule, one representative is elected from each district, while two to four
representatives are elected with the medium size district system.

The score is the mean effective number of parties. The score is calculated based on
districts that have a district magnitude of 3.

3 The Lower House used a regular medium size district system with the SNTV rule, until it
adopted a mixed electoral system in 1994. The effective number of parties for the Lower
House Election in 1986 is 4.4. There is a .5 difference between the regular medium size
district system that is used in the Lower House before 1994 and the medium size district part

that researchers have missed a factor inherent in mixed systems that
produces more than the number of parties predicted by Duverger’s Law.

The purpose of this dissertation then is to show what is missing in
current assessments of mixed electoral systems and explains why these
systems will produce more than the number of parties that are specified in
Duverger’s Law and the M+1 Rule. I argue that Duverger’s Law and the M+1
Rule do not hold under the plurality component of mixed electoral systems
because the effects of the SMD elections or medium size district elections
(hereafter, I will refer the SMD and medium size district elections as plurality
elections) are contaminated by that of the PR elections.

The logic behind the contamination effect is quite intuitive. Voters have
two ballots under a mixed system, casting one ballot for a candidate under
the plurality section of the system and another for a party under the PR
component. Because voters make choices in the two parts of the system
simultaneously, the choice made in one part of the election is contaminated
by the choice of the other.

This simultaneous decision-making on the part of voters, in turn,
influences how parties expect voters to perceive their popularity. More
specifically, political parties have an incentive to place more candidates in the
plurality component of an election system because placing a candidate in
such districts will help them gain extra votes in the corresponding PR districts.

This is because political parties achieve more visibility in the PR districts by

 

of the mixed system. The difference seems to be important since the difference of .5 means
that we will have an additional small to medium size party under the mixed electoral system,

having candidates stand in the plurality portions of those same PR districts.
The only prerequisite for making individual candidates help a party in the PR
districts is that these candidates expose voters to their respective party
labels.“

For example, Mizusaki and Mori (1998) found that in the 1996
Japanese Lower House election, the placement of a candidate in a single
member district improved parties” performance in corresponding PR districts.
Mizusaki and Mori found such an improvement in party performance by
comparing a party’s results in districts in which it placed its candidate with
those in which it did not. Herron and Nishikawa (2001) extended the work of
Mizusaki and Mori by adding other controlling variables to the model. Herron
and Nishikawa also improved the previous studies by adding Russia’s Lower
House elections. Herron (2000) further developed the study of contamination
effects by applying the idea to a larger number of countries.

What we may conclude from this research is that an incentive exists for
parties to put up candidates in SMDs even if they have no chance of winning.
The reasoning behind such an incentive is that party leaders believe that
placing a candidate will give voters the impression that their respective parties
are performing well in the district and may also perform well in the election as
a whole. Contrarily, party leaders believe that if voters do not see a candidate

from their respective organizations, they may interpret this as a sign of

 

compared to the regular medium size district system.
“ Some researchers have applied this idea to mixed electoral systems that combine SMD
and PR rules found that the placement of a candidate in an SMD did indeed help that

10

weakness. In short, the placement of a candidate has an advertising effect for
that particular party. The creation of an incentive also inflates the number of
candidates because of the change in strategic entries of marginal
competitors.15

While we have learned much from existing research, it is necessary to
point out that previous studies have shown only that the number of
candidates who are placed in the SMD part of a mixed system will be larger
than Duverger's Law suggests. In light of this, existing research can be
improved in two ways. First, previous research has been carried out only on
mixed systems with an SMD rule. As a result, the idea that the plurality
section of a mixed system will violate Duverger’s Law has not been
generalized to those mixed systems that combine a medium size districts with
an SNTV rule and a PR rule. We should be able to observe the contamination
effect in other mixed systems that have two separate components that work
independently, because the violation of expected equilibria should be
applicable to other mixed systems.

Secondly, and perhaps more importantly, previous research has not
examined what role voters play in shaping party systems under these
condition. Existing research has only shown that the number of candidates

will be inflated because of the change in strategic entries of marginal

 

particular party to gain an extra vote shares in a corresponding PR district (Mizusaki and Mori
1998, Herron and Nishikawa 2001, Herron 2000).

'5 Mizusaki and Mori (1998) may not agree with this idea. Mizusaki and Mori may not agree
with this idea. They suggest that this tendency to improve the performance in the PR election
only helps large parties since small parties cannot place many candidates in the plurality
election.

11

competitors in the plurality section of a mixed system. It would be most useful
to determine how voters under such conditions cast their ballots for these
marginal competitors. As mentioned above, mixed systems create an
incentive for parties to place candidates in the plurality component of the
system. This change in incentive produces an opportunity to vote for a
marginal candidate who would not be running if an election was carried out in
a regular plurality system. If voters, who happen to prefer one of these extra
candidates, actually cast their ballots for these candidates, we can expect that
the number of parties will be inflated, leading to the violation of Duverger’s
Law. If voters do not vote for these candidates, the number of parties would
conform to the Duvergerian expectation.

The literature on strategic voting does not confirm that voters
necessarily cast their votes for these extra candidates. Strategic voting refers
to the voting pattern in which a voter casts her ballot for a less preferred
candidate.16 The voting literature suggests that voters do not waste their
votes by casting ballots for candidates who are manifestly unable to win.
Where voters prefer a candidate who is only marginally performing, they are
likely to either abstain from voting or vote for other candidates who are more
likely to win. This pattern of behavior is especially found in plurality systems.

Although no studies have been carried out to find out how voters play a
role in shaping party systems in mixed systems, a significant number of

scholars have argued that a substantial amount of strategic voting is also

 

1° The definition of strategic voting will be discussed more thoroughly later.

12

present in mixed systems (Barnes et al. 1962, Fisher 1973, Jesse 1988,
Bawn 1993 and 1999, Kohno 1997, Reed 1999). If we follow their argument
for now and assume that a significant amount of strategic voting is present, it
can lead us to believe that the number of parties may not be as inflated as we
may expect, even though mixed systems tend to increase the number of
candidates in plurality districts. This is because candidates who are more
likely to win get most vote shares, while marginally performing candidates
receive little to no support, adhering to Duverger’s Law or the M+1 Rule.

However, we wonder if we can accept above conclusion without
analyzing voters’ behavior further. This is especially true because we find
some problems in the approach that the existing studies followed. One
problem with existing research is that it assessed strategic voting without
considering how the two components in mixed systems may interact, so that
voting behavior in mixed systems may be different from that in regular
plurality systems. More specifically, existing research ignored the
contamination effect of the PR district elections on the plurality elections. It
would be more appropriate to assess voters’ behavior by taking the
contamination effect into account.

The other problem with previous studies is that it tested strategic voting
only by looking at district level data. Since strategic voting is an individual
level phenomenon, analysis with district level data may result in an ecological
inference problem. The main purpose of my dissertation then is to assess

voting behavior in mixed systems by taking the contamination effect into

13

account and assess voting behavior with both individual level and aggregate
level data. By doing so, I will show that voters do not vote strategically in the
plurality component of mixed systems as originally thought.

To address the problems that have been described above, this
dissertation will be organized as follows. Chapter 2 provides an overview of
Japan’s two mixed electoral systems. The first system refers to Japan’s Lower
House (House of Representatives), which is a mixed system composed of a
plurality SMD rule and a PR rule. The second system refers to Japan’s Upper
House (House of Councilors), which is a mixed system that uses an SNTV
rule and a PR rule. This chapter will provide the description of Japanese
mixed systems and the empirical descriptions of the results produced under
these rules with the idea being to show how Duvergerian expectations never
accurately described election results produced in these mixed rule systems.

Chapters 3 and 4 will describe the theory that explains why the plurality
portions of Japan’s mixed superposition systems violate Duverger’s Law and
the M+1 Rule. In these two chapters, I will discuss the idea behind
contamination effects in mixed systems in more detail. I will do this by
discussing how the behavior of political parties and voters in a mixed system
deviates from the behavior that Duverger’s Law predicts. I will show that
these deviations take place because of the contamination effects by the
presence of PR districts. These deviations in voter behavior in turn lead

mixed electoral systems to produce a higher number of political parties.

14

The strategic behavior of parties will be discussed in Chapter 3 while
the strategic behavior of voters will be discussed in Chapter 4. In Chapter 3, I
will discuss how the presence of the PR districts contaminates the behavior of
political parties in the plurality districts. I will especially focus on the fact that
extra vote shares will be obtained in the corresponding PR district by placing
a candidate in the plurality portion of the system. This promotes political
parties to place their candidates in the plurality section of the system. For
party decisions on candidate entries, it is important to focus on candidates
who are only performing marginally since the placement of marginal
competitors plays a key role in increasing the opportunity for voters to spread
their votes to minor parties. I will also generalize the idea that parties with
marginal competitors develop an incentive to place these competitors in the
plurality component of a mixed system with a medium size district rule.

In Chapter 4, I will show that strategic voting is not a dominant choice
among voters in mixed electoral systems. I will do so especially by describing
how two features of a mixed system prevent voters from behaving
strategically. The first is that mixed electoral systems create voting conditions
in which voters cannot easily engage in strategic voting. Briefly, one of the
most important factors that specifically reduces the occurrence of strategic
voting in mixed systems is that they create the condition in which some voters
have an opportunity to vote for their most preferred candidate, while others do
not. This will make strategic voting impossible to occur among the latter group

of voters.

15

The second is that mixed systems encourage voters to vote for
marginal competitors even when the voting environment allows voters to
behave strategically. In particular, I will argue that the presence of PR districts
encourages voters to cast their ballots for a marginal candidate in systems.

In Chapter 5, I will derive two hypotheses based on the theory of
strategic entry of political parties from Chapter 3. The first hypothesis will deal
with parties’ strategic entries in the mixed system that combines an SMD rule
and a PR rule. This hypothesis applies directly to Japan’s Lower House,
which uses a mixed system with SMD and PR rules. The second hypothesis
will deal with strategic entries in a mixed system with an SNTV rule and a PR
rule. This is one that relates to Japan’s Upper House elections. After deriving
the hypotheses, I will describe the methods and data that will be used to test
these two hypotheses for parties’ strategic entries. Specifically, electoral
returns at the district level will be utilized for these hypotheses. The results of
empirical tests conducted in this chapter will show that parties have an
incentive to place their candidates in the plurality portions of mixed systems,
even when these candidates have little chance of winning a seat.

In Chapter 6, I will empirically support that the plurality component of
the systems create a voting environment that prevents voters from engaging
in strategic behavior. I will also show that the occurrence of strategic voting is
rare in mixed electoral systems, so that they do not reduce the number of

parties as Duverger’s Law and the M+1 Rule suggest.

16

Chapter 7 will be devoted to testing the impact of sincere voting on
party systems. Although Chapter 6 will be informative in showing that voters
rarely vote strategically in mixed systems, it is not sufficient to say that a
significant number of voters cast their ballots sincerely so that the number of
parties will be inflated.17 In Chapter 7, I will show how the increase in the
number of candidates is directly related to the inflation of the number of
parties in mixed electoral systems.

Chapter 8 will conclude this dissertation by summarizing its findings
and discussing their implications. In particular, I will discuss potential impact
of mixed systems on democratic performance. Finally, I will end the
discussion by considering the generalizability of the study findings in Japan to

other countries.

 

'7 Sincere voting refers to the voting pattern in which a voter casts her ballot for her most
Preferred competitor.

17

Chapter 2

Mixed Electoral Systems in Japan

Although studying Japan’s mixed systems in the Lower and Upper Houses is
attractive, we need to be cautious about the comparability of the two systems.
This is especially true because the electoral rules of mixed systems are rather
complex, and moreover, the rules have been changed over time in Japan. In
addition to the description of Japan’s mixed systems and the comparison of
the systems between the Lower and the Upper Houses, they should be
compared with mixed systems in other countries. There are many types of
mixed systems in the world today, and a significant number of studies have
been carried out to study these systems. This suggests that it is necessary to
clarify what type of mixed systems Japan uses and how they are classified in
the literature on mixed systems.

For example, Japan’s mixed systems are different from the system that
has been traditionally used in Germany. Japan’s systems do not use the PR
component of the system to compensate for disproportionality that is created
in the plurality section, whereas the German system corrects for such
disproportionality. Moreover, there are other rules that vary across different
countries, and as a result, other factors that determine the classification of
Japan’s mixed systems along with whether or not Japan’s mixed systems

have a compensatory mechanism should be examined cross-nationally.

18

Another important task that needs to be completed before we start to
analyze the impact of Japan’s mixed systems is the description of election
outcomes. In Chapter 1, I briefly sketched election outcomes in terms of the
effective number of parties and argued that Japan’s mixed system has
produced the results that do not follow our expectation based on Duverger’s
Law and the M+1 Rule. However, a more complete discussion of election
outcomes is essential. This is especially because since there are some
scholars who argue that the plurality component of the Lower House elections
does not violate a Duvergerian equilibrium (Kohno 1997, Reed 2001). For
example, Kohno states that a mixed electoral system as a whole produces a
multiparty system. However, the single member district (SMD) part produces
a two party system, suggesting that the system does not violate the
Duvergerian expectation.

Under such circumstances, it is important to describe both the plurality
component of the election outcomes and the proportional representation (PR)
section more carefully and show that the plurality component has produced a
party system that does not conform to a Duvergerian equilibrium. In particular,
I will show the tendency that the plurality section of a mixed system produces
a larger number of parties is maintained even when it is compared with the
election results under regular plurality systems.

There is another reason why the description of election outcomes
should be carefully done before we begin the analysis on how mixed systems

influence party systems. This has to do with the fact that election rules take a

19

while in shaping party systems. As was mentioned earlier, only two elections
were carried out under a mixed electoral system in the Lower House,
including the 1996 and 2000 elections. The outcomes so far support the idea
that mixed systems inflate the number of parties. Arguably, however, this
does not necessarily lead us to conclude that mixed systems will lead to the
inflation of the number of parties. This is because the mixed system may
produce a two party system if enough elections are carried out in the future.

Under the situation, it is important to analyze the Upper House election
outcomes more carefully since mixed electoral systems have been used a
number of times in the Upper House elections. If the outcomes in these
elections show that the number of parties is inflated, it would be stronger
evidence to support that the plurality component of the mixed system inflates
the number of parties.

Given the concerns stated above, I will describe Japan’s systems in
more detail including the background on how they were adopted and will add
the information on where Japan’s system fit in the world’s mixed systems.
Second, I will describe Japan’s election outcomes in the Lower and Upper

House elections.

History of Mixed Electoral Systems in Japan
Mixed electoral systems have been used in the most part of Japan’s
postwar history, during which Japan experienced variations of mixed systems.

The first mixed system was used in the 1950 Upper House election. The

20

mixed system at the time combined a nation-wide multimember district
system with a plurality rule, i.e. single non-transferable vote rule (SNTV), and
forty-seven multimember prefecture districts with a plurality rule. The nation-
wide district elected 56 councilors, while the prefecture districts elected 76
councilors (Asahi Nenkan 1951).

In 1981, the Liberal Democratic Party (LDP) proposed to change the
plurality rule that was used for the nation-wide multimember district rule to a
closed-list PR rule (Asahi Nenkan 1982). The LDP proposed to change the
old system because many politicians complained that the system was one of
the worst electoral rules because of the nonpartisan nature of the election.
Moreover, the old system required an unusual amount of funding since
candidates had to cover the entire nation-wide district in their campaign, while
the system promoted intra-party competition (Ramseyer and Rosenbluth
1993).

Despite of the disadvantage of the old mixed system, some small
parties opposed to the proposal to change the nation-wide multimember
district. These parties included the Clean Government Party (CGP), Minshato,
Japan Communist Party (JCP), Niin Club, and lchinokai. These parties were
against the proposal for a number of reasons. First, they argued that the party
oriented aspect of the closed list goes against the idea of practicing
bicameralism (Asahi Nenkan 1982). They argued this because they thought

that the elections of Lower House at the time promoted party based

21

elections.18 As a result, making the Upper House election more party base
elections would produce election results that are similar to those of the Lower
House elections. Second, small parties argued that the proposed new system
would constrain candidacy of particular group of candidates including those
who are unaffiliated with a political party and those who belong to a small
party that does not fulfill the requirement to become an official party. As a
result, the proposed rule violated “fairness under the law” and “freedom of
organization” (Asahi Nenkan 1982). Despite the opposition by the small
parties, the proposal was passed in 1982 (Asahi Nenkan 1983). This was
made possible because the LDP made an alliance with the New Liberal Club
(NLC), which was composed of the former LDP members.

By the end of the 1990s, the LDP again proposed to change the
system in the Upper House. This time, the LDP produced to use an open list
PR system instead of the closed list PR system, while the multimember
prefecture districts with an SNTV rule was retained. The LDP argued that the
closed list PR rule deprived voters of the right to choose individual candidates.
However, if an open list PR rule is adopted, it would give voters a chance to
select a candidate since voters can choose a candidate on a party list. These
votes in turn are used to determine the ranking of the candidates within the

party list. Moreover, an open list PR rule can promote party based elections,

 

'8 This may be a debatable argument since the Lower House of the time used a multimember
district system with an SNTV rule where average district magnitude was between three and
four. As a result, the SNTV system also promoted intra-party competitions, since major
parties needed to place more than one candidate in each district to win a majority.

22

since these ballots for a particular candidate were counted toward a party that
the candidate has an affiliation.

Although the argument sounded plausible, the real purpose of the LDP
proposing the open list PR was to save the LDP from losing seats in future
elections. The LDP was suffering from a declining level of party support in the
19903, and the LDP leaders thought that they could save the LDP seats in the
Upper House if the election was carried out under an open list PR rule. The
LDP leaders expected that the new system would be beneficial to them since
they could nominate well-known candidates, including professional sports
players, academics, and commentators.

The proposal was passed in both houses in 2000, and the first election
under an open list PR system was carried out in 2001. As a result, the Upper
House currently uses a nation-wide district with an open-list PR rule and forty-
seven multimember prefecture districts with an SNTV rule.

The Lower House started to use a mixed electoral system more than
four decades after it was first used in the Upper House. The Lower House
had used the SNTV system most of Japan’s postwar period. Despite the use
of the SNTV system for such a long period of time, it was changed because it
was said to promote corruption. The problem with the SNTV system was that
it promoted intra-party competition which was led to promote corruption since
politicians depended on pork barrel type politics to win an election. The LDP
leaders originally preferred to adopt the SMD system to replace the SNTV

system since the SMD system would distribute more seats to the LDP relative

23

to their vote shares. Because of the benefits with the SMD system, the LDP
leaders tried to adopt the SMD system as early as 1956.

As we can imagine, however, the political parties that would not benefit
from the use of the SMD system opposed to adopt the SMD system. These
parties include the Japan Socialist Party (Social Democratic Party after 1996)
and the Japan Communist Party (JCP). Because of the resistance by the
opposition parties, the LDP could not pass the bill at the time.

Being unable to pass the bill for the SMD system, the LDP leaders
started to consider the use of a mixed system in 1962 along with a PR system
(Reed and Thies 2001) since minor parties can secure their seats through PR
elections. One of the mixed systems under the consideration was a system
without a compensatory mechanism, while the other was with a compensatory
mechanism. The former type was mainly supported by the LDP, while minor
parties supported the latter type. However, some parties including the Clean
Government Party (CGP) opposed any of these proposals. The First Advisory
Committee was formed to promote the electoral reform. However, the parties
never came to a consensus.19 Despite the effort, the SNTV rule was kept
unchanged, except that its districts were redrawn from time to time.

The desire to change the SNTV system, however, became prominent
again in the early 19905. This was especially because the on-going corruption
was reducing the popular support of the LDP and was eroding political trust

among the general public. Under the situation, the opposition parties,

 

1° Since mixed systems were first considered1956, eight Advisory Committees were formed
until 1992.

24

including the socialist party and the CGP came to a consensus that they
would support the mixed electoral system with a compensatory mechanism.
It is especially meaningful that the CGP agreed to adopt a mixed system with
a compensatory mechanism this time.

However, the cabinet of the time that was led by the Prime Minister
Miyazawa introduced the bill for the SMD plurality rule, even though the
cabinet members expected the opposition by the socialist and the CGP
members. It was also expected that some of the LDP members would oppose
the bill. As a result, the bill did not pass. Moreover, some significant number
of the LDP members left the party in protest and formed new parties. This led
to the creation of the non-LDP alliance in the Diet. The only party that did not
join the alliance was the JCP. Under the non-LDP coalition, the Lower House
successfully adopted a mixed system that combines an SMD plurality rule and
a PR rule in 1994. The first election under the mixed system was carried out
in 1996.

The Lower House has experienced some changes in the rules since
1996. Before the second election in 2000, the LDP proposed to reduce the
number of seats that were allocated by the PR section of the system from two
hundred to one hundred and eighty. This was carried out in part to increase
the level of public trust toward politicians. Japan has been in recession since
the beginning of the 1990s, and a large number of businesses have been
cutting the number of employees to lower the cost and make their businesses

work more efficiently. The government needed to appeal to the public by

25

showing that it also was suffering with the public and was trying to make the
government function more efficiently by cutting the number of the Diet
members. Although this argument sounds convincing, it was proposed
because it would increase the proportion of seats that was held by the LDP

members.

Description of Japan’s Mixed Electoral Systems

Japan's mixed electoral systems have been labeled in many ways. For
example, they are often referred to as parallel systems. This seems to be a
direct translation of the Japanese name, Hein'itsusei, in which Hein’tsu means
parallel, while Sei means a system. The name, Heiritsusei, is used basically
among Japan specialists.

The Japanese mixed systems are also called mixed superposition
systems (Massicotte and Blais 1999)20 or mixed member majoritarian
systems (Shugart and Wattenberg 2001). Superposition systems are defined
as rules in which voters are subjected to two different electoral formulae that
involve two independent sets of districts (Massicotte and Blais 1999). On the
other hand, mixed member majoritarian systems are defined as systems that
combine two different types of rules, such as a plurality rule and a PR rule
(Shugart and Wattenberg 2001).

Between the two definitions, I use Massicotte and Blais’s definition and

call the Japanese mixed systems superposition systems for the purpose of

26

simplicity. This is because their definition encompasses all of the Japanese
mixed systems into one group, while the definition by Shugart and
Wattenberg does not.21 According to the definitions above, the first mixed
system that the Upper House adopted in 1950, as well as the system that was
adopted in 1983, would be considered as a superposition system. On the
other hand, the rule adopted in 1950 is not classified as a mixed member
majoritarian system, while the rule that is adopted in 1983 is a mixed member
majoritarian system. This is because the system of 1950 used two plurality
rules, including the nation-wide multimember district system and the
prefecture districts. Since both parts use a plurality rule, it will not be classified
as a mixed member majoritarian system, even though the 1950 system
combines two sets of rules that greatly differ in terms of the district
magnitude.

Table 2-1 shows the electoral rules of Japan’s Lower and Upper
Houses. Among the mixed superposition systems that were used in Japan’s
postwar history, I will focus on two systems. The first is the 1996 Lower
House election. The 1996 election is chosen mainly because of the
availability of public opinion data. A survey for the 1996 election is publicly

available while that for the 2000 election has not been available yet.

 

2° Massicotte and Blais (1999) define superposition systems in which "all electors are
subjected to the two formulas and independent sets of districts are established for each
formula.”

2' See Appendices 2-1 and 2-2 for the countries that fit into the two definitions.

27

Table 2-1: Electoral Rules for the 1996 and 2000 Lower House Elections

 

 

 

 

 

 

 

 

 

 

 

LOWER HOUSE UPPER HOUSE
ELECTION ELECTION
Term of Office Four years Six years (A half of the
Councilors’ Membership
expires every three years)

Number of SMDf: 300 MSW; 47

Districts PR: 11 PR: 1

No Seats SMD: 300 MSD: 76
PR: 200c PR: 50

District SMD: 1 MSD: 1 to 4

Magnitude PR: 11 PR: 50

Ballot One ballot for a candidate One ballot for a candidate
Other ballot for a party Other ballot for a party

Double Allowed for candidates who Not allowed

Candidacy belong to a party

Threshold for a None None

Seat

Qualifications for SMD: MSD:

Political Party 1. Five representatives 1. Five representatives
belong to the organization; belong to the organization;
or or
2. Has more than 2% of 2. Has more than 2% of vote
vote share in any plurality share in any plurality or PR
or PR election in either election in either Lower or
Lower or Upper House. Upper House.

PR: PR:

Qualifications 1 and 2 Qualifications 1 and 2 above,

above,or or

3. The number of cands. in 3. The number of cands. on

each district exceeds 20% the PR list in the Upper

of the district magnitude. House election exceeds 10.
Date Adopted 1994 1983

 

 

 

a: SMD: Single member district. b: MSD: Medium size district. 0: The number
was reduced to 180 before the 2000 election.
Source: Senkyoseido Kenkyukai

28

 

The second is the system that was used in the Upper House elections
between 1983 and 1993. As discussed earlier, the mixed system in this
period applied a closed list PR system. This period is chosen for the Upper
House instead of the 2001 election for the same reason that I chose the 1996
election in the Lower House. A public opinion survey is available for an
election in this period, while no survey is accessible for the 2001 election in
which the Upper House started to use an open list PR system.

The table shows that even though certain features of electoral rules
differ between the Lower and the Upper Houses, they seem to cause no
serious problems for us to examine how Duverger's Law and the M+1 Rule
may work in the two houses. More specifically, the seat allocation mechanism
of the two houses is similar in how it utilizes a plurality rule and a PR rule. The
Lower House allocates 300 seats by an SMD rule, while 200 seats are
decided by a PR rule.22 On the other hand, the Upper House allocates 75
seats by the multimember districts with an SNTV rule, while 50 seats are
decided by a PR rule. In both houses, a plurality rule determines 1.5 times
more than a PR rule. What this may suggest is that the ability of a PR rule to
inflate the number of parties may be similar in the two houses.

One of the features that can cause the two houses to develop a
different incentive pattern is in the use of double candidacies. Double
candidacies refer to the practice in which candidates are allowed to run in

both a plurality election and a PR election. The Lower House allows double

 

2’ The number of seats in the PR portion in the Lower House has been reduced from two
hundred to one hundred and eighty seats.

29

candidacies, while the Upper House does not. We can expect that the cost of
parties in placing candidates in plurality districts would be lower if double
candidacy is allowed. This may suggest that the Lower House is more
inclined to inflate the number of parties than the Upper House.

In the Lower House elections, most of the political parties practiced
double candidacy. Of those parties, the Liberal Democratic Party (LDP) and
the Democratic Party of Japan (DPJ) utilized double candidacy the most in
the 1996 election. About 80% of the LDP candidates who are on the PR list
also entered into SMD elections, and 90% of the DPJ candidates did so.

Another feature that we should be aware of concerns district
magnitude. The number of districts for the PR component in the Lower House
is 11, while that in the Upper House is 1. What this may suggest is that a
larger number of parties may exist in the PR section of the Upper House,
compared to that of the Lower House. As a result, the inflation in the number
of candidates in the plurality section of the system may be larger in the Upper
House than the Lower House. This suggests that the intensity of the
contamination effect may be stronger in the Upper House. However, since the
Upper House uses an SNTV system for the plurality component of the
election, whereas the Lower House uses an SMD plurality rule, the intensity
of the contamination effect from the use of the nationwide district in the Upper
House may cancel out. This is because the ability to reduce the number of

parties in the SNTV election is lower than the SMD elections.

30

Election Results

There may be a number of ways to describe election outcomes and to
show what party systems Japan’s superposition systems may have produced.
I will mainly depend on two ways to describe election outcomes. One is to
describe the election outcomes by looking at seat shares of each party as we
see in regular news reports. Observing an election outcome in this way will be
able to give us a basic understanding on what party systems are produced.

Although such basic reports are important, they are difficult to be used
when we need to compare a large number of elections. The second way may
make the comparisons easier by using an indicator such as the effective
number of parties (Laakso and Taagepera 1979).23 I will calculate both the
effective number of legislative parties and the effective number of elective
parties. If we use seat shares, the effective number of legislative parties is
calculated, while the use of vote shares calculates the effective number of
elective parties.”

The use of both types of effective number of parties is useful in this
study since they can reveal the effects of mixed superposition systems in a
meaningful way, since it is important to analyze how voters behave and how
their votes are translated into seats. The idea supported in this study is that

political parties have an incentive in placing candidates in the plurality section

 

23 The effective number of parties is calculated with the following formula: N=1/}Zp,2 where p.
stands for fractional shares of each party. The effective number of parties is created based on
the inverse of Herfindahl- Hirschman concentration index, 2p.2 , where p. stands for the
fractional share of .

2"The effective number of parties is often used to measure party systems. Another indicator
is the Molinar index.

31

of superposition systems, and as a result, the number of candidates is
inflated. The increase in the number of candidates leads to the inflation of
number of parties since voters are willing to vote for marginal candidates.

If the idea is true, the number of legislative parties in the plurality
component of superposition systems may not be greatly different from other
regular plurality systems, although we may still see small differences. On the
other hand, we should see a substantial gap between the plurality section of
superposition systems and regular plurality systems when we use the
effective number of elective parties. This is because the use of plurality rule in
distributing seats in a superposition system will be unfriendly to marginal
parties, and it creates relatively a large number of wasted votes in the plurality
component of the system. The share of wasted votes is large in the plurality
section of a superposition system because voters are more willing to cast
their ballots for marginal candidates since marginal parties can survive
through the plurality section of the system.

What this process suggests is that the number of legislative parties will
not be inflated greatly in the plurality component of superposition systems,
even when voters cast their ballots for marginal candidates. On the other
hand, the number of elective parties as an indicator should show a substantial
difference between the plurality section of a superposition system and a
regular plurality system because it is calculated based on vote shares.

The election outcomes described with basic reports and the effective

number of parties supports the idea that the plurality component of

32

superposition systems produces a larger number of parties than other regular
plurality systems. This tendency is especially true when we look at the
effective number of electoral parties as we expect.

Tables 2-2a and 2-2b and Tables 2-3a and 2-3b show the raw election
results for the Lower and Upper House elections respectively. In Tables 2-23
and 2-2b, Japan’s Lower House elections of 1996 and 2000 are reported,
while Tables 2-3a and 2-3b report the results for the Upper House elections
between 1983 and 1998. For the Lower House, both elections are important
to be reported since the 1996 election will be used to carry out individual level
analysis in later chapters. It was also the first superposition system in the
Lower House. The 2000 election is important since it is the most recent
election for Japan’s Lower House. The report on the 1996 and 2000 elections
together should give us a basic idea of how the superposition system have
shaped the party systems to date. More importantly, however, these are the
only two elections that were carried out under a superposition system in the
Lower House.

The two years for the Upper House are selected for the same reason.
The 1983 election is important since it is the first election under the
superposition system with an SNTV rule and a closed list PR rule. The 1983
election is also important since it will be used for individual level analysis in
later chapters. The 1998 election is selected because it is the last election

under the closed list PR rule. As is the case for the Lower House elections,

33

Table 2-2a: The 1996 Lower House Election Results

 

 

 

 

 

 

 

 

 

 

 

I SEAT TOTAL SEAT PR SEAT SMD
ILDP 239 70 169

INFP 156 60 96

IDPJ 52 35 17

“JCP 26 24 2

IDSPJ 15 11 4

INPH 2 2
[Minkairen 1 1
Unaffiliated 9 9

|Total 491 200 300

 

 

 

 

 

34

Table 2-2b: The 2000 Lower House Election Results

 

 

 

 

 

 

 

 

 

 

 

 

 

FOLITICAL SEAT SEAT SEAT
PARTIES TOTAL PR SMD
[LDP 233 56 177
IDPJ 127 47 80
"CGP 31 24 7
Liberal Party 22 18 4

CP 20 20 0
SDPJ 19 15 4
[Conservative Party 7 0 7
riyu Rengo 1 0 1
lMushozokunokai 5 0 5
lUnaffiliated 15 0 15
[Total 480 180 300

 

 

 

 

 

 

35

Table 2-3a: Number of Seats by Political Parties in the 1983 Upper
House Election Results

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

POLITICAL
PARTIES SEAT TOTAL SEAT PR SEAT SNTV
IILDP I68 19 49
"JSP 22 9 13
CGP 14 8 6
IIMinsha 6 4 2
JCP 7 5 2
Shinjiyukurabu 2 1 1
IlMinshurang
lNiinkurabu 1 1
Fukushi ‘I 1
ISalaryman 2 2

thers 2 2
IUnaffiliated 1 1
ITotal 126 50 76

 

 

 

 

 

36

 

Table 2-3b: Number of Seats by Political Parties in the 1998 Upper
House Election Results

 

 

 

 

 

 

 

 

 

“LPOLITICAL

PARTIES SEAT TOTAL SEAT PR SEAT SNTV
ILDP 44 14 30

“DPJ 27 12 15

ICGP 9 7 2

SDPJ 5 4 1

JCP 15 18 7

“Liberal Party 6 5 1

Unaffiliated 26 0 20

 

 

 

 

 

37

 

the observation of the two elections should give us a basic idea on how party
systems, or have not, changed over time.

In both 1996 and 2000 Lower House elections, a large number of
parties were successful gaining seats in the Lower House, and we can
observe that the superposition systems produced multiparty systems. This
tendency seems to be kept unchanged between the two time points. Seven
parties are successful obtaining a seat in the 1996 election, while there are 9
successful parties in the 2000 election. This tendency to produce a multiparty
system is also true even when we look at the two parts of the system
independently. The number of parties that were successful in obtaining seats
in the plurality section of the system is 6 in the 1996 election, while it
increased to 7 in the 2000 election. For the SMD plurality section, the number
of parties with seats is 7 parties in 1996, while that in the 2000 election is 8
parties.

Table 2-3a and 2-3b also reveal that superposition systems produced
a large number of parties in the Upper House. The system with SNTV and PR
rules together produced 11 parties in the 1983 election, while 9 parties were
produced in the 1998 election. The tendency to produce a large number of
parties is seen even when we look at each part of the system separately as it
is the case in the Lower House elections. The numbers of parties with seats
in the plurality section are 9 and 6 parties in the 1983 and 1998 elections
respectively. The multimember district with a plurality rule produced 8 in the

1983 election and 6 parties in 1998 election.

38

Table 2-4 gives us an idea where the Japanese superposition systems
are located in terms of their legislative number of parties compared to those
of other systems in the world. The comparison between Japan’s superposition
systems with the superposition systems of the world shows that Japan’s
Lower House produced results that are slightly larger than the world’s
average. The effective numbers of legislative parties in Japan’s Lower House
is 2.9, while that of the world is 2.8. The effective number for the Upper
House is also larger than the world average. The figure for Japan is 3.1. The
Upper House produced a larger number probably because it uses a
multimember district system with an SNTV rule instead of the SMD plurality
rule.

It is also worthwhile noting that the average effective numbers of
legislative parties of corrective systems and PR list systems are the only
systems that produced a substantially larger number of parties compared to
superposition systems. The figures for corrective systems and PR list systems
are 3.6 and 3.7.

The election outcomes in the table also suggest that the SMD part of
Japan’s Lower House produced a larger number of legislative parties than
regular SMD systems of the world. However, the difference is not so large as
we expected. A small gap is seen because the measurement is based on
seat shares instead of vote shares. The effective number of legislative parties

in Japan’s Lower House is 2.2, while that in regular SMD systems is 2.0.

39

Table 2-4: Mean Effective Number of Legislative Parties for Different
Electoral Systems

 

 

 

 

 

 

 

 

 

 

 

 

TYPES OF ELECTORAL MEAN NS
SYSTEMS
Japan’s Superposition Lower House 1996, 2000 2.9
Systems (Two Parts)
Upper House 1983~1998 3.1
(Two Parts)
Lower House 1996, 2000 2.2
(SMD plurality section)
Mixed Systems from Other Superposition (Two parts) 2.8
Countries
Corrective (Two parts) 3.6
Standard Electoral Systems SMD plurality 2.0
Majority runoff 2.9
Single nontransferable vote 2.5
Single transferable vote 2.4
List proportional representation 3.7

 

 

 

 

 

Note: N3 is the effective number of parties based on seat shares (Taagepera
and Shugart 1989).

Source: The results for the mixed systems from other countries and standard
electoral systems are from (Herron and Nishikawa 2000). Others are
calculated by the author.

40

Although the difference of 0.2 is not so great, it is worthwhile noting
that it still reflects a substantial difference in party systems. This becomes
clear if we compare how a typical system with 2.0 is different from a system
with 2.2. Table 2-5 is the election outcomes for the US. Congressional
elections in 2000. This election produced two effective number of legislative
parties for both House and Senate elections. When the effective number of
parties is 2.0, seat shares are more evenly distributed between two parties.
The US. case is a representative case of such a party system. Both the
Republican Party and the Democratic Party obtained 50% seat shares.
However, when the number of parties is 2.2, we tend to see some extra
parties with a substantial number of seats in the legislature, as it is the case in
Japan’s Lower House. This is shown in Tables 2-2a and 2-2b.

The rest of the tables summarize election outcomes based on the
effective number of elective parties. The difference between the plurality
component of superposition systems and regular plurality systems is much
more prominent when we look at the effective number of elective parties.
Table 2-6a shows the results for the Lower House elections of 1996 and 2000
in Japan and those for the US. Congressional Elections. The figures are 2.9
and 2.7 for the Japanese case, while they are only 1.8 and 2.0 in the US.
case. These figures reveal that the SMD plurality section so far produced a
larger effective number of elective parties than other typical SMD plurality

system, such as the US. case?5

 

25 The US. case is used in the analysis to represent the election results under the regular
SMD plurality rule. It would be more appropriate, if I include cases from other countries. For

41

Table 2-5: Seat Distribution and Effective Number of Legislative Parties
in the U.S. 2000 Congressional Election

 

 

 

 

 

 

POLITICAL PARTIES HOUSE SENATE
Republicans 220 18
Democrats 21 1 15
Others 2

Total 433 33

EN LP 2.0 2.0

 

 

 

 

 

Note: ENLP stands for the effective number of legislative parties. The figure
for ENLP is the same with or without the category of “others.”

 

example, the Canadian election of 2000 produced about 2.7 effective number of electoral
parties at the district level. The value in the Canadian case is closer to the Japanese case,
compared to the U.S. case. This seems to be because the Canadian political environment is
more prone to produce a larger number of parties because of the existence of French
speaking people. As such, I avoid using the Canadian case to compare with the Japanese
case.

42

Table 2-6a: Effective Number of Electoral Parties in the SMD Plurality
Section of Japan’s Lower House and U.S. Congress

 

ELECTIONS ENEP

 

Japan’s Lower House- 1996 2.9

 

Japan’s Lower House- 2000 2.7

 

U.S. House 1.8

 

US. Senate 2.0

 

 

 

 

Note: ENEP stands for the effective number of electoral parties.

43

Table 2-6b shows the effective number of electoral parties for Japan’s
Upper House broken down by district magnitude. When the district magnitude
is 1, the number of parties tends to be larger than 2. The figure for 1986 is 2.4,
and it is 2.3 for the years of 1989 and 1992. The election returns from the
Upper House also confirm that the effective number of elective parties is
larger when a plurality election is carried out with a PR election.

The rest of election results also manifest that the number of parties in
the plurality section of superposition systems are tend to be larger than other
regular systems. Table 2-6c includes the results for Japan’s Lower House
under the SNTV system before it introduced a superposition system. The
inclusion of these results makes comparisons of the number of parties
between the plurality component of the superposition system and the regular
plurality system possible since it controls for the differences in district
magnitude.

When the district magnitude is 3, the smallest number of parties is
equal to 4.8 in the superposition system, whereas the largest is 5.2. On the
other hand, the smallest is 4.4 while the largest is 4.6 in the regular SNTV
system (Table 2-60). This shows that the difference is at least 0.2. It is also
important to note that the regular SNTV election of 1993 when it produced 4.6
is the time when Japan was under a new party boom, and the election in 1993
should be considered to have produced a larger number of parties than

typical years of elections.

Table 2-6b: Effective Number of Electoral Parties in the Plurality Section

of Japan’s Upper House

 

 

 

 

DISTRICT 1983 1986 1989 1992
MAGNITUDE

1 2.4 2.4 2.3 2.3

2 3.5 3.3 3.4 3.2

3 5.2 4.9 4.8 4.9
4 a a a a

 

 

 

 

 

 

 

Note: The figures are calculated without the vote shares of independent

candidates.

a: There are only two districts.

45

 

Table 2-6c: Effective Number of Electoral Parties in the Single
Nontransferable Vote Systems - the Japanese Lower House Between
1983 and 1996

 

 

 

 

 

 

DISTRICT 1983 1986 1990 1993
MAGNITUDE

1 b b d b

2 a c b 3.4

3 4 4 4.4 4 4 4 6
4 5 6 5.5 6 2

5 6.3 6 3

 

 

 

 

 

 

 

Note: The figures are calculated without the vote shares of independent
candidates. Some districts have the district magnitude of 6. However, they
are not reported in the table since there are not enough cases

a: There is only one district

b: There are only two districts.
c: There are only four districts.
d: There are no cases.

46

Part II.

Explaining Japanese Party System

47

Chapter 3

Strategic Behavior of Political Parties

The French political scientist, Maurice Duverger was the first to identify
mechanisms by which electoral rules affect the number of political parties. In
his classic work on political parties, Duverger states that “the simple-majority
single-ballot system favors the two-party system” (217). Duverger claims that
the association is strong and the relationship comes close to “a true
sociological law.” Similarly, the extension of Duverger’s Law, the M+1 Rule, is
supported widely in the literature (lshikawa 1981, Reed 1990, Cox 1997). The
M+1 Rule generalizes Duverger's Law in terms of district magnitude, M, and
states that the number of parties will approach M+1.

The effectiveness of Duvergerian equilibria, however, is not without a
limit. The plurality component of superposition systems inflates the number of

parties more than Duvergerian equilibria suggest.26

When a single member
district (SMD) system with a plurality rule is combined with a proportional
representation (PR) rule, the SMD part of the system tends to produce a
multiparty system instead of a two party system. Similarly, the use of

multimember districts with a single nontransferable vote (SNTV) rule leads to

the creation of the number of parties that is greater than M+1.

 

2° The argument that Duverger’s Law does not work in mixed superposition systems can be
extended to other mixed systems, such as corrective systems. However, the inclusion of
corrective systems at this point will be beyond the reach of this dissertation and will not be
discussed here.

48

The inflation of the number of parties originates in the change of
behavior that takes place among political parties in the plurality component of
superposition systems. Political parties place a greater number of candidates
than the equilibria suggest. The number of candidates will be larger than two
when superposition systems adopt an SMD rule, while it will exceed M+1 with
an SNTV rule.

In this chapter, I will discuss the mechanism in which superposition
systems change the behavior of political parties, thereby creating the origin of
the violation of Duverger's expectations. I will do so especially by focusing on
a certain aspect of Duverger’s Law, namely a psychological effect. The
explanation of Duverger’s Law is composed of two factors, including
mechanical and psychological factors. The mechanical effect refers to how an
electoral rule translates votes into seats, while the psychological effect
reflects how political actors may react to the mechanical effect. Since
Duverger's Law works based on the two effects, I believe that the mechanism
of how mixed superposition systems violates Duvergerian equilibria can be
revealed by analyzing how the two effects work differently between the
plurality component of superposition systems and regular plurality rules.

To achieve the above, I will have a detailed discussion on the two
effects under regular electoral systems first. I will then extend the analysis to
superposition systems and show that the psychological effect does not work
in the same way in the plurality section of mixed electoral systems as it does

in regular plurality systems. I will especially focus on the presence of the

49

contamination effect and show how the psychological effect on political
parties is altered by the contamination effect.

There is another important task that needs to be carried out in this
chapter. A particular assumption of the contamination effect needs to be
satisfied. The contamination effect works based on an assumption that
political parties can place candidates to plurality districts where they have little
chance of winning or that candidates voluntarily enter into these districts. The
basic argument of the contamination effect is that political parties create an
incentive to place candidates in the plurality districts of superposition systems
since political parties can gain extra vote shares in corresponding PR districts.
By doing so, political parties can maximize their party’s seat gains collectively.

However, the attempt of political parties to place candidates may be
resisted by some candidates. It is often argued that politicians are ambitious
and seek to maximize their chance of winning (Dawns 1957, Schlesinger
1991). If this describes a true characteristic of candidates, we notice that it is
irrational for them to enter into a district where they have little chance of
winning. Instead, candidates will try to run in a district where they would have
more popular support, or they may simply refrain from entering.

More specifically, a candidate’s party may be popular in urban areas
than rural areas. This candidate may want to enter into a district that is
located in an urban area where he has a better chance of winning and wants
to avoid a district, which is in an rural area. There may also be other

candidates from the same party who may want to enter into an urban district.

50

Under such circumstances, the party would have many candidates who would
want to enter into an urban district, while they have no candidates who would
want to enter into a rural district. Under the condition, the political party may
have to reduce the number of candidates to one to avoid infra-party
competition in urban districts, while it has to encourage some candidates to
enter into rural districts. The above suggests that it is necessary to reveal how
political parties overcome the conflict with their candidates.

The purpose of this chapter is two fold. First, I will discuss how the
psychological effect on political parties is contaminated by the presence of the
PR component, and as a result, political parties try to maximize their vote
shares by placing more candidates than Duverger’s Law expects. Second,
because of the contamination effect, political parties try to place some of their
candidates into a plurality district where they have little chance of winning and
may face a resistance by the candidates, party leaders have a mechanism
that helps to place candidates, so the contamination effect is maintained.
Before I get into these two aspects, I will discuss how Duverger’s Law works

in general. I will also extend the discussion to the M+1 Rule.

Duverger’s Law

As it was briefly discussed earlier, Duverger explains that simple-
majority SMD systems favor two party systems because of two factors: a
mechanical and a psychological factor. The mechanical effect refers to how

votes are translated into seats. Because of the use of a plurality rule and

51

single member districts, small parties’ seat shares are likely to be smaller than
their vote shares, and as a result, minor parties are usually “under-
represented.” For example, if there are three parties, one being a minor party
and the other two being major parties, the minor party has difficulty in winning
a seat since their votes may be spread across different plurality districts.

The presence of such a mechanical effect influences political actors’
behavior, leading to the creation of a psychological effect; another effect
which hurts small parties even further. The minor party will find that the
system is mechanically working against the party because of under-
representation. Under such circumstances, the minor party would either try to
merge with one of the major parties or refrain from entering. It is also
conceivable that political contributors, such as interest groups, stop
contributing to the minor party or the members of the party, since they realize
that the candidates from the minor party have little chance of winning a seat.

Even when marginal competitors enter into a district by overcoming
these weaknesses, plurality SMD systems will further give disadvantages to
minor parties. More specifically, voters in the simple SMD system soon find
out that their votes are wasted every time they vote for the third party. This in
turn leads voters to start transferring their ballots to “the less evil of its two”
(226). Thus voters are polarized into two camps. As a result of the
mechanical and psychological effects working together, simple SMD systems
create a political dualism in a party system. Duverger also points out that the

process of polarization may take place overtime since voters are at first do

52

not know what party may be trailing. However, as soon as voters find out who
is trailing, they start transferring their votes to one of the two major parties.

The above argument can be extended to other plurality systems, such
as an SNTV rule. With a multimember district rule, the number of candidates
should approach M+1. The political parties that are ranked lower than M+1 in
their popularity should be considered marginal competitors. As an illustration,
ifthe district magnitude is 3, the number of candidates in districts should be
close to 4. Other candidates who are ranked 5 or lower would have difficulty
in surviving in the system because of the mechanical and psychological
effects.

Duverger (1954) identifies another relationship between electoral
systems and party systems. Duverger states that “the simple-majority system
with second ballot and proportional representation favor multi-partism” (239).
Although how a simple majority rule with the second ballot system and a PR
rule create a multiparty system is different, we notice that they both do not
force political actors to create a political dualism in a party system that the
simple SMD system does.

For example, a majority is required to win an election under the two-
ballot SMD system. If no candidates obtain more than 50% vote share, the
system requires another round of the election between the top two candidates.
With this process, minor parties do not have to be concerned about under-
representation by the mechanical effect in the first round of the election, since .

the second round takes place frequently with a two-ballot system. Moreover,

53

they can always regroup in the second round of the election, if it is necessary.
Voters also do not have to be concerned about wasting their votes under the
system. Voters who prefer a minor party can vote for the minor party in the
first round. If the second round takes place, they can then consider voting for
one of the two major parties.

Under PR systems, the mechanical effect is also less of a concern for
minor parties since the system translates votes into seats more proportionally.
As is the case under the second ballot SMD system, voters do not have to
worry about wasting their votes in PR rules. Moreover, political contributors
may not stop supporting small parties in these systems. As a result, PR
systems create a multiparty system.

Duverger’s findings in the above are now recognized as Duverger’s
Law and Hypothesis: the former refers to the association of the plurality SMD
rule with two party systems, and the latter refers to the association of the
second ballot SMD rule and a PR rule with multiparty systems. These
propositions by Duverger and his explanation of how the propositions work
can especially give us insights into how mixed superposition systems may
shape party systems. If we can find out how the mechanical and
psychological effects work in each part of mixed electoral rules, we will have a
better understanding in how the electoral process under these rules
influences party systems. In the following section of this chapter, I will discuss
how Duverger’s Law is violated in the plurality component of superposition

systems with more detail.

54

Duverger’s Law and Mixed Superposition Systems

As suggested earlier, how the psychological effect is influenced in the
plurality component of superposition systems is the key to the explanation of
the increased number of parties. This is evident, if we look at each part of
superposition systems and compare the effects of each part with regular
electoral rules. More specifically, the mechanical effect works similarly
between the PR component of a mixed system and a regular PR system.
Votes are translated into seats proportionately in the PR section of a mixed
system and in a regular PR rule. The mechanical effect between the plurality
component of superposition systems and regular plurality systems also works
in the same way. They both translate votes into seats in a way that is not
friendly to small parties.

The psychological effect in the PR component of a mixed system and a
regular PR system also works similarly. Votes are translated into seats
proportionately in the PR section of mixed systems. Under these conditions,
minor parties do not have to be concerned about under-representation.
Voters who prefer a minor party also do not have to transfer their votes to
other parties that are more likely to win an election.

Only the psychological effect in the plurality section of mixed systems
works differently from the standard plurality systems. More specifically, the
psychological effect is inhibited from working in the plurality part of the mixed

systems in the same way as it does in a regular plurality system. This is

55

because the PR component of the system is carried out simultaneously with
the plurality section, while the presence of the PR election changes political
parties’ strategies in placing candidates.

The changes in parties’ strategies happen because political parties
perform better in the PR election by placing candidates in plurality districts. If
political parties place their candidates in plurality districts, their candidates will
be physically present in the plurality districts during the election campaign.
Their presence would be reported in local newspapers and W3. A large
number of posters would also be put up on the streets of the districts. More
importantly, each candidate would carry out a campaign with a party label. As
a result, parties that place their candidates are more likely to be recognized
by voters in PR districts. This recognition leads to an improvement of party
performance.

Moreover, some voters may pay attention to the plurality component of
an election first and decide whom to vote for in the PR component. They may
then select a party in the PR component that is consistent with whom they
voted for in the plurality race.

It is possible that political parties or party leaders may not realize the
interactive effect between the two parts of the mixed system in the first
election. However, they would soon recognize the effect and try to place their
candidates even when their candidates have little chance of winning an

election.

56

A significant number of politicians indeed have noticed the effect
already. For example, Hiroshi Kumagai, the deputy secretary general of the
Democratic Party of Japan who was the head of the election committee for
the 1996 Lower House election, states that by placing a candidate in an SMD
plurality district, a vote share for the corresponding PR district would be
increased substantially.27 Moreover, Kumagai states that the PR district
without a candidate in the corresponding SMD district did the worst in gaining
a vote share compared to other PR districts.28 Kumagai also suggests that a
similar effect exists in the Upper House election in which a mixed system with
an SNTV rule is applied. Kuniji Toda who is a councilor from the Liberal Party
also agrees that political parties can gain an extra share of votes by placing a
candidate in plurality districts.

In short, the presence of a PR rule is expected to contaminate the
psychological effect of Duverger’s Law in mixed superposition systems.
Political parties develop an incentive to place candidates even when their
candidates have little chance of winning a seat. As a result, the number of
candidates in the plurality section of mixed systems is inflated. This suggests
that the number of candidates will also exceed an M+1 number of candidates,

when an SNTV rule is used.

 

27 28The comment was obtained by conducting a face-to-face interview with Hiroshi Kumagai.
28The comment was also obtained by an interview with Kuniji Toda.

57

Contamination Effects and Candidate Nomination Process

The contamination effect works based on the idea that political parties
develop a motivation to place candidates in plurality districts in which
candidates may have little chance of winning. This suggests that some
candidates have to enter into a district where they may not want to enter. The
argument of the contamination effect then needs to be reconciled with the
idea that candidates are rational actors and try to maximize their chance of
winning a seat.29 I argue that the entrance into these districts do not
necessarily eliminate the chance of winning for these candidates. Instead, it
may help their career as a politician. There are a number of reasons why the
entrance into a plurality district may help candidates. The first reason has to
do with the fact that elections are repeated over and over and are not a one-
shot game. Under the condition, candidates may understand a loss in a
plurality district election as a necessary step to become a successful politician
who will be able to win in succeeding elections.

The notion of repeated games in rational choice theory seems to
support this idea. A number of scholars who practice rational choice theory
suggests that games are usually repeated in politics (Morrow 1994). As a

result, political actors may make a decision that would look as if they made an

 

2° The information on how candidates are put up is found in different ways. The easiest way is
to find a written document that describes a candidate nomination process in each party.
There are three parties that have the information in their constitution. These parties include
the Liberal Democratic Party (LDP), Social Democratic Party of Japan (SDPJ), and
Democratic Party of Japan (DPJ). For other parties, the information was obtained by writing
to the party's headquarter directly. The information for the Japan Communist Party (JCP) was
obtained in this way. Additional information was found in the documents that are available
through each party’s home page. The information is difficult to be obtained for some parties

58

irrational choice. However, if we consider the fact that games are repeated,
their action does not look as irrational. We can apply the notion of repeated
games to elections since elections are carried out repeatedly. If candidates
see that the chance of getting elected in a succeeding election increases by
entering into a plurality district, candidates may be more willing to enter.

This may be especially applicable to young politicians. A young
ambitious politician may be told to enter into a plurality district where there is
a popular incumbent. She may think that there is little chance of winning a
seat for the first time she enters. However, the incumbent politician may be
old and will be replaced by another candidate sooner or later. When that
replacement takes place, the young politician may have a great chance of
winning a seat, if she has carried out an election campaign In previous
elections. This is because the level of recognition by her constituents may be
higher.

Kuniji Toda, a councilor from the Liberal Party in Japan, suggests that
young politicians are indeed more willing to enter into districts where they
have little chance of winning.30 Toda mentions that this is because they may
understand that they would have to go through a loss before they actually win
a seat. If this is how candidates see winning an election, they may perceive a
benefit in entering a plurality district.

The above does not mean that the concept of repeated elections is

nonexistent in regular plurality elections. The difference between the plurality

 

since these parties do not exist anymore and the documents that give us the information are
not available.

59

component of mixed systems and regular plurality elections rests in the fact
that small parties are more likely to survive in a longer term in mixed systems
than in regular plurality systems because of the use of PR rules. As a result,
candidates in mixed systems do not have to worry about their party’s survival
as much as in regular plurality systems.

The second reason why political parties can place candidates in the
plurality component of mixed systems has to do with the fact that election
outcomes are not always easy to predict and may magnify the perceived
chance of winning in an election. For example, there is always a possibility
that even a strong incumbent may lose. We can see such an uncertainty of an
outcome in recent elections, including the 2000 Lower House election.

Masayoshi Takemura, who is a former Minister of Finance, lost his seat
in the 2000 Lower House election. Takemura was also the chairman of a
party called the New Harbinger Party and was a well-respected politician. His
loss was reported in news media as a surprise. If a politician like Takemura
can lose in an election, candidates may see a chance of winning, even in a
district where there is a strong incumbent. Moreover, the perceived chance of
winning will be magnified when we think about the fact that elections are
repeated, since an unexpected outcome may take place in one of the many
following elections.

The third aspect that promotes candidate entries to plurality districts

has to do with how political parties increase the chance of winning for their

 

2° The comment is obtained by an interview.

60

candidates by using double candidacy. The Public Office Election Law
provides political parties an ability to promote the placement of candidates.
Candidates who belong to a political party are allowed to run in both parts of
the system under the law. With the use of double candidacy, political parties
can promote candidates” entrance into plurality districts since political parties
usually decide the candidates’ ranking on the PR list. Because parties have a
control over the ranking, candidates who are on the PR lists especially have
an incentive to follow their parties’ suggestion that they should enter a
plurality district.

Hiroshi Kumagai, the deputy general secretary of the Democratic Party
of Japan, states that political parties have difficulty in placing candidates
where they have little chance of winning. However, double candidacy helps
parties to place candidates in these difficult districts by using double
candidacy.2’1 In fact, a large number of political parties practices double
candidacy. Kumagai’s party, the Democratic Party of Japan (DPJ) is one of
them. The examples of other parties would be the Liberal Democratic Party
(LDP), the Liberal Party, the New Progressive Party, the New Harbinger Party,
and the New Japan Socialist Party.

Finally, political parties have tremendous power over the decision on
how candidates are placed in plurality districts. If political parties have control
over candidate actions, political parties can force some candidates to enter

into districts. It is true that candidates can run as an independent candidate

g

3' The comment is obtained by an interview.

61

who does not belong to a political party, so that they can threaten their parties.
However, if candidates do not belong to a political party, they will face many
disadvantages in elections. These candidates do not have as much
government support as other candidates who belong to a political party.

In particular, how much individual candidates can campaign is more
limited if they are not endorsed by a party. For example, the government
provides a certain length of airtime on television to candidates who belong to
a party for their campaign, whereas candidates without a party are not given
such a privilege. Similarly, the number of posters that candidates can put up
will be reduced for those candidates who run as an independent candidate
(Ozawa 1997). Under such circumstances, candidates are expected to stay
with a political party and, as a result, these benefits give political parties a
leverage over the placement of candidates.

What is more, how political parties can exercise control over
candidates are also suggested in party constitutions, which specify party
activities in elections. For example, the Committee for Election Strategy and
the Permanent Secretariat in the Democratic Party of Japan (DPJ) direct a
wide range of election related activities.32 The Committee for Election
Strategy in the DPJ nominates candidates and the Permanent Secretariat

confirms the nominationsf"3 The Social Democratic Party also has the

 

32 The translation, the Executive Standing Committee, is made by the author and is not an
official translation. It is originally called “Jyoninkanjikai.”

33 The Committee for Election Strategy is set up by the DPJ’s National Convention, and the
members of the committee are selected by the chair of the Committee who is appointed by
the DPJ's Diet members. For the Lower and the Upper House elections, the Permanent
Secretariat decides the ranking of the nominees for the PR section. The Permanent

62

Committee for Election Strategy.2’4 The final approval for the selection of
candidates is made by the Executive Standing Committee of the SDP.

For the case of the Japan Communist Party (JCP), the party’s control
over the candidate nomination process is not specified in the constitution.
However, the constitution states how the party is organized and suggests that
the party has a control over the candidate nomination process. The
Constitution of the JCP states that the party follows the principle of
“democratic centralism” (Article 14). The centralization of the organization
gives power to the JCP. More specifically, the Central Committee is the
executive organization of the JCP that has control over wide range of party
activities (Article 25). The area of control also includes the candidate
nomination process.35

The constitution of the LDP does not state who finally confirms the
nomination, even though it states that the LDP has the Election Strategy
Headquarters (Article 51- 54) in its constitution. However, the LDP has local
organizations that coordinate candidate placement. The LDP has prefectural
branches that coordinate candidate nomination and give advice to candidates.

When they do so, they follow the LDP Headquarter’s request.

 

Secretariat also has the authority to cancel any of the nominations (Article 20). The
translation, “National Convention” is done by the author. The original is “Totaikai”. The
translations, “National Federation" and “the Committee for Election Strategy" are made by the
author and are not official translations, either. The original in Japanese are “Zenkokukaigi”
and “Senkyo Taisaku linkai.”
2“ The Committee for Election Strategy is set up by the party’s National Federation
(Zenkokurengo). The translation, “the Committee for Election Strategy” is made by the author.
The original in Japanese is “Senkyo Taisaku linkai.”

The Executive Standing Committee makes decisions on candidate nomination for both the
Lower and the Upper Houses. The translation, the Executive Standing Committee, is made
by the author. It is originally called “Jyoninkanbukai.”

63

For example, Asahi Shimbun Senkyo Hombu (1997), a division of the
Asahi Newspaper that specializes in elections, reports that in the 1996 Lower
House election, the Shimane prefecture had four potential LDP candidates,
while there were only three SMD districts in the prefecture. The candidates
were Yoshio Sakurauchi, Hiroyuki Hosoda, Noboru Takeshita, and Hisamoto
Kamei. The LDP Shimane Prefectural Conventiona’6 (Jimintokenren)
negotiated with Yoshio Sakurauchi and requested that he pulled out of the
SMD district race and run in the PR race. Since Sakurauchi pulled out of the
SMD race, the remaining candidates could run in the three SMD districts in
the Shimane prefecture.

Furthermore, the evidence of the LDP’s control over the candidate
nomination through its local organization is seen in Weekly Jimin, the
magazine that the LDP publishes weekly. It reports a meeting that was held in
Kyushu, one of Japan’s four main islands, which is located in a southern part
of Japan and corresponds with one of the twelve PR districts. The meeting
was held by Kyushu Block Chief Secretaries and Kyushu Block Secretary
General (Kyushu Block Kanjicho, Jimuchokucho Kaigi). The meeting was
joined by three LDP members who were from the LDP headquarter. The
report suggests that the LDP’s headquarter has the ultimate control over the
candidate nomination process. It describes the struggle the LDP's local

organizations went through because they often have to coordinate LDP

 

2° The translation, “the LDP Shimane Prefectural Convention" is not an official translation.

64

candidates in their districts and suggest some of them to pull out of a district,
as it was the case in the Shimane prefecture.

It is also important to note that a coalition among some parties can
reduce the number of parties. This is especially true if political parties have a
control over the candidate placement. If the creation of a coalition happens
successfully, the number of candidates can be reduced to the level where
Duvergerian equilibria predict. However, creating a coalition can be difficult
since it requires coordination and may not be achieved.

For example, the alliance between the LDP and the CGP in the 2000
Lower House election is characterized as a failure of such an attempt. The
LDP and the CGP agreed not to place candidates in seventeen districts if the
other party decides to place their candidate. The LDP placed three
candidates in these districts, while the CGP placed ten candidates. Despite
the effort to coordinate between the two parties, the CGP candidates did not
win a seat in five districts. The reason why this occurred was that some of the
LDP candidates who could not get support from the LDP ran as unaffiliated
candidates and won. Kensaku Morita, a former actor, who ran as an
unaffiliated candidate in Tokyo is a representative case that destroyed the

coordination between the two partiesf"7

In summary, mixed superposition systems create an incentive among

political parties to place candidates, even when they have little chance of

 

”See Yomiuri Shimbun (June 26, 2001).

65

winning. By placing candidates in plurality districts, political parties can gain
extra vote shares in the corresponding PR districts. Candidates may resist
against parties’ attempt to place them into a plurality district where they have
little chance of winning. Even though small parties may have a limit in their
resources to place candidates, the fact that superposition systems motivate
political parties to place more candidates contribute to the increase in the
number of candidates. Under the conditions, if voters cast their ballots in
these extra candidates who would have not been placed, if an election is
carried out in a regular plurality system, the number of parties will be larger

than Duverger’s Law or than the M+1 Rule may predict.

66

Chapter 4

Strategic Behavior of Voters

Chapter 3 showed how political parties inflate the number of candidates in the
plurality section of mixed superposition electoral systems. To support the idea
that the plurality component of a superposition system produces a larger
number of parties than Duverger’s Law predicts, the effects of the increased
number of candidates on voters” behavior are shown in this chapter. If the
increased number of candidates leads more voters to behave strategically by
not voting for marginal candidates, we can conclude that the plurality
component of the system will produce a Duvergerian outcome. On the other
hand, if the increase leads them to vote for marginal candidates, it will
produce a non-Duvergerian outcome.

The concept of strategic voting is described by Duverger (1954) as the
voters’ tendency to avoid “wasting” their votes. Duverger’s argument is
especially relevant to elections under single member district (SMD) systems.
According to Duverger, constituents often avoid “wasting” their votes in a
plurality SMD system by casting a ballot for a candidate who has manifestly
no chance of winning. By engaging in strategic voting, voters transfer their
ballots from the most preferred competitor who has little chance of winning to
a less preferred but more competitive candidate. In this way, voters can

minimize the probability of creating the worst outcome.

67

By studying mixed electoral systems, many scholars conclude that
strategic voting exists in countries that use a mixed system. The studies of
Japan’s mixed systems are no exceptions to this rule. Kohno (1997) finds
evidence of strategic voting in Japan’s superposition system by comparing
how voters behave within the two segments of the system. This approach is
followed because how voters behave in the PR section can be used to assess
the existence of strategic voting in the plurality component. Voters behave
sincerely in the PR section of the system since they are less likely to waste
their votes under that system. Contrarily, voters are more likely to vote
strategically in the plurality section since they may waste their ballots if they
vote sincerely. Under such assumptions, the assessment of strategic voting is
possible by comparing how voters behave between the two systems.

Vlfith this approach, Kohno studies three political parties, including the
Liberal Democratic Party (LDP), New Frontier Party (NFP) and Japan
Communist Party (JCP) and finds that a major party, such as the LDP, gains
more vote shares in the SMD part of the system, while there is no such
tendency is found in a minor party like the JCP.

Similarly, Reed (1999) analyzes the existence of strategic voting in the
1996 Japanese Lower House election. Reed also compares voters’ behavior
between the PR section and the plurality component and finds “strong

evidence for strategic voting: candidates with a good chance of winning

68

receive more SMD than PR votes, and those with little chance of winning
receive more PR votes” (257)?8

As these studies may demonstrate, it is tempting to support the idea
that strategic voting exists in mixed superposition systems. This is because
the existence of strategic voting is supported in regular SMD plurality systems
both theoretically (Farquharson 1969, McKelvey and Ordeshook 1972) and
empirically (AbramSon et al. 1992 and 1995, Kim and Fording 2001). If a
mixed system is composed of an SMD plurality rule and a PR rule, it may look
rather natural that the SMD part produces strategic voting. Moreover,
Scholars have found strategic voting in other electoral rules, including
majority runoff systems (Cox 1997), multimember district systems with a
single non-transferable vote system (SNTV) (Reed 1991, Cox 1997), and
approval voting (Cox 1984, Niemi 1984).

Perhaps, the most important evidence that leads scholars to support
the occurrence of strategic voting in superposition systems, though, is that a
significant number of studies support the presence of strategic voting in other
mixed systems, including the one that is used in Germany (Barnes et al.
1962, Fisher 1973, Jesse 1988, Bawn 1999, Cox 1997). As discussed in an
earlier chapter, the German mixed system is called a corrective system
(Massicotte and Blais 1999) since the PR section of the system is used to
“correct“ a disproportionality that is caused by the use of a plurality rule.

Finding strategic voting under various electoral rules, including corrective

 

3"Onizuka (2002) also finds the presence of strategic voting in the Japanese mixed
superposition system.

69

systems, seems to lead scholars to think that strategic voting should be
present in mixed superposition systems. Indeed, Reed (1999) states that the
existence of strategic voting “should come as no surprise” (258).

Although existing studies on strategic voting in mixed systems
consistently report that strategic voting is evident, we wonder if strategic
voting takes place in mixed systems as much as the previous studies suggest.
This occurs for two reasons. First, these studies have not considered how the
two components of the mixed systems may interact with each other so that
they may produce outcomes that were not expected originally. Specifically,
scholars have not considered how the presence of PR in the system may
contaminate the effect of plurality systems that we usually see. Hence, the
underlining logic of strategic voting in regular plurality systems may not be
applicable to the plurality component of superposition systems. This is true for
both superposition systems and corrective systems.

Second, the previous studies on mixed superposition systems
measured strategic voting only at the aggregate level, even though there are
some studies with individual level data on regular plurality systems (Cain
1978, Abramson et al. 1992 and 1995). For example, to study strategic voting
in a superposition system, Kohno uses election returns that are aggregated at
the prefecture level in Japan, while Reed uses those that are aggregated at
the candidate level. Since strategic voting is an individual-level phenomenon,

assessing strategic behavior at an aggregate level needs a caution since it

70

may cause an ecological inference problem (Kramer 1983, King 1997).39
Moreover, the studies on mixed corrective systems were also carried out by
using aggregate level data (Fisher 1973, Barnes et al. 1962, Cox 1997, Bawn
1999). This suggests that the empirical support of strategic voting in mixed
systems may not be as firm as we originally thought.

One of the specific purposes of this chapter then is to investigate
voting behavior in superposition systems and improve on previous studies by
especially focusing on the contamination effect by abandoning the
assumption that the two segments of mixed systems do not interact with each
other in influencing how voters behave. By doing so, I will show that the
contamination effect creates conditions in superposition systems that reduce
the occurrence of strategic voting. The other problem which should be dealt
with is the ecological inference problem. However, I will deal with the question
in one of the empirical chapters (Chapter 6) since it is more related to
measurement issues.

In the following sections, I will discuss how the contamination effects in
superposition systems create voting conditions that prevent voters from
behaving strategically and how the use of aggregate level data can
overestimate the occurrence of strategic voting. Before turning to the primary
discussion, however, I will review political parties’ behavior briefly and
develop how voters may make decisions given specific conditions. Voters’

behavior will be preconditioned by how political parties place their candidates,

 

22 Both Reed (1999) and Kohno (1997) point out that there is a potential problem in using
aggregate level data to assess strategic voting.

71

so that it is important to analyze the condition in which voters may make
decisions.

More specifically, as Chapter 3 shows, the expected pattern of
candidate placement by political parties changes, when we take the
contamination effect into account. The expected number of candidates
increases in the plurality component of the system when the contamination
effect is present. The increase in the number of candidates in turn is expected
to influence what candidates are available for voters to choose and thus
influence voters’ behavior as well.

In addition to analyzing voters’ decisions to cast their ballots for
marginal or top running candidates, I will also analyze voters’ decisions to
abstain from voting. An abstention as a choice needs to be investigated since
the idea that voters would vote sincerely in the plurality section of a mixed
system may be undermined if we consider the fact that an abstention as a
choice is also available to voters. For voters whose most preferred candidate
happens to be trailing in the plurality component of the election, an abstention

as a choice may be dominant over sincere voting.

Strategic Voting and Contamination Effects

If we assume that the two components of a mixed system work
independently, voters are expected to behave strategically in the plurality
component of the system while they would vote sincerely in the PR

component of the system. That is, voters would behave as if they vote in two

72

separate elections. In other words, voters make a decision under the
conditions which work against marginally performing candidates. Under such
circumstances, a voter engages himself in strategic voting by casting his
ballot for a less preferred candidate, even when he happens to prefer a
marginal candidate.

More specifically, a voter can maximize his expected utility by voting
for one of the top candidates in the plurality election, while he votes for the
most preferred competitor in the PR section of the election. The notion of
expected utility is widely used in political science. Utility measures actors’
preferences over outcomes. An expected utility is the probability of producing
an outcome multiplied by its associated utility (Morrow 1994). If expected
utility for a particular competitor is high, voters are more likely to cast their
ballot for the competitor.

As an illustration, if there are three competitors whose popularity is C1>
Cz>C3, and a voter’s preference ordering is C3>Cz>C1, the particular voter
would vote for 02 to avoid the worst outcome if the plurality section of the
system applies SMD system. However, the voter will vote for C3 in the PR
district election.

Similarly, if the plurality component uses a multimember district system,
and if there is an M+2 number of candidates in a plurality district, a voter will
avoid voting for the M+2“d running candidate, if she is trailing in the district.
Thus, strategic voting takes place among voters who happen to prefer the

M+2”1| running candidate, although these voters are still expected to vote

73

sincerely in the PR component of the system. What the above suggests in
terms of mixed systems is that voters would engage in strategic voting in the
plurality part, although they vote sincerely in the PR component.

However, if we take the contamination effect into account, our
expected model of how voters behave will be altered. The contamination
effects of mixed systems create some unique conditions that may not be
observed if each component of the system is carried out separately. More
specifically, voters will be influenced by being in the condition that a mixed
system may create. This suggests that we need to consider the effects of
mixed systems by two phases. The first is to consider what kind of voting
conditions mixed systems may create. The second is to analyze how voters
may behave under different voting environment.

For the voting conditions, as Chapter 3 suggests, the number of
candidates will be increased in the plurality section of mixed systems. If a
mixed system combines a plurality SMD system, the number of candidates
will be larger than two, while it will be larger than M+1 if a mixed system is
composed of multimember district system. Such an increase in the number of
candidates should influence the patterns of voting that are available to voters
and how they choose to vote.

More specifically, the increase in the number of candidates creates two
patterns of voting conditions. Some voters find a candidate who is from their
most preferred party, while others do not. As it was suggested, political

parties create an incentive to place their candidates in the plurality districts.

74

However, this does not mean that all parties can place their candidates in all
districts since their ability to place their candidates varies from party to party.
Large parties are more likely to place candidates in most of the districts,
whereas small parties cannot do so because their resources are more limited.
Moreover, small parties may make an arrangement with other parties in
deciding where to place their candidates, so that they can secure a seat in
some districts. As a result, voters who prefer a small party may not see a
candidate who is from their most preferred party.

The above does not suggest that the existence of such two patterns of
voting conditions do not exist in regular plurality systems. However, an
increase in the number of candidates is expected to spread voters’ party
preferences. This will lead to the increase in the number of voters who do not
find their most preferred candidate, and this, in turn, magnifies the importance
in differentiating the two voting conditions in mixed systems.

If voters are in the condition in which they do not find their most
preferred candidate, they are forced to transfer their votes to other parties. If
this happens, voters are more likely to transfer their votes to the most
preferred candidate who is available. Although this may look as if they voted
strategically at the aggregate level, this pattern of behavior does not fit with
the definition of strategic voting since voters should be considered to have
voted for the most preferred candidate who is available in their district.

The other pattern of candidate placement for voters is that voters find a

candidate who is from their most preferred party. This pattern actually creates

75

an opportunity for voters whose most preferred candidate is less likely to win
to cast their ballots strategically. Indeed, the increase in the number of
candidates suggests that a higher portion of voters will prefer one of the
marginal candidates, compared to the case in which a small number of
candidates are running in districts. This may lead us to think that a substantial
number of voters may vote strategically.

However, in the plurality component of the mixed system, such
strategic voting is less likely to occur. Cox (1997) specifies the conditions
under which strategic voting is unlikely to occur. One of the conditions is
when voters are not oriented with a short-term instrumental motivation. Some
voters may think that their most preferred candidate may have a chance of
wining when elections are repeated, even though the candidate does not
have a chance of wining in a current election. If this is how voters perceive,
they are more willing to vote for marginal candidates.

If we consider a mixed system with the contamination effect, it is likely
that the system creates a longer-term incentive among voters. This is
because small parties can survive because of the presence of the PR election,
so that small parties would not at least disappear quickly as they may do in a
regular plurality SMD election. As a result, voters may be more willing to
support their most preferred candidate to keep them in the plurality section of
the system, hoping that the candidate would win in a future election.

There is another reason why mixed systems may promote sincere

voting for those voters who have an opportunity to engage in either sincere or

76

strategic voting. Cox suggests that when voters have difficulty in guessing
who is trailing in the district, voters are more likely to vote sincerely. If we
apply this idea to the plurality component of the mixed systems, we can
expect that voters are more likely to vote sincerely than in a regular plurality
system, since a large number of candidates will run in plurality districts. With
an increased number of candidates, compared to regular plurality systems,
the system will create close races between top running candidates and
marginal candidates in plurality districts. If a marginal candidate who a voter
prefers most is perceived to gain a significant amount of vote share, she
would be more willing to vote for the candidate.

For example, compare the following two situations. One is that an
election is held under an SMD plurality rule, and voter’s most preferred
candidate is expected to gain about 20% of the total vote share, while the top
two candidates are perceived to gain about 40% each. The other is that
voter’s most preferred candidate is expected to have 10%, while the top two
candidates are expected to gain 45%. The voter who is under the first
condition is more likely to vote sincerely compared to the voter under the
second condition.

It is also important to notice the increase in the number of candidates
can lead to the promotion of strategic voting, while it can also discourage
strategic voting depending on the situation. This is because the increase in
the number of candidates may create a close race between the top running

candidates, while it also creates a close race between the top candidates and

77

marginal candidates. More specifically, when the race between the top
candidates is close, voters who happen to support a marginal candidate is
more likely to engage in strategic voting and cast a ballot for one of the
candidates who they prefer more. On the other hand, if the race between the
top two candidates is not close, and if the race between the top and marginal
candidates is close, voters are more likely to vote for the marginal candidate
who is their most preferred candidate. What this suggests is that the increase
in the number of candidates can have positive or negative pressure on
strategic voting. At this point, whether or not the increase in the number of
candidates creates a positive or negative pressure should be left as an
empirical question.

Nonetheless, if we observe the entire picture of what takes place in
the plurality section of superposition systems, mixed systems seem to
produce a large number of voters who do not have a chance to vote
strategically. Moreover, the voters’ tendency to cast their ballots sincerely
should be substantially greater than that in regular plurality systems. The
tendency for sincere voting would be larger in the SMD plurality component of
a mixed system than in the regular SMD plurality rule. The tendency should
also be larger in the multimember plurality component of a mixed system than

in a regular multimember plurality system.

78

Abstention and Contamination Effects

What is the relationship between abstention and the contamination
effect? It is possible that abstention as a choice may influence how voters
behave. To support the argument that a substantial number of voters in the
plurality section of a superposition system vote sincerely even further, I will
discuss the relationship between abstention and the contamination effect and
show that the presence of an abstention as a choice does not undermine the
idea that the likelihood of sincere voting is larger in the plurality component of
the mixed system than in a regular plurality system.

The literature on turnout suggests that voters are more likely to turn out
in PR rules than in SMD systems (Reynolds and Reilly 1997). This is because
fewer voters are alienated in PR rules than in SMD rules. Voters whose most
preferred candidate has no chance of winning are less likely to turn out since
their votes will not give an impact on election outcome.

The formal literature also supports why PR rules creates a higher
likelihood of turnout. According to Riker and Ordeshook (1968), one of the
key factors in influencing turnout is the probability to cast a decisive vote.40 If
a voter perceives that the chance of casting a decisive vote is high, she is
more likely to turn out. The work done by Riker and Ordeshook can guide us

in determining how the plurality section of a mixed system influences turnout.

 

‘° Riker and Ordeshook operationalize by using the following formula.

R = PB + D - C
where R is the utilities that an individual benefits from act of voting, P is the probability of
casting a decisive vote, and B is the differential utilities that are associated with an outcome.
0 is the selective benefits which are obtained from casting a vote, such as the pleasure they
get by fulfilling the duty of voting as a citizen. Finally, C is the cost of voting, such as to gather

79

The probability of casting a decisive vote is a factor that varies
significantly across different electoral systems. For example, the probability of
casting a decisive vote is larger in a PR rule than in an SMD rule. If we
assume that the contamination effect is not present, the rate of turnout will be
similar between the plurality component of the system and a regular plurality
system. The PR section of the system will also maintain a similar rate of
turnout compared to standard PR systems.

Assume that there is no contamination effect on voters but political
parties are influenced by the contamination effect. The increase in the
number of candidates occur in the plurality section of the election, while the
plurality component of the election would produce a lower level of turnout
than in a regular plurality system. As it was suggested earlier, the increased
number of candidates in plurality districts spreads voters’ support to marginal
candidates. However, these voters who prefer these extra candidates in the
plurality districts do not have a chance in influencing an election outcome.
Since a larger number of voters may happen to prefer marginal candidates in
the plurality section of a mixed system compared to a regular plurality election,
the probability of abstention may be greater in the plurality section of a mixed

system.41

 

information about candidate and to physically get to the poll. If R is positive, voters will turn
out to vote. If R is negative, voters abstain.
‘1 McKelevey and Ordeshook (1972) generalize the formula for a multicandidate race. The
formula for multicandidate competition is as follows:

E"- E°= B"‘P"‘ + B"2P +
Based on the equation above, McKelvey and Ordeshook (1972) compare a two way race with
a three way race that apply a regular plurality rule and suggest that adding another
competitor to a race does not necessarily increase turnout, although it may be counter
intuitive at first. The addition of a competitor may decrease the likelihood of turnout among

80

However, if we consider the contamination effect even further and
think that voters are also influenced by it, the turnout in the plurality
component of the election is not necessarily lower than the plurality
component of the system. This is because rational voters will cast their ballots
in the PR component of the system, and, as a result, voters would cast their
ballots in the plurality component of the system, rather than to abstain in the
plurality section. This happens because the cost of voting, such as going to a
poll on the election day and paying attention to campaign and gaining
relevant information, will be lowered to vote in the plurality component of the
system since rational voters gather the information to vote in the PR section
regardless of their decisions in the plurality section. This suggests that an
abstention as a choice does not refrain voters who happen to prefer a
marginal competitor from going to a poll, as much as we may expect,
otherwise. If these voters turn out, the choice that they make is between
strategic voting and sincere voting. Between the two choices, voters are

expected to vote sincerely.

Impact of Sincere Voting on Number of Parties
The above discussion suggests that some voters would not have their

most preferred candidate in their plurality district, even though political parties

 

certain groups of voters. For example, if a voter is indifferent between competitors 1 and 2
and prefers these two competitors to competitor 3, the likelihood of turnout may decline, if we
assume that voters are uncertain about who will win. This is because the voter is not
concerned about breaking a tie between candidates 1 and 2. The voter is only concerned
about breaking a tie between candidates 1 and 3 or 2 and 3. Since the probability in breaking
a tie between these candidates is smaller than that of breaking a tie between candidates 1

81

develop an incentive to place their candidates in the plurality component.
This leads us to think that Duverger’s Law may be maintained in the plurality
component of the mixed system if there are not too many constituents who
have a chance to vote for a marginal candidate. If no voters cast their ballots
for a marginal candidate, the number of parties will not be inflated by voters,
and thus lead the system to produce a Duvergerian equilibrium. In other
words, to support the idea that the number of parties is inflated in
superposition systems, I need to show that there are enough voters who
contribute to the inflation of the number of parties by voting for marginal
candidates. The final section of this chapter addresses whether or not voters
will contribute to the increase in the number of parties by voting for a marginal
candidate.

The argument here is a rather simple one. If a significant number of
voters cast their ballots for a marginally performing candidate given a larger
number of candidates in plurality districts, then the number of parties will be
larger than that in a regular plurality election. That is, even When a number of
voters do not have an opportunity to vote for the most preferred candidate, as
long as some significant number of voters cast their ballots for marginal
competitors, the plurality section of the system would produce an inflated
number of parties, leading to the violation of a Duvergerian equilibrium. As it
was mentioned earlier, voters’ tendency to cast their ballots sincerely creates

a violation of the Duvergerian equilibrium.

 

and 2, the addition of the third candidate discourages voters who have the above preference
to turn out.

82

In the following three chapters, we will turn to empirical analysis.
Chapter 5 demonstrates the change in the incentive structure of political
parties in superposition systems leads to an inflation of the number of
candidates. Chapters 6 and 7 focus on how voters contribute to shape party
systems. Chapter 6 demonstrates how the increased number of candidates
changes voting conditions and how the change in conditions lead to prevent
voters from engaging in strategic voting. I will also return to a question of
ecological inference problem in Chapter 6 and empirically show that the use
of aggregate level data will overestimates the number of parties. Chapter 7
examines whether or not a superposition system creates a significant number
of sincere voters so that the number of parties is inflated to a level that is

larger than what Duvergerian equilibria expect.

83

Part III.

Empirical Assessment of Strategic Behavior in Mixed Superposition

Electoral Systems

Chapter 5

Testing Parties’ Strategic Entries

The idea of the contamination effects between the plurality component and
the proportional representation (PR) component of a mixed superposition
system suggests that political parties would perform better in the PR
component of an election by placing their candidates in plurality districts. The
contamination effects have actually been found in the system that combines
single member district (SMD) and PR rules."2 Japan’s Lower House elections
and Russia’s Duma elections are the examples of such cases.

As it was mentioned earlier, Japan practices two types of mixed
systems. The Lower House uses a mixed system that combines an SMD rule
and a PR rule. The Upper House applies a mixed system with a plurality
multimember district rule and a PR rule. If we find the contamination effects in
the Upper House elections, the idea that political parties gain more votes in
the PR section by placing candidates in the plurality districts will be
strengthened since the results are robust, even when it is tested in a different
type of superposition system. The purpose of this chapter, then, is to test the
robustness of the contamination effects by testing the idea in Japan’s Upper
House elections.

Although testing the contamination effects in Japan’s Upper House is

the main goal of this chapter, the empirical pattern of how candidates are

 

‘2 For example, the study carried out by Mizusaki and Mori (1997) is on the Japanese Lower
House elections. Herron and Nishikawa (2001) also studied Japan’s Lower House elections.

85

placed will also be included. This is because the map of how candidates are
placed in different districts will give us a preliminary understanding on whether
or not the number of candidates is inflated in the plurality section of
superposition systems. However, a careful description of how candidates are
placed has not been carried out in previous studies.

The organization of this chapter is as follows. First, I will show that the
number of candidates is inflated substantially in the plurality component of
Japan’s superposition systems, including the Lower and the Upper Houses.
Second, I will test the contamination effects in both the Lower and the Upper
House elections. The results for the Lower House will be a reproduction of the
test done by Herron and Nishikawa (2001). Although such a reproduction of
the results may be redundant, it will be added to the chapter. In this way, the
flow of the analysis will be smoother. Moreover, the comparison of the results
between the Lower House and the Upper House will be more thorough, if the
results for both the Lower and the Upper Houses are presented.

For testing the contamination effects, two hypotheses will be used, one
for the Lower House and the other for the Upper House. I will then describe
how Herron and Nishikawa (2001) tested the hypothesis for the Lower House,
followed by their empirical finding. After the description of the results for the

Lower House, I will carry out the analysis for the Upper House in a similar

manner.

 

Empirical Patterns of Candidate Placement

Table 5-1 shows the number of candidates in the SMD and PR
components in the 1996 and 2000 Lower House elections. It shows that the
number of candidates is substantially larger than what a Duvergerian
equilibrium suggests. The number of candidates is larger than two in the
plurality component of the Lower House elections since they use an SMD
plurality rule. Only 1.7% of the districts has two candidates in the 1996
election. The result is the same for the 2000 election, showing that only 1.7%
of the districts has two candidates. The typical number of candidates in these
districts are higher than two and falls somewhere between three and five. The
districts with three to five candidates make up more than 85% of the cases. In
both the 1996 and 2000 election, the mean number of candidates is four.43

Table 5-2a presents the number of districts broken down by district
magnitude. The information on district magnitude is added before the
presentation of the patterns of candidate placements, since candidate
placements are largely influenced by district magnitude.

The table shows that the largest frequency (i.e. mode) of district
magnitude is one. Twenty-six districts have the magnitude of 1 between 1983
and 1992."4 The next largest is the district magnitude of 2 followed by 3 and 4.
There are only two districts that produce four seats between 1983 and 1992.

These districts are Hokkaido and Tokyo prefectures.

 

“The number of candidates broken down by political parties is shown in Appendix 5-1.
“ The same district magnitude was kept between 1983 and 1992. However, it was adjusted
before the 1995 election.

87

Table 5-1: Patterns of Candidate Placement in the Lower House

Elections — 1996 and 2000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NUMBER OF CANDIDATES 1996 2000
2 1.7 1.7
3 27.3 34.3
4 38.0 37.0
5 20.7 18.7
6 8.3 7.0
7 3.0 1.3
8 1.0 0
Total 100% 100%
Mean 4.21 3.99
N 300 300

 

88

 

Table 5-2a: District Magnitude in the Plurality Section of the Upper
House Elections Between 1983 and 2001

 

 

 

 

 

DISTRICT MAGNITU DE
Year 1 2 3 4 Total
1983-1992 26 15 4 2 47
1995 24 18 4 1 47
2001 27 15 4 1 47

 

 

 

 

 

 

 

 

Note: The total number of seats is 76.

89

 

Before the 1995 election, the number of seats in Hokkaido was reduced to
two.

Table 5-2b presents the patterns of candidate placement in the Upper
House elections between 1983 and 2001 in the same way as the table for the
Lower House elections.” Since the district magnitude varies from district to
district, it is difficult to determine whether or not the numbers of candidates
are higher than what Duvergerian equilibria suggest in this table. However,
the table reveals that only a few districts have two candidates, althOugh there
are more than twenty-four districts that produce one seat.

One of the interesting features that the table reveals is that the number
of candidates has not decreased over time, although we see a structural
break between 1992 and 1995 when the district magnitude is adjusted in
some districts. These results are rather counterintuitive if we follow the idea
that Duvergerian equilibria are reached over time; that is, the number of
parties will be reduced over time. The table, as a result, may suggest that an
important factor may be interfering with the system to reach to a Duvergerian
equilibrium.

Since Table 5-2b is not adequate to see how the number of
candidates is larger than Duvergerian equilibria suggest, the next table (Table
5-20) shows the number of candidates controlled for district magnitude. The

columns represent district magnitude, while the rows reflect the number of

 

‘5 The same superposition system was used between 1983 and 1998. They used a closed list
PR rule. The system in 2001 uses an open-list PR rule.

90

Table 5-2b: Patterns of Candidate Placement — the Upper House

Elections Between 1983 and 2001

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

NUMBER OF 1983 1986 1989 1992 1995 1998 2001
CANDIDATES

2 2.1 2.1 0 2.1 2.1 0 0

3 42.6 40.4 34.0 42.6 17 6.4 2.1
4 23.4 34.0 21.3 25.5 27.7 21.3 27.7
5 14.9 2.1 14.9 8.5 25.5 36.2 34

6 6.4 6.4 2.1 4.3 12.8 2.1 10.6
7 4.3 4.3 12.8 6.4 2.1 6.4 6.4
8 O 6.4 2.1 2.1 2.1 10.6 6.4
9 O 2.1 4.3 0 0 2.1 0

10 or more 6.4 2.1 8.5 0 10.6 14.9 12.8
Total 100% 100% 100% 100% 100% 100% l 100%
N 47 47 47 47 47 47 47

 

91

 

Table 5-2c: District Magnitude and Patterns of Candidate Placement -

the Upper House Elections Between 1983 and 2001

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

DISTRICT MAGNITUDE
Excess 1 2 3 4 Total
Candidates
0 2.2 3.6 0 0 2.4
1 46.4 30.0 3.4 0 35.6
2 29.1 23.6 17.2 9.1 25.5
3 20.7 13.6 24.1 27.3 18.8
4 0.6 9.1 3.4 9.1 4
5 or more 1.1 20.0 51.7 54.5 13.7
Total 100% 100% 100% 100% 100%
N 179 110 29 11 329

 

92

 

excess candidates. The number of excess candidates equals the number of
candidates minus M+1. For an illustration, if there are three candidates when
the district magnitude is 1, the number of excess candidates is 1 since 3-2=1.
The table shows that the numbers of candidates in the Upper House
elections tend to be larger than the M+1 Rule predicts. The tendency that the
number of candidates reaches the level that is higher than the M+1 Rule
predicts are more prominent when the district magnitude is larger. More
specifically, the mode is one excess candidate (46.4%) when the district
magnitude is 1. When the district magnitude is 4, none of the districts are
close to the level that Duvergerian equilibria suggest. Fifty-five percent of the

districts have five or more excess candidates.46

Hypotheses for Strategic Entries
The first hypothesis focuses on the strategic party entries into the SMD

part of a mixed system in Japan’s Lower House elections. If the idea of

advertisement effect is correct, we expect parties who place candidates in

SMD districts to do better in the PR districts compared to others who do not

place candidates. Based on the idea, the following hypothesis is tested.
Hypothesis 1. The performance of political parties in the PR
portion of the mixed system is improved by placing a candidate

in conesponding single member districts (Herron and Nishikawa
2001).

 

‘5 The number of candidates broken down by political parties is shown in Appendix 5-2.

93

If Hypothesis 1 is supported, we can expect that parties’ candidate
nomination strategy in the SMD portion of the system will be contaminated by
the PR portion. If this is the case, we can predict that the SMD elections,
under a superposition system will not follow Duverger’s Law in terms of party
entries.

As is the case for the Lower House elections, we expect that parties
who place candidates in the plurality component of the Upper House election
perform better in the corresponding PR districts. If this is true, a following
hypothesis should be supported empirically.

Hypothesis 2. The performance of political parties in the PR

portion of the mixed system is improved by placing a candidate

in corresponding multi-member districts.
Hypothesis 2 is similar to Hypothesis 1 in that both hypotheses examine the
impact of placing a candidate in the plurality portion of the system on parties”
performance on the PR portion. Hypothesis 1 examines the impact in the
mixed system with the SMD rule, while Hypothesis 2 examines the effect in
the mixed system with the medium size district rule. However, the hypotheses
are different in terms of the district magnitude in the plurality section of the
election. As it was mentioned, Hypothesis 1 has already been tested by
Herron and Nishikawa (2000). If Hypothesis 2 is supported, the idea of the

contamination effects will be bolstered further.

94

Testing Hypotheses for Mixed Superposition Electoral System with SMD
and PR

The first hypothesis is that the performance of political parties in the
PR portion of the superposition system is improved by placing a candidate in
corresponding single member districts, demonstrating contamination effects.
The dependent variable in the analysis is performance of the party in the PR
component. The data from the 1996 Lower House election are used to
measure each party’s performance in the PR component of the election. The
dependent variable reflects the proportion of PR vote received by the given
party in the corresponding single member district. The unit of analysis is
districts. The number of cases is 300 since there are 300 single member
districts.

There are three independent variables in each model. The primary
independent variable is candidate placement in SMDs. It is coded 1 if a
candidate of the given party participated in the district, and it is coded 0 if the
party did not field a candidate. If it is statistically significant, party performance
in PR is improved by candidate placement, indicating that there is a
contamination across the two components of the mixed system.

The second and the third independent variables provide a control for
the main hypothesis. More specifically, while we anticipate that the
advertisement effect, in part, explains party performance in the PR section,
one could argue that careful candidate placement is a more appropriate
explanation. That is, a party will place candidates in districts where it knows

that the party enjoys a high level of popularity.

95

To account for this alternative, two variables are introduced to the
model. One is the variable for incumbency. Parties can expect that
incumbents will do well in general, and thus, adding information on
incumbency can be a proxy for popularity. A dummy variable, INCUMBENT,
is added to all models. The variable is coded 1 if the candidate running in the
district is an incumbent who won her seat in the previous election and 0 if the
candidate is not an incumbent.47 The other control variable is how well parties
performed in different districts in the 1995 Upper House elections.48 The
Upper House election has fifty districts, and we can acquire detailed
information regarding in which district each party is popular.

The Ordinary Least Squared (OLS) estimates are used to find out
whether or not the SMD variable has a positive impact on parties’
performance in the PR section of the election. It is ideal to test the hypothesis
with all of the parties. However, some parties will be excluded from the
analyses. For example, the Liberal Democratic Party (LDP) and the Japan
Communist Party (JCP) will not be included since they place candidates in
most of the single member districts. As a result, the independent variable will
not have enough variance, and the performance of OLS estimation will be

impossible.

 

‘7 INCUMBENT was coded 1 if the candidate was an incumbent. That is, candidates are
coded 1 even when they are not from the district in which she ran in the election under the
analysis.

‘2 The 1995 Upper House election results are used instead of the 1993 Lower House election
results. This is because some new parties were created or some old parties are merged into
different parties between 1993 and 1996. As a result, it is difficult to use the 1993 Lower
House election to assess the popularity of party in 1996.

96

Similarly, some parties are excluded from analyses since they place
their candidates in only a few districts. Placing only a few candidates also
prevent the independent variable from creating enough variance. Some
parties are excluded since they do not run in the plurality component of the
election.

In a similar manner, the addition of a control variable cannot be carried
out for some parties, since these parties do not have enough variation in a
control variable. For example, some parties are new and were created just
before the 1996 election. As a result, these parties do not have an electoral
history to create the POPULARITY variable. For these parties, the analysis is
carried out without the POPULARITY variable. The results are also shown
separately from other parties that have both INCUMBENT and POPULARITY
variables as control variables.

The results that include both INCUMBENT and POPULARITY
variables are shown in Table 5-3. The parties that are included in this table
are the New Frontier Party (NFP) and the Social Democratic Party (SDP).
The table supports the idea of contamination effects. The variable of our
interest, SMD, is significant, even after controlling for INCUMBENT and
POPULARITY. For example, for the NFP, the coefficient is 3.174, and it is
statistically significant at the 99% confidence level. Substantively, this means
that, on average, the NFP gained an extra 3 percentage points in a
corresponding PR district by placing a candidate in a single member district.

Similarly, the coefficient for the SDP is 2.264 and statistically significant.

97

Table 5-3: OLS Results for the 1996 Lower House Election with

INCUMBENT and POPULARITY Variables

 

 

 

 

 

 

 

 

 

 

NFP SDPJ
Constant 4.720*** .610*
(1.278) (.335)
SMD 3.174*** 2264*“
(.687) (.347)
Incumbent 1 .995*** 7.052***
(569) (570)
Popularity .643*** .331 ***
(.043) (.019)
Adjusted R2 .565 .758
N 299 286

 

Dependent Variable: SMD vote share for the specified party in the PR
component of the ballot. Standard errors are shown in parentheses.

*** = significant at the .01 level
** = significant at the .05 level
* = significant at the .10 level

 

Note: The problem of collinearity is not substantial since the Variance Inflation
Factor (VIF) for each variable was close to 1.00 in each equation. The
elimination of outliers does not affect the output substantively.

Source: Herron and Nishikawa (2001)

98

This suggests that the SDP gained 2 percentage points more in PR districts
by placing candidates in the SMD part of the election.

Table 5-4 shows the results for parties that do not include
POPULARITY as a control variable. The results for six parties are shown in
the table. These parties include the Democratic Party of Japan (DPJ),
Jiyurengo (JlYU), the National Frontier Party (NFP), the New Harbinger Party
(NHP), and Shinshakaito. The results support that the parties gain an extra
vote share by placing candidates in the SMD part of a mixed system. The
coefficients for the SMD variable are positive for all the parties and statistically
significant at the 99% confidence level. These positive coefficients suggest

the presence of contamination effect in mixed electoral systems.

Testing Hypotheses for Mixed Electoral Systems with SNTV and PR

The hypothesis for the mixed system with a single nontransferable vote
(SNTV) rule is tested by using the data from election returns, as it was the
case for the first version of the mixed system. Hypothesis 2 tests whether or
not the performance of political parties in the PR portion of the mixed system
is improved by placing a candidate in corresponding medium size districts. If
they improve their performance in the PR component of the system, we can
consider that the improvement was caused by advertisement effects. This, in

turn, suggests the existence of contamination effects.

99

Table 5-4: OLS Results for the 1996 Lower House without POPULARITY

 

 

 

 

 

 

Variable
DPJ JIYU NFP NHP SDPJ SHINSH
AKAITO
Constant 1 1 .859*** .732*** 22.166*** 1 .285*** 6.129*** 1 632*”
(.379) (.104) (.706) (.080) (.141) (.091)
7.903*** .802*** 5.737*** 3.501*** 2202*“ 1.164***
SMD (.615) (.218) (.881) (.670) (.496) (.274)
3208*” 1 5.271 *** 2838*“ 8.91 3*“ 8.893*** 5.684***
Incumbent (.867) (1.196) (.750) (.781) (.561) (1.080)
Agijusted .471 .451 .236 .663 .505 .160
R
N 299 232 299 153 286 299

 

 

 

 

 

 

 

 

Dependent Variable: SMD district-level percent of vote for the specified party
in the PR component of the ballot. Standard errors are in parentheses.

*** = significant at the .01 level
** = significant at the .05 level
* = significant at the .10 level

Note: The problem of collinearity is not substantial since the Variance Inflation
Factor (VIF) for each variable was close to 1.00 in each equation. The
elimination of outliers does not affect the output substantively.

Source: Herron and Nishikawa (2001)

100

 

The dependent variable will be the PR vote shares of parties that correspond
with the medium size district in the 1983 Upper House election. The unit of
analyses is each medium size district. The number of cases is 47 since there
are forty-seven districts.” The independent variable of our interest is whether
or not candidates are placed in the plurality districts.

The two independent variables are again used to give controls for the
primary independent variable. The control variables are added to the
analyses for the same reason as for the Lower House. We can expect that
parties may know in what districts they have a chance to win a seat. That is,
according to their knowledge, they place their candidates. The vote shares in
these particular districts are expected to be higher than other districts. In
other words, parties do better in these districts, not because they place their
candidates, but because they are popular in these districts.

The first control variable is whether or not the nominated candidates
are incumbents. The other control variable is the popularity variable. The
popularity variable is created based on how well each party performed in
each district in the 1980 Lower House election.

As was the case for the mixed system with an SMD rule, the JCP is
excluded from the analysis. This is because the JCP placed their candidates
in all of the plurality districts, and the independent variable will not have any

variance at all. As a result, the OLS estimation cannot be carried out. For a

 

‘9 These districts correspond with prefectures.

101

similar reason, the analysis on New Party Harbinger (NPH) does not have the
INCUMBENT variable.

Table 5-5 shows the results for parties that have both INCUMBENT
and POPULARITY variables as control variables. The table basically supports
the contamination effects, even though the results are not as consistent as
the ones for the mixed system with an SMD rule (Table 4-1). More specifically,
the table shows that the coefficients for the medium size district system
(MSD) variable are positive and statistically significant for the Japan Socialist
Party (JSP) and the New Frontier Party (NF P), while it is not significant for the
Liberal Democratic Party (LDP). The coefficients for the JSP and NFP are
2.955 and 3.184 (the former is statistically significant at the 95% confidence
level, while the latter is significant at the 90% confidence level). These
coefficients suggest that the JSP and NF P on average gain extra 3
percentage points in the PR section of the system by placing their candidates
in the medium size districts.”

The results for a party which does not have the INCUMBENT variable
are shown in Table 5-6. The party that is analyzed here is the NPH. The
results support the hypothesis that the contamination effect exists. The
coefficient for the SMD variable is 4.49. The association is statistically

significant at the 99% confidence level. The coefficient suggests that the

 

5° The effects on the LDP do not exist after controlling for other independent variables. This is
only a speculation but the reason for this is because the LDP has been a ruling party of

Japan for most of its postwar history. As a result, the advertisement effect is not as effective
and does not give an extra vote share to a large party like the LDP, as it does to small parties.
In other words, voters in Japan know enough about the LDP, and the placement of a
candidate does not help the party as much as it helps other parties.

102

placement of a candidate improves NPH’s performance in the PR component

of the system by 4.5 percentage points.

Summary

The findings in this chapter support that the number of candidates is
larger in the plurality component of a mixed system, compared to that of a
regular plurality rule. This is especially because small parties develop an
incentive to place their candidates to gain an extra vote share in the PR
component of a mixed system. As a result, even when small parties expect
that their candidates would not win a seat, they would chose to place
candidates.

Overall, the results for Hypotheses 1 and 2 support that the number of
candidates will be larger in the plurality component of a mixed system. These
results suggest that the number of candidates is more likely to be larger than
2 in the SMD part of the mixed system (Hypothesis 1), while the number of
candidates would be larger than M+1 (Hypothesis 2).

What these suggest is that the change in parties’ behavior will give
increased opportunity for voters to spread their votes. In other words, this
chapter showed that parties created the foundation for violating Duverger’s
Law and the M+1 Rule. If voters indeed spread their votes to small parties
ranked lower than two in the SMD part of a mixed system or M+1 in the
medium size district part, we can expect that the plurality section of a mixed

system has a tendency to violate Duverger’s Law and the M+1 Rule.

103

Table 5-5: OLS Results for the 1995 Upper House Election with

INCUMBENT and POPULARITY Variables

Dependent Variable: Vote share for the specified party in the PR component

 

 

 

 

 

 

LDP JSP NFP
Constant 16.348*** 5.955*** 19.606***
(2.929) (2.159) (2.062)
MSD 0.251 2955* 3.814"
(2.138) (1.491) (1.717)
incumbent 1.713 1.639 .265
(.569) (1.632) (2.708)
Popularity .313*** .613*** .290***
(058) (136) (070)
Adjusted n2 .377 .727 .346
N 47 47 47

 

 

 

 

 

 

of the ballot. Standard errors are shown in parentheses.

*** = significant at the .01 level
** = significant at the .05 level

it

= significant at the .10 level

104

 

Table 5-6: OLS Results for the 1995 Upper House Election without
INCUMBENT Variable

 

 

 

NPH
Constant 1 .755***

(299)
MSD 4.490***

(1 .089)

 

Popularity .378***
(.051)

 

Adjusted R2 .728
N 47

 

 

 

 

 

Dependent Variable: Vote share for the specified party in the PR component
of the ballot. Standard errors are shown in parentheses.

*** = significant at the .01 level

** = significant at the .05 level
* = significant at the .10 level

105

Chapter 6

Voting Environment and Occurrence of Strategic Voting

As was discussed in Chapter 4, the previous studies on voting behavior find
the evidence of strategic voting in mixed electoral systems. However, if we
take the existence of the contamination effect into account, the occurrence of
strategic voting should not be as frequent as the literature suggests. In this
chapter, I will empirically show that strategic voting rarely occurs in mixed
superposition systems.

I will carry out two tasks to achieve the above. First, I will show that few
voters are in an environment that promotes strategic voting in superposition
systems. As it was argued in the theoretical section for voting behavior, voters
need to be in a certain voting environment for strategic voting to take place.
Superposition systems tend to create conditions that prevent voters from
behaving strategically. As a result, the occurrence of strategic voting is
reduced more than we may expect. Second, I will show that the proportion of
voters who engage in strategic voting is actually small in superposition
systems. In particular, I will classify voters into different groups, based on their
voting pattern, and show that the proportion of voters who belong to the group
of strategic voters is small.

In the following section of this chapter, I will first discuss how certain
voting environment reduces the occurrence of strategic voting in more detail. I

will especially focus on two aspects, including the opportunity factor and the

106

expected utility factor. I will also discuss the reason why the use of aggregate
level data may overestimate the occurrence of strategic voting in
superposition systems. I will especially discuss how the contamination effects
relate to an overestimation of strategic voting when aggregate level data are

used.

Contamination Effects and Voting Conditions

The use of aggregate level data overestimates the occurrence of
strategic voting because it ignores the existence of voting conditions that
prevents strategic voting from occurring. I argue that the occurrence of
strategic voting is reduced by two major factors, including the opportunity
factor and the expected utility factor. Briefly, the opportunity factor refers to
whether or not voters’ most preferred candidate is available in their plurality
district, while the expected utility factor refers to whether or not voters can
increase their utilities by voting for a less preferred candidate.

For the opportunity factor, it plays an important role in creating a
different pattern of voting behavior between the plurality component of
superposition systems and regular plurality systems. The opportunity factor
leads superposition systems to reduce the occurrence of strategic voting.
Superposition systems create a condition in which voters face different choice
sets in terms of political parties.51 Political parties create an incentive to place

candidates in the plurality districts of a superposition system. However, this

 

5‘ The idea that voters’ choice sets influence voting behavior is introduced in Blais and
Nadeau (1996).

107

does not mean that parties can always place their candidates in all of the
districts. As a result, voters face different patterns of candidate placement
across districts. This creates the condition in which some voters have an
opportunity to vote for a candidate who is from their most preferred party in
their district, while others do not. If we ignore these different voting
environments by measuring strategic behavior only at an aggregate level
data, it will lead to an overestimation of the occurrence of strategic voting. As
an illustration, the definition of strategic voting is to transfer a vote to a less
preferred candidate to maximize a voter’s expected utility. More specifically, a
voter’s preference ordering is C3>Cz>C1, where candidates C1 and C2 are
front runners while candidate C3 is a marginal candidate. If the voter transfers
his ballot to 02 to maximize his expected utility, his behavior is considered to
be strategic.

An important assumption involved in the definition is that a marginal
candidate who the voter prefers most runs in his district. If his most preferred
candidate is not available in the district, the voter is forced to transfer his vote
to one of the other two candidates who are available in the district. In the
above example, the voter is expected to vote for 02 since he prefers C2 to C1.

This pattern of voting may look as if the voter engaged in strategic
voting, if we only look at aggregate level data. For example, Reed (1999)
supports the presence of strategic voting by showing that the first and second
finishers get more SMD votes compared to PR votes, while the third and

fourth finishers get more PR votes than SMD votes. However, a problem with

108

this approach is that the opportunity factor is not taken into account. Voters
who do not have an opportunity to vote for their most preferred candidate are
expected to transfer their ballots to other candidates. If they do so, they look
as if they voted strategically, even though transferring their vote from a
marginal candidate to a front runner was out of their control. If a large number
of voters transfer their ballots to one of the front runners instead of a marginal
candidate, this will look as if a large number of voters behaved strategically at
aggregate level data.52

The analysis of two public opinion surveys, including the 1996 Japan
Election and Democracy Study (JED896) and the 1983 Japan Election Study
(JES83), suggest that researchers need to be cautious about the opportunity
factor. The surveys show that there are a significant number of voters who do
not have an opportunity to vote for their most preferred candidate. There are

feeling thermometer questions on political parties in the surveys. The party

 

52 Reed (1999) states that the JCP candidates gain more votes in the PR component than in
the SMD part. The fact that the JCP places candidates in most of the districts suggests that
the above pattern is created by the presence of strategic voting. However, the difference is
only 0.18 percentage points. This may suggest that JCP supporters vote sincerely, but does
not necessarily support the presence of strategic voting. Another evidence that Reed uses is
that the first and the second finishers get more SMD votes than PR votes. This may have an
ecological inference problem. This is because there is a variation in candidates' placement
across districts, and voters who do not have an opportunity to vote for their most preferred
candidate may choose to vote for one of the front runners. For example, there are voters who
prefer the Democratic Socialist Party (DSP). If these voters do not have an opportunity to
vote for a candidate from the DSP, they have to transfer their votes to one of the other
candidates. If the voters are more likely to transfer their votes to a front runner instead of a
marginal candidate, the overall percentage of vote shares among the first and the second
finishers in the plurality component will be larger in the SMD part than the PR component.
Indeed, the JED896 survey reveals that 82% of the DSP supporters with no opportunity,
transferred their votes to a non-JCP candidate, while only 18% transferred to the JCP
candidate. Given that JCP candidates are usually marginal candidates, this suggests that
more voters transfer their ballots to a front running candidate than marginal candidates.

109

that has the highest value is considered to be a voter’s most preferred party.53
If a voter prefers a particular political party, we can project that the voter is
likely to prefer a candidate who is from that party.“ Since we can locate what
district respondents came from, we can find out whether or not they have an
opportunity to vote for their most preferred candidate.

According to the JEDS96, 91% of the electorate had an opportunity to
vote for their most preferred candidate in the Lower House election, while 9%
did not. The result is similar in the 1983 Upper House election, where-92%
had an opportunity and 8% did not. The figures show a substantial reduction
in the number of potential strategic voters (i.e. 9% in the Lower House
election and 8% in the Upper House election).

The second aspect of voting conditions that we need to pay attention
to is the expected utility term. In addition to finding out if voters’ most
preferred candidate is running in their districts, it is necessary to see if they
actually need to vote strategically. More specifically, another assumption of

strategic voting is that a voter happens to prefer a marginal candidate. A voter

 

5" Even though there may be some problems in using thermometer questions in surveys,
using these questions is considered to be one of the best ways to measure voters’
preferences and is widely used.

It is also possible to rank candidates instead of political parties since candidates play an
important role in Japan’s elections (Miyake 1989; Rochon 1981). However, it nonetheless
seems to be appropriate to depend on the preference ordering of parties as the proxy for the
preference ordering of candidates for two reasons. First, since the SMD rule was only
introduced in the 1996 Lower House election, and thus about 50% of the candidates in each
district were new to voters, it is impossible for voters to rank the popularity of candidates
apart from their party membership. The second reason applies to the Lower and Upper
House elections. Some researchers have shown that political parties play a substantial role in
Japan’s elections (Kobayashi 1991; Miyake 1995; Richardson 1988). In other words, ranking
parties may be a better solution than ranking candidates. Moreover, to rank candidates in
terms of the likelihood of winning a seat is a complicated task since we always see some
mobility in the entry of new candidates in an election. Using actual election results in 1983 is
one of the best solutions to solve this problem, even though the ranking is done ex ante.

110

who prefers a front-runner does not have to transfer her vote to maximize her
expected utility. Only those voters who prefer a marginal candidate have to
choose whether or not to behave strategically. When a voter prefers a front
runner, she can simply cast her ballot for the most preferred candidate since
their sincere preference and sophisticated choice coincide. When she casts
her ballot for her most preferred candidate, she is called a straightfonrvard
voter (Farquharson 1969).

The existence of straightforward voters suggests that it is important to
consider the marginality of candidates.“’5 We need to detect who happens to
prefer a front running candidate and exclude these voters from the group of
potential strategic voters. Since the previous studies use aggregate level
data, they cannot find the size of straightfonivard voters in the electorate. By
identifying who prefers a marginal candidate, we can measure the occurrence
of strategic voting more precisely.

According to the two surveys, including the JEDS96 and the JES83,
the marginality factor reduces the occurrence of strategic voting greatly.56 A
vast majority of the electorate happens to support a front running candidate

and does not have to vote strategically to maximize their expected utility. In

 

55 This does not suggest that it is not necessary to consider the existence of straightforward
voters in regular plurality systems since the marginality factor can play a role in reducing the
number of potential strategic voters in both regular plurality systems and superposition
systems. However, since detecting the size of straightforward voters is important, the analysis
on the marginality factor is included.

56 We can create the marginality variable by looking at how voters answered feeing
thermometer questions on political parties and how candidates performed in the election.
Candidates will be considered marginal when their vote shares are the third largest or lower
in Japan’s Lower House election since it applies a superposition system with an SMD plurality
district. When candidates’ vote shares are M+2"d or lower in Japan's Upper House election,

111

the Lower House election, 77% of the electorate belongs to this category,
while 72% belongs to this category in the Upper House election.

The reduction in the occurrence of strategic voters is even more
substantial, if we consider how the opportunity and marginality factors
together decrease the size of potential group of strategic voters. Tables 6-1 a
and 6-1b show the results. As described earlier, the 9% does not have an
opportunity to vote for their most preferred candidate in the Lower House.
Moreover, 77% belongs to the voting condition in which they do not have to
vote strategically since they happen to prefer a front runner. This means that
86% (9%+77%) of the electorate is not in the condition that promotes
strategic voting (Table 6-1a). In the Upper House, the figures are 8%
and72% respectively. This shows that 80% of the electorate in the Upper
House is not in the condition to vote strategically (Table 6-1b).

Besides the marginality factor, there are other factors that relate to
expected utilities. Voters’ perceptions of candidates’ probability of winning
and utility differences also need to be in a certain condition to promote
strategic voting. More specifically, it is not rational to vote strategically even
when the opportunity and the marginality factors promote them to vote
strategically. This is especially because of voters’ perceptions on candidates’
probability of winning and utility differences. As an illustration, a voter’s
preference ordering is C3>C2>C1. The voter perceives that candidate C1 is

leading the race in an SMD plurality election. However, the competition

 

they will be considered marginal since the election utilizes a multimember plurality district
rule.

112

Table 6-1a: Reduction in the Size of Potential Strategic Voters by

Opportunity and Marginality Factors - Lower House

 

 

 

 

 

 

 

 

 

 

OPPORTUNITY
MARGINALITY Yes No Total
Front Candidate 77% 0% 77%
Marginal 14% 9% 23%
Candidate
Total 91 % 9% 100%
N 963

 

113

 

Table 6-1b: Reduction in the Size of Potential Strategic Voters by
Opportunity and Marginality Factors - Upper House

 

 

 

 

 

 

OPPORTUNITY
MARGINALITY Yes No Total
Front Candidate 72% 0% 72%
Marginal 20% 8% 28%
Candidate
Total 92% 8% 100%
N 1329

 

 

 

 

 

 

114

between Cg and C3 is close, even though Cg may be leading slightly. Under
such circumstances, it is more rational for the voter to behave sincerely
because it is rather difficult to determine which candidate (Cg or C3) is trailing
in the district (Cox 1997).

Similarly, there may be some voters who prefer a particular candidate
with a strong intensity and do not perceive much difference among other
candidates in terms of utilities. In this case, the voter is not expected to vote
strategically since casting a decisive ballot among the rest of the candidates
yields similar utilities in the election outcome. Under the condition, voters are
discouraged to vote strategically (Cox 1997).57

Table 6-2 shows the percentage of people who belong to a group in
which voters would not increase their expected utility by transferring their
ballots. The analysis includes respondents who have an opportunity to vote
for their most preferred candidate and happen to prefer a marginal candidate
in the Lower House election (i.e. 14% of the electorate). The respondents are
categorized based on the values on the expected utility term for the most
preferred candidate minus the second most preferred candidate that is

calculated based on an equation developed by McKelevey and Ordeshook

 

57McKelevey and Ordeshook (1972) developed a formula for a multicandidate race in a
plurality SMD rule. It summarizes different conditions that may encourage or discourage
strategic voting. The formula is as follows. E"- E°= 2..., [ p... (u., - u.) ] where E" is the
expected utility that a voter will gain by voting for candidate k, and E° is the expected utility
that is associated with abstention. The term p», is the probability that the voter will cast a
decisive vote in between the candidates i and k. The term (u. — u.) shows the utility of voting
for the candidate k instead of i.

115

Table 6-2: Reduction in the Size of Potential Strategic Voters by
Candidates’ Probability of Winning and Utility Differences

 

 

 

 

 

 

 

MOST PAREFERRED %
- SECOND PREFERED

-7.99~-1.00 14.0
-0.99~0.93 15.1
0.93~2.99 34.9
3.00~25.00 36.0
Total 100%

N 86

 

 

 

 

116

 

(1972).58 Respondents who have a negative value are the ones who tend to
increase their expected utilities by engaging in strategic voting, whereas
respondents who have a positive value should vote sincerely to maximize
their expected utility. Only about 30% of those who are included in this
analysis have a positive value.59 This means that only about 4% (30% x 14%)
of the electorate is in the condition that encourages voters to behave
strategically in the 1996 Lower House election.

Furthermore, the reduction in the number of strategic voting may not
stop here. In the plurality component of the mixed system, strategic voting is
less likely to occur even when above three conditions promote strategic
voting. This also has to do with the contamination effect by the presence of
PR elections. More specifically, strategic voting requires more conditions in
order to make voters behave strategically. One of these conditions is that
voters need to be oriented with a short-term instrumental motivation (Cox

1997). Some voters may think that their most preferred candidate may have a

 

58Based on the work done by McKelevey and Ordeshook (1972), the following equation is
introduced by Black (1978). The formula states: E1 — E2 = 2 p13 ( u, - ug ) + p13 ( U1 — u3 ) -
pg3 ( ug - u3 )2. Black states that if E"-E° > 0, the rational voter casts a vote for candidate 1,
when E1 - E > 0. It leads to sincere voting. If E1 - E2 < 0, the voter will vote for candidate 2
and is classified as a strategic voter. The formula suggests that candidates’ probability of
winning and utility differences should fit in a certain pattern to promote strategic voting. For
example, the probability of casting a decisive vote for the first preferred candidate (p1g or p13)
should be small to promote strategic voting. The utility difference between the first preferred
candidate and the second, (u1-ug), or the third, (u1-u3), should also be small. Moreover, the
utility difference between the second preferred candidate and the third preferred candidate
(ug-u3) should be substantially large as well as their race should be close, is. pg3 should be
large. If these conditions do not meet, it Is not rational for voters to engage in strategic voting.
We can assess E1 - E2, if we know the values for the term plk. and the term (uk — u.). I will
use actual vote shares of candidates to create the term p..,. It may be better to use public
Opinion survey data to find out voters’ perceptions. However, I will use election outcomes
since no surveys have enough cases in each plurality district to project how voters perceive
candidates’ probability of winning. To reflect (uk — u.), feeling thermometer questions in a
public opinion survey will be used. For example, a voter answers 100 degrees for competitor

117

chance of winning when elections are repeated, even though the candidate
does not have a chance of winning in a current election. If this is how voters
perceive a candidate’s chance of winning, they are more willing to vote for a
marginal candidate.

If we consider a mixed system with the contamination effect, it is likely
that voters under mixed system have a longer-term incentive. This is because
small parties can survive due to the presence of the PR election, so that small
parties would not at least disappear as they may do in a regular plurality SMD
election. As a result, voters may be more willing to support their most
preferred candidate to keep them in the plurality section of the system, hoping
that the candidate would win in a future election.

Another reason why mixed systems promote sincere voting is because
the contamination effects increase the number of candidates in plurality
districts. The increase, in turn, can decrease the frequency of strategic voting.
This is true even when all of the conditions that are discussed above (i.e. the
opportunity and the expected utility factors) promote strategic voting.

When voters have difficulty in guessing who is trailing in the district,
voters are more likely to vote sincerely (Cox 1997). If we apply this idea and
compare the plurality section of superposition systems and regular plurality
systems, we can expect that voters are more likely to vote sincerely in
superposition systems than in regular plurality systems because the system is

expected to create close races.

 

C1, 90 degrees for competitor Cg and 80 degrees for competitor C3. The voter's utility
distance between the first and the second candidate is 10 since (u, — ug) = 100 - 90.

118

What the above discussion suggests is that mixed systems produce a
large number of voters who are in the condition that prevents or discourages
voters from behaving strategically. The occurrence of strategic voting is rare
even some factors may promote strategic voting. In the following section of
this paper, I will classify voters, based on their voting patterns, and show that
the percentage of strategic voting is actually small in the plurality part of

superposition systems.

Classification of Voters

To measure the frequency of strategic voting, it is necessary to
investigate how voters actually behave in the plurality section of superposition
systems. The goal here is to detect voters who cast their ballots
strategically.60 To do so I will classify voters into several groups depending on
their voting pattern. The first group of voters is strategic voters who happen to
prefer a marginal candidate and transfer their votes to a less preferred
competitor who is one of the front runners in their district.

The second group of voters is sincere voters. Sincere voters have a
similar condition as the first group in terms of the opportunity and marginality
factors. Voters have an opportunity to vote for their most preferred candidate,
and their most preferred candidate is a marginal performer. The difference

between the first group and the second group is that voters in the second

 

2° It is also possible to classify voters in terms of candidates' probability of winning and utility
differences. However, since it is difficult to apply the equation for expected utility to the Upper
House, these factors will not be considered in classifying voters in this part of the analysis.

119

group vote for the most preferred candidate instead of transferring their votes
to a less preferred candidate.

The third group is straightfonNard voters. These voters happen to
prefer a front running candidate and vote for their most preferred candidate.
The fourth group is the voters who do not have an opportunity to vote for their
most preferred candidate and, therefore, vote for one of the front running
candidates. These voters should probably be categorized as sincere voters
since they vote for their most preferred candidate among those who are
available in their district. However, I will call them non-opportunity voters to
distinguish from strategic and sincere voters.

The fifth group of voters is irrational voters. Voters who are categorized
into this group are not necessarily irrational despite its label. For example, a
voter’s preference ordering is C1>Cg>C3>C4. The candidates’ popularity is
C4>03>C2>C1. Under the situation, the voter can improve her expected utility
by voting for Cg, although Cg may be considered as a marginal candidate. It is
especially rational to vote for Cg when the race between Cg and C3 is close.
However, the voter will be classified as irrational since she did not transfer her
ballot to one of the front runners. In this way, we do not have to make a large
number of categories to classify voters in a more complicated way and can
focus on the analysis on the occurrence of strategic voting in superposition
systems.

Similarly, when a voter does not have an opportunity to vote for her

most preferred candidate and transfer her vote to a less preferred candidate

120

who is also a marginal candidate, she will be classified as an irrational voter,
even though she did not necessarily behave irrationally. It is possible that the
voter behaved rationally under the condition, since she might have transferred
her ballot to another marginal candidate based on her perception on
candidates’ probability of winning and utility differences.‘51

The final group is the voters who abstained from voting in the plurality
component of the superposition system. There may be a variety of reasons
why voters abstain. For example, a voter may abstain since her most
preferred candidate is not present in her district. It is also possible that a voter
is indifferent toward the candidates. As a result, the voter is not motivated to
turn out. Although there may be many reasons for abstention, voters who
abstain from voting are classified as one single group in this analysis.

To create the outcome variable based on the above six categories, I
will use the opportunity and marginality factors as well as the information on
how voters actually voted. The opportunity and marginality factors are created
based on feeling thermometer questions and election returns. Feeling
thermometer questions are used to find voters’ preferences, while election
returns are used to gain the information on who is placed in plurality districts

or who may be trailing.62 For the information on voting behavior, the questions

 

61There are still other voters who are classified as irrational voters. Some voters prefer a
marginal candidate and may transfer their ballot to a less preferred candidate who is also a
marginal candidate. These voters are not necessarily irrational since it is possible that they
may be able to improve their expected utility. However, as mentioned in the text, I will avoid
creating a large number of categories to keep the analysis simple.

‘2 It is also important to note that the JEDS96 survey provides thermometer questions for six
political parties. Some parties, such as Shinshakaito and Jiyurengo, are not included even
though they place candidates in some plurality districts. As a result, these parties are dropped

121

on which party candidate the survey respondents voted for in the plurality
portion of the elections will be used.

To classify voters into the six categories, I will first identify respondents
who have an opportunity to vote for their most preferred candidate, but the
candidate is only performing marginally. If a voter casts her ballot for a less
preferred candidate who is one of the front running candidates in the district,
she is classified as a strategic voter. If the voter casts her ballot for the
marginal candidate, she is classified as a sincere voter. When the voter casts
her ballot for a less preferred candidate who is not a front running candidate,
she is classified as an irrational voter.

Second, I will identify respondents who do not have an opportunity to
vote for their most preferred candidate. If these voters cast their ballots for
one of the front runners, they are categorized as no opportunity voters. If they
vote for one of the marginal candidates, they will be classified as irrational
voters.

In the next step, I will identify voters who prefer a front running
candidate. If these voters cast their ballot for their most preferred candidate,
they are categorized as straightforward voters. If not, they are categorized as
irrational voters. These voters should be counted as irrational voters since
they do not have to transfer their ballot to another candidate to maximize their

expected utility.

 

in calculating expected utilities. Since these parties' vote shares are minimal, the impact of
dropping the parties should not influence the results substantially.

122

Finally, I will identify voters who did not turn out to vote. Detecting
respondents who did not vote is achieved by using two sets of questions in
the surveys. 63 The surveys ask whether or not respondents went to the poll at
all. If respondents report that they did not go, they are categorizedas voters
who abstained. Another question asks what competitor respondents voted for
in the plurality component of the system. If respondents answered that they
did not vote in the plurality component, they are also categorized as

respondents who abstained.“

Occurrence of Strategic Voting

Table 6-3 shows the frequency of strategic voting and other types of
voting patterns. The result suggests that a dominant strategy among voters
who prefer a marginal candidate is to vote sincerely. The result shows that
only 3% of the electorate engages in strategic voting in the Lower House
election, while only 1% is categorized as strategic voters in the Upper House
election. Sincere voting is 9% in the Lower House election, while it is 6% in

the Upper House election.

 

“3 A potential problem with this method is that voters have a tendency to report that they
voted when they did not actually vote. Moreover, we can expect that respondents who are
indifferent over competitors or who perceive that their votes do not have a chance to
influence an election outcome are more likely to abstain from voting. This may suggest that
an overestimation of turnout may be more likely to be found among respondents who fit a
particular characteristic, and as a result, it may create a systematic error in the analysis.
Indeed, if one compares between how the respondents in the JEDS96 survey reports with the
actual turnout rate, she can estimate the overestimation to be about 13 percentage points.
However, I will carry out the analysis with survey questions since they will be the best
available method to detect who did not vote in an election.

The question that asks what competitor respondents voted for in the PR part of the election
is not used in the analysis since the focus should be on the plurality part of the election.

123

Table 6-3: Occurrence of Strategic Voting

 

 

 

 

 

 

 

 

 

 

 

 

CATEGORY LOWER UPPER
HOUSE HOUSE
Strategy 3% 1%
No opportunity 5% 4%
Sincere 9% 6%
Straightforward 50% 55%
Absent 22% 30%
Irrational 1 1% 4%
Total 100% 100%
N 963 1329

 

124

 

The category for no opportunity voters is 5% in the Lower House and 4% in
the Upper House. This shows that a substantial number of voters cast their
ballots for one of the front runners, when they do not find a candidate from
their most preferred party. The largest category is straightforward voting. Half
(50%) of the electorate belongs to this group in the Lower House, while more
than half (55%) belongs to the category in the Upper House. The values for
abstention in the Lower House and the Upper House are 22% and 30%
respectively. Irrational voting is 11% in the Lower House and 4% in the Upper
House.

It is also important to note that there are some possibilities that the
above estimates of strategic voting in the plurality section of superposition
systems are overestimated, even though an individual level of data are
applied instead of an aggregate level data. One is that the percentages of
strategic voters in Table 6-3 include voters who do not have to vote
strategically to maximize their expected utility in terms of candidates’
probability of winning and their utility differences. We discussed how
candidates’ probability of winning and utility differences need to fit a certain
condition to promote strategic voting. Some voters cast their ballots for a less
preferred candidate and fit the pattern of strategic voting, but they may fall
into the pattern even when they are not voting strategically.

The result also suggests that the percentages of strategic voting in the
plurality component of superposition systems seems to be substantially

smaller than the percentages that are produced in other regular SMD plurality

125

systems. The literature on strategic voting suggests that the percentages of
strategic voting in regular SMD plurality systems are between 5% and 15% of
the electorate (Black 1978, Kim and Fording 2001).‘55 For example, Cain
(1978) suggests that there is about 15% of the electorate voted strategically
in the 1970 British General Election. Blais and Nadeau (1996) report that the
percentage of strategic voting in the 1988 Canadian election in the entire
electorate is at least 6%.

If we only focus on those who turn out in the plurality component of
superposition systems, strategic voting in a superposition system is also
smaller than the values that are reported in other countries with a regular
SMD plurality system. Black (1978) reports that 12% of those who voted in
the 1968 and 1972 Canadian elections voted strategically. This percentage
seems to be substantially larger than the percentages obtained from Japan’s
Lower House election. Strategic voting among those who turn out in the 1996

Lower House election is only about 3%.

Conclusion

The findings reported in this chapter show that strategic voting in the
plurality component of superposition systems is rare since it creates voting
conditions that prevent voters from voting strategically. The contamination
effect in superposition systems is shown to create conditions that reduce the

occurrence of strategic voting in superposition systems compared to other

 

‘5 Kim and Fording (2001) report that the percentage of strategic voting is somewhere
between 5% and 10%.

126

regular plurality systems. Among these conditions, the opportunity factor
plays an important role in reducing the occurrence of strategic voting.
Moreover, even when voters have an opportunity to vote for their most
preferred candidate, the dominant choice of voting pattern is to vote sincerely
rather than strategically.

If we analyze voters’ behavior by using only aggregate level data, non-
opportunity voters may be unknowingly counted as strategic voters. Although
it is rather difficult to calculate exactly how much an overestimation of
strategic voting takes place with aggregate level data, the opportunity factor
seems to be playing a key role in producing a lower number of strategic
voting compared to regular plurality systems.

For future research, strategic voting in mixed corrective systems
should be reanalyzed by taking the contamination effect into account. The
findings in this study suggest the importance of taking the contamination
effect into account when we study superposition systems. The argument can
be generalized to the studies of corrective systems. In fact, the contamination
effect influences voters’ behavior more in corrective systems than
superposition systems, because how seats are distributed is not independent
between the two parts of the election in corrective systems, while they are
allocated independently in superposition systems. The reanalysis of
corrective systems seems to be especially important since previous studies

only used aggregate level data to measure strategic voting.

127

Chapter 7

Impact of Sincere Voting

Chapter 5 empirically showed that the number of candidates in the plurality
section of superposition systems is increased compared to regular plurality
systems. Chapter 6 showed that only a few voters engaged in strategic voting
in superposition systems. This also seemed to be true, even when they were
compared to other regular plurality systems. These results lead us to expect
that vote shares are spread across a number of candidates including
marginally performing candidates when plurality elections are carried out
under superposition systems. Chapters 5 and 6 together then can project that
an increase in the number of candidates will lead to an inflation of the number
of parties and that the number of parties will be greater than Duverger’s Law
and the M+1 Rule suggest.

Although the above effect of superposition systems on party system
seems to exist, the connection is only indirectly shown so far. For this reason,
I will carry out two sets of more direct analysis to make the connection explicit
in this chapter. First, I have been arguing that one of the key factors that can
explain the increase in the number of parties in superposition systems is how
the increase in the number of candidates in plurality districts creates an
opportunity for some voters to cast their ballots for a marginal candidate.
What this suggests is that if the likelihood of voting for marginal candidates is

larger when voters’ most preferred competitor is present in their plurality

128

districts, the number of parties is expected to be inflated. This seems to be
especially true since the increase in the number of candidates also inflates
the number of voters who prefer marginal candidates.

The second part of the chapter brings the discussion to a district level
hypothesis where Duverger’s Law is originally designed to work. An earlier
analysis focused on individual level data to overhaul the mechanism of how
superposition systems inflate the number of parties. In the second part of this
chapter, I will focus on district level hypotheses to see an overall impact of
superposition systems on party systems by utilizing a bimodality analysis.
VIfith a bimodality analysis, I will show that a violation of a Duverger’s
equilibrium is caused by an increase in the number of candidates.

Cox (1997) developed a bimodality test to analyze the impact of
electoral rules on party systems. A bimodality test uses SF ratios to find the
impact by using district level data. An SF ratio is the ratio of the second loser
to the first and shows the relative strength of the competitors. It can show
whether or not a district is in a Duvergerian equilibrium or a non-Duvergerian
equilibrium. I will apply the SF ratio tests in the second part of this chapter
and discuss how the increase in the number of candidates lead to the

creation of a non-Duvergerian equilibrium by using aggregate level data.

Significance of Opportunity Factor

We need to determine whether or not the opportunity factor, i.e.

whether or not a voter’s most preferred candidate is present in her district,

129

plays a significant role in increasing the number of voters who cast their
ballots for marginally performing candidates. If the opportunity factor plays a
significant role for voters to place ballots for marginal candidates, the number
of parties will be inflated in the plurality component of superposition systems.

Rational voters have three choices in an election. The first is to vote for
a marginal candidate. The second is to vote for a front running candidate.
The third is to abstain from voting. Under such circumstances, if voters who
have an opportunity to vote for their most preferred candidate actually cast
their ballots for a marginal candidate significantly more than other voters who
do not have an opportunity, we can expect that the number of parties will be
inflated. On the other hand, if there is no significant difference in the behavior
of voters with and without an opportunity, the number of parties will not be
inflated. That is, if voters cast their ballots for a front runner or simply abstain
from voting, the number of parties will not be inflated.

As mentioned earlier, the Lower House uses a superposition system
with single member district (SMD) and proportional representation (PR) rules.
The Upper House uses single nontransferable vote (SNTV) and PR rules.
The hypotheses are as follows:

Hypothesis 1a: The likelihood of voting for marginally performing

candidates instead of abstaining in the Lower House election increases

when voters’ most prefemed competitors are present in their districts.

Hypothesis 1b: The likelihood of voting for marginally performing

candidates instead of abstaining in the Upper House election increases

when voters’ most preferred competitors are present in their districts.

Hypothesis 2a: The likelihood of voting for marginally performing

130

candidates instead of voting for one of the top two candidates in the

Lower House election increases when voters’ most preferred

competitors are present in their districts.

Hypothesis 2b: The likelihood of voting for marginally performing

candidates instead of voting for one of the top M+1 candidates in the

Upper House election increases when voters’ most preferred

competitors are present in their districts.
Methods and Data

The dependent variable contains three categories: voting for a
marginally performing candidate, voting for a front running candidate, and
abstaining from voting. The dependent variable is created based on the
outcome variable that was created in the previous chapter. If a respondent is
coded as a strategic voter, he is coded as a voter who casts a ballot for a
front running candidate. If a respondent is coded as a non-opportunity voter,
he is also coded as a voter who votes for a front runner. A voter who belongs
to the category of sincere voters, he is categorized as a voter who casts his
ballot for a marginal candidate. If a voter is categorized into the group of
abstention, he is coded as a voter who abstains from voting. For voters who
are categorized as irrational, I will categorize them as voters who vote for a
marginal candidate or voters who vote for a front runner, depending on how
they voted. StraightfonNard voters are excluded from this part of the analysis

since the analysis should focus on voters who prefer a marginal candidate. In

the case of the Lower House, voters who prefer the third ranked competitor or

131

lower are selected. In the Upper House, voters who prefer the competitor who
is ranked lower than M+1 are selected.66

The independent variable of interest is whether or not voters’ most
preferred competitors are present in their districts, so that they have an
opportunity to vote for their most preferred candidate. If a voter has an
opportunity to vote for the most preferred competitor, this particular voter is
coded 1. Otherwise, the voter is coded 0. As was the case in the previous
chapter, voters’ preferences over candidates are determined based on their
party preferences. If there is a candidate who is from their most preferred
party, the voters are coded as having an opportunity.

I will also add control variables. These variables include factors that
may be related to strategic choices of voters. For example, if voters have a
sense of citizen duty to participate in elections, they are less likely to abstain
from voting, even when they are indifferent over multiple candidates. The
questions that are asked in the two surveys ask respondents whether or not
voting is one of a citizen’s duties. There are five answer categories, including

“agree, somewhat agree”, “neither,” “somewhat disagree,” and “disagree.”67
The second control variable is whether or not respondents support a
party. If voters support a party, they are more likely to turn out to vote

compared to other voters who do not support a party. This is because these

 

°° For the category of voting for a front running candidate, the classification will be different
between the Lower House and Upper House. Since the Lower House uses an SMD rule for
the plurality section of the mixed system, those voters who cast their ballots for the top two
most popular candidates are included in this category. On the other hand, the Upper House
uses a medium size district rule, so voters who cast their ballots for the top M+1 candidates
will be classified into this group.

132

voters tend to prefer one candidate to others with more intensity. These
voters are more likely to care about who is elected into an office. Similarly,
respondents who support a party may be less likely to vote strategically. A
stronger intensity may discourage to desert their most preferred candidate
even when the candidate is trailing in their district. Other usual socio-
economic variables are also included in the models.

Multinomial logit models are used to test the hypothesis for both lower
and upper house elections since the vote choice variables (the outcome
variables) are nominal level variables and have three categories.68 The three
categories include voters who voted for a marginally performing competitor,
voters who voted for a competitor with a larger chance of winning, and voters

who abstained.69

 

67 See Appendixes 6-1 and 6-2 for question wordings.
6" There are multiple ways to test the hypotheses. One way is to perform two logit models
with sincere voting and abstention as two outcomes followed by sincere voting and strategic
voting as two outcomes. Even though coefficients and test statistics may vary, this approach
produces substantively the same results as a multinomial logit model. Another way is to
perform multinomial probit models. Since multinomial probit models do not assume the HA
assumption, it is more appropriate than multinomial logit models. However, since there is no
access to a statistical software that can perform multinomial probit models, I will use
multinomial logit models. Moreover, it is often mentioned that the probit and logit models
produce substantively similar results. The probit models are also much more difficult and time
consuming to produce results than the logit models.
2° The model is formally expressed as follows:

Probrv. =1) = EXPrmxo I 2 Bax. (1)

133

Results

The results from both the Lower and Upper Houses are shown in Table
7-1.70 The results that are shown in the table support the two sets of
hypotheses. The upper portion of the table shows the results for Hypothesis
1a and 1b that test whether or not the likelihood of voting for marginally
performing candidates instead of abstaining increases when voters’ most
preferred competitors are present in their districts. Hypothesis 1a is the test
for the Lower House election, while Hypothesis 1b is that for the Upper House
election. The coefficient is —2.006 for the OPPORTUNITY variable in the
Lower House (p<0.01).

The coefficients in multinomial logit models can be converted into an
odds ratio by raising e to the coefficient. For example, e'2'°°6 yields 0.13.
Substantively, this suggests that after controlling for the other variables, the
likelihood of abstention among voters who have an opportunity to vote for
their most preferred competitor is about one tenth compared to voters who do
not have an opportunity. Similarly, the coefficients can be expressed as
changes in predicted probabilities after controlling for other variables.71 Table

7-2 summarizes the results. According to the table, the probability of

 

where Y and x represent the dependent and independent variables, (3 represents coefficients
(Greene 1990) .

° There is no multicollinearity since the values for bivariate correlation between the variables
are smaller than or equal to 0.4 for both houses.
7‘ See Appendix 7-1 for the calculation of predicted probabilities.

134

Table 7-1: Impacts of Presence of Most Preferred Competitor

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

OUTCOME VARIABLES Bs LOWER Bs UPPER
HOUSE HOUSE
Abstention OPPORTUNITY -2.006*** -1 .618***
(0.516) (0.516)
DUTY -0.807*** 0.228
(0.274) (0.156)
PARTY -1 .093“ -1 .254"
SUPPOERT (0.510) (0.605)
SEX 0.354 -0.603
(0.469) (0.552)
AGE -0.024 -0.042*
(0.023) (0.022)
EDUCATION -0.094 -0.911**
(0.184) (0.402)
INCOME 0.028 0.134
(0.129) (0.112)
UNION 0.051 0.074
(0.602) (0.622)
CITYSIZE 0.191 0.105
(0.159) (0.253)
Intercept 5.417 4.179
(1 .903) (2.068)
Front OPPORTUNITY -2.402*** -2.788***
runners (0.482) (0.508)
DUTY -0.268* -0.152
(0.313) (0.144)
PARTY -0.610 -0.110
SUPPORT (0.455) (0.630)
SEX 0.362 -0.013
(0.440) (0.515)
AGE 0.019 0.002
(0.021) (0.020)
EDUCATION -0.180 -0.070
(0.177) (0.368)
INCOME -0.030 -0.018
(0.120) (0.107)
UNION -0.264 -0.636
(0.617) (0.663)
CITYSIZE 0.146 0.511**
(0.149) (0.246)
Intercept 1 .945 0.508
(2.000) (1 .952)

 

135

 

Table 7-1 Continued

 

 

 

|| LOWER HOUSE UPPER HOUSEJ
|| Pseudo R2 .376 .434 I
I N 166 145 l

 

 

 

Note: Dependent variables are composed of three categories, including
individuals who voted for a marginal competitor, voted for a front runner, and
abstained from voting. Voters who voted for competitors who are marginally
performing (i.e., sincere voters) are treated as the base category. Voters who
voted for one of the top two candidates in the SMD or one of the top M+1
ranked competitors in the medium size district (i.e., strategic voters) are
considered as front runners. These front runners are coded 2. Voters whose
most preferred competitors are present are coded 1, otherwise 2. The
explanatory variables do not have a problem of multicollinearity since the
values for bivariate correlations between the variables are smaller than .32 for
Lower House and smaller than .039 for Upper House. F ifty-four percent of the

cases are correctly predicted for Lower House, while 54% is correctly
predicted for Upper House.

Source: JEDS96, JESB3

136

Table 7-2: Predicted Probabilities- 1996 Lower House Election and 1983
Upper House Election

OWER HOUSE PPER HOUSE

inal inal
.08 .18
.46 .68

.32 .26
.24 .19

ront runners ront runners
.60 .56
.30 .13

 

137

abstention decreases from 0.32 to 0.24, when voters’ most preferred
candidate is present in their district.

The result for the Upper House election yields similar results. The
coefficient for the OPPORTUNITY variable is —1 .618 and is statistically
significant (p<0.01). This means that the likelihood of abstention is one fifth
when voters’ most preferred candidate is available compared to voters whose
most preferred candidate is not available (6'1'5‘°=0.20). Substantively, this
means that the probability of voting for a front runner decreases from 0.26 to
0.19 (Table 7-2).

The lower portion of the table shows the results for testing Hypotheses
28 and 2b. Hypotheses 2a and 2b test whether or not the presence of most
preferred competitor increases the likelihood of voting for marginally
performing candidates, instead of voting for a front running candidate. The
coefficients for the OPPORTUNITY variable are -2.402 (p < 0.01) for the
Lower House and —2.788 (p < 0.01) for the Upper House. The coefficient of
-2.402 suggests that the likelihood of voting for a front runner is about one
tenth when voters’ most preferred competitor is present compared to voters’
most preferred candidate is not present (e‘2'4°2=0.09). Similarly, the
coeffecient of —2.788 means that the likelihood of voting for a front runner is
less than one tenth (e’2'788=0.06). According to Table 7-2, the probability of
voting for a front runner decreases from 0.6 to 0.3 in the Lower House and

from 0.56 to 0.13 in the Upper House.

138

The result for hypotheses 1 and 2 suggest that the presence of voters’
most preferred competitors encourage them to vote for marginally performing
competitors, instead of abstaining from voting or voting strategically. Table 7-
2 supports these tendencies clearly. Moreover, these tendencies are shown
to be statistically significant (Table 7-1). These suggest that the plurality
section of a mixed system has tendencies to inflate the number of parties,
and as a result, Duverger’s Law or the M+1 Rule is violated.

The effect of other variables is not consistent across the two houses
except the PARTY SUPPORT variable. The coefficients for the party support
variable for abstention are —1.093 for the Lower House and —1.254 for the
Upper House and are statistically significant (p<0.1 and p<0.05 respectively).
These coefficients mean that voters are less likely to abstain when they
support a political party. As mentioned, this may be related to voters’ intensity
in preferences. Voters who support a party are more likely to prefer a party
more intensely than other voters. As a result, these voters are more likely to

turn out.

SF Ratio Test

Cox ( 1997) applies bimodality tests to assess the existence of strategic
voting. The bimodality test is done by calculating an SF ratio, the ratio of the
second loser to the first, for each district. When the contest between the first
loser and the second is close, the ratio approaches 1. On the other hand,

when the contest is not close, the ratio becomes close to 0. Under a

139

Figure 7-1a: Bimodal Distribution of SF Ratios

12

 

 

Count

 

.05 .15

SF Ratio

Figure 7-1 b: Non-bimodal Distribution of SF Ratios

 

 

 

Count

 

140

Duvergerian equilibrium, the SF ratio will be near 0, while under a non-
Duvergerian equilibrium, the ratio should approach 1. This is because
Duverger’s Law and the M+1 Rule suggest that in plurality districts, the
second loser or other candidates who are ranked lower than the second loser
have difficulty in obtaining votes. Under such circumstances, the ratio of the
second loser to the first loser will approach 0. However, when Duverger’s Law
or the M+1 Rule is not at work, the vote shares between the first loser and the
second will be similar. The SF ratio then should approach 1. If we classify
districts based on SF ratios, we will see a large number of districts with the SF
ratio of 0 or near 0, when a Duvergerian equilibrium is maintained among
these districts. By using the approach above, Cox assesses the existence of
strategic voting. Cox suggests that strategic voting is evident when the
distribution of SF ratios is bimodal. As an illustration, Figure 7-1a and 7-1b
show two distributions of districts. Figure 7-1a has a bimodal distribution,
while 7-1 b does not. In Figure 7-1a, there are a large number of districts with
SF ratios of 0 or near 0, suggesting that many districts are under a
Duvergerian equilibrium. Figure 7-1 b has a relatively large number of districts
with an SF ratio of 1 or near 1. This suggests that more districts tend to be
under a non-Duvergerian equilibrium.

The analysis with SF ratio tests is informative but the assessment of
strategic voting in this method can be problematic even though Cox applies
them to assess strategic voting. This is because the use of aggregate level

data to analyze an individual level phenomenon may cause an ecological

141

fallacy. As it was suggested in an earlier chapter, it is important to take the
opportunity factor into account. The application of a bimodality hypothesis to
assess strategic voting ignores the existence of different voting environment.
That is, there are some voters who find a candidate who is from their most
preferred candidate in their plurality districts, while there are others who do
not find their most preferred candidates. Under such circumstances, the latter
group of voters will transfer their votes to a candidate who is available. If they
transfer their votes to a front runner, it will make the SF ratio closer to 0, while
the true cause of the creation of the above pattern of voting behavior may be
how political parties place their candidates.

Even though the use of a bimodality hypothesis to assess strategic
voting could be problematic, SF ratios are a useful tool to assess the impact
of superposition systems on party systems as a whole. We can analyze how
the increase in the number of candidates impacts the creation of a
Duvergerian equilibrium because a Duvergerian equilibrium is a district level
phenomena.

If we classify districts according to the number of candidates in plurality
districts and calculate SF ratios, we can examine how the increase in the
number of candidates may Influence a Duvergerian equilibrium. For example,
we should see a relatively large number of districts with the SF ratios of 0 or
near 0 when the number of candidates is small. On the other hand, when the
number of candidates in districts is large, we should see more districts with a

ratio of 1 or near 1. The above pattern should be true because the increase in

142

the number of candidates tend to create more voters who prefer a marginal
candidate. Moreover, these voters’ likelihood of voting for marginal
candidates increases when their most preferred candidates are placed in their
districts. Since political parties create an incentive to place candidates in the
plurality component of superposition systems, the number of candidates is
larger. This in turn leads to a violation of Duverger’s Law.

We can apply the same method to the plurality component of the
Japanese Upper House elections. For the Japanese Upper House,
multimember districts are used instead of single member districts. To make
Japan’s Upper House elections comparable with the Lower House elections, I
will classify districts based on how many excess candidates are present. For
example, if there are five candidates, where the district magnitude is two, the
excess number is two given 5-(2+1), based on the M+1 Rule. When there are
few excess candidates, districts have a tendency to be in a Duvergerian
equilibrium. This leads to create a large number of districts with SF ratios of 0
or near 0. When there are more excess candidates in a plurality district, we

should see a larger number of districts with an SF ratio of 1 or near 1.72

Results
The results of the SF ratio tests support the idea that the increase in
the number of candidates will lead to the creation of a non-Duvergerian

equilibrium. Tables 7-28 through 7-2d show the results for Japan’s Lower

143

Figure 7-2a: SF Ratios in the 1996 Lower House Election - 2-3
Candidates

 

Count

 

  

 

0.05 0.15 0.25 0.35 0.45 0.55 0.65 0.75 0.85 095
SF Ratio

Figure 7-2b: SF Ratios in the 1996 Lower House Election - 4 Candidates

30

 

 

Count

 

 

0.15 0.25 0.35 0.45 0.55

0.65 0.75 0.85 0.95
SF Ratio

72 Cox ( 1997) uses district magnitude to classify districts. However, since my focus here is to

find the impact of the increased number of candidates in plurality districts of a superposition
system, | classify based on the number of candidates.

144

Figure 7-2c: SF Ratios in the 1996 Lower House Election - 5 Candidates

 

 

Count

 

 

SF Ratio

Figure 7-2d: SF Ratios in the 1996 Lower House Election - 6 Candidates

 

 

 

Count

 

SF Ratio

145

Figure 7-3a: SF Ratios in the Upper House Election (1983-1992) - 0 to 1
Excess Candidates

 

 

 

  

Count
O

0.05 0.l5 0.25 0.35 0.45 0.55 0.65 0.75 0.85 0.95

SF Ratio

Figure 7-3b: SF Ratios in the Upper House Election (1983-1992) - 2
Excess Candidates

 

-__l

Count

 

146

Figure 7-3c: SF Ratios in the Upper House Election (1983-1992) - 3 or
More Excess Candidates

 

 

 

Count

 

0.05 0.25 0.35 0.55

0.65 0.75 0.85 0.95
SF Ratio

147

House election. When there are two to three candidates in a district, we see
more districts with an SF ratio of 0 or near 0, suggesting that these districts
are in a Duvergerian equilibrium (Table 7-3a). However, the distribution of SF
ratios shifts as the number of candidates increases. When there are six
candidates in districts, a large number of districts produce an SF ratio that is
close to 1.

Similarly, Figure 7-3a through 7-3c show the results for the Upper
House elections between 1983 and 1992. The figures also support that the
increase in the number of candidates lead to a violation of the M+1 Rule.
When there is only one excess candidate, the SF ratios are close to 0 among
a large number of districts. However, the tendency to produce the SF ratios of
0 changes, as the excess number of candidates increases. This is similar to
what happened in the Lower House election. When there are three or more
excess candidates, districts are more likely to produce the SF ratios that are

closer to 1.

Conclusion

The plurality component of superposition systems is shown to produce
a larger number of parties than Duverger’s Law and the M+1 Rule suggest.
This is true because parties have an incentive to place their candidates in the
plurality component of a superposition system since they can get an extra
share of vote in a corresponding PR district. As a result, a large number of

candidates is placed. This leads to an increased opportunity for voting for

148

voters’ most preferred competitor. This is especially true for voters whose
most preferred competitor happens to be performing marginally.

The first part of this chapter showed that the existence of the
relationship between the opportunity factor and the vote choice variable. It
suggests that the increased number of candidates in the plurality section of a
superposition system is important in explaining an inflated number of parties.
It also shows that voters play a major role in inflating the number of parties.
However, voters do so only under the condition that parties place a larger
number of candidates in districts. The second part of this chapter also
supports that an increase in the number of candidates leads to a violation of a
Duvergerian equilibrium. This leads us to conclude that the plurality
component of superposition systems tends to inflate the number of parties
than regular plurality systems do. If an SMD plurality rule is combined with a
PR rule, the SMD part is more likely to produce a number of parties that is
larger than a standard SMD plurality rule would produce. If a multimember
plurality system is used with a PR rule, the number of parties will be larger
than the number of parties that may be produced by a regular multimember

plurality rule.

Implication
Perhaps one of the most important implications of this dissertation is
that it will help answer the question of whether or not voters play a significant

role in shaping party systems. A number of researchers have argued that

149

political elites, such as politicians, are the ones who play the principle role in
shaping party systems (Meehl 1977; Riker 1982; Cox 1997). This is called
“the elite-level hypothesis,” and its proponents begin with the notion that it is
political elites who are experts in politics. This means that it is they who know
the electoral potential of candidates who are in district elections, and, as a
result, it is they who decide whether or not to enter a district contest.
According to the literature, party elites make such choices based on their
evaluations of a candidate’s potential for victory. The point here is that
decisions on entry by political elites shape party systems, which then
relegates any role voters may play in shaping a party system to obscurity.
What this point of view misses is that parties’ decisions on candidate
entry must consider how many votes a potential candidate is likely to receive.
Specifically, elites must consider whether or not electors will avoid wasting
their votes and cast their ballots strategically. The view that voters behave in
this way is called “the voter based hypothesis” (Cox 1997), and its proponents
argue that electors cast their ballots for candidates who are more likely to win
a seat by in fact deserting their most preferred candidate. There is evidence
that voters behave in this way (Barnes et al. 1962; Fisher 1973; Jesse 1988;
Bawn 1993; Johnston and Pattie 1991; Heath et al. 1991; Niemi, Whitten and
Franklin 1992; Crewe 1987; Catt 1989; Butler and Kavanagh 1988). What I
showed is that voters rarely engage in strategic voting. This suggests that

voters do not play a role in reducing the number of parties, since voters are

150

more likely to vote sincerely, although they contribute to inflate the number of
parties.

It is also important to note that the support for contamination effects
will imply that researchers need to be more cautious when they study mixed
electoral systems. More specifically, many researchers carry out their studies
by assuming that each component of the system works independently of the
other (Jesse 1988; Bawn 1993; Cox 1997; Moser 1997, Reed 2000). That is,
as if the plurality component of a mixed system works independently of the
PR section of the system. Jesse (1988), Bawn (1993), and Cox (1997) use
the German mixed system in their studies, but as stated above, the German
mixed system is different from the Japanese style mixed systems in that the
PR section of the system is used to compensate for the disproportionality that
is caused by the plurality rule. However, the argument on the independence
assumption is also relevant for the German mixed system, since
contamination effects are also expected to be at work in the German mixed
system. It may be fine to assume that the two components of a mixed system
work independently since researchers often need to start building a theory
with assumptions that make their theory simple. However, it is also important
to be aware that the two components of a mixed system do not work
independently and thus may influence conclusions.

What the above also leads to is that mixed electoral systems will create
a large number of wasted votes in the plurality component of the system, and

thus produces a more disproportionate result, compared to regular plurality

151

systems. This happens especially because the mixed system will promote
political parties to place candidates. The increased number of candidates in
turn attracts voters to spread their votes across candidates. However, the
seat distribution of the plurality section does not reflect parties’ vote shares.

As a result, the system produces more disproportionate results.

152

Part IV.

Conclusion

153

Chapter 8

Conclusion: Party Systems in Democracies with Mixed Superposition
Systems

In this dissertation, mixed superposition systems are shown to inflate the
number of parties. This happens because political parties place their
candidates even when these candidates have little chance of winning, and
voters cast their ballots for these candidates. As a result of how these political
actors behave, the plurality component creates a number of parties that is
greater than Duvergerian equilibria suggest.

What policy implications do these findings have on Japan’s party
politics? Are there any implications in terms of how superposition systems
affect democratic performance? Are the effects generalizable to other
countries? In this concluding chapter, I assess these questions and consider

the implications of their findings.

Impact of Superposition Systems on Democratic Performance

Although superposition systems influence democratic performance in
many ways, three aspects stand out more important than others. First,
superposition systems influence politics by dispersing political power among
multiple political parties. Second, these systems promote political durability in
which a party would maintain a ruling party status for a long period of time.
Third, superposition systems seem to promote party-based politics than

candidate-based politics. This can lead the Japanese politics to diminish the

154

occurrence of corruption. In this section of this chapter, I discuss how
superposition systems influence democratic performance with regard to these
three aspects in more detail.

The first impact of superposition systems on democratic performance
is how they disperse political power among multiple political parties. With the
use of a superposition system, political power would be shared by multiple
parties since it is more likely to create a multiparty system, instead of a two
party system.73 The tendency that superposition systems create a multiparty
system has a profound implication on how a democracy with a superposition
system may function. In particular, Lijphart (1999) states that democracies
can be classified into two types: a majoritarian and a consensus type. In the
majoritarian model, political power tends to concentrate on the bare majority
whereas it is dispersed among multiple parties in the consensus type.

According to Lijphart, there are several factors that make democracies
more majoritarian or consensus. However, one of the most important factors
that classify different democracies into the two types is how party systems are
set up." More specifically, having a two party system is one of the conditions
that pushes a democracy to become a more majoritarian type because the
two party system tends to concentrate political power into a single party. This
is because the majority of the seats in the legislature tends to be held by a

single party, which, in turn, creates a strong cabinet leadership. On the other

 

73 It may be possible to produce a two party system if there is only one issue dimension.

2‘ There are other factors that influence what type of democracy may be created. These
include factors such as whether or not constitutions are written and whether or not a federal
or unitary government is exercised

155

hand, with a multiparty system, the cabinet leadership tends to be weak since
a coalition of multiple parties is often necessary to keep a majority status in
the legislature. As a result, a compromise between parties is likely to take
place when political parties attempt to pass a bill in the Diet. The necessity of
a compromise between parties in turn diffuses political power to the hands of
multiple parties.

What the above means in terms of the impact on a democracy is that it
may have an ability to push a democracy to become a consensus type rather
than a majoritarian type. This dissertation shows that a superposition system
with single member district (SMD) plurality rule and proportional
representation (PR) rules creates a multiparty system by using the Japanese
Lower House elections. Similarly, by using the Japanese Upper House
elections, a superposition system with a single nontransferable vote (SNTV)
rule and a PR rule was shown to create a multiparty system. These results
support that superposition systems promote a consensus type democracy by
dispersing political power among multiple parties. Under such circumstances,
a cabinet and the Diet is more likely to be controlled by a coalition of parties,
which will also force parties to share political power.

If we observe the election results in the Japanese case, it is evident
that Japanese politics reveals to function like a consensus democracy. More
specifically, the politics in Japan has been run by the coalitions of parties
since it started to use a superposition system. After the 1996 Lower House

election, the Liberal Democratic Party (LDP), the Social Democratic Party

156

 

 

(SDP), and the New Party Harbinger (NPH) formed a coalition. Although the
NPH only had two seats, it played a pivotal role in the Diet. For example,
Masayoshi Takemura, the head of the NPH, was the minister of the Finance,
although his party was a marginal party.

Currently, the coalition among the LDP, the Clean Government Party
(CGP), and the Conservative Party exists, and as a result, power sharing is
continuously present in Japanese politics today. For example, the CGP has
been pushing to increase the district magnitude in the plurality component of
the Lower House elections. On the other hand, the LDP has been trying to
decrease the number of seats that are allocated by the plurality component of
the system. The LDP successfully decreased the number of seats that are
distributed through the PR component after the 1996 election. It was
decreased from 200 to 180. Currently, the two parties have been
compromising and discussing about increasing the district magnitude in city
areas.

The second influence by superposition systems on democratic
performance has to do with the political durability. How long the government
lasts is one of the important factors that indicates how healthily a democracy
is functioning. If a democracy has a series of short-lived governments, it may
not be functioning effectively. This is because the implementation of policies
takes a certain period of time. If the government does not last long enough,

policy implementation will not be carried out effectively. More importantly,

157

political instability may create an opportunity for the rise of a group that
supports an extreme political view.75

Although the production of a stable political environment is necessary
for an effective democracy, the government that lasts too long is not
considered a healthy democracy, either. For example, postwar Japanese
politics is characterized as a predominant party system where the LDP took a
control of the government for about fifty years. Because of the predominant
party system, Johnson (1982) calls Japan a “soft authoritarian” state.
Similarly, Pempel (1990) raises the question of the quality of democracies in
other countries where single party dominance is seen. These countries
include Sweden and Israel. The above suggests that a democracy needs to
be stable but it also requires alternation of the party in power.

The use of a superposition system has an implication on political
durability because they influence the configuration of party systems. Two
party systems are said to alternate the ruling party between two major parties.
Fragmented party systems may risk political instability. A one party
predominant system may keep the same ruling party over some decades. I
argue that superposition systems may promote political durability since they
discourage a fragmentation of party systems. However, this positive effect on
political stability may promote a political stagnation by keeping the same

political party in power for a long period of time.

 

75 An example of this is the case under the Nazis Germany.

158

This happens because superposition systems use both a plurality rule
and a PR rule. The plurality section discourages political parties from splitting
since the plurality section of the system works unfriendly to marginal parties.
That is, the presence of a plurality rule discourages candidates from defecting
their parties. On the other hand, the PR rule encourages candidates to join a
party. The reason for these empirical patterns is rather simple. Candidates
cannot run in the PR section of the election if they do not belong to a party.
As a result, candidates try to join a party to maximize their chances of winning.
The above suggests that superposition systems may be effective in creating a
stable political environment.

However, the promotion of such a stable environment may mean that
superposition systems create a political stagnation by keeping the same
political party as a governing party for many years. At first glance, the
presence of a plurality rule in a superposition system seems to make the
alternation of ruling parties frequent. However, since superposition systems
have the contamination effects, the total amount of voters who swing from a
ruling party to another in an election may not be substantial.

More specifically, if voters have a tendency to cast their ballot
strategically, it is possible to see a substantive number of swing voters who
change their voting patterns. This is the case because strategic voters are the
group of people whose most preferred candidate is a marginal performer.
Since their most preferred candidate does not have a chance to win, voters

transfer their ballot to a candidate who is more likely to win. This means that

159

strategic voters should be considered as a group of voters who are not
committed to the candidate whom they vote for. In other words, strategic
voters can swing from one party to another in an election and can give an
impact on the election outcome. In a superposition system, since a large
number of voters cast their ballots sincerely In the plurality section of
superposition systems, such a substantive amount of swing is less likely to
occur

The above logic on the frequency of the alternation of ruling parties
leads us to believe that Japan’s superposition system may not be so helpful
to improve Japan’s democratic performance, other things being equal. The
use of an SMD plurality rule in the Lower House does not necessarily
increase the occurrence of the alternation of power compared to the old
SNTV rule that was used before the adoption of the superposition system.
This leads us to conclude that the new rule may not be so helpful to promote
the creation of a responsive party system that is attentive to the public.

Similarly, the alternation of power in the Upper House would not be
frequent. Indeed, the number of occurrences would even be less than that of
the Lower House. Instead of using an SMD plurality rule, the Upper House
uses an SNTV rule, which is the same as the old Lower House system. The
frequency of the alternation is less likely under the conditions, since the
amount of strategic voting is smaller than that of the SMD plurality rule.

The third is how superposition systems promote party based politics

and diminish the occurrence of corruption. One of the goals of the electoral

160

reform in implementing a superposition system in Japan’s Lower House was
to regain political trust among the public by eliminating corruptions. It seems
that the adoption of a superposition system in Japan has helped to improve
the corruption problem.

More specifically, since the adoption of a superposition system in the
Lower House, corruption has been more controlled, although it has not
eliminated corruption.76 We saw series of corruption one after another
throughout the post war Japanese politics." It is said that the corruption
problem was so serious that few other industrial countries were worse than
that in Japan (Reed 1996). Although corruption among bureaucrats and some
politicians is still seen after the implementation of a superposition system,
they seem to be the ones who still practice distributive politics that were
developed under the old SNTV system.78 This leads us to believe that the
adoption of the mixed system might have helped to improve the situation,
although one can argue that the diminished level of corruption is coincidental
with the change in other aspects of electoral laws and that the superposition
system did not play a part.

Nonetheless, the fact that we see an improvement after the
implementation of a superposition system in the Lower House may suggest
that the electoral reform probably helped to diminish the occurrence of

corruption. This argument is especially plausible if we reconsider how a

 

76 This does not mean that corruption has been eliminated. Indeed, we may still see
corruption since the change in institution cannot improve on the problem over night.
77 See Fukui and Fukai (1996) and Inoguchi (1997).

7° For example, Helweg (2000) states that corruption is dying.

161

superposition system influences the behaviors of political actors. This has to
do with the fact that superposition systems discourage intra-party
competitions. Before the adoption of a superposition system in the Japanese
Lower House, Japan used a multimember district system with an SNTV rule,
in which major parties had to place more than one candidate in a district to
win a majority of seats in the Lower House. This encouraged the LDP to
create factions and candidates to develop a tie with local businesses,
including construction companies. The promotion of pork barrel politics then
was inevitable, while broad national interests were downplayed.

The decreased level of intra-party competitions under the
superposition system is evident empirically if we observe how political parties
placed their candidates. In particular, in the plurality component of the system,
political parties placed only one candidate, and, as a result, candidates can
promote themselves in elections by emphasizing on policy differences based
on party platforms. In the plurality component, voters would choose their most
preferred party, and as a result, inter-party competitions were promoted
instead of intra-party competitions. It is also said that each faction in the LDP
has been raising less money since the mixed system was adopted in 1994
and has less political power. We can find some evidence that supports the
above. The Yomiuri Newspaper suggests that the factions that existed under
the old SNTV system has weakened, while political parties such as the LDP is

strengthening its party leadership (January 6, 2002).

162

It is worthwhile to point out that the superposition system in Japan’s
Upper House also promotes inter-party competitions, compared to the
standard SNTV rule, even though it partly uses the SNTV rule. Despite the
use of the SNTV rule, intra-party competitions are not encouraged in the
Upper House because the district magnitude is small in most of the districts.
As Chapter 3 reveals, about forty-two districts out of forty-seven districts have
a district magnitude of either one or two. Under such circumstances, it is less
frequent to see candidates from the same party compete against each other
in the same district, although this does happen sometimes. The point is that
the frequency is much less compared to that in the old SNTV rule in the
Lower House elections.

Recently, the politicians in Japan have been discussing about
increasing the district magnitude in the plurality component of the system in
the Lower House elections. While this change may concern some observers
since the change may promote intra-party competitions, the impact on
corruption will not be substantial. This is especially because the district
magnitude will be increased from one to two only in certain city areas. It is
designed in this way since one of the coalition partners of the LDP, the CGP,
can win some seats in the plurality component of the election since their
supporters mainly live in the city areas. Moreover, although the district
magnitude may increase in the city areas, it is unlikely that the LDP will place
more than one candidate in these areas since the performance of the LDP

candidates are not strong. If the LDP places more than one candidate in

163

 

these areas, it is possible that both candidates may lose, because of over-
placement of candidates. Additionally, placing one candidate made the party
more likely to secure one seat.79 As a result, corruptions are less likely to
arise. As long as two candidates from the same party do not enter into a
single district, politicians do not have an urge to create a strong tie with local
businesses.

However, it is important to note that the above does not necessarily
mean that superposition systems can push the major party to pass bills that
may reflect general national interests. Politicians may place importance on
national interests instead of local interests under the new system, however,
these emphasis do not necessarily lead to the effective production of national
public policies. In particular, superposition systems create a multiparty system.
Under these conditions, a coalition government is likely to be formed. What
this means is that compromises are necessary among the political parties in
the coalition to promote their nationally based issues. As a result, politics may

not be changed dramatically.

Generalizability of the Japanese Findings
To consider the generalizability of the findings in the Japanese case to
other countries, two sets of issues stand out more importantly than others.

The first set of issues is how political or social environment may interfere with

 

79 Brown and Patterson (1999) suggest that the Japanese political parties engage in rational
behavior, in which they place candidates strategically so that they can maximize the number
of seats. This suggests that the LDP is less likely to place more than one candidate in the city
areas.

164

the tendency of superposition systems to create a multiparty system. The
second is how the arrangement of electoral institutions, such as the
percentage of seats distributed by a PR rule, may interfere.

Although both sets of issues look potential in influencing the effects of
superposition systems, I argue that only the second set is influential in
inhibiting a superposition system to promote a multiparty system. Regarding
the first set of issues, a large number of countries use a mixed superposition
electoral system today. These countries are geographically spread around the
world. For example, there are some countries that are located in Asia, such
as South Korea and Taiwan. Many countries in the former Soviet Union use a
superposition system. These countries include Armenia, Azerbaijan, Georgia,
Russia, and Ukraine. We can also find some countries with a superposition
system in Africa. Guinea, Senegal and Seychelles are the examples.80
Moreover, there are other countries outside of these areas that use a
superposition system. These countries include Croatia and Mexico.

Since these countries geographically spread around the world and are
greatly different in their social and political background, we may have an
impression that the differences in background may change the outcome that a
superposition system creates a multiple number of parties. However, I
suggest that the difference in social or political background may not be so

influential in reducing the number of parties.

 

5° Most of these countries adopted a superposition system in the 19905.

165

More specifically, Duverger’s propositions state that a two party system
is produced if an SMD plurality rule is applied while a multiparty system is
created if a PR rule is used. However, this tendency does not hold if other
factors interfere. One of the most powerful factors that interferes is social
cleavages (T aagepera and Grofrnan 1985, Ordeshook and Shvetsova 1994).
If multiple social cleavages exist, the number of parties may not be reduced to
two even when an SMD plurality rule is used. This seems to be especially true
if multiple cleavages are salient in society.

However, since the argument in this dissertation is that superposition
systems produce a multiparty system, the above argument is not necessarily
applicable. That is, the presence of multiple cleavages may not interfere with
the impact of superposition systems as much as they interrupt the impact of
SMD plurality rules in reducing the number of parties. When the number of
salient issues in society is large, multiple parties will be created under the
SMD plurality rule. Similarly, when a superposition system is applied, the
number of parties will be inflated regardless of the number of salient issues.
This happens because superposition systems inflate the number of parties
even without the help of the presence of multiple social cleavages.81

Another factor that may interfere in terms of social or political
background is whether or not a country has a well-established democracy.
Some countries with a superposition system have a shorter democratic

history than Japan and have not come to practice full-fledged democracy.

 

8' Canada and India may be representative cases that have produced more than two parties,
even though their elections are carried out under SMD plurality rules.

166

More specifically, Japan established democracy after the Second World War
and has democratic history that has lasted more than 50 years. Many of the
countries with a superposition system other than Japan are the former Soviet
Union Countries. These countries include Armenia, Azerbaijan, Georgia,
Russia, and Ukraine. Their democracies just began in 1991 and are still not
full-fledged democracy. Even though their political systems made a transition
from being a communist state to a democratic state, their political system is
not as free as that of Japan. Seychelles became independent in 1993, and
the country is considered only partly free.82

How these variations in the level of democracy can influence the
impact of superposition systems has to do with the level of information that
the electorate possesses. Although such differential in the level of information
"may influence the effectiveness of Duverger's equilibria, it is not a concern
when we consider the impact of superposition systems to create a multiparty
system.

Duverger’s Law works because the electorate has the information on
who is trailing in elections. That is, if voters have the information on who is
more likely to win or lose, they may engage in strategic voting. As a result, the
number of parties may be reduced. If voters do not have the information, they
have a difficulty in engaging in strategic voting. The connection between
Duverger’s Law and democracies is that stable democracies are a

prerequisite to give the electorate information on who is more likely to win.

 

‘2 According to the Freedom House, Seychelles is classified as a part free country.

167

There are two ways to acquire information on who is trailing in a district.
One way is to look at the history of how a party performed over time. If a
candidate is from one of the major parties, it is easy for voters to project that
the particular candidate has a good chance of winning. Projecting who will
perform well may be more difficult when a country has not repeated elections
in a number of times. This suggests that Duverger’s Law requires a stable
democracy that carries out multiple elections over time.

The other way is to look into public opinion survey results. If public
opinion surveys show that certain political parties are trailing in terms of their
party support level, voters may have a good understanding of who may win in
their districts. Public opinion surveys can be successfully conducted in a
country which has achieved democracy. As a result, voters in a non-
democratic country may have a more difficult time projecting who is trailing in
an election. If a society does not have a well-grounded democracy, it is very
difficult for voters to project who may win. What this means is that the number
of parties may be inflated more in countries with a low level of democracy.

Although it may sound that the influence of superposition systems may
be affected depending on whether or not the electorate is informed, the
difference in the level of information to influence the number of parties under
a superposition system is weak. This is because the presence of information
does not play a role in reducing the number of parties in superposition
systems. As earlier chapters show, the dominant choice for voters is to cast

their ballots sincerely in superposition systems. What this amounts to is that

168

voters would vote sincerely regardless of whether or not they know who are
marginal candidates.

The second set of factors that may influence party systems differently
from the Japanese case is how countries arrange their electoral rules besides
selecting a category of systems, such as a PR, an SMD or a mixed system.
Countries with a superposition system can arrange their electoral rules in their
own ways, and none turn out to be exactly the same as those of Japan. As a
result, these differences may be crucial in inhibiting to produce the same
results as the Japanese case, even though their systems as a whole are
categorized as superposition systems as the Japanese systems.

There are several factors that can increase or decrease the number of
parties. The first is to have a different proportion of PR or plurality seats.
Japan uses a PR rule to allocate about 40% of the seats, whereas South
Korea uses a PR rule to distribute 15% of the seats. This suggests that the
impact of the plurality component on the party system outcome is larger in
Japan than South Korea. As a result, a lesser amount of contamination effect
may be at work in South Korea than in Japan. On the other hand, there are
countries that allocate a larger portion of seats through a PR rule than Japan
does. In Guinea, 67% of the seats are allocated through a PR rule. This
suggests that the impact of the PR section may be larger in Guinea than in
Japan.

It is clear that when the number of seats in the plurality section of the

system is O, the system is reduced to a standard plurality system, while the

169

system is reduced to a regular PR rule, if the number of seats in the SMD part
is 0. It is difficult to project exactly how many seats in the PR component are
necessary to create a contamination effect. However, it may be safe to argue
that the SMD part of the superposition system dominates in South Korea,
while the PR component dominates in Guinea's case.

Second, if the district magnitude is increased in the plurality section of
the election, the number of parties will be larger than that of the Japanese
Lower House elections. As the findings from Japan show that the number of
parties in the Lower House is slightly smaller than the Upper House elections.
The effective number of legislative parties in the Lower House is 2.9, while
that in the Upper House is 3.1 (See Table 2-4). The difference seems to be
created by the fact that the district magnitude in the Lower House elections is
one, whereas it is equal to or more than one in the Upper House elections.83

The third factor has to do with double candidacy. Japan’s Lower
House uses double candidacy where candidates can run in both parts of the
system. On the other hand, Japan’s Upper House does not use double
candidacy. If double candidacy is not used, the number of parties will be
smaller, compared to the system without double candidacy. This is because
double candidacy cuts down the cost of placing candidates.

The findings in this dissertation are based on the idea that parties
working in a superposition system create an incentive to place candidates

even when candidates have little chance of winning a seat. Although the

 

83 Similarly, if a majority runoff system is used instead of an SMD plurality rule, the number of
parties will be larger than that of Japan’s Lower House case.

170

 

incentive to place a candidate exists, this does not necessarily mean that
each party can place a candidate to all of the plurality districts, because of the
cost involved. Political parties need to find a large number of qualified
candidates. This is not an easy task, especially for a small party. Moreover,
running a campaign for a candidate is always costly. Under such
circumstances, double candidacies help small parties to place candidates.
They can save time to look for qualified candidates and save money to run a
campaign. What these mean is that without a double candidacy, the number
of parties will not be inflated as much as Japan’s Lower House elections.
Overall, a superposition system, as a choice of electoral rule, can
influence party politics substantially. Although variation in the subcategories
of mixed electoral rules may have an influence on electoral outcomes, the
idea that superposition systems violate Duvergerian equilibria seems to be
generalizable. Mixed electoral systems also determine what type of

democracy a country practices and how long the government may last.

171

Bibliography

Abramson, Paul R., John H. Aldrich, Phil Paolino and David Rohde. 1992.
“'Sophisticated’ Voting in the 1988 Presidential Primaries.” American
Political Science Review. 86:55-69.

Abramson, Paul R., John H. Aldrich, Phil Paolino, and David W. Rohde. 1995.
“Third-Party and Independent Candidates in American Politics: Wallace,
Anderson, and Perot.” Political Science Quarterly. 110: 349-367.

Amorim Neto, Octavio and Gary W. Cox. 1997. “Electoral Institutions,
Cleavage Structures and the Number of Parties.” American Joumal of
Political Science. 41 2149-174.

Asahi Shimbunsha. 1951. Asahi Nenkan. Osaka-shi: Asahi Shimbunsha.
Asahi Shimbunsha. 1982. Asahi Nenkan. Osaka-shi: Asahi Shimbunsha.
Asahi Shimbunsha. 1983. Asahi Nenkan. Osaka-shi : Asahi Shimbunsha.

Asahi Shimbun Senkyo Hombu. 1997 Asahi Senkyo Taikan. Tokyo: Asahi
Shimbunsha.

Barnes, Samuel, Frank Grace, James Pollak and Peter Sperlich. 1962. “The
German Party System and the 1961 Federal Election” American Political
Science Review. LVI: 909-10.

Bawn, Kathleen. 1993. “The Logic of Institutional Preferences: German
Electoral Law as a Social Choice Outcome.” American Journal of Political
Science. 37:965-989.

Bawn, Kathleen. 1999. “Voters Responses to Electoral Complexity: Ticket
Splitting, Rational Voters and Representation in the Federal Republic of
Germany.” British Joumal of Political Science. 29: 487-505.

Black, Jerome. 1978. “The Multicandidate Calculus of Voting: Application to

Canadian Federal Elections.” American Journal of Political Science. 22:
609-638.

Blais, Andre and Richard Nadeau. 1996. “Measuring Strategic Voting: A Two
Step Procedure.” Electoral Studies. 15: 39-52.

172

Browne, Eric and Dennis Patterson. 1999. “An Empirical Theory of Rational
Nominating Behavior Applied to Japanese District Elections.” British
Journal of Political Science. 29: 259-289.

Budge, Ian et al. 1997. The Politics of the New Europe. New York: Addison
Wesley Longman Limited.

Butler, David, and Dennis Kavanagh. 1988. The British General Election of
1987. London: Macmillan.

Cain, Bruce E. 1978. “Strategic Voting in Britain.” American Joumal of
Political Science. 22: 639-655.

Campbell, John C. 1981. Parties, Candidates, and Voters in Japan: Six
Quantitative Studies. Michigan Papers in Japanese Studies No. 2. (edt.)
Ann Arbor: The University of Michigan.

Catt, Helena. 1989. “Tactical Voting in Britain.” Pariiamentary Affairs 42:548-
59.

Chhibber, Pradeep and Kenneth Kollman. 1998. “Party Aggregation and the
Number of Parties in India and the United States.” American Political
Science Review. 92:329-342.

Choi, Jungug. 1999. “Sophisticated Voting, Plurality Rule, and
Presidentialism: An Individual-Level Test.” The paper prepared for the
annual APSA meeting.

Christensen, Raymond V. and Paul E. Johnson. 1995. “Toward a Context-
Rich Analysis of Electoral Systems: The Japanese Example.” American
Joumal of Political Science. 39:575-598.

Cox, Gary. 1997. Making Votes Count. New York: Cambridge University
Press.

Cox, Gary. 1984. “Strategic Electoral Choice in Multi-Member District:
Approval Voting in Practice?” American Joumal of Political Science, 28:
722-738.

Dawns, Anthony. 1957. An Economic Theory of Democracy. New York.

Duverger, Maurice. 1954. Political Parties: Their Organization and Activity in
the Modem State. New York: Vlfiley.

Enelow, James M. and Melvin J. Hinich. Advances in the Spatial Theory of
Voting. Cambridge: Cambridge University Press.

173

Farquharson, Robin. 1969. Theory of Voting. New Haven: Yale University
Press.

Fenno, Richard F. 1973. Congressmen in Committees. Berkeley, California:
Institute of Governmental Studies Press.

Fey, Mark. 1997. “Stability and Coordination in Duverger’s Law: A Formal
Model of Preelection Polls and Strategic Voting.” American Political
Science Review. 91:135-147.

Fish, M. Steven. 1995. “The Advent of Multipartism in Russia, 1993-1995."
Post-Soviet Affairs. 1 1:340-383.

Fisher, Stephen L. 1973. “The Wasted Vote Thesis.” Comparative Politics.
293-299.

Fukui, Haruhiro and Shigeko Fukai. 1996. “Pork Barrel Politics, Networks and
Local Economic Development in Contemporary Japan.” Asian Survey 36:
268-286.

Gebethner, Stanislaw. 1996. “Proportional Representation Versus
Majoritarian Systems: Free Elections and Political Parties in Poland, 1989-
1991.” In Arend Lijphart and Carlos H. Waisman, eds. Institutional Design
in New Democracies. Boulder: Westview Press.

Green, William H. 1990. Econometric Analysis. New Jersey: Prentice Hall
International, Inc.

Haspel, Moshe, Thomas F. Remington and Steven S. Smith. 1998. “Electoral
Institutions and Party Cohesion in the Russian Duma.” Journal of Politics.
60:417-439.

Heath, Anthony, Roger Jowell, John Curtice, Geoff Evans, Julia Field, and
Sharon Witherspoon. 1991. Understanding Political Change: The British
Voter 1964-1987. Oxford: Pergamon Press.

Helweg, Diana. 2000. “Japan: A Rising Sun?” Foreign Affairs. 79: 26-39.

Herron, Erik. 2000. Mixed signals: Incentive structures and party strategies in
mixed electoral systems Dissertation

Herron, Erik and Misa Nishikawa. 2001. “Contamination Effects and the

Number of Parties in Mixed Superposition Electoral Systems.” Electoral
Studies. 20: 63-86, 2001.

174

Herron, Erik and Misa Nishikawa. 2000 “A Cross-National Evaluation of Mixed
Electoral Systems and Their Effect on the Number of Political Parties.”
presented at the annual meeting of the Midwest Political Science
Association, Chicago, Illinois, April 19-22, 2000.

Holler, Manfred. 1982. Power, Voting, and Voting Power. (edt.)Wurzburg;
Wren: Physica-Verlag.

Horie, Fukashi. 1989. “Senkyoseido Kaikakuron no Rigaitokushitsu.” In
Yomiuri Shimbunsha Chosakenkyu Honbu, eds. Nihon no Senkyo, Sekai
no Senkyo. Tokyo: Yomiuri Shimbunsha.

lnoguchi, Takashi. 1997. “A Step Toward One-Party Predominance: Japan’s
General Election of 20 October 1996.” Govemment and Opposition. 32:
48-64.

Institute of Electoral Research. 1967. A Review of Elections of the
World:1963 and 1964. London: \filliers Publications Ltd.

lshikawa, Masumi. 1981. Nihonser'ji No Ima. Gendaino Rironsha.

Jesse, Eckhard. 1988. “Split-Voting in the Federal Republic of Germany: An
Analysis of the Federal Elections from 1953 to 1987.” Electoral Studies.
72109-124.

Johnson. 1982. Johnson, Chalmars. 1982. MIT! and the Japanese Miracle:
The Growth of Industrial Policy, 1925-1975. Stanford University Press,
Stanford, California

Jones, Mark P. 1999. “Electoral Laws and the Effective Number of
Candidates in Presidential Elections.” Journal of Politics. 61 :171-184.

Kato, Shujiro. 1998. Senkyoseido no Shiso to Riron. Edited. Tokyo:
Ashishobo.

Kim, Heemin and Richard C. Fording. 2001. “Does Tactical Voting Matter?:
The Political Impact of Tactical Voting in Recent British Elections.”
Comparative Political Studies. 34: 294-311.

King, Gary. 1997. A Solution to the Ecological Inference Problem:
Reconstructing Individual Behavior From Aggregate Data. Princeton:
Princeton University Press

175

Kobayashi, Yoshiaki. 1991. Gendai Nihon No Senkyo. Tokyo: Tokyo Daigaku
Shuppan.

Kohno, Masaru. 1997. “Voter Turnout and Strategic Ticket-splitting Under
Japan’s New Electoral Rules.” Asian Survey. 37: 429-40.

Kramer, Gerald. 1983. “The Ecological Fallacy Revisited: Aggregate- Versus
Individual-Level Findings of Economics and Elections, and Sociotoropic '
Voting.” American Political Science Review 77: 92-11 1.

Laakso, Marku, and Rein Taagepera. 1979. “Effective Number of Parties: A
Measure with Application to West Europe.” Comparative Political Studies
12:3-27.

Lijphart, Arend. 1984. Democracies: Pattems of Majoritarian and Consensus
Governments in Twenty-One Countries. New Haven: Yale University
Press.

Lijphart, Arend. 1994. Electoral Systems and Party Systems. Oxford: Oxford
University Press.

Massicotte, Louis and Andre Blais. 1999. “Mixed Electoral Systems: A
Conceptual and Empirical Survey.” Electoral Studies. 18: 341-366.

McKelvey, Richard D. and Peter Ordeshook. 1972. “A General Theory of the
Calculus of Voting.” In J.F. Herndon and J.L. Bemd, eds., Mathematical
Applications in Political Science, VI. Charlottesville: University Press of
Virginia, pp. 32-78.

McLean, lain. 1987. Public Choice: An Introduction. Cambridge,
Massachusetts: Basil Blackwell, Inc.

Meehl, P. E. 1997. “The Selfish Voter Paradox and the Thrown-away Vote
Argument.” American Political Science Review. 61:11-30.

Merrill, Samuel. 1988. Making Multicandidate Elections More Democratic.
Princeton: Princeton University Press.

Mizusaki, Tokifumi and Hiroki Mori. 1998. “Tokuhyo Data kara Mita Heiritsusei
no Mekanizumu.” Senkyo Kenkyu 1998.

Miyake, lchiro. 1989. Tohyokodo. Tokyo: Tokyo Daigaku Shuppan.

Miyake, lchiro. 1995. Nihon No Seiji To Senkyo. Tokyo: Tokyo Daigaku
Shuppan.

176

Morrow, James D. 1994. Game Theory for Political Scientists. Princeton, New
Jersey: Princeton University Press.

Moser, Robert G. 1997a. “The Impact of Electoral Systems on New
Democracies: The Russian Experience in Comparative Perspective.”
Paper presented at the American Political Science Association
Conference.

Moser, Robert G. 1997b. “The Impact of Parliamentary Electoral Systems in
Russia.” Post-Soviet Affairs. 13:284-302.

Moser, Robert G. 1999. “Independents and Party Formation: Elite
Partisanship as an lntervening Variable in Russian Politics.” Comparative
Politics. 31:147-165.

Mueller, Dennis.1979. Public Choice. Cambridge: Cambridge University
Press.

Murakami Eichu. 1995. “Youyaku Kecchaku Shita Seijikaikaku: Fuhai Boshi
Ha Fujyubun.” In Asahi Nenkan 1995. Tokyo, Japan.

Niemi, Richard. 1984. “The Problem of Strategic Voting Behavior under
Approval Voting.” American Political Science Review. 78: 952-958

Niemi, Whitten and Franklin 1992. "The two faces of tactical voting" British
Journal of Political Science 24:549-557.

Ordeshook, Peter. 1986. Game Theory and Political Theory: An Introduction.
Cambridge: Cambridge University Press.

Ordeshook, Peter and Olga Shvetsova. 1994. “Ethnic Heterogeneity, District
Magnitude and the Number of Parties.” American JoumaI of Political
Science. 38:100-123.

Ozawa, Hideyuki. 1997. “Shinsenkyoseido” in Asahi Senkyo Taikan.
Tokyo: Asahi Shimbunsha

Palfrey, Thomas. 1989. “A Mathematical Proof of Duverger’s Law.” In Peter C.
Ordeshook, ed., Models of Strategic Choice in Politics. Ann Arbor:
University of Michigan Press.

Pempel, T.J. 1990. “Uncommon Democracies: The One-Party Dominant

Regimes.” In Uncommon Democracies: The One-Party Dominant
Regimes. Edited by T.J. Pempel. Ithaca: Cornell University Press.

177

Petrenko, Viktor, Olga Mitina and Ruth Brown. 1995. “The Semantic Space of
Russian Political Parties on a Federal and Regional Level.” Europe-Asia
Studies. 47:835-857.

Rae, Douglas W. 1967. The Political Consequences of Electoral Laws. New
Haven: Yale University Press.

Ramseyer, J. Mark and Frances McCall Rosenbluth. 1993. Japan’s Political
Marketplace. Cambridge: Harvard University Press.

Reed, Steven. 1990. “Structure and Behavior: Extending Duverger’s Law to
the Japanese Case.” British Journal of Political Science. 20:335-356.

Reed, Steven. 1996. “Political Corruption in Japan.” International Social
Science Journal 48: 453-466

Reed, Steven. 1999. “Strategic Voting in the 1996 Japanese General
Election.” Comparative Political Studies. 32:257-270.

Reed, Steven. 2001. “Duverger’s Law is Working in Italy.” Comparative
Political Studies. 34:312-327.

Reed, Steven and Michael F. Thies. 2001. “The Consequences of Electoral
Reform in Japan.” In Mixed-Member Electoral Systems: The Best of Both
Worlds? Edited by Matthew Soberg Shugart and Martin P. Wattenberg.

Reeve, Andrew and Alan Ware. 1992. Electoral Systems: A Comparative and
Theoretical Introduction. New York: Routledge.

Remington, Thomas F. and Steven S. Smith. 1996. “Political Goals,
Institutional Context, and the Choice of an Electoral System: The Russian
Parliamentary Election Law.” American Journal of Political Science. 40:
1253-1279.

Reynolds, Andrew and Ben Reilly. 1997. The Intematial IDEA Handbook of
Electoral System Design. Stockholm, Sweden: lntemational Institute for
Democracy and Electoral Assistance.

Richardson, Bradley M. 1988. “Constituency Candidates Versus Parties in
Japanese Voting Behavior” American Political Science Review 82: 695-
71 8.

Riker, William H. 1982. "The Two-Party System and Duverger’s Law: An

Essay on the History of Political Science.” American Political Science
Review 76: 753-766

178

Riker, William H. 1986. “Duverger’s Law Revisited.” In Bernard Grofrnan and
Arend Lijphart, eds. Electoral Laws and their Political Consequences.
New York: Agathon Press.

Riker, William H. and Peter C. Ordeshook. 1968. “A Theory of the Calculus of
Voting” American Political Science Review 62: 25-42

Rochon, Thomas R. 1981. “Electoral System and the Basis of the Vote: The
Case of Japan.” Parties, Candidates, and Voters in Japan: Six
Quantitative Studies. Michigan Papers in Japanese Studies No. 2. Edited
by Campbell, John C. Ann Arobor: The University of Michigan.

Sangiin Jimukyoku. 1998. Dai 18kai Sangiin Senkyo lchiran (Heise 10nen
7gatsu 12nichi shikko). Tokyo: Okurasho lnsatsukyoku

Sato, Seizaburo. 1997. ”Senkyokaikauronsha Wa Haibokushita” Shokun
1gatsugo.

Schlesinger, Joseph A. 1991. Political Parties and the Winning of the Office.
Ann Arbor: University of Michigan Press.

Senkyoseido Kenkyukai. 2000. Shugiin Senkyo Yoran: Heisei 12nen
Saishinban. Tokyo: Kokuseijoho Senta- Shuppankyoku.

Shubik, Martin. 1984. A Game-Theoretic Approach to Political Economy.
Volume 2 of Game Theory in the Social Sciences. Cambridge,
Massachusetts: The MIT Press.

Shugart, Mathew S. and Martin P. Wattenberg. 2001. Mixed-Member
Electoral Systems. The Best of Both Worlds? Oxford: Oxford Press.

Siavelis, Peter and Arturo Valenzuela. 1994. “Electoral Engineering and
Democratic Stability: The Legacy of Authoritarian Rule in Chile.” In Arend
Lijphart and Carlos H. Waisman, eds. Institutional Design in New
Democracies. Boulder: Westview Press.

Smyth, Regina. 1997. Building Democracy by Winning Votes? Candidates
and Parties in Transitional Russia. Dissertation: Duke University.

Taagepera, Rein and Bernard Grofrnan. 1985. “Rethinking Duverger’s Law:
Predicting the Effective Number of Parties in Plurality and PR Systems —

Parties Minus Issues Equal One.” European Journal of Political Research
13: 341-352

179

Taagepera, Rein and Markku Laakso. 1980. “Proportionality Profiles of West
European Electoral Systems.” European JoumaI of Political Research.
82423-446.

Taagepera, Rein and Matthew Soberg Shugart. 1989. Seats and Votes: The
Effects and Determinants of Electoral Systems. New Haven: Yale
University Press.

Tullock, Gordon. 1998. On Voting. Cheltenham, UK: Edward Elgar
Yomiuri Shimbun 2001. June 26. 2002. Tokyo Japan

Yomiuri Shimbun. 2002. “Nagatacho, Atarashii Kaze: Tetsu no Danketsu
Kuzushita Hanninmae.” January 6, 2002. Tokyo, Japan.

Yomiuri Shimbunsha Chosakenkyu Honbu. 1989. Nihon no Senkyo, Sekai no
Senkyo. Tokyo: Yomiuri Shimbunsha.

Wada, Junichiro. 1995. “Shosenkyoku Hireidaihyo Heiritsusei Ni Kansuru
Geimuronteki lchikousatsu.” Senkyo Kenkyu. No.10 1995.

Wada, Junichiro. 1994. The Japanese Election System: Three Analytical
Perspectives. New York: Routledge.

White, Stephen, Richard Rose and Ian McAIlister. 1997. How Russia Votes.
Chatham: Chatham House Publishers.

White, Stephen, Matthew Wyman and Olga Kryshtanovskaya. 1995. “Parties
and Politics in Post-Communist Russia.” Communist and Post-Communist
Studies. 28:183-202.

White, Stephen, Matthew Wyman and Sarah Oates. 1997. “Parties and

Voters in the 1995 Russian Duma Elections.” Europe-Asia Studies.
49:767-798.

180

Appendix 2-1: Classifications of Mixed Systems without Compensatory

Seats

Massicotte and Blais (1999)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

COUNTRIES YEAR ADOPTED
Andora 1993
Armenia 1995
Azerbaidjan 1995
Croatia 1995
Georgia 1992
Guinea 1991
Japan Lower House 1994‘
Japan Upper House 1982
Mexico Senate 1996
Russia 1993
Senegal 1998
Seychelles 1993
South Korea 1994
Taiwan National Assembly 1991
Current National Legislative Yuan 1992
Ukraine 1997
Shugart and Wattenberg (2001)
COUNTRIES YEAR ADOPTED
Armenia 1995
Georgia 1992
Hungary
Japan Lower House 1994
Japan Upper House 1982
Lithuania
Macedonia
Mexico Senate 1996
Philippines 1995
Russia 1993
South Korea 1994
Thailand
Ukraine 1997

 

 

Venezuela

 

 

181

 

 

Appendix 2-2: Classifications of Mixed Systems with Compensatory

Seats

Massicotte and Blais (1999)

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

COUNTRIES YEAR ADOPTED
Albania 1997
Bolivia 1996
Germany 1990
Italy, Chamber of Deputies 1993
Italy, Senate 1993
Mexico, Chamber of Deputies 1986
New Zealand 1993
Phillipines 1995
Tunisia 1993
Venezuela, Chamber of Deputies 1993
Shugart and Wattenberg (2001)
COUNTRIES YEAR ADOPTED
Bolibia 1996
Germany 1990
New Zealand 1993
Venezuela 1993

 

 

 

182

 

 

Appendix 5-1: Political Parties and Candidate Placement in the Lower
House Election - 1996 and 2000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

POLITICAL PARTIES 1996 2000

SMD PR SMD PR
LDP 288 327 269 325
NFP 235 133
DPJ 143 159 239 259
DSPJ 43 48 63 76
JCP 299 53 300 67
NPH 13 1 1
Jiyurengo 88 50 1 19 33
Shinshakai 37 26
Minkairen 2 1
CGP 21 63
Jiyu 55 72
Conservative 10 2
Mushozokunokai 9
Kaikaku Club 3
Other parties 28 0 5
Unaffiliated 85 - 79

 

Note: The numbers of candidates for the 1996 election are found in Asahi
Senkyo Taikan (1997). The numbers for the 2000 election are counted by the

author based on The Yomiuri Newspaper (2000).

183

 

Appendix 5-2: Political Parties and Candidate Placement in the Upper
House Election - 1996 and 2000

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

POLITICAL PARTIES 1998
SMD PR
LDP 57 30
NFP
DPJ 23 25
DSPJ 19 17
JCP 46 25
NPH 0 3
Jiyureggo 43 9
Shinshakai 1 1 3
Minkairen
CGP 2 18
Jiyu 9 12
Conservative
Mushozokunokai
Kaikaku Club
Joseito 7 3
Niin Club 0 3
Sports Heiwato 6 3
Seinen Jiyuto 12 5
Ishinseito Shinpu 7 2
Other parties 3
Unaffiliated 60

 

 

 

 

 

Note: The number of candidates for the 1998 election is counted by the
author.
Source: Sangiin Jimukyoku (1998)

184

Appendix 6-1: The Survey Question Used to Test Hypotheses Under the
Mixed Electoral System with SMD and PR

Japan Election and Democracy Study (JEDS) 1996

Q3 Please consider 50 degrees as the feelings toward the individuals and
organizations who you do not have any feelings. If you like the individuals
and organizations, please pick a degree that is between 50 and 100
depending on the strength of your feelings. If you have negative feelings
toward the individuals or organizations, please pick a degree that is between
0 and 50 degrees depending on the strength of your feelings.
Liberal Democratic Party (
Japan Socialist Party (
Clean Government Party (
Democratic Socialist Party (
Japan Socialist Party (
New Liberal Club (

vvvvvv

Q13 What party’s candidates did you vote for in the Lower House election?
1. Liberal Democratic Party
2. Japan Socialist Party
3. Clean Government Party
4. Democratic Socialist Party

5. Japan Communist Party

6. New Liberal Club

7. Shaminren

8. Others

9. Forgot

10. Did not vote

Q18 By the way, do you have a party which you feel close to?
1 Yes
0 No

Q30 SQ3 Going to vote is a voter’s duty
1 Agree
2 Lean toward agree
3 Neither
4 Lean toward disagree
5 Disagree

F1 Sex
1 Male
2 Female

F2 What year, month and date did you born? What is your age?

185

F6 Which level of school did you last attend, or what level of school are you
currently attending?

1 Primary school

2 Junior highschool

3 High school

4 Technical or vocational school

5 Two-year college, higher technical school

6 Four-year college

7 Graduate school

F7 What was your household income, including everybody’s in your home, in
last year (From January to December Heisei 7)? Please include in your
answer pre-tax income and all temporary income.

Less than 2,000,000 yen

From 2,000,000 to 4,000,000 yen

From 4,000,000 to 6,000,000 yen

From 6,000,000 to 8,000,000 yen

From 8,000,000 to 10,000,000 yen

From 10,000,000 to 12,000,000 yen

From 12,000,000 to 14,000,000 yen

More than 14,000,000 yen

QNOJUI-wa—k

F1OS7 SQ 7 Do you belong to a labor union?
1 Yes
0 No

B2 SQ1 What candidate did you vote for in your single-member district?
( ) Candidate code
( ) Party if only party mentioned
( ) Candidate from outside of the district

8285 SQ5 What party did you vote for in the proportional representation
election?
1 Liberal Democratic Party
2 National Frontier Party
3 Democratic Party of Japan (Hatoyama, Kari New Party)
4 Social Democratic Party of Japan
5 Japan Communist Party
6 New Party of Harbinger
7 Other party ( )
8 R did not vote in the proportional representation district
9Rmmm
10 Don’t know
11 No answer
Note: The above is not an official translation.

186

Appendix 6-2: The Survey Question Used to Test Hypotheses Under the
Mixed Electoral System with Medium Size District and PR

Japanes Election Study (JES) 1983

Q35 People have a variety of feelings toward individuals and organizations
who are influencial in politics. We came up with a question that can measure
such feelings. Please consider 50 degrees as the feelings toward the
individuals and organizations who you do not have any feelings. If you like
the individuals and organizations, please pick a degree that is between 50
and 100 depending on the strength of your feelings. If you have negative
feelings toward the individuals or organizations, please pick a degree that is
between 0 and 50 degrees depending on the strength of your feelings.
Liberal Democratic Party (
Japan Socialist Party (
Clean Government Party (
Democratic Socialist Party (
Japan Socialist Party (
New Liberal Club (

vvvvvv

Q28 SQ2. What party’s candidate did you vote for in the district election?
Liberal Democratic Party
Japan Socialist Party
Clean Government Party
Democratic Socialist Party
Japan Communist Party
New Liberal Club
Jiyuchotohakulabu
Katteren

. Dainiinkulabu

10. Sararimanshinto

1 1. Nihonkokuminsejirengo
12. MPD,Heiwatominshuundo
13. Zatsuminto

14. Kyoikuto

15. Mutohashiminrengo

16. Sekaijoreikai

17. F ukushito

18. Nihonyonaoshito

19. Others ( )

mmsewewwe

029 What Party did you vote for in the PR list election?
1 Liberal Democratic Party
2. Japan Socialist Party
3 Clean Government Party
4 Democratic Socialist Party

187

Japan Communist Party
New Liberal Club
Jiyuchotohakulabu
Katteren

Dainiinkulabu

10 Sararimanshinto

11 Nihonkokuminsejirengo
12 MPD,Heiwatominshuundo
13 Zatsuminto

14 Kyoikuto

15 Mutohashiminrengo

16 Sekaijoreikai

17 Fukushito

18 Nihonyonaoshito

19 Others ( )

COQNODU'I

Q15 Now, changing the subject somewhat, what political party do you
suppon?

1 If a party is mentioned by the respondent
0 Others

DUTY Q10 SO 3 Going to vote is a voter’s duty

1 Agree

2 Lean toward agree

3 Neither

4 Lean toward disagree
5 Disagree

UNION Do you or your family member belongs to a union?

CITY

SEX

1 If the respondent belongs to a union
0 Others.

1 10 large cities,

2 Cities over 100 thousand in population,
3 Other Cities

4 Counties

1 Male
2 Female

INCOME

1 Less than 1,000,000 yen

2 1,000,000 to 2,000,000 yen
3 2,000,000 to 3,000,000 yen
4 3,000,000 to 4,000,000 yen
5 4,000,000 to 5,000,000 yen

188

6 5,000,000 to 6,000,000 yen
7 6,000,000 to 7,000,000 yen
8 7,000,000 to 8,000,000 yen
9 8,000,000 to 9,000,000 yen
10 9,000,000 to 10,000,000 yen
11 More than 10,000,000 yen

EDUCATION

1 Primary school, Junior high school
2 High school
3 College

Note: The above is not an official translation.

189

Appendix 7-1: Predicted Probabilities

Given the following formula, the model can be reexpressed as the following:
Probm = i) = EXP(mxo I: Bax. (1)

The following equation is formulated for the results from Lower House.

Prob(Yi = Abstention) = [EXP(5.417 - 2.006x1i - 0.807x2; - 1.093x3,
+0.354x4, - 0.024x5,- 0.094x5i + 0.028», +0.051x3, +0.191X9i)]
r {1+[EXP(5.417 - 2.006x1i - 0.80m, - 1.093x3; - 0.354x4;
- 0.024X5i- 0.094Xei ‘I’ 0.028X7i + 0.051X3i+ 0.191Xgi)
+ EXP(1.945 - 2.402X1i - 0.268X2i - 0.610X3i +0.362X4i + 0.019X5i
- 0.180x6; - 0.030x7, - 0.264x3;+ 0.146X9i)]}

where yi = VOTE CHOICE, x1 = OPPORTUNITY, X; = DUTY, x3 = PARTY
SUPPORT, x4 = SEX, x5 = AGE, x5 = EDUCATION, x7 = INCOME, x3 =
UNION, and x9 = CITY SIZE

By varying the value of x1 from 0 to 1 while keeping other variables constant
(i.e. mean, medium or mode), we obtain the followings.

When x1=0
Prob(Yi = Abstention)
= [EXP(5.417 - 2.006(0) - 0.807(5) - 1.093(0) + 0.354(2) - 0.024(47)-
0.094(3) + 0.028(4) +0.051(0) + 0.191(3))] l{1+[EXP(5.417 - 2.006(0)
- 0.807(5) - 1.093(0) +0.354(2) - 0.024(47)- 0.094(3) + 0.028(4) +
0.051(0)+ 0.191(3)) + EXP(1.945 - 2.402(0) - 0.268(5) - 0.610(0) +
0.362(2) + 0.019(47)- 0.180(3) - 0.030(4) - 0.264(0)+ 0.146(3))1}
=0.32

When x1=1
Prob(Yi = Abstention)
= [EXP(5.417 - 2.006(1) - 0.807(5) - 1.093(0) + 0.354(2) - 0.024(47)-
0.094(3) + 0.028(4) +0.051(0) + 0.191(3))] I {1 +[EXP(5.417 - 2.006(1)
- 0.807(5) - 1.093(0) +0.354(2) - 0.024(47)- 0.094(3) + 0.028(4) +
0.051(0)+ 0.191(3)) + EXP(1.945 - 2.402(1) - 0.268(5) - 0.610(0) +
0.362(2) + 0.019(47)- 0.180(3) - 0.030(4) - 0.264(0)+ 0.146(3))1}
=0.24

The above means that the probability of abstention will decrease from 0.32 to
0.24 when voters’ most preferred candidates are present. Similarly for (Yi =
Strategic), we can obtain predicted probabilities.

190

The values of control variables are shown below.

 

 

 

 

 

 

 

 

 

LOWER UPPER
HOUSE HOUSE
DUTY 5 4
PARTY SUPPORT O 1
SEX 2 2
AGE 47 34
EDUCATION 3 2
INCOME 4 5
UNION 0 0
CITYSIZE 3 3

 

 

 

 

 

191

M IIIIIIIIIIIIIIIIIIIIIIIIIIIIII

lllllllllllllllllllIIIll/Illllllllllllllllll

_ 3 1293 02372 0604 _