m» :25 :1. : my. .. a? 19...)? 11124:“ ‘ .100 u... .. .31: a) v‘... (in .5 :3. .1 >5... 1? .13. 1:35.. 2’... ‘ :3 \1. ‘3. ’20... 11$ 1 ) . xl .0 . .. «2|!va 1‘... J; .3317. & ix .. z , up 5.....n:!.. 123.93 Y .C .1 ‘ £2 dgfigfi agfigg rug. 5!. . :1; L $51: LIBRARY -‘~ Michigan State a o o b/ University This is to certify that the dissertation entitled INSTITUTIONS, OPPOSITION BEHAVIOR, AND THE FATES OF GOVERNMENTS IN PARLIAMENTARY DEMOCRACIES presented by K0 MAEDA has been accepted towards fulfillment of the requirements for the PhD. degree in Political Science 777% 3771“) Major Professor’s Signature may / 9 3&05 Date MSU is an Affirmative Action/Equal Opportunity Institution 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 EST 3 o 2007 111808 2105 c:/CIRC/DateDue.lnd¢p.15 INSTITUTIONS, OPPOSITION BEHAVIOR, AND THE FATES OF GOVERNMENTS IN PARLIAMENTARY DEMOCRACIES By Ko Maeda A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 2005 ABSTRACT INSTITUTIONS, OPPOSITION BEHAVIOR, AND THE FATES OF GOVERNMENTS IN PARLIAMENTARY DEMOCRACIES By Ko Maeda Analyzing party competition in the world’s parliamentary democracies, this re- search seeks to find out the mechanism that makes some governments stay in power for a long time and others for a short time. I argue that we should not view all par- ties in a country as homogenous entities but pay closer attention to the differences between governing parties and opposition parties in their behavioral patterns and strategies. I first demonstrate that the governing parties are benefited in elections if the opposition camp is fragmented into many parties. The focus is then shifted to the question of what determines the degree of opposition fragmentation. I show that the various institutional rules—both electoral and parliamentary—give varying incentives to opposition parties and make the life of smaller parties either easier or tougher. I then take on another important question about opposition parties: why some countries have had opposition parties whose platforms were too radical for moderate voters, which is also a factor that prolongs the tenure of governments. Analyzing the case of the Japan Socialist Party, I demonstrate how Japan’s elec- toral system gave an incentive to some socialist politicians to wish their party not to become larger and hence to resist the proposed change of their platform. Copyright by K0 Maeda 2005 Ryoshin ni sasageru iv ACKNOWLEDGEMENTS In completing my six years of graduate work, I cannot help being amazed at how fortunate I have been in having received an enormous amount of help from so many people. Since there are too many individuals to mention in this space, I want to give special thanks to a few special ones. First of all, I would like to thank my advisor, Mark Jones, for all of his advice and guidance. He has always been a great mentor who shaped my professional skills. Without the deadlines he sometimes gave me during the process of writing this dissertation, I would not have been able to complete it. I also owe a great deal of thanks to other members of my dissertation committee: Burt Monroe, Brian Silver, Eric Chang, and Mark Elder. Other professors at MSU, in particular, Tom Hammond, Bill Jacoby, Dennis Patterson, Dave Rohde, and Ken Williams were significantly important in my professional development, and I am thankful for the opportunity of having met them here and received their help and guidance. I also would like to thank the Political Science Department at MSU for providing me with ample financial aid, many teaching opportunities, and a great office and computer facilities. My thanks also go to former and current colleagues at the MSU Political Sci- ence Department. In particular, Misa Nishikawa, Jamie Carson, Chris Bonneau, Kirk Randazzo, Becky Tothero, Mark Souva, Chuck Finocchiaro, Scott Szymendera, Gina Lambright, Michele Kuenzi, Bob Moore, Wonjae Hwang, Philip Bretton, Mike Crespin, Bill Bewick, Paloma Bauer, Wonbin Cho, RoHan Williams, Stewart French, Jeremy Duff, Dave Dreyer, Tetsuya Fujiwara, Kris Grady, Sara Gubala, Dan Lewis, Akitaka Matsuo, Carl Snook, and Fred Wood created both a fun and an intellectu- ally stimulating environment in which to work. In deciding to come to the United States to pursue a Ph.D., the professors I had at the University of Tsukuba were greatly influential to me. I would like to express my gratitude to the following professors: Yutaka Tsujinaka, the late Yutaka Akino, Mikiko Iwasaki, Ken’ichi Ohno, Keiji Maegawa, and Hiroshi Komai. At the State University of New York at Oswego where I spent two semesters as an international exchange student from T sukuba, I was fortunate to get to know Walter Opello and Stephen Rosow, who later wrote me letters of recommendation when I applied for graduate schools. My good experiences at Oswego were very important in deciding to continue my work in the United States. I would also like to thank my newlywed wife, Ayumi, whom I married at the very last stage of my graduate career. Her encouragement and support—both mental and logistical—enormously helped me during the hectic days of dissertation writing and job hunting. I hope I can repay her by promising much more peaceful days in our new life that is going to start now. Finally, I would like to give a special thanks to my parents, grandparents, and bothers. Their unconditional support throughout my 23 years of school life made it possible for me to pursue my interests and make my dream come true. I dedicate this dissertation to my parents. vi Table of Contents LIST OF TABLES ix LIST OF FIGURES x 1 Introduction 1 2 Political Institutions and Party Competition 4 2.1 Studies of Party Systems ......................... 4 2.2 The Determinants of Party Systems ................... 5 2.2.1 The Effects of the Single-Member District System ....... 5 2.2.2 The Effects of the Proportional Representation System . . . . 8 2.2.3 Institutional and Social Determinants of Party Systems . . . . 9 2.3 The Dynamic Aspects of Party Systems ................. 11 2.4 Real World Relevance .......................... 15 3 Opposition Hagmentation and the Electoral Fortunes of Governing Parties 17 3.1 Introduction ................................ 17 3.2 Theory ................................... 19 3.3 The Data and the Method ........................ 20 3.4 Results from Separate Samples ..................... 24 3.5 Results from the Pooled Sample ..................... 27 3.6 Conclusion ................................. 29 3.7 Appendix: Notes on the calculation of vote changes .......... 31 4 Party Competition in Regional Politics 33 4.1 Introduction ................................ 33 4.2 Analysis .................................. 33 4.2.1 Cost of Ruling ........................... 33 4.2.2 Opposition Fragmentation .................... 35 vii 4.2.3 The Effects of Opposition Fragmentation ............ 35 4.3 Conclusion ................................. 41 4.4 Appendix ................................. 42 5 Institutional Rules and Opposition Fragmentation 44 5.1 Introduction ................................ 44 5.2 The Measurement of the Opposition Party System Evolution ..... 45 l' 5.3 Institutional Determinants of the Opposition Party System Evolution 48 5.4 Empirical Analysis ............................ 50 5.4.1 Sample Cases ........................... 50 5.4.2 Independent Variables ...................... 51 5.4.3 Statistical Model ......................... 53 5.4.4 Results ............................... 53 5.5 Conclusion ................................. 55 6 Position Change of Opposition Parties: The Case of the Japan Socialist Party 57 6.1 Introduction ................................ 57 6.2 The Policy Position of the J SP and the Legislators’ Re-election Quest 59 6.2.1 A Short History of the JSP and its Platforms ......... 59 6.2.2 The Structural Reform Plan and its Demise .......... 61 ' 6.2.3 The Incentive Structure Induced by the SNTV System . . . . 64 6.3 Empirical Investigation .......................... 71 6.3.1 Identifying the JSP Legislators’ Attitudes ........... 71 6.3.2 Cross Tabulation Analysis .................... 75 6.3.3 Maximum Likelihood Estimation ................ 79 6.4 Conclusion ................................. 83 7 Conclusion 85 Bibliography 89 viii 3.1 3.2 4.1 5.1 6.1 6.2 6.3 6.4 6.5 List of Tables The Determinants of the Governments’ Vote Change (Separate Sam- ples) .................................... 25 The Determinants of the Governments’ Vote Change (Pooled Sample) 28 Determinants of the Governments’ Vote Change ............ 39 Regression Analysis ............................ 54 The Number of the JSP Candidates in the 1960 Election, Classified by the District Magnitude and the Number of the JSP Candidates in the District ................................ 66 The J SP incumbents and their competition against other opposition parties in the 1960 election. ....................... 70 Cross-Tabulation Analysis ........................ 77 Ordered Logit Estimation of the J SP Members’ Attitudes toward the Structural Reform ............................. 81 Predicted Probabilities .......................... 82 ix 1.1 3.1 3.2 3.3 4.1 4.2 4.3 4.4 5.1 6.1 6.2 List of Figures Overview of the Argument ........................ The Distribution of the ENOP ...................... Conditional Effects of Unemployment (Separate Samples) ....... Conditional Effects of Unemployment (Pooled Sample) ........ The Distribution of the Percent Vote Change of the Governments Effective Number of Opposition Parties ................. EN OP and the Governments’ Vote Change ............... Partial plot of ENOP and the Governments’ Vote Change ....... Calculation of AENOP ......................... Ideological Locations of the Japanese Political Parties in the Early 19603 .................................... Factional Realignment within the JSP juring the Conflict over the Structural Reform Plan .......................... 2 22 26 29 34 36 37 40 69 73 Chapter 1: Introduction Modern democracies operate with political parties. This research seeks to con- tribute to our understanding of modern democratic politics by examining how po- litical institutions affect the manner in which political parties compete. The big question upon which this dissertation is built is why some parties stay in power for a long time and others for a short time. Japan’s Liberal Democratic Party continuously stayed in power from 1955 to 1993. Italy’s Christian Democratic Party, likewise, dominated the country’s post-World War II politics until the early 19903. On the other hand, some countries have experienced frequent alternations in the party control of the executive branch, such as Belgium and Denmark. I be— lieve that such a difference in the frequency of party control turnovers is politically consequential since frequent changes in the executive branch would make the imple- mentation of long-term policies difficult. At the same time, if a single party controls the government for an exceptionally long time, the politics may become stagnant or corrupt (also see Maeda and Nishikawa forthcoming). This study focuses on the strategies and behaviors of political parties, especially those of opposition parties, as the factors that affect the length of time governing parties can stay in office. I also argue that institutional rules, such as electoral systems, partly determine how parties behave. Figure 1.1 shows the relationships of the factors I suspect to exist. It is now widely accepted that political institutions make differences in political processes and outcomes. In particular, in this research I argue that the preferences and strategies of both leaders and members of opposi- tion parties are influenced by the characteristics of political institutions. Then, the preferences and strategies of opposition politicians are reflected on the behavioral patterns of opposition parties. Among many aspects of opposition behavior, this research pays particular attention to the fragmentation of opposition (i.e., the num- Figure 1.1: Overview of the Argument Patterns of Party Competition T’I‘T Parties’ Strategies and Behaviors TTT Institutional Rules ber and sizes of opposition parties) and ideological positions of opposition parties. These factors, I argue, make differences in the electoral performances of incumbent parties and, therefore, how frequently the alternations in party control of executive branches take place. In this dissertation, my analysis is solely focused on parliamentary systems. Although whether the mechanism found in this research is applicable to presidential systems is a fascinating research topic, I leave it for an agenda for future research. The structure of this dissertation is as follows. In the next chapter, I will dis- cuss general issues related to the theme of this research while reviewing the relevant literature. Chapters 3 to 6 present empirical analyses that are related to each other and collectively form the core component of this dissertation. Chapter 3 exam- ines how the degree of opposition fragmentation affects the electoral performance of incumbent parties by analyzing the election data from advanced parliamentary democracies in the world. A similar analysis is conducted in Chapter 4 with regional election data. The focus is then shifted to the question of why some countries have fragmented oppositions. Chapter 5 seeks to answer this question by focusing on the effects of institutional rules on the behaviors and strategies of opposition parties, which then lead to the emergence of certain types of opposition party systems. In Chapter 6, I explore another important question about opposition parties: why do some opposition parties have extreme or radical ideological positions that are electorally disadvantageous to them? By analyzing the case of the Japan Socialist Party, I argue for the importance of paying attention to intra-party politics and individual politicians’ preferences and strategies in examining a party’s behavior. Chapter 7 will conclude this dissertation by summarizing its findings and discussing their implications. Chapter 2: Political Institutions and Party Competition Before presenting the empirical analyses in Chapters 3 to 6, in this chapter I provide general discussion as to what has been done in this area of research, what innovation I am making, and why this research is relevant. 2.1 Studies of Party Systems The study of party systems is one of the well studied sub-fields in comparative politics. In particular, there is a large literature on the classifications of party systems. Duverger’s (1954) seminal book is often considered as the harbinger of the modern studies of party systems. Duverger’s classification was to categorize party systems into one-party, two-party, and multiparty systems. The focus of party system typologies in early years, such as Duverger’s, was mostly on the number- based classifications (also see Blondel 1968; Rokkan 1968). One of the innovations of Sartori’s (1976) classification of party systems was its attention to the ideological aspect of party systems. His seven categories of party systems are: one party system, hegemonic party system, predominant party system, two party system, limited pluralism system, extreme pluralism system, and atomized system. The differences between limited pluralism and extreme pluralism are not only in the number of parties but also in the ideological distances among parties. Specifically, an extreme pluralism system has many parties (six or more), and they are ideologically far apart (Sartori 1976, 132-140). Sartori’s typology involves both numerical and ideological aspects of party systems, but his general argument is that the two dimensions are correlated. Almost three decades have passed since Sartori (1976), and there have been more recent attempts of party system classifications (e.g., Siaroff 2000). Yet, Sar- tori’s (1976) typology is still arguably the most influential work and is frequently cited and discussed (see Mair 1996; Ware 1996). 2.2 The Determinants of Party Systems As well as the classifications of party systems, much research has been con- ducted on the determinants of party systems. Duverger’s (1954) early contribution is still greatly influential in this literature. Duverger observed that plurality systems tend to create two-party systems and proportional representation (PR) or majority run-off systems tend to produce multi-party systems. Later, Riker (1986) argued that Duverger’s statement about two-partism was a law and the statement about multi-partism was a hypothesis because the latter had little empirical support. In the following, I will review the existing studies on these propositions. 2.2.1 The Effects of the Single-Member District System Duverger’s law works by two factors: Mechanical and Psychological. The me- chanical factor refers to “the ‘under-representation’ of the third, i.e. the weakest party, its percentage of seats being inferior to its percentage of the poll” (Duverger 1954, 226). Also, the first party’s seat share tends to be overrepresented. Since larger parties gain bonus and smaller parties are penalized, the difference in votes is amplified when seats are allocated to parties. This mechanism reduces the number of parties—it is mechanical and no human strategy is involved. Mechanical effects are studied in, for example, Taagepera and Shugart (1989; 1993). The psychological factor, on the other hand, stems from the strategic behavior of political actors. Duverger (1954, 226) argued, “In cases where there are three parties operating under the simple-majority single-ballot system the electors soon realize that their votes are wasted if they continue to give them to the third party: whence their natural tendency to transfer their vote to the less evil of its two adversaries in order to prevent the success of the greater evil.” This kind of voters’ behavior is also called “strategic voting.” In addition to the voters’ level, it has been observed that the psychological factor works also in the candidate level, although Duverger himself did not mention this. Lijphart (1999, 165) explains, “the psychological factor operates at the level of politicians, whose natural tendency is not to waste their energy by running as third-party candidates but instead to join one of the large parties.” Blais and Carty (1991) find that the psychological effect is as strong as the mechanical effect. It has been repeatedly argued that the district level and the national level should be considered separately (Wildavsky 1959; Riker 1986). Reed (2001, 313- 314) nicely summarizes the issue: “the effects of Duverger’s law should be thought of as two distinct but related processes. First, competition is reduced to no more than two candidates per district. Second, the candidates in each district coordi- nate their efforts into two parties in parliament and two parties contesting elections nationwide.” Cox (1997, 185-186) further argues that nothing in Duverger’s theo- retical arguments predict two—party system in the national level. “Duverger’s Law at the district level is a theoretical proposition, while Duverger’s Law at the national level is an empirical generalization, and one to which there are many exceptions at that”(Cox 1997, 186). As for the district level, Chhibber and lKollman (1998) Show that even India, which is generally seen as an exception to Duverger’s law, has two-party systems in most districts; and Reed (2001) argues that most districts in Italy have moved toward bipolar competition after the introduction of the new electoral rule. The first step process appears to be omnipresent. The question is then on the second step: national level aggregation. How can a bipolar competition in the district level lead to a two-party system in the national level? If politicians from different districts have no incentive to make groups or coalitions, the number of parties in the national level will be quite large. In order for a local bipartism to be “projected” into a national bipartism, some incentives must push politicians to link together in national parties. Cox (1997, ch.10) attributes this process to the logic of economics of scale. When “a task that requires the help of a large number of legislators or legislative candidates” (186) is attractive for politicians, the pursuit of it will be the incentive for linkage. Cox mentions five such tasks: enacting laws, electing a president, electing a prime minister, securing seats in a national upper tier, and securing campaign finance (186-187). From the analysis of India and the United States, Chhibber and Kollman (1998) argue that national level aggregation occurs when national governments centralize political and economic powers, thereby making locally competitive but nationally noncompetitive parties less relevant. This argument is related to that of Cox since the centralized power should increase the incentive for politicians to gain power by forming large group or coalition. In summary, Duverger’s law predicting the emergence of two-party system in plurality electoral rules needs some reformulations. First, district and national levels should be considered separately. Duverger’s theoretical explanation works well in the district level, and it has strong empirical support. However, we need another theory for the national level. Second, as suggested by Cox (1997) and Chhibber and Kollman (1998), national level aggregation occurs when there are attractive things in the national level that can only be gained by a large group of politicians. It should be added here that a generalized form of Duverger’s law, the M+1 rule, is now widely recognized (Reed 1990; Cox 1997). The M+1 rule states that the equilibrium of the effective number of candidates in a district is the number of seats plus one. Hence the equilibrium is two in single-member-district plurality rule. Cox (1994) formally proved the equilibrium of Duverger’s law and the M+1 rule. 2.2.2 The Effects of the Proportional Representation Sys- tern Unlike the relationship between plurality rule and bipartism, Duverger’s hy- pothesis about PR or runoff system and multi-party system has not been supported strongly by empirical evidence. Riker (1986, 27-28) mentions four counterexam- ples: Australia, Austria, Germany, and Ireland. Although a strong probabilistic association exists, it is shown to be much weaker than the law. As Rae (1971) demonstrates, the penalty effect for smaller parties indeed ex- ists in any electoral system. However, the plurality rule gives enormously stronger discouraging effects to smaller parties than the PR or runoff system does. In fact, the strength of the penalty effect is determined not only by the electoral formula (plurality, PR, or runoff) but also by many other factors such as district magnitude, legal thresholds, and assembly size. Hence, even under the PR rule, small parties are difficult to be represented if the district magnitude is small, the threshold is high, and /or the assembly is small. A problem of Duverger’s hypothesis, then, may be the fact that it did not take those factors into consideration. However, a more important problem in Duverger’s hypothesis would be that the absence of those penalizing factors does not mean it actively encourages the emer- gence of new parties. Instead, it just works as a permissive condition for smaller parties to survive. Whether smaller parties continue to exist and whether the num- ber of parties increases are different questions. PR and runoff systems are generally less discouraging to smaller parties than the plurality system is; it only means that there is little pressure for smaller parties to disappear, and it does not mean that a two-party system is pressured to transform to a multi-party system. The reason for the relatively weak association found for Duverger’s hypothesis appears to be that PR and runoff systems give only a permissive condition for multi-partism. The number of parties resulting from other factors therefore remain unchanged under the PR or runoff system. Taagepera and Shugart (1989, 65) modify the hypothesis in the following way: “PR rules tend not to reduce the number of parties, if the number of issue dimensions favors the existence of many parties.” 2.2.3 Institutional and Social Determinants of Party Sys- tems Duverger’s argument is institutional in that the electoral system is the deter- minant of party systems. The alternative approach focuses on social structure. A well-known example of this approach is Lipset and Rokkan (1967b). The authors of this approach consider that social cleavages, ideological distribution, and so forth are the determinants of party systems. Some even assert that electoral systems are irrelevant and only social factors determine party systems. Cox (1997, 15-16) gives a good review of those arguments. The divide of institutional and social explanations of party systems somewhat reflects the divide in the whole political science community between economic and sociological approaches. Those who are influenced by economics value such ideas as rational actors and institutional effects; and those who are influenced by sociology tend to appreciate macro-social explanations. The “social determinist” (Cox 1997, 15) argues that both long-term multi-party and two-party systems are explained by the number of social or ideological cleavages and that electoral systems play an inconsequential role. These arguments, however, do not seem convincing in light of the massive empirical support for Duverger’s law we observe. Cox (1997, 20-23) further refutes the social determinist by showing the difference in party systems in lower and upper houses of many countries where the two houses adopt different electoral rules. If social cleavage is the sole determinant, two chambers of a country should show the same party system in spite of different institutional rules; but it is not the case. On the other hand, pure institutional explanation would also be insufficient especially for Duverger’s hypothesis because it works only as a permissive condition. It would therefore be plausible to consider that both institutional and social factors jointly determine party systems. The relative strength of the two is an empirical question. Powell (1982) is perhaps the first empirical analysis of the number of parties where both institutional and social factors are considered. Powell specified the num- ber of parties as the result of additive effects of the two forces. Later, Ordeshook and Shvetsova (1994) showed that the number of parties is determined by interac- tive, not additive, effects of institutional and social effects; and Amorim Neto and Cox (1997) demonstrated that the interactive specification holds even when more countries are included in the analysis, operationalization is modified, and more vari- ables are controlled for. Taagepera (1999), in agreement with the interactive model, developed a single equation that expresses the relationship between the variables: N : 16M“ +1 where N is the effective number of assembly parties, I is the number of politicized issue dimensions, and M is the effective magnitude. What the interactive model asserts is that (1) two-party systems emerge either because the electoral system penalizes smaller parties or because few social cleav- ages exist and (2) multi-party systems arise as the product of both multiple social cleavages and permissive electoral systems. 10 2.3 The Dynamic Aspects of Party Systems The approaches to the study of party systems can be roughly classified into two groups: static and dynamic. The static approach is to see the configuration of the parties in a country, such as the number and the relative sizes of the parties and the ideological distances among them. The studies reviewed in the last few pages are mostly static analyses of party systems. The dynamic approach, on the other hand, focuses on the change or behavior of parties over time. The topics of this latter type include party system change (e.g., Laver 1989; Mair 1997; Pennings and Lane 1998), parties’ position changes (Harmel and Janda 1994; Janda et a1. 1995), cabinet duration (a good review is found in Laver 2003), and the “structure” of party competition (Mair 1996). In the existing literature, the dynamic aspect of party systems is studied far less than the static aspect. Since this dissertation is concerned about a longitudinal issue (e.g., how long a governing party remains in office), my analyses have to consciously take the dynamic approach. I pay particular attention to competition among political parties, as opposed to the static “snapshot” characteristics of party systems (e.g., how many parties exist and where they are located in the ideological spectrum). In analyzing the manner by which parties compete, it is important not only to look at a party system as a whole but the strategies and behaviors of individual parties that make up a system of parties. In the following, I will discuss two issues that I believe are particularly important in analyzing party competition dynamically but have not been taken seriously in the literature. (1) The Behavioral Differences of Governing and Opposition Parties In the simple forms of the study of party politics, parties in a country are often seen as if they all have similar behavioral patterns, goals, and organizational features. In some countries, this way of viewing parties is indeed plausible and the 11 parties are homogenous in many respects. Hence, we often seek to characterize the parties in a specific country, e. g., “the parties in country X are cartel parties,” while implicitly stating that the parties in country X are homogenous. However, parties can be quite heterogeneous within a system. Especially in the systems where alternation in power does not take place regularly, government parties are accustomed to be in power and their behavioral patterns would reflect the fact that they are used to being in oflice. Opposition parties in such systems, on the other hand, do not have much experience in ruling, and their behaviors inevitably reflect the fact that they have been out of power for a long time. This dissertation consciously pays attention to the distinction between gov- erning parties and opposition parties in political systems. The next two chapters examine how the characteristics of opposition parties affect the electoral perfor- mances of incumbent parties. In Chapter 5, I will analyze the mechanism by which “opposition party systems” change. Chapter 6 will examine the behavior of the Japan Socialist Party, which was often called “a permanent opposition party” (e.g., Stockwin 1986) and demonstrate that its behavior stemmed from how the members felt about being in opposition. In the study of political parties and party systems, it appears that the strategies and behaviors of opposition parties have not received much attention. They were studied as political parties but not as opposition parties; it has not been common to study them with a clear distinction between governing parties and opposition parties—with a notable exception of Dahl (1966). The reason the studies of opposition parties became less common since the time of Dahl (1966) would probably be the same as the reason for the emergence of “cartel parties.” Mair and Katz (1997) argue that parties in many European countries now became “semi-state agencies” as many parties gained experience in being in office. “[T]he set of ‘governing parties’ is no longer as limited as it once was. At the 12 risk of over-generalization, almost all substantial parties may now be regarded as governing parties” (Mair and Katz 1997, 107). Indeed, when Dahl’s (1966) book was published, some European countries had parties that had never lost power in their post-World War II politics, such as West Germany’s Christian Democrats and Sweden’s Social Democrats. In these situations, distinctions between governing parties and opposition parties would be more noticeable. Yet, party competition in these advanced European democracies became more institutionalized, and more and more parties experienced being in power since then. However, I believe the perspective that emphasizes the behavioral and strategic differences between governing and opposition parties is still relevant. First, even in Europe, not all countries have cartel parties that are collusive and cooperative to one another and have similar organizational and behavioral characteristics. For example, the parties in Britain do not seem to be becoming cartel parties as Mair and Katz (1997, 108) admit. Party competition in Britain is not characterized by cooperation and accommodation among parties, and the behavior of a party would- be strongly affected by whether it is in power or in opposition. The radical reform of the party platform of the British Labour Party in 1994 would not have been possible if the party had not been out of power for a long time. Also, there are still many instances of parties in the world that never took office or have been out of power for a long period of time. Their behavior would reflect the fact that they were in opposition. In sum, the process of “cartelization” of parties would still be at an early stage, as Mair and Katz (1997, 108) note, or be taking place only in a limited part of the world. Second, whether the distinction between governing and opposition parties in terms of their behaviors and strategies is outdated or not is an empirical question. If political parties are indeed becoming cartel parties, probably such a distinction is no longer useful in understanding party competition. Yet, we should not easily 13 accept propositions without testing them empirically. My analyses presented in the following chapters will show whether my approach that emphasizes the behavioral and strategic differences between governing and opposition parties is useful or not. (2) Party Behavior and Intra-party Politics In the literature of party competition, it has been common to assume that a party is a single actor. This unitary actor assumption would indeed be plausible depending on time and place; yet it is not always the case. When a party makes an important decision about its political behavior, it often involves fierce intra-party politics in which multiple actors with respective purposes interact and compete. Borrowing Shepsle’s (1992) expression, a party is a “they,” not an “it.” I believe that the study of party competition will significantly advance if our theory can incorporate both intra-party politics and inter-party competition. Laver (1998, 22) acknowledges the need for going beyond the unitary actor assumption and calls this task “One of the biggest unclaimed prizes in the field.” Stro‘m (1990) argues that minority cabinets are formed because potential coali- tion parties choose not to join the coalition in pursuit of future electoral gains. I find this argument interesting and plausible, but I believe that the behavior of a party in joining or not joining a coalition will be better analyzed if we pay closer attention to the different incentives of the various actors within a party. For elec- torally vulnerable rank-and-file legislators, increasing the possibility of re-election would be the first and largest concern. On the other hand, party leaders are usually secure in their own re—election and hence their goals would be, first, to keep the position as the party leader and, second, to seek more power. Since their goals are typically different, the party leader and the rank-and-file members may well have different preferences about the behavior of the party (e.g., whether to join a coalition, whether to change the policy position, or whether to merge into another party). I believe that analyzing party behaviors from intra-party politics would give 14 us a deeper understanding than what we would obtain from the analysis with the unitary actor assumption. One of my main arguments is that political parties are not just seat / vote max- imizers. Parties’ behaviors sometimes seem irrational, and they cannot be well un- derstood in the framework that views parties as unitary actors. But I believe that many of seemingly irrational behaviors of political parties can be explained consis- tently if we go beyond the unitary actor assumption and analyze parties in richer contexts that incorporate intra—party politics. In Chapter 6, I will demonstrate that the behavior of the Japan Socialist Party, which was self-defeating electorally, is explained as an end result of its internal politics in which leaders and members of the party pursued their respective goals. 2.4 Real World Relevance I believe that this research is not only theoretically interesting but also im- portant in its relevance to real world politics. As I briefly noted in the previous chapter, how long a government stays in office would make significant differences in political and economic outcomes. Maeda and Nishikawa (forthcoming) report that the variation in the length of governments is much larger in parliamentary systems than in presidential systems; some parties are thrown out of office after only a few months, but some parties stay in power for exceptionally long periods (e.g., Italy’s Christian Democrats, Japan’s Liberal Democrats, and India’s Congress Party). In presidential systems, alternations in power take place more regularly. Frequent turnovers of ruling parties are not usually considered a desirable char- acteristic of party politics. A government that stays in power for a short time cannot implement long-term policies, and frequent policy changes may damage the effec- tiveness of public policies. Also, there have been the cases where excessive instability 15 of party politics led to breakdowns of political regimes, such as the French Fourth Republic. On the other hand, it is also problematic if ruling parties do not alternate. If one party dominates politics and opposition parties are weak, the incumbent government will not have much incentive to do what the voters want, and the voters do not have a real alternative to choose. The responsiveness of governments may decline under such circumstances. As Pempel (1990) called “uncommon democracies,” one-party dominant systems appear to be an anomalous type of democracies, which can bring about unaccountable governments. (It may not be a coincidence that some of the countries that have had one-party dominance also had extensive political corruption, such as Japan and Italy.) We have recently observed the emergence of one-party dominance in some new democracies such as South Africa, Namibia, and East Timor. The establishment of one-party dominance and the absence of strong opposition in those countries are making the future of their democracy questionable. As this research seeks to find out the mechanism of how political institutions affect the electoral performances of governing parties, the findings of this dissertation would have important implications for real world politics. Institutional rules, such as electoral systems, are factors which people can change, as opposed to sociological characteristics of human societies. Since much attention is currently paid to the designs of political institutions in such countries as Afghanistan and Iraq, analyzing the institutional mechanisms that influence the behavior of parties and the patterns of party competition appears to be an endeavor that has much practical relevance in the present day. 16 Chapter 3: Opposition Fragmentation and the Electoral Fortunes of Governing Parties 3. 1 Introduction The determinants of parliamentary election results have been extensively stud- ied. In particular, many scholars have attributed the electoral fortunes of governing parties to the country’s economic performance. More recently, much attention has been paid to political contexts, such as whether the government is composed of one party or a coalition of parties. In this chapter, I extend this argument by fo- cusing on opposition parties. Specifically, I will examine the effects of opposition fragmentation on the electoral fortunes of governing parties. Unlike U.S. congressional elections, where incumbency works as a strong advan- tage for candidates seeking re-election, incumbency in other advanced parliamentary systems is often a disadvantage for parties in government (Rose and Mackie 1983). Ster (1990) calls this phenomenon the “incumbency effect,” which is the core con- cept in his explanation of the formation of minority cabinets. This incumbency dis- advantage stems from the fact that governing parties receive much media scrutiny while in office and their performance is critically checked against their previous promises (Strem 1990). This factor is also called the “cost of ruling” (Paldam 1986). Of course, incumbency status is not the sole determinant of the election result. In particular, the country’s economic situation is considered an important factor that affects the election results, and the relationship between economic performance and the electoral fortunes of the governing party (parties) has received much attention from scholars. Although the majority of such studies examined the elections in advanced countries (Paldam 1981; Strem and Lipset 1984; Lewis-Beck 1988; Host and Paldam 1990; Paldam 1991), effort has also been made to extend the focus 17 to other parts of the world.1 Some use country-level aggregate data in which the dependent variable is the change in vote share of the incumbent party (parties), and others use individual-level survey data in which the dependent variable is each voter’s vote choice (for the incumbent or not). The results of these analyses vary, as Powell and Whitten (1993, 391) note, “Despite the large literature analyzing economic effects over time within countries, it has proved surprisingly difficult to demonstrate consistent effects in cross-national studies.” There have also been attempts to take political contexts into account when examining election results. For example, Lewis-Beck (1988, 108-09) argued that coalition governments tend to suffer less from anti-incumbent economic voting be- cause the government’s responsibility is diffused. Generalizing this argument, Powell and Whitten (1993) introduced the concept of clarity of responsibility and demon- strated that economic performance significantly affects a government’s electoral for- tune where the clarity of responsibility is high. They created an index of clarity by combining five indicators: voting cohesion in government, the nature of the com- mittee system, strength of the bicameral opposition, minority government status, and coalition government status. There have since been several follow-up studies of this thesis using both survey data (Anderson 1995; Anderson 2000; Nadeau, Niemi and Yoshinaka 2002) and aggregate data (Whitten and Palmer 1999; Chappell and Veiga 2000; Royed, Leyden and Borrelli 2000). In the next section, I will explain my theory and present a testable hypothesis. The data and the variables will then be described in Section 3, followed by the empirical analysis in Section 4. Section 5 concludes this chapter. 1Remmer (1991; 1993) on Latin America; Pacek (1994) and Fidrmuc (2000) on former commu- nist Eastern European countries; Pacek and Radcliff (1995) on developing countries in general; and Wilkin, Haller and Norpoth (1997) on a sample from all over the world. Lewis-Beck and Stegmaier (2000) is an extensive review article on this literature. 18 3.2 Theory AS described in the previous section, in recent years the literature on economic voting has begun to pay closer attention to the political context. In this study, I introduce another aspect of the political context: fragmentation of opposition parties. The responsibility hypothesis that originates from Powell and Whitten (1993) states that voters who are not satisfied with the government are more likely to vote against the governing party (parties) if responsibility for policy outcomes is clear. This argument implicitly assumes that discontented voters will vote for opposition parties to punish the government. However, whether these voters choose to vote for the opposition depends not only on how dissatisfied they are with the government, but also on the attractiveness of the opposition parties. Specifically, if many oppo— sition parties exist and compete with each other, the discontented voters who look for an alternative to the incumbent government may not see the opposition camp as a realistic and credible alternative. Phrther, where multiple opposition parties exist, they often have different and specific policy agendas (e.g., agrarian party, commu- nist party, and ecological party). Those particularistic opposition parties may have stable support bases, but they would not attract moderate swing voters who previ- ously voted for the current incumbent. On the contrary, a unified opposition party may have a more general and centrist stance and hence will attract more voters than fragmented ones could. Therefore, in addition to the clarity of responsibility, the characteristic of op- position parties should also be considered as a part of the political context. I agree with the recent literature that when “clarity” is high, the effects of economic per- formance on governments’ electoral performance would be stronger. Yet, I argue that among the “high clarity” cases, the degree of opposition fragmentation would further influence the strength of the economic determinants of the electoral results. 19 I am not the first one to pay attention to this factor. Anderson (2000, 155) indeed argues, “fragmented party systems . . . should make it more difficult for voters to identify a clear alternative to the incumbent government” and includes a variable for “clarity of available alternatives” in his regression equation. Yet, I have two reasons to believe that this study will still be a significant contribution. First, while Anderson’s (2000) research uses survey data, I use aggregate data. Second, I believe there is room for improvement in the operationalization of the variable. Anderson measures his concept “clarity of available alternatives” by the effective number of parties (Laakso and Taagepera 1979) in the legislature. Hence it measures the degree of party system fragmentation in the whole legislature but not opposition fragmentation which is more appropriate in this context. Hypothesis: When the opposition parties are unified and the clarity of responsibil- ity is high, the eflect of economic performance is strong in determining the governing party’s (parties’) vote change. 3.3 The Data and the Method The units of analysis are parliamentary elections in the world’s industrialized parliamentary systems. The observation period begins in 1963 because of a lim- itation in the sources of economic variables. The end of the observation period is 1997, the last year for which the government composition data are available in Woldendorp, Keman and Budge (2000). The countries included in the analysis are Australia, Austria, Belgium, Canada, Denmark, Finland, Germany, Greece, Ire- land, Italy, Japan, the Netherlands, New Zealand, Norway, Sweden, and the United Kingdom.2 The dependent variable is the percent change in the governing party’s 2Although France, Switzerland, and the United States have been often included in these types of studies (e.g., Powell and Whitten 1993), I chose to exclude them to focus only on parliamentary systems. The following elections are also excluded from the analysis because the situation makes it difficult for the voters to evaluate the incumbent government: the cases where a non-party cabinet 20 ( parties ’) vote share from the previous election. The details about this variable are described in the Appendix of this chapter. The main independent variable in this analysis is economic performance, which is represented by the unemployment rate. Unemployment is generally found to have the strongest impact on voting behav- ior among other economic variables; and Powell and Whitten (1993) also find this variable the most significant.3 In the first set of my analysis, the countries are classified into two groups ac- cording to the “clarity of responsibility.” I borrowed Powell and Whitten’s (1993) dichotomous classification of the countries, which is based on their additive scor- ing of five features of the political systems that clarify or blur the responsibility of economic performance, namely, whether parties are internally cohesive, whether opposition parties share committee chairs, how strong the bicameral opposition is, whether minority governments are frequently formed, and whether coalition cab- inets are frequently formed. Australia, Austria, Canada, Greece, Ireland, Japan, New Zealand, Sweden, and the United Kingdom are the “high clarity” group, while Belgium, Denmark, Finland, Germany, Italy, the Netherlands, and Norway are the “low clarity” group. Later in this chapter, I will present an analysis using a pooled sample. It has been shown that where “clarity” is high, economic performance has a significant effect on the government’s electoral fortune. I hypothesize that the effects of economic performance depend on whether the opposition parties are fragmented or unified. The degree of opposition fragmentation is quantified by Laakso and existed before the election (Italy 1996, Greece 1989 (November) and 1990, and Finland 1975); the cases where the all parties in government had been in power less than a year before the election (11 observations); and the case where the party system changed so fundamentally that it is impossible to identify the incumbent party (Italy 1994). Lastly, Canada 1993 and the Netherlands 1994 are dropped from the sample as extreme outliers, where the governing parties lost 27% and 21% of votes, respectively, which are more than three standard deviations away from the average. 3I repeated my analysis by adding the inflation rate and the GDP growth rate, but the results remained unchanged. The data source of the unemployment rate is Bulletin of Labour Statistics by the International Labour Office, and the value is averaged over four quarters including the quarter the election took place. 21 Taagepera’s (1979) “effective number of parties” to the number of seats of all non- government parties.4 It shows the number and relative sizes of opposition parties; the larger this value is, the more parties of similar size exist. I call this variable the “Effective Number of Opposition Parties (ENOP).” Figure 3.1 shows the distribution of this variable. Figure 3.1: The Distribution of the ENOP I l J I l l l 25- Frequency LI] 0‘— \l—i ENOP To see whether the degree of opposition fragmentation changes the effects of unemployment on the results of the elections, I created the multiplicative interaction term between the unemployment rate and opposition fragmentation (Unemployment X 4Since the sources of election results usually lump small parties into an “other” category, we cannot precisely know what parties received how many seats. I used the method recommended by Taagepera (1997) to approximate the effective number of parties from incomplete data. 22 ENOP) and added this to the model. To be sure, the number and relative sizes of opposition parties do not solely determine the extent to which the opposition camp can attract the voters who are dissatisfied with the incumbent government. As I discussed earlier, ideological po- larization of opposition parties should also be important: swing voters would be less likely to vote for parties of extreme ideology. However, as Sartori (1976) argued while contrasting limited multipartism and extreme multipartism, the numerical as- pect and the ideological aspect of party fragmentation would surely be correlated; and Ware (1996, 173) presents empirical evidence for this thesis. Hence I believe that the variables described above would be enough to capture the effects of the frag- mentation of opposition parties on the relationship between economic performance and election results. The previous vote percentage of the governing party (parties) is included in the model as a control variable (Previous Vote), as is common in the literature. A negative coefficient is expected for this variable because if the governing parties received a large share of the votes in the previous election, they have a large “base” of votes to lose in the next election. Another control variable is whether the government controls 50% of the seats in parliament (Majority Status). This is a dummy variable that takes the value of one if the cabinet before the election is a majority cabinet and zero if it is a minority cabinet. The logic is that voters may not punish the government that does not control the parliament. The natural log of median district magnitude (DistMag) is also included in the model, and this variable controls for the effects of electoral systems. Since the sample of parliamentary elections consists of repeated observations of many countries, the assumption of spherical errors of the Gauss-Markov theorem will not probably hold. In particular, the presence of “panel heteroscedasticity” is highly suspected because the degree to which the electoral outcomes are predictable 23 should differ between countries. In addition, serial correlation within each country may also exist. To solve these problems, I employed the cross—national time-series model using panel corrected standard errors (PCSE), and the Prais-Winsten method takes care of serial correlation (Beck and Katz 1995). 3.4 Results from Separate Samples Table 3.1 shows the results of the regression analysis. Models (1) and (2) do not include EN 0F and the interaction term and thus evaluate the un-interacted effects of the unemployment rate on the electoral performance of governments. Model (1) is for the “High Clarity” countries, and Model (2) for the “Low Clarity” ones. Consistent with Powell and Whitten’s (1993) findings, the unemployment rate has a significant and negative impact on the dependent variable in Model (1) but not in Model (2). In Models (3) and (4), ENOP and the interaction term (Unemployment x ENOP) are included. Since the interactive effect is specified in the models, the estimated effects of the economic variable and its significance have to be interpreted with the value of the EN 0P variable. For example, the coefficients of Unemployment and Unemployment x ENOP are -0.803 and 0.326, respectively, in Model (3). Hence, the slope coefficient of Unemployment is: —0.803 + 0.326 x ENOP Since higher values of ENOP mean more fragmented opposition, this expression shows that the tendency for high unemployment rates to lower the vote share of the governing parties is intensified when opposition parties are less fragmented, as predicted by my hypothesis. For example, when there is only one opposition party (EN OP 2 1), this expression reduces to —0.803, which means, when the 24 Table 3.1: The Determinants of the Governments’ Vote Change (Separate Samples) Dependent Variable: Government's vote change (%) Model (1) (2) (3) (4) Clarity High Low High Low Unemployment -O.231 * -0. 150 -0.803 -0.419 (0.116) (0.121) (0.418) (0.342) ENOP -O. 731 -0. 895 (0.846) (1.051) Unemp. x ENOP 0.326 0.102 (0.222) (0.114) Previous Vote -O.141"‘ -0.286** -0.162** -O.294** (0.057) (0.059) (0.059) (0.059) Majority Status -0.413 2.640 0.175 2.150 (1.294) (1.434) (1.346) (1.536) District Magnitude 0.162 -O.627 0.161 -0.539 (0.458) (0.372) (0.464) (0.372) Intercept 4.945 12.610“ 6907* 15.353" (2.554L (2.928) (3.507) (4.543) N 72 62 72 62 p -0. 184 -0.048 -0. 158 -0.044 R-squared 0.131 0.336 0.159 0.345 Panel corrected standard errors are given in parentheses. * significant at 5% level; ** significant at 1% level unemployment rate increases by 1%, the vote share of the government is expected to decrease by -0.803%. On the other hand, when ENOP is 3, the slope coefficient is: —0.803 + (0.326 x 3) = 0.175. The significance of the slope coefficients of the economic variables also varies with ENOP, and the standard error of the unemployment variable in Model (3) is calculated in the following way (Friedrich 1982): .\/0.174 + (0.049 x ENOP2) + (2 x ——0.089 x ENOP) where 0.174 is the variance of the coefficient of Unemployment, 0.049 is the variance 25 Figure 3.2: Conditional Effects of Unemployment (Separate Samples) High Clarity Countries Low Clarity Countries 2 J l l l I l p 2 1J 1 EL . l l *- 1.5~ ~ 1.5~ ,g 8 1 . m E ‘3 R .5 ~ 8 t: ._ 1:: .9 :5 3:: ‘ 8 E U U . ‘ -1.’ T j I I l 2 3 4 5 of the coefficient of the interaction term, and -0.089 is the covariance between these two coefficients. Since both the coefficient and the standard error of the unemployment vari- able change with the value of ENOP, graphical presentation is useful in evaluating the effect and the significance of this variable. The two graphs in Figure 3.2 show them. The left panel is for the high clarity countries (Model (3)), and the right panel for low clarity countries (Model (4)). The solid lines represent the coefficient of the unemployment variable, and the dotted curves show the 95% confidence in- tervals. Hence, where the zero line does not fall within the confidence interval, the unemployment variable is statistically significant. The left panel of Figure 3.2 shows that the smaller ENOP is, the more votes the government loses due to a certain level of unemployment. Also, when EN OP is smaller than 1.8 (relatively unified opposition), the unemployment rate has a signif- icant and negative impact on the vote change of the incumbent parties; and when the opposition is more fragmented, unemployment does not affect the dependent variable significantly. The right panel of Figure 3.2 shows a much flatter line of the 26 coefficient. That is, in the low clarity context, ENOP does not change the effects of unemployment on the dependent variable much. Further, the unemployment vari- able is insignificant in the entire region. In other words, unemployment does not matter when the clarity is low, regardless of the value of EN OP. This result supports my hypothesis. 3.5 Results from the Pooled Sample In the above analysis, the sample cases were divided into two groups based on the dichotomous classification of the clarity of responsibility. Yet, since the original “clarity” variable by Powell and Whitten (1993) is not a dichotomous but a continuous variable, we can probably analyze the sample observations without dividing them into two groups. Table 3.2 shows the results of the regression analysis of the pooled sample. All 134 cases are analyzed in this model. Two interaction terms— Unemployment x ENOP and Unemployment x Clarity—are included in the model, and thus the coefficient and the standard error of unemployment depend on the two variables, ENOP and Clarity5 Figure 3.3 shows how the effect and the significance of unemployment change according to the values of ENOP and Clarity. In the left panel, Clarity is held at its mean (1.61) and the x-axis represents ENOP. The graph shows that when the opposition is unified, the government loses more votes due to a certain level of unemployment. When EN OP is smaller than 2.5, unemployment has a negative and significant impact on the dependent variable. This finding is interesting because the Clarity is set at its mean. Hence, this result suggests that economic voting takes place if the opposition is unified, and the clarity of responsibility does not have to 5In the original “clarity” variable, smaller values indicate higher clarity. In my analysis, I flipped the values for easier interpretation. 27 Table 3.2: The Determinants of the Governments’ Vote Change (Pooled Sample) Dependent Variable: Government's vote change (%) Model (5) Unemployment -0.442 (0.293) ENOP -0.773 (0.653) Unemp. x ENOP 0.120 (0.085) Clarity -O.874 (0.633) Unemp. x Clarity -0.010 (0.082) Previous Vote -0.209** (0.043) Majority Status 0.381 (0.999) District Magnitude -O.380 (0.336) Intercept 1 1.877" (3.330) N 134 p -0.091 R-squared 0.226 Panel corrected standard errors are given in parentheses. * significant at 5% level; ** significant at 1% level 28 Figure 3.3: Conditional Effects of Unemployment (Pooled Sample) 141 4 I l J— 1“ l l 1 l ‘5 .5: 8 .5- 6) d.) a a s s 8 S -.5‘ U U -11I I I I I '- -1-4' I I I I l 2 3 4 5 1 2 3 4 5 ENOP Clarity Clarity is held at its mean (1.61) ENOP is held at its mean (2.28) be particularly high. In the right panel of Figure 3.3, ENOP is held at its mean (2.28) and the x-axis represents Clarity. It is shown that the higher the Clarity is, the more votes the government loses; but the slope is quite flat, and the effect is not significant in most of the region. Thus, it appears that opposition fragmentation is more important than the clarity of responsibility in determining the economic effects on the government’s electoral fortune. 3.6 Conclusion The analyses presented in this chapter have shown that when the clarity of responsibility is high and the degree of opposition fragmentation is low (New Zealand before the electoral reform and Britain, if we ignore the Liberal Democrats, would be representative examples), unemployment rates have especially strong effects on the electoral fortunes of the incumbent parties. I argued that a unified opposition camp would attract the voters who are dissatisfied with the government by showing a 29 clear alternative choice at the ballot box. On the other hand, fragmented opposition parties do not induce as much vote transfer even when the government is not popular. The desire to continue in government provides governing parties with a strong incentive to try to do what the citizens want, which is the mechanism by which democracies stay accountable to the people. The findings from this chapter suggest that how well this mechanism works would depend partly on the characteristics of opposition parties. Specifically, a government that is confronted by a unified opposi- tion camp will face a more realistic chance to lose power in the next election than a government surrounded by fragmented opposition parties. Indeed, Japan and Italy, where a single party had stayed in power for an exceptionally long time in the post- World War II era, are also the countries where the opposition parties were highly fragmented. Hence, the degree of opposition fragmentation would be an important characteristic of democracies that may influence the longevity of governing parties and also the accountability of governments. How democracies function has been explained mostly from the institutional systems perspective (e.g., electoral systems, presidential vs parliamentary systems, legislative procedures, and federal vs unitary system) and types of governments (e. g., partisanship, coalition cabinet or not, and minority cabinet or not). However, little attention has been paid to the role of opposition parties in democratic systems. The findings from this chapter imply that the characteristics of opposition parties are also likely an indispensable component in understanding the mechanism of democratic, governance. 3O 3.7 Appendix: Notes on the calculation of vote changes The data of the vote shares of the political parties were obtained from many sources, namely, Mackie and Rose (1991), Gorvin (1989), Day, German and Camp- bell (1996), Siaroff (2000), Ishikawa (1984), various issues of Keesings Record of World Events, Facts on File, Political Handbook of the World, Europa Year Book, and Electoral Studies, and several web sites such as Parties and Elections in Europe (http : //www . part ies-and-elections .de/indexe.htm1), Election Results Archive (http://cdp.binghamton.edu/era/), and Elections New Zealand (http://elections.catalyst.net.nz/) In calculating the changes in parties’ vote percentages, a complication comes from mergers and splits of parties. When two or more parties merge together or form an electoral alliance, the vote change in the next election is calculated by subtracting the sum of the vote shares of the old parties in the previous election from the new party’s current vote share. For example, in the 1985 election in Sweden, the Center Party and the Christian Democrats formed an electoral alliance and received 12.4% of the total votes. In the previous election in 1982, the Center Party and the Christian Democrats had obtained 15.5% and 1.9% of votes, respectively. Hence, the alliance’s vote change in 1985 is calculated by subtracting the sum of 15.5 and 1.9 from 12.4, that is, -5. In the cases of splits, the vote changes of the parties that experienced a split since the previous election are evaluated against the vote shares of the pre—split party in the previous election, which are then reduced according to the proportion of the legislators that left the party. For example, 5 members of the Freedom Party in Austria, which obtained 33 seats and 16.6% of the votes in the 1990 election, left 31 the party in 1993 and founded the Liberal Forum. In other words, a party with 33 seats split into two parties with 28 and 5 seats, respectively. Then, in the next election in 1994, the Freedom Party’s vote share was 22.5% and the Liberal Forum received 6.0%. Their vote changes are calculated as follows: Votes Votes Vote in 1994 in 1990 Change Freedom Party: 22.5 - 16.6 x (28/33) 2 8.4 Liberal Forum: 6.0 - 16.6 x (5/33) : 3.5 For a few cases, I was unable to obtain the exact numbers of the legislators who split the parties. In such cases, I tentatively used the vote shares of the post-split elections to adjust the vote levels of the pre-split elections. In the above example, instead of (28/33) and (5/33), {22.5/(22.5 + 6.0)} and {6.0/(22.5 + 6.0)} would be used if the information on the number of legislators at the split was not available. 32 Chapter 4: Party Competition in Regional Politics 4. 1 Introduction In the previous chapter, I demonstrated the relationship between the character- istics of opposition parties and the electoral performances of governing parties. In this chapter, I analyze the electoral data from subnational governments to explore how that mechanism works in the subnational level. Party politics in subnational governments is an under-researched field in polit- ical science. But since political parties compete not only in the national level but also in subnational levels, studying subnational politics would help us understand the mechanism of party competition better, particularly because of the large number of cases. If the argument provided in the previous chapter is really robust, it should also be supported by analyses of subnational party politics. Because of the difficulty in obtaining the subnational election data, the sample observations I analyze in this chapter are quite limited. I have 119 elections from Australia, 114 elections from Canada, and 162 elections from Germany. The more detailed description of the sample cases is found in the Appendix of this chapter. 4.2 Analysis 4.2.1 Cost of Ruling Before conducting multivariate analysis, let us first look at the distribution of the dependent variable. As mentioned in the previous chapter, it is known that the parties in power tend to lose votes in elections (“cost of ruling”) probably because their performance is critically evaluated by the citizens. Will the same pattern appear in the regional level? 33 Figure 4.1: The Distribution of the Percent Vote Change of the Governments (a) Subnational Elections (b) National Elections 66- r 1 I *1 1 r — 58 _ l l l r“! 1 l .— - i >3 - F >, - — O O s - — cr - » H 8 FJ-i Iii-r 0‘ r" O _ (:3 r: I D: _ -3'0 -2'0 -1T0 () 1b 2b -30 -20 -10 0 lb 20 Governments' Vote Change (%) Govemments' Vote Change (%) Figure 4.1 (a) is a histogram of the governments’ vote change in the subnational level, and Figure 4.1 (b) is from the national level for comparison.’ The two show a quite similar pattern, but the tails of the distribution is fatter in the subnational level, meaning that the electoral changes tend to be more dramatic in subnational elections than in national elections. Perhaps this is because regional governments are closer to the people and hence regional elections reflect new movements and dynamism in the society more sensitively than national elections, in which party competition may be more institutionalized and the barriers to new entrants may also be higher. The mean value is higher in the subnational level (-1.98) than in the national level (-2.44). However, the difference between them is not significantly different from zero (the respective standard deviations are 7.30 and 5.13). 1The data are. from 16 countries analyzed in the previous chapter. The result from the national elections in Australia, Canada, and Germany is comparable to the one presented here. 34 4.2.2 Opposition Fragmentation Next, let us examine the pattern of opposition fragmentation. As I did in the previous chapter, I use the Effective Number of Opposition Parties (ENOP) to measure the degree of opposition fragmentation. The distribution of ENOP in the subnational governments is shown in Figure 4.2. Figure 4.2 (a) is the sum of the three countries, and (b) to (d) are for Australia, Canada, and Germany, respectively. All three countries show a similar pattern of distribution, but Canada has the largest concentration at the left end of the spectrum. Average values are: 1.60 for all countries, 1.58 for Australia, 1.72 for Canada, and 1.53 for Germany.2 Figure 4.2 (e) is for the EN OP values of the national level elections in the three countries (average: 1.48).3 We can clearly detect that ENOP in the regional level is more diverse than the national level. This may also be because subnational party systems are less institutionalized than national party systems and thus the entries of new parties are easier in the subnational level. 4.2.3 The Effects of Opposition Fragmentation In this subsection, I examine the eflects of opposition fragmentation on the electoral fortunes of governing parties. As I argued in the previous chapter, a unified opposition would present a clear alternative to the voters; and hence we can expect that the votes transfer from the government to the opposition more smoothly if the opposition is unified than in the case where it is fragmented into many parties. In Chapter 3, I demonstrated that anti-incumbent economic voting is intensified if the opposition is unified. In the subnational level, the economic performance would not affect the election results as much as it does in the national level since most important economic 2Without the four outlying cases, the average value of Canada is 1.56. 3The graph for the national level elections of 16 advanced countries is found in Figure 3.1. 35 Figure 4.2: Effective Number of Opposition Parties (a) All 3 (Zombies ‘ (b) Australia 2: 5 6 7 AI 5 7' ENOP ENOP (0) Canada ((1) Germany 44].?...m._43._... L 41...... l3 . :‘s I: '5 6 7' ’5 i ENOP ENOP (e) 3 counhies' national level ’8"? I l I l I 3‘ 8 '5'“ i 2 m . 0, ilfh , 234567 ENOP 36 Figure 4.3: ENOP and the Governments’ Vote Change (a) Subnational Level Cb) National Level 204' l J_ l l l 1 20 I l A l l I Ir .a . a) . a) O E" 10",“; - ,5 10- o if“); . Ca), 93 0 ltéér; - - . *6 0 o ,1 v", . a :> ’.> ”5‘4"; , ’ 33 3404:»??- , . . . L $-10- 5 523;” g $-20.” ~ §-20: 3 . ' o o -30 ‘ I I I F I IT -30 I I I I I fl I 1 2 3 4 5 6 7 1 2 3 4 5 6 7 ENOP ENOP policies are made by the national governments. Also, it is difficult to obtain regional economic data. Therefore, in my analysis of subnational elections, my focus will not be on the interactive effects of the economy and opposition fragmentation but rather on the effects of opposition fragmentation itself. Figure 4.3 are the scatterplots of the ENOP variable against the percent vote change of the governing parties in both subnational (a) and national levels (b). Comparing the two graphs, it is noticeable that the variety along the y-axis changes with the level of ENOP in (a) but it is more constant in (b). Specifically, in (a), the smaller the ENOP value is, the larger the magnitude of vote change is. This is consistent with my expectation that a unified opposition magnifies the amount of the votes that transfer between government and opposition. In (b), on the contrary, EN OP does not seem to affect the degree of variation in the y-axis. Possibly this is due to the tendency that, as I demonstrated in Chapter 3, opposition fragmentation and economic performance interactively affect the election results in the national level, and hence the effects of EN OP do not appear in a bivariate graph. 37 Let us further examine the data of the subnational elections. Table 4.1 shows the regression results of the determinants of the percent vote change of the gov- ernments. As I did in Chapter 3, the cross-national time-series model with panel- corrected standard errors (PCSE) was used in this analysis. 377 elections in 33 subnational units are included in the model. Other than the ENOP variable, the following three control variables are included in Model (1). The first is the vote percentage of the governing party (parties) in the previous election (Previous Vote), which is the basis of the electoral change in the current election. Second, I added the variable of whether the government controls 50% of the seats in the parliament (Majority Status). Voters may not punish a government that does not control the legislature and cannot push through its policy plans. Third, the number of parties in government is included (#GovParties). The government’s responsibility would not be clear if the government consists of a coalition of parties. In addition, country dummy variables are added in Model (2). The two Models show quite similar results. The majority status and the num- ber of governing parties are statistically significant. A government is expected to perform well in elections if it is a minority government and is made up by many parties. EN OP does not appear to have any systematic effects on the dependent variable; but it comes as no surprise because we have no theoretical reasons to ex- pect that opposition fragmentation by itself increases or decreases the votes for the incumbent. Figure 4.4 is a partial plot of the dependent variable and ENOP, which shows the relationship between the two variables after the effects of all other variables are controlled. It is based on Model (1), but the result from Model (2) is quite similar. In this graph, the points located far from zero along the y-axis are the observations that were not well predicted by the statistical model, and it is clearly seen that those observations are found in the left side in the graph. That is, the graph tells 38 Table 4.1: Determinants of the Governments’ Vote Change Dependent Variable: Govemment’s vote change (%) Model (1) (2) ENOP -0.025 -0.014 (0.570) (0.568) Previouos Vote 0069 -0.078 (0.050) (0.052) Majority Status 2531 * -2.532 * (1.092) (1.102) #GovParties 2.478 ** 2.239 ** (0.727) (0.792) Canada 0061 (1.020) Germany 0.636 (0.932) Intercept 0.539 1 .032 (2.392) (2.486) N 377 377 p 0.006 0.003 R-squared 0.049 0.050 Panel corrected standard errors are given in parentheses. * significant at 5% level; ** significant at 1% level 39 Figure 4.4: Partial plot of ENOP and the Governments’ Vote Change 00 o omfifio 20> $859850 .20 -1 I4 40 us that, even after the effects of other variables are controlled, when the opposition is unified, the vote change of the government tends to be large. In other words, for whatever reason the incumbent government wins (loses), its gain (loss) is magnified if it is confronted by a unified opposition. 4.3 Conclusion The analysis presented in this chapter supports the basic conclusion of the previous chapter. When the opposition camp is fragmented into many parties, the voters’ evaluation of the incumbent government would not be reflected in the election outcome as clearly as it would when there is a unified opposition. The choice at the ballot box is more straightforward if the voters can see a clear alternative in the unified opposition, as opposed to fragmented opposition parties that usually have particularistic policy agendas and stable support bases. In this chapter, I showed that, in subnational elections, a unified opposition camp makes the electoral gain or loss of the governments larger. If a government wins for any reason, it tends to win by a lot if the opposition is unified. At the same time, the vote loss of a government is also amplified if there is a unified opposition. The degree of opposition fragmentation by itself does not increase or decrease the votes for the governing parties. In this and the previous chapters, I argued and demonstrated with empirical evidence that the degree of opposition fragmentation affects the pattern of party competition and the fate of governing parties in parliamentary democracies. In the next chapter, the focus is shifted to the determinants of opposition fragmentation. 41 4.4 Appendix The following is the list of the subnational units included in the analysis (the number of elections is in parentheses): 0 Australia (states & territories): Australian Capital Territory (5), New South Wales (18), Northern Territory (7), Queensland (20), South Australia (18), Tasmania (17), Victoria (18), and Western Australia (16) 0 Canada (provinces): Alberta (23), British Columbia (24), Manitoba (2), New- foundland and Labrador (15), Nova Scotia (12), Ontario (4), Prince Edward Island (7), Quebec (4), and Saskatchewan (23). 0 Germany (Lander): BadenWiirttemberg (11), Bayern (14), Berlin (13), Bran- denburg (3), Bremen (14), Hamburg (16), Hessen (15), MecklenburgVorpom- mern (3), Niedersachsen (14), NordrheinWestfalen (12), Rheinlandealz (13), Saarland (11), Sachsen (3), SachsenAnhalt (3), SchleswigHolstein (14), and Thiiringen (3). The temporal ranges vary across countries and subnational units due to the varying difficulties in obtaining the election data. The oldest election included is the 1903 election of the province of British Columbia (Canada), and the most recent one is the election in the Australian Capital Territory that took place in October 2004. The data of the elections and the cabinet composition were obtained from the following sources: 0 Australia. http : //elect ions .uwa.edu.au/e1ectionsearch. lasso 0 Canada. O’Handley (2003), http://www.swishweb.com/Politics/ Canada/,http://www.electionsalberta.ab.ca/pastelections.html, 42 http://www.elections.bc.ca/elections/elections-results.htm, http://www.electionsmanitoba.ca/, http://www.cbc.ca/manitobavotes2003/,http://www.gov.f1.ca/ elections/electionarchive/1999/provincia1/stats.asp, http://www.gov.ns.ca/elo/elections/OBstats/votes.pdf, http://www.electionsontario.on.ca/en/home;en.shtml, http://www.gov.pe.ca/photos/original/generalstat866t.pdf, http://www.e1ectionsquebec.qc.ca/en/res_generales.asp, zuulhttp://www.e1ections.sk.ca/history.php Germany. Andersen and Woyke (2000), http: //www.wahlrecht .de/ ergebnisse/,and.http://www.e1ection.de/ 43 Chapter 5: Institutional Rules and Opposition Fragmentation 5. 1 Introduction In September 2003, the second largest opposition party in Japan, the Liberal Party, joined the largest opposition party, the Democratic Party of Japan (DPJ); and the enlarged DPJ further increased its seats dramatically in the general election in November. In Canada in December 2003, the two largest opposition parties, the Progressive Conservative Party and the Canadian Alliance, merged together to form a new party, the Conservative Party of Canada, which also increased its number of seats in the June 2004 general election although it failed in winning power. In both instances, the intention of the mergers was to compete with the govern- ing parties—the Liberal Democrats in Japan and the Liberals in Canada—both of which had been continuously in power for about 10 years. If opposition parties really seek to compete with the government and win power, it may be a rational action for them to merge together. However, we do not often observe such events in real politics; and in some countries, opposition parties are highly fragmented while the ruling party (parties) continues in office. What, then, makes this difference? This chapter explores the patterns and determinants of the evolution of the opposition party systems. Although there is a large literature on party system change, many of the empir- ical studies have focused on whether the party system has been “frozen” or volatile (e.g., Lipset and Rokkan 1967a; Pedersen 1983; Bartolini and Mair 1990), but not much attention has been paid to how the party systems change or evolve. To be sure, there are researches on the rise and decline of specific types of parties;1 yet, 1See, for example, Jackman and Volpert (1996) and Knigge (1998) on extreme right parties 44 the evolution of party systems in a general framework, not to mention the evolu- tion of the opposition party system which I will examine in this paper, is greatly understudied. I argue that this is an often-neglected but politically important re- search question. As I demonstrated in previous chapters, a united opposition party has a better chance to bring down the governing party than fragmented opposition parties. Hence the behavior of opposition parties and the degree of opposition frag- mentation partly determine how long the governing parties can remain in office and how frequently alternations in power take place. 5.2 The Measurement of the Opposition Party System Evolution Since the concept this chapter deals with—the evolution of the opposition party system—is quite new, in this section I explain how it is operationalized so that the following discussion on what factors determine it will be clearer. First of all, I should note that the political systems in which the cabinet composition changes frequently are not considered in this analysis. In these systems, the parties often move in and out of government and change their status, and the behavioral patterns and strategies of all parties may be able to be analyzed in a common framework. My interests are rather on the systems where governing parties and opposition parties are more clearly separable, and this analysis focuses on how the opposition party system changes while the same government continues in office. Operationally, the cases where the cabinet composition changed between two elections are removed from the sample, and I focus only on the cases in which all opposition parties continuously stayed in opposition during the whole electoral cycle. and Kitschelt (1988) and Franklin and Rudig (1995) on environmental or new-left parties. Also, there are studies on the party system changes that stem from electoral law reforms (Shugart 1992; Lijphart 1994; Barker and McLeay 2000; Reed 2001). 45 Figure 5.1: Calculation of AENOP Case A Case B Case C I: t-I-l t t+l I: t+l 30 PartyG o -Gov't-)I30 -Gov't-)l40 /° PartyA 3,00 ‘GOV'I'N 60 PartyD Z/O ............ o PartyH 30 43°“? 20 p t B 20 """""" 40 30 ary 3U partyE 20 PartyI 30 . . 10 PartyC 20 10 PartyF 10 ........... 20 Party] 10 10 PartyK AENOP =-0.4 AENOP = 0 AENOP = 0.67 The change of opposition fragmentation is calculated by taking the difference between (I) the effective number of opposition parties (ENOP), measured by the vote shares, at t and (2) the effective number of the parties that were in opposition between t and t + 1, measured by their vote shares at the election t+ 1.2 Note that it does not matter whether or not some of the oppositions during the electoral cycle between t and t + 1 become governing parties after t + 1. This variable is denoted as AE N 0P. Hypothetical examples may clarify this matter. Three possible cases are pre- sented in Figure 5.1. In Case A, Party A received 60% of votes in the two consecutive elections at t and t + 1. It was the governing party between t and t+ l, and it suc- ceeded in remaining in office after t + 1. In this case, AENOP is calculated by measuring the change in the level of fragmentation of the parties that stayed in opposition during the electoral cycle, that is, Parties B and C. At t, both E and C 2If new parties appear at t + 1, they are also taken into calculation. 46 received 20% of votes. The level of fragmentation (the ENOP at t) is: 1 : 2 20 . 20 (:20+'20)‘2 + (20+20)2 At t + 1, they obtained 30% and 10% of votes, respectively. The level of fragmen- tation is: 1 = 1.6 30 10 (30+10)2 + (30+10)2 Hence, AENOP is 1.6 — 2 = —0.4. A negative value indicates that the opposition became less fragmented during the electoral cycle. Some voters may have abandoned Party C and switched their support to Party B. In Case B, the ruling Party D decreased its vote share from 70% to 40%. Two opposition parties, E and F, increased theirs from 20% to 40% and 10% to 20%, respectively. In this scenario, although the vote shares of the opposition parties changed from t to t + 1, the ratio of E to F stayed the same, and the level of fragmentation remained at 1.8. AENOP is: l l 402_202— 202 10 (40+20) I (40+20) (20+10) +(201L10)2 :0, which means that the level of opposition fragmentation did not change during this electoral cycle. Note that it does not matter whether either—or both—of Parties E and F became governing parties after t+ 1 since my dependent variable is calculated by the sizes of the parties that were in opposition between t and t + 1. In Case C, a coalition cabinet of Parties G and H was in power during the electoral cycle. There were two opposition parties, I and J, both of which received the same vote shares in t + 1 as they did in t. However, a new party, K, entered the competition in t + 1 and received 10% of votes. The emergence of a new party shows that the opposition became more fragmented during the electoral cycle. In 47 this situation, AENOP is calculated as: l I — 2 0.67 30 2 10 10 30 10 (30+10+10) + (30+10+10)2 + (30+10+10)2 (304.10)2 + (304(10)2 In my sample, which will be explained in detail later, 89% of the AENOP values fall within the range of -1 and l. The mean value is —0.058, and the median is 0.017. In 53% of the cases, the opposition became more fragmented (positive AE N OP) while it became less fragmented (negative AE N OP) in 47% of the cases. 5.3 Institutional Determinants of the Opposition Party System Evolution AE N OP deviates from zero in two types of situations. The first type is when the relative sizes of opposition parties change due to a new election. Specifically, AE N 0P takes a positive value when larger opposition parties lose votes and become smaller and / or smaller ones become larger; and it takes a negative value when larger opposition parties become even larger and/or smaller ones become even smaller. The second type is when the identity of the parties change. Specifically, AEN OP is positive when an existing opposition party splits or a new party emerges; and it is negative when existing opposition parties merge together or a party disappears. I argue that institutional rules make differences in the behaviors and strategies of parties and thus the opposition party systems. In particular, I focus on the effects of electoral systems and parliamentary rules in the following discussion. Both significantly affect the environment in which parties exist and compete. Electoral systems determine the difficulty for small parties to survive. Where the district magnitude is small or the electoral threshold is high, small parties have difficulty in getting represented and receive disproportionately fewer seats than their 48 vote shares. This “punishment” effect which small parties have under dispropor- tional electoral rules has been studied extensively since Duverger’s ( 1954) seminal work. In such an environment, life is tough for minor parties, and the benefit of being a major party is large. As a result, defections from established parties are deterred, and mergers of parties are facilitated. Also, minor parties may fail to pass the threshold in elections and die out. Furthermore, entries of new parties are diflicult if the threshold is high. Therefore, it is expected that the degree of opposition fragmentation becomes smaller or stays small where the electoral system is disproportional. On the other hand, in proportional electoral systems, the survival of small parties and the entry of new parties are relatively easier. Since the environment is lenient for small parties, splits of parties may be seen more than in the systems with disproportional electoral laws. Thus, the values of AE N OP would be larger where the electoral system is more proportional. Parliamentary rules also affect the evolution of the opposition party system by influencing the strategies of opposition parties. In some countries such as Britain, the ruling party dominates the legislative process, and opposition parties have little influence on policy making. In other systems, on the other hand, the parliaments are more deliberative in that opposition parties have influence on policies through the legislative process. These parliaments have various rules that promote deliberation in policy making—for example, committees have strong powers; committee chairs are proportionally distributed to parties; and the upper house exists and has real powers (see, for example, Strom 1990). If opposition parties can influence policies to some extent, they may be able to please their supporters and secure a support base by partially realizing their promises. Hence, staying in opposition would not give an immediate threat to their existence. On the contrary, opposition parties that cannot influence policies would 49 feel stronger pressure to take over the government because staying in opposition gives nothing to the party and the supporters. When the opposition parties seriously seek to win power, they may consider forming an electoral alliance or even merging together, thus decreasing the degree of fragmentation. Under such less deliberative parliamentary rules, the survival of small opposition parties is especially difficult because voters may not think that these parties have realistic chances to win power and thus support instead for major opposition parties. As a result, similar to the cases with disproportional electoral laws, splits of parties are deterred, fusions of parties are facilitated, and entries of new parties are discouraged. All of these lead to lower values of opposition fragmentation. If the above arguments are correct, the following hypotheses should be sup- ported empirically: Hypothesis 1: The opposition tends to become less fragmented (small AENOP values ) if the electoral system is disproportional. Hypothesis 2: The opposition tends to become less fragmented (small AEN OP values) if the parliamentary rules do not allow opposition parties to influence policy making. 5.4 Empirical Analysis 5.4.1 Sample Cases The empirical test of the hypotheses is conducted with the sample from 16 industrial parliamentary democracies.3 The units of analysis are electoral cycles, and the observation period is between 1945 and 1997 due to the availability of the data. The sample consists of 155 observations. As discussed above, the dependent variable 3Australia, Austria, Canada, Denmark, Finland, Fiance, Germany, Greece, Ireland, Japan, New Zealand, the Netherlands, Norway, Spain, Sweden, and the United Kingdom. 50 is the change of the degree of opposition fragmentation, denoted as AE N OP.4 5.4.2 Independent Variables (1) Electoral Systems The electoral system characteristics are measured by two variables, which are also used in Amorim Neto and Cox (1997). One is the median district magnitude in the lowest electoral tier (MedMagnitude). The larger this variable is, the more proportional the electoral law is. The other is the percentage of seats allocated in the upper tier(s) (Upper). Large upper tier districts generally enhance proportionality of the electoral system (Li jphart 1994). The data for electoral systems were obtained from Golder (2004). (2) Parliamentary Rules The characteristics of parliamentary rules in terms of opportunity for opposition in- fluence are quantified by two variables created by Powell (2000). The first variable represents the nature of the committee system (Committee). This is an index vari- able that combines two factors: strength of the committee system in legislation and whether the ruling parties monopolize committee chair posts or major opposition parties receive some share of them. This variable varies from 0 to .25, and a higher value indicates that the committee system allows more influence of the opposition. The second variable is the opportunity of opposition parties in bargaining with the government (Bargaining). This variable reflects such information as majority status of the government, the control of the important upper house by the oppo- sition, and whether a two-thirds majority is required for legislation. A high value in this variable indicates that opposition parties have a better opportunity in ne- gotiating with the government for policy decisions, and a low value means that the government dominates the policy making process. This variable varies from .1 to .5 4The data sources are the same as the ones reported in Chapter 3. 51 and does not take zero because the opposition can at least “use the legislative forum to try to shape the government’s actions by arousing public opinion” (Powell 2000, 105). (3) Control Variables The following control variables are also included in the statistical model. First, the ENOP after the previous election (lagENOP) is added. Since our dependent variable is the change in the ENOP values, the base value of the ENOP should be controlled. Second, the number of governing parties during the electoral cycle (#GovParties) is included. Whether the cabinet consists of a single party or a coalition of parties may affect the strategies of opposition parties since opposition parties may think that they have a better chance to join the government if a coalition cabinet is in power. Third, I controlled for the change of the vote shares for the governing parties from the previous to the current election (A Gov Votes) because whether the opposition parties increased or decreased their total size may make a difference in their fragmentation level. Fourth, the number of previous elections the governing party (parties) has won (PrevTerms) is included. The incentives and strategies of opposition parties would be different depending on how long they have been out of office, and this variable captures the impacts these differences make. Fifth, I added the variable of whether the governing parties controlled a majority of the parliament (Majority). If a minor- ity cabinet is in office, opposition parties may feel that they may be invited to join the cabinet, and thus their strategies may be aflected. Finally, the length of time, measured in months, between the previous and the current elections (EleCycle) is controlled. 52 5.4.3 Statistical Model Since the 155 observations are drawn from 16 countries, the assumption of in- dependence of the observations may not hold; cases within countries are probably more similar than cases across countries (e.g., some countries have stable party sys- tems while others do not). The violation of this assumption will lead to inaccurate estimation of the standard errors. I obtained the robust standard errors by employ- ing the clustering method in which I treated the observations within a country as a group. The panel method is not appropriate in this analysis because, as noted earlier, the observations in which the composition of the ruling parties changed dur- ing the electoral cycle are dropped from the sample. Consequently, the observations within a country are not necessarily serially ordered but rather sporadic, and thus my sample cases do not have the shape of a panel data set. 5.4.4 Results The results of the analysis are presented in Table 5.1. The variables of Med- Magnitude and Bargaining are found to have significant and positive effects on the dependent variable, which is consistent with my hypotheses. Substantively, this means that the opposition party system tends to become more fragmented if the district magnitude is large (a proportional electoral system) and the opposition par- ties have a bargaining power against the government in legislation. The variable of the strength of the committee system in the parliament (Committee) has a positive coefficient as the hypothesis predicts. Although its effect is not statistically signif- icant at the .05 level, its p-value is 0.055, which is quite close to the significance level. The Upper variable does not appear to have any systematic effects on the dependent variable. The ENOP after the previous election (lagENOP) has a negative and signifi- 53 Table 5.1: Regression Analysis Dependent Variable: The change of opposition fragmentation MedMagnitude 0.003 * (0.001) Upper -0.001 (0.004) Committee 1.625 (0.783) Bargaining 1.211 * (0.480) lagENOP -0.238 ** (0.055) #GovParties -0.290 ** (0.053) AGovVotes -0.048 ** (0.014) PrevTerm 0.057 * (0.021) Majority 0.161 (0.199) EleCycle -0.009 * (0.004) Intercept 0.459 * (0.192) N 155 R2 0.362 SER 0.579 Robust standard errors are given in parentheses. * p < 0.05, ** p < 0.01. 54 cant effect. This result is quite natural because if the opposition is already unified, it is unlikely to become more unified. The number of parties in government (#Gov- Parties) has a negative effect, meaning that the opposition tends to become less fragmented if the cabinet is made of many parties. The change of the vote shares of the governing parties (A GovVotes) is negatively associated with the dependent variable. The PrevTerms variable has a positive coefficient that is significant. This means that the longer the same government has been in power, the more fragmented the opposition becomes. The majority status of the government (Majority) is not significant. The length of time between the previous and current elections (Ele- C'ycle) has a negative effect, meaning that the opposition is likely to become more fragmented if the time between the elections is short. 5.5 Conclusion My analysis demonstrated that institutional rules affect the evolution of the opposition party system. Specifically, when the electoral system is disproportional and the parliamentary rules do not allow opposition parties to influence the policy making process, the opposition tends to become less fragmented. Since the degree of opposition fragmentation affects the electoral fortune of governing parties, as I demonstrated earlier in this dissertation, the findings of this paper suggest that institutional rules indirectly influence the length of time the governments stay in office. When a disproportional electoral system and less delib- erative parliamentary rules are used, the opposition party system is likely to evolve into a more unified one, through the fusion of parties and/or the disappearance of minor parties. A unified opposition, then, poses a stronger threat to the govern- ment in elections, and the tenure of the government may be shortened. On the contrary, proportional electoral systems (under which the survival of small parties 55 is easier) and deliberative parliamentary rules (in which opposition parties can af- fect the policy decisions without becoming a governing party) facilitate opposition fragmentation, which makes the re-election of the government easier. If opposition parties are allowed to influence policy making, it may seem as if the government is doing a favor to the opposition. However, the findings from this research imply that allowing opposition influence may in fact help the government’s own survival. For ruling parties that wish to continue in office, it is of course important to keep their popularity high; yet it is also beneficial to them if the opposition parties do not seriously seek to win power. I argued in this chapter, with supportive empirical evidence, that opposition parties that are deprived of political influence will be relatively more serious in winning power than the ones that can affect policies and partially achieve their electoral promises. It may be a government’s rational action to “tame the opposition” by letting the opposition parties participate in the policy making process. Whether opposition parties are “tamed” and settle as opposition or they se- riously compete with the government in pursuing power should make significant differences in the functioning of democracy. When oppositions do not pose serious threats and the ruling politicians do not feel a realistic chance of being thrown out of office, the governmental accountability may decline. Thus, the existence of strong and credible opposition parties would be an indispensable component of democratic politics. In the next chapter, I will look at an example of a weak opposition party that failed to pose threats to the ruling party and how institutional rules shaped its behaviors. 56 Chapter 6: Position Change of Opposition Parties: The Case of the Japan Socialist Party 6. 1 Introduction In the previous chapter, I examined the mechanism of opposition fragmentation. As I argued in earlier chapters, when many opposition parties exist, voters may not see the opposition camp as a credible alternative to the incumbent government and hence fragmented oppositions may make the life of governments longer. Another possible situation in which opposition parties help governments stay in power is when opposition parties have ideologies that are too radical or extreme for moderate voters to accept. In such situations, similar to the cases of fragmented opposition, voters who are not satisfied with the government and are looking for an alternative would not be attracted by those radical opposition parties. Why do some opposition parties cling to extreme policy positions that are disadvantageous in pursuing power? This is the question I now turn to ask. In this chapter, I analyze the case of the Japan Socialist Party (JSP) as an example of opposition parties that embraced radical ideologies. Japan is a well- known case where a dominant. party—the Liberal Democratic Party (LDP)—stayed in power for an exceptionally long time. The LDP is a center-right party that was established in 1955 by the merging of the Liberal Party and the Democratic Party. Since then, the LDP continuously governed the country until 1993, almost always without a coalition partner. The LDP has also been in power since 1994, usually with coalition partners. During the LDP dominance from 1955 to 1993, the JSP was always the largest opposition party. It is now widely accepted that the JSP’s failure to moderate its strong leftist policy position was one of the factors that made the LDP’s long 57 tenure possible (see, for example, Kohno 1997). There have been several explana- tions of why the JSP failed to become a less radical, center-left party. Perhaps the most widely accepted explanation is the organizational dependency hypothesis (e. g., Flanagan 1984 and references cited therein). In this view, the JSP did not abandon its leftist policies because it heavily relied on the leftist labor unions for votes and funds Another explanation focuses on the JSP’s disastrous experience in the late 19403. After participating in two coalition governments with two conservative par- ties between 1947 and 1948, the JSP suffered a devastating setback in the 1949 election, losing two-thirds of the seats it gained in the previous election. Stockwin (1992, 87) argues that this experience brought about “a steady shift in the balance of power [within the party] from the Right, which was initially dominant, toward the Left.” The memory of losing supports after cooperating with conservative parties, according to this view, kept haunting the party, making it reluctant to lean toward the middle of the ideological spectrum (See also Tani 1992, 80-84). Yet another explanation emphasizes the role of the Japanese Communist Party (JCP). Kohno (1997) argues that the JSP’s position change was difficult since its ideological posi- tion was located between the JCP on its left side and a few small centrist parties on its right side. I admit that each of these explanations is plausible to some extent. Yet there are two points about which I am not content. First, the existing explanations tend to be abstract because they try to account for the process of 30 or 40 years as a whole. As I will discuss later, there was a crucial period in the early 19603 during which the party underwent a severe internal conflict; and the decisions made at that time subsequently restricted the behavior of the party for decades. In my analysis, I will focus on this crucial period and provide a concrete account of the JSP’s ideological immobility. 58 Second, the previous approaches have not paid much attention to the individual actors’ behaviors in determining the party’s policy positions. Instead, many studies have seen the J SP as a coherent single entity that has its own preference and strategy. This “unitary actor assumption” is indeed useful depending on the circumstances; yet, as will be shown, there were instances of fierce internal conflicts within the J SP over the party platform, and individual legislators were the important actors during the conflicts. I contend that the behavior of a party is an end result of its internal politics, and I analyze the JSP’s behavior from the individual level. My argument is that Japan’s electoral system gave an incentive to some socialist politicians to wish their party not to become larger. Strangely, there were incumbent JSP members who were better off if the party had failed to become more popular, and hence they had the incentive to resist the proposed change of the party platform. How could this ever happen? 6.2 The Policy Position of the J SP and the Leg- islators’ lRe-election Quest 6.2.1 A Short History of the JSP and its Platforms Like many leftist parties in the world, the JSP has a history of internal debates and fights between those who embrace the pure Marxist ideal and those who have more moderate and realistic ideas. The JSP was established in 1945, soon after the end of World War II, and participated in two coalition cabinets during the early years of the post-war politics (1947-48). In 1951, the party split into two: the Leftist JSP and the Rightist JSP. The two re-united in 1955—the same year as the LDP was established by the merger of two conservative parties. Since then, the LDP and the JSP continued to be the 59 two largest parties for four decades, and the LDP continuously stayed in power and the JSP in opposition until 1993—~this party system came to be known as “the 1955 regime.” Even after the re-unification, the internal rivalry within the JSP between the leftists, who seek to pursue the Marxist ideal and refuse to compromise their policy position, and the rightists, who take a more realistic attitude, was always intense. In 1959, only four years after the re-unification, a rightist faction led by Suehiro Nishio left the party and established the center-left Democratic Socialist Party (DSP).l In the early 19603, as I will discuss in detail shortly, there was an attempt to moderate the party’s platform (the “Structural Reform” plan); but it failed after an intense intra-party conflict, and in 1964 the party adopted a hardliner leftist document “The Road to Socialism.” Once adopted, “The Road” strongly bound the behavior of the party (Mori 2001). While Japan’s economy and the people’s living standards were rising rapidly under the LDP governments and the market economy, the JSP kept criticizing the capitalist regime and arguing for transforming the country into a socialist system. The JSP basically acted as an anti-system party that opposed everything the government did; and it rarely constructed its own policy proposals or alternative ideas to the government bills. “The Road” was finally discarded in 1986, and the party increased the number of its legislative seats in the late 19803 when the LDP government was unpopular for a tax reform plan and corruption scandals. However, once several new opposi- tion parties were established and started competing in elections in the early 19903, the support for the JSP dramatically declined. It participated in several coalition cabinets during the 19903 while becoming smaller and smaller. In the 2003 lower house election, the JSP gained only six seats out of 480.2 Perhaps largely because of its unrealistically leftist platform, many voters did 1The DSP existed until 1994 and was usually the third or fourth largest party. 2The JSP changed its name to the Social Democratic Party in 1996. 60 not see the J SP as a credible alternative to the LDP but a party that should monitor what the LDP does and prevent the LDP from abusing the power. The studies by Inoguchi (1983) and Kabashima (1988) demonstrated the existence of the voters called “buffer players,” who want the LDP to stay in power but do not want the LDP ’3 seats to be far more than 50% of the total seats so that checks and balances would work. Those “buffer players” sometimes voted for the J SP or other opposition parties to show their dissatisfaction to the LDP government but did not want a J SP government. The votes for the J SP from those “buffer players” made the popularity of the JSP look far larger than the size of its real support base. The JSP was never seen by the public as a party that can run the government, and the LDP’s long dominance was thus made possible. 6.2.2 The Structural Reform Plan and its Demise Before “The Road” was adopted and started binding the party, there was one occasion in the early 19603 when the moderation of the party platform was on the agenda and seriously considered within the party. It was arguably the moment when the J SP came closest to crossing the Rubicon and becoming a party that could really compete for power. It was Saburo Eda who championed this issue. Eda was originally a leftist within the party and belonged to the Leftist J SP while the party was divided into two between 1951 and 1955. He became the party secretary (the number two position) in March 1960 and soon brought up the “Structural Reform” plan (hereafter referred to as SR) that sought to moderate the party. This plan is often compared with the position change of the German Social Democratic Party at the Bad Godesberg Conference in 1959 (e.g., Ishikawa 1984, 150; Soga 1989). Yet, the largest difference between the JSP and its German counterpart was, of course, that the JSP failed to adopt the SR and instead clung to the hardliner leftist position—and failed to take 61 over the government. Since Eda became the party secretary, he argued for the SR in various occasions, such as speeches at party meetings and articles published in magazines. For example, in a speech presented at a party meeting in July 1962, he praised “America’s high living standards, the Soviet Union’s thorough social welfare, Britain’s parliamentary democracy, and Japan’s pacifist constitution” and said these things should be seen as the image of “the new socialism in the modern society, which is different from the ones in the Soviet Union and China” (Ishikawa 1984, 67). Even for its living standards, praising the United States was truly a bold step for a socialist. Eda’s argument split the party. Although some people favored it, some fiercely criticized Eda for being a revisionist and accepting “the current capitalist regime” (Ishikawa 1984, 67). At the 2lst party congress in January 1962, there was a heated debate over the SR, and a motion was adopted that prohibited the party to imme- diately accept the SR as the party’s principle. Later, at the 22nd party congress in November 1962, the anti-SR group intensified its attack on Eda, and a motion to censure Eda was adopted by a narrow margin.3 Eda then resigned from the party secretary post, and the conflict within the party subsided after this event (Mori 2001, 57). As I noted earlier, the JSP then adopted a hardliner leftist document “The Road to Socialism” in 1964 which subsequently bound the party’s behavior for a long time. Why did Eda and his SR fail? This should be the central question in examining the JSP’s behavior during the LDP dominance in the post-World War II Japanese politics. For the Socialist members, whether they maintain the party’s Marxist platform or moderate it would of course be a sensitive issue, and their beliefs and ideologies— 3The delegates to the party congress were elected from local organizations and labor unions, but many of the delegates were actually the legislators in the national parliament. Hence the legislators and the factions were the important actors during this event. 62 how earnestly they believed in Marxism——should no doubt be an important factor in determining their attitudes toward the SR. We can easily speculate that the leftist people in the party opposed the SR; indeed it was generally the case. However, a careful observation of the behaviors of the then JSP legislators shows that some members of leftist factions supported the SR and some from rightist factions opposed it. Also, not a few JSP members switched factions during this period.4 These facts suggest two things. First, ideology by itself does not explain much about JSP members’ attitudes toward the SR. Second, although the JSP was made up by factions and power struggles within the party were typically fought between factions, individual legislators appear to be the important actors in this event. It can therefore be thought that there existed another factor other than the ideology that affected individual JSP members’ attitudes toward the plan to moderate the party’s position. I argue, in essence, that the Single Non-Ttansferable Vote (SNTV) electoral system created various incentives for JSP politicians and made differences in their behaviors. Specifically, because of the complex district situations created by the SNTV system, the adoption of the SR would have increased some members’ re- election prospects but have decreased others’. If the JSP had adopted the SR and become a center-left party, it would probably have grown to a party that was larger in size and could have competed with the LDP for power in the long run. Yet, I argue that the legislators whose re-election quests would have been jeopardized by the position change had opposed the SR. Why could someone’s re-election become difficult if his or her party becomes larger? This should sound strange and implausible, but it was made possible by the SNTV system. The mechanism of this seemingly impossible situation is elaborated below. 4The detail will be shown in the next section. 63 6.2.3 The Incentive Structure Induced by the SNTV Sys- tern The SNTV system was used to elect the members of the House of Representa- tives (Shugiin) of Japan’s National Diet from 1947 to 1993. The district magnitude varied across districts, but it typically was between three and five. Each voter cast one ballot for an individual candidate, and top N candidates in an N-member district were elected. In the 1960 general election, which elected the legislators who were incumbents when the J SP dealt with the SR issue, there were 40 three-member districts, 39 four- member districts, and 38 five-member districts, in addition to an exceptional single member district that covered some remote islands of the Kagoshima prefecture. In total, there were 467 members of the lower house, and 145 of them belonged to the JSP. Under the SNTV system, a candidate’s chance of winning a seat depends on many different factors. First, the size of the personal and partisan support base in the district is, of course, important as it is in any systems. Second, the number of seats in the district (district magnitude) significantly changes the difficulty of election. In a three-member district, one needs 25% of the total votes to secure a seat, but 20% is needed in a four-member district, and 16.67% in a five-member district. Third, whether there are candidates from the same party running in the same district makes a big difference. If a party runs two candidates in a district, they have to coordinate or compete in gaining the votes from the party’s supporters. Regardless of whether a coordination succeeds, the partisan votes are divided and the election of the candidates becomes more difficult than if only one candidate runs from the party. Fourth, in a multi-party system, which parties field candidates in the district also makes a difference. If two parties that have similar positions compete with each other in a district, their candidates have to compete for the same type of 64 voters and thus it becomes more difficult to obtain many votes. Eda’s SR plan was intended to make the JSP a more moderate party that can attract a wider support. If the JSP had increased its popularity, it would then naturally have tried to run more candidates in elections. The number of the JSP candidates in the 1960 election was 186, which means that even if every one of them had won, the party would have controlled only 40% of the 467-seat House of Representatives. In order to become a party that can really compete against the LDP for power, the JSP needed to increase both its popularity and the number of its candidates. However, fielding more candidates would have made the J SP incumbents worse off in terms of their re-election prospects. As noted above, if two or more JSP candidates run in a district, the votes of the party supporters are spread out, making the campaigns more difficult. Hence, some JSP incumbents at the time were in a dilemma in which the plan that was supposed to make the party more popular would jeopardize the possibility of their re-election. It should be noted that not all JSP incumbents were in this dilemma. The extent to which the SR would lower the chance of re—election was different depending on the situation of each incumbent’s home district. For instance, there were 13 members who were elected from three-member districts where the J SP nominated two candidates. Since it is unthinkable that the JSP runs three candidates in a three-member district—unless it becomes an extremely popular, giant party—these 13 members did not have to worry about having another, third JSP candidate in their respective home districts. Rather, they might indeed have welcomed the SR because it would have increased the party’s popularity and made their re-elections easier. Hence, the possible effects of the adoption of the SR on the re—election quests of the J SP incumbents depended on the district magnitude and the number of JSP 65 Table 6.1: The Number of the JSP Candidates in the 1960 Election, Classified by the District Magnitude and the Number of the JSP Candidates in the District Number of J SP candidates 1 2 3 ... g l 1 (0.0) $5.5 3 29 (82.8) 22 (59.1) E 5 4 18 (100.0) 42 (71.4) 2 5 9 (100.0) 44 (79.5) 21 (76.2) Note: the numbers in parentheses show the winning percentages. candidates who previously ran. There were members elected from the districts where the JSP was unlikely to increase the number of candidates and hence the SR would not have jeopardized their re-election prospects. On the other hand, there were members from the districts where the JSP ran relatively few candidates and an increase of the number of candidates was more realistic—and needed if the party were to become larger. Table 6.1 shows the number of JSP candidates in the 1960 election, classified by the district magnitude (row) and the number of JSP candidates running in the district (column).5 The percentages of the JSP candidates that were elected are reported in the parentheses. Quite intuitively, the fewer the number of JSP candi- dates running in the districts, the higher the probabilities of their winning were.6 In the 27 districts where the district magnitude was four or five and the JSP ran only one candidate, every JSP candidate won a seat. Since the JSP candidates can mobilize labor unions, it would not be too difficult to consolidate the size of votes that was needed to secure a seat in larger districts. These 27 sole candidates in four- or five-member districts were enjoying the easiest campaign among the all 5The electoral data used in this chapter were kindly provided by Ross Schaap. 6This trend is observed not only in the 1960 election but in most elections as well. 66 JSP candidates. As long as the party did not nominate second candidates in their districts, they could have probably retained their “safe seats” easily, perhaps until they decided to retire. For the party, it was a safe strategy to nominate only one candidate in a large district because it was almost certain that the candidate would win. Yet, if the JSP were to seriously compete for power, it would have had to abandon the safe strategy and start nominating more candidates in large districts; and that was what would have happened if the SR had been adopted and the party’s popularity had increased. Hence, the party’s (and perhaps party leaders’) seat maximizing incentive col- lided with some incumbents’ re—election incentives, which would not typically happen under other electoral systems. Those with “safe seats” may well have resisted the change to the status quo if their foremost concern was protecting their careers as legislators rather than realizing a Socialist government. On the other hand, the members other than the “safe seaters” would generally welcome the change that might increase the party’s popularity and make their re-elections easier. Some may find my argument excessively cynical and question if the JSP legis- lators, many of whom were labor union activists before becoming politicians, were really concerned about their own re-elections. Yet, it is not entirely new to View the JSP members as re-election seekers. Reed and Bolland (1999) indeed find a “downward ratchet” mechanism in the JSP’s candidate nomination pattern. That is, “whenever the J SP runs one less candidate than in the previous election, it finds it difficult to increase the number of candidates thereafter” (Reed and Bolland 1999, 214), and they argue that this tendency is caused by “the power of incumbents in the nomination process” (217). Also, I am not arguing that the re-election incentive was the sole determinant of the J SP members’ attitudes toward the SR. I admit that the members’ ideologies, of course, were one of the most important factors. What I 67 argue is that, after controlling for the effects of the ideology, whether they had safe seats that would have been lost by the adoption of the SR made a difference in their attitudes. Hypothesis 1: The JSP legislators who were the sole JSP candidates in four- or five-member districts resisted the Structural Reform (SR) plan more than the ones who were not. Another possible factor that I suspect made differences in JSP incumbents’ attitudes toward the SR is whether they were competing against other opposition parties’ candidates in their home districts. In the early 19603, there were two op~ position parties other than the JSP: the Japanese Communist Party (JCP), which was at the far left, and the Democratic Socialist Party (DSP), which was located between the JSP and the LDP. Figure 6.1 shows the ideological locations of the Japanese political parties at that time. Located between two other opposition parties, the JSP was in a difficult sit- uation in choosing its ideological position. If the party had moved closer to the center, some leftist JSP supporters might have switched their support to the JCP; but if it had moved more leftward, centrist voters would have abandoned the JSP and voted instead for the DSP (or even for the LDP because the DSP was so small then). Ramseyer and Rosenbluth (1993, 42) suggest that the JSP’s adherence to the leftist policy was a “niche strategy” to locate itself between the other parties. Kohno (1997) elaborates this thesis and shows empirical evidence. I consider that this “niche” hypothesis is a plausible explanation of the JSP’s failure to moderate its policies in general and in a longer perspective. However, for this particular event in the early 19603 with which this chapter deals, individual legislators’ behaviors were crucially important as I discussed earlier, and hence we should examine the happenings within the party and where the individual JSP members wanted to locate their party, rather than where the JSP wanted to locate 68 Figure 6.1: Ideological Locations of the Japanese Political Parties in the Early 19605 Figure 1 Ideological locations of the Japanese political parties in the early 19603. JCP JSP DSP LDP (3) (145) (17) (296) Left < A, Right The numbers in parentheses show the numbers of seats the parties obtained in the 1960 election. 69 Table 6.2: The JSP incumbents and their competition against other opposition parties in the 1960 election. (1) The number of JSP incumbents who... Did not have a J CP candidate 0 Had a JCP candidate who lost 142 Had a JCP candidate who won 3 total 145 (2) The number of JSP incumbents who... Did not have a DSP candidate 16 Had one or two DSP candidates who lost 110 Had a DSP candidate who won 19 total 145 itself. I argue that whether the individual JSP incumbents were in competition with other opposition candidates in their home districts, especially with the DSP ones, made a difference in their attitudes toward the SR. Table 6.2 classifies the JSP members in terms of the pattern of competition against other opposition candidates in the 1960 election. The JCP ran its candidates in all 118 districts, as it kept doing so in subsequent years, but won only three seats in 1960. Hence all JSP incumbents were in competition against a JCP candidate. As the table shows, there was not much variance among the J SP members in terms of the pattern of their competition against communist rivals. Indeed, 88% of the JCP candidates in the 1960 election received less than 5% of votes in their respective districts. Due to the small variance in the strength of J CP candidates across districts, it does not appear to be a major discriminating factor in the JSP politicians’ attitudes toward the SR. On the other hand, there was a significant difference among the J SP incumbents 70 in terms of the level of competition they had with DSP candidates. 16 J SP members did not have to compete with DSP candidates, 110 members had DSP candidates who lost in their districts, and 19 members had DSP candidates who were elected. Since the DSP was located between the JSP and the LDP, the presence of a DSP candidate in a district should have made a big impact on the strategy of the J SP members running in that district. Without a DSP candidate, a J SP candidate could gain support from center and center-left voters if the JSP moved toward the center. On the contrary, if a DSP candidate was present, the move of the J SP toward the center would not give much electoral benefit to the JSP candidate because the centrist votes would be split between the DSP and the JSP. At the same time, since everyone had a J CP candidate in his or her home district, they all would have lost some leftist votes to the JCP if the SR had been adopted. Hence I argue that whether the JSP members had to compete with DSP candidates or not made a considerable difference in the amount of the electoral benefit the SR was supposed to bring to them, and I speculate that this difference divided the JSP members’ attitudes toward the SR. Hypothesis 2: The JSP legislators who were in competition against DSP candidates in their home districts resisted the Structural Reform (SR) plan more than the ones who were not. 6.3 Empirical Investigation 6.3.1 Identifying the J SP Legislators’ Attitudes To test the hypotheses, we first have to operationalize the dependent variable (the J SP members’ attitudes toward the SR), and we can obtain useful information from the factional affiliation of the JSP legislators. Figure 6.2 shows the sizes of the 71 J SP factions in 1961 and 1963 and the transition of the members during the two—year period between them. The sources are the 1961 and 1963 editions of K okkai Binran (Nihon Seikei Shinbunsha 1961; 1963). The 1961 edition was published in February 1961, and it would thus reflect the initial factional affiliations of the JSP members who were elected in the November 1960 election. Although the 1962 edition also shows the factional affiliations, it is quite similar to the 1961 edition and hence it does not reflect the factional realignment that took place then. The 1963 edition indicates that each of the four largest factions—Kawakami, Wada, Suzuki, and Heiwa7—had an internal cleavage. The Kawakami faction, led by Jotaro Kawakami who became the party leader in March 1961, consisted of the mainstream group and the group under Eda’s influence. The Wada faction, led by Hiroo Wada, contained two groups that followed Eda and Seiichi Katsumata, respectively, in addition to the mainstream group. The Suzuki faction was led by former party leader Mosaburo Suzuki but at this time split into two groups led by Eda and K020 Sasaki, respectively. The Heiwa faction contained two members who were under Sasaki’s influence and 13 members of the mainstream group. As is shown in Figure 6.2, there were quite a few faction switchers at this time, and the Kuroda faction and the Leftist Group disappeared by 1963. Since the factional realignment between 1961 and 1963 took place concurrently with the fierce intra—party conflict within the JSP over the SR and the rivalry be- tween Eda and Sasaki, it would be safe to assume that the factional affiliation in 1963 reflects the members’ attitudes toward the SR. In the rightmost part of Figure 6.2, I indicated whether each faction or group of 1963 was for or against the SR. As described earlier, Eda and Sasaki were the main figures during this conflict. Thus, there is no doubt that the members of Kawakami-Eda, Wada-Eda, and Suzuki-Eda 7T he names of the Kawakami, Wada, and Suzuki factions show the respective leaders of the factions. The real name of the Heiwa faction is “Heiwa doushi kai,” meaning “The Comrades for Peace.” 72 Figure 6.2: Factional Realignment within the J SP juring the Conflict over the Struc- tural Reform Plan 1961 Kawakami (27) Wada (35) Suzuki (45) Leftist Group (8) Heiwa (18) Kuroda (4) Nomizo (7) Independent (1) 1963 Kawakami-main (15) Kawakami-Eda (10) Wada-Eda (8) Wada-Katsumata (5) Wada-main (22) Suzuki-Eda (9) Suzuki-Sasaki (45) Centrists (3) Heiwa-main (13) , Heiwa-Sasaki (2) Nomizo (4) Attitude toward the SR favor favor favor favor athehween favor oppose upbenween oppose oppose oppose (Bold arrows indicate five or more people.) Source: Nihon Seikei Shinbumsha 1961; 1953. 73 were supportive of the SR, and those of Suzuki-Sasaki and Heiwa-Sasaki opposed it. As for other factions and groups, journalistic evidence suggests that Katsumata was one of the proponents of the SR and was supportive of Eda,8 and hence we can classify the Wada-Katsumata group as pro-SR. Similarly, the Kawakami-main can be considered as a pro-SR group.9 On the other hand, the Heiwa faction and the Nomizo faction—both were leftist—were against the SR and supported Sasaki.10 The Wada-main group is somewhat difficult to classify. Although it is said that most members of the Wada faction supported Eda and the SR,11 it was also reported that Wada himself took an ambiguous or neutral stance.12 Thus it can be considered that the mainstream group within the Wada faction was at least less supportive of the SR than the members who followed Eda or Katsumata. I therefore placed the Wada-main group between the pro-SR people and the anti-SR people. The three legislators of the “centrists” are also considered as “in-betweens.” Lastly, the nine JSP members whose factional affiliations are not listed in the 1963 K okkai Binran are excluded from the analysis. The 136 JSP legislators are now classified according to their attitudes toward the SR. There are 47 pro-SR members, 25 in-betweens, and 64 anti-SR members. In what follows, I will test my hypotheses by two methods: cross tabulation and maximum likelihood estimation. 8Asahi Shinbun (one of the leading newspapers), October 30, 1961, p.1; January 15, 1962, p4; June 20, 1962, p.1; July 10, 1962, p.2. 9After the Nishio faction left the party in 1959, the Kawakami faction had been the only true rightist faction in the JSP (Soga 1989). Asahi Shinbun repeatedly reported that the Kawakami faction was supportive of Eda in this period (January 23, 1962, p.2; April 39, 1962, p.2). According to Asahi Shinbun (January 23, 1962, p.1), Sasaki, after losing to Eda in the election for the party secretary post, said, “Although the Kawakami faction was claiming, ‘We are neutral,’ they supported Eda at the last moment. While we, the Suzuki faction, have endorsed Mr. Kawakami for the party leadership post after Mr. Asanuma died and have supported him, we hereby sever relations with Mr. Kawakami.” 10Asahi Shinbun, October 20, 1961, p.1; January 15, 1962, p4; January 23, 1962, p.2; November 30, 1962, p.1. Also see Shiota (1994). ”Asahi Shinbun, January 23, 1962, p.2; April 30, 1962, p2. 1"zAsahi Shinbun, January 23, 1962, p.2; July 10, 1962, p.12. 74 6.3.2 Cross Tabulation Analysis In order to investigate whether J SP legislators’ situations in their home districts affected their attitudes toward the SR, we have to control for their ideology since it is quite plausible that those who strongly believed in Marxism tended to oppose the SR more than the rightist (center-left) people in the party. I contend that much of this problem can be avoided by looking at each faction in turn since the factions in the JSP were aligned from left to right and thus those who belonged to the same faction may have more or less similar ideologies. For example, the Suzuki faction was known as one of the leftist factions, and hence its members in 1961 are thought to have had leftist ideologies. By examining these members’ positions in 1963, therefore, we can find out the determinants of their attitudes toward the SR while controlling for the effects of ideologies. In this subsection, I cross-tabulate the members of each faction according to the variables of interest. Among the seven factions that existed in 1961, three of them are not suitable for this analysis. Everyone who belonged to the Kawakami faction in 1961, except for the two whose factional affiliations were unknown in 1963, supported the SR. Likewise, the Kuroda faction unanimously opposed it. Therefore, the behavior of the members of these two factions do not give us any insights in terms of what determined their attitudes toward the SR. me the Nomizo faction, one member moved to the mainstream Wada faction which is considered as an “in- between” group while all the other members opposed the SR. Since a deviation of only one person does not give us much systematic knowledge, this faction is also excluded from the subsequent analysis. The cases of the other four factions are examined below. (1) The Suzuki Faction The faction that was affected the most by the conflict over the SR was definitely 75 the Suzuki faction, the largest faction in 1961. Both Eda and Sasaki, the leaders of the pro-SR and anti-SR camps of the party, respectively, belonged to this faction; and the rivalry and animosity between them was so intense that the internal cleav- age within this faction eventually grew to the point where the two groups became recognized as separate factions.l3 Out of the 43 members who belonged to the Suzuki faction in 1961 and whose factional affiliations in 1963 were documented in K okkai Binran, 33 (76.7%) were anti-SR and were with Sasaki, 7 were pro-SR with Eda, and 3 were neutral. The left-hand panel of Table 6.3 ( 1) classifies them by whether they were the sole JSP candidates in four— or five-member districts. As discussed earlier, I expect that those who enjoy their safe seats by being the only JSP candidates in large districts would oppose the attempt to enlarge the party. As shown in the table, none of those “safe-seaters” were pro-SR, and 8 out of 9 were against the SR. On the other hand, 20.6% of the “non-safe—seaters” supported the SR, and the percentage of anti-SR among them (73.5%) was much smaller than that among the “safe-seaters” (88.9%). The gamma value for this relationship is -0.52, which indicates a strong relationship.14 This result clearly shows that the Suzuki faction members who were sole J PS candidates in four- or five-member districts in the 1960 election were more likely to have opposed the SR; and it supports Hypothesis 1. The right-hand panel of Table 6.3 (1) classifies the members by the level of the presence of the DSP. 8 members of the Suzuki faction had a DSP candidate who won a seat in their respective districts. These 8 were going to have to compete against a DSP incumbent in the next election unless their DSP rivals were to retire. Hypothesis 2 states that these members should have opposed the SR, and indeed all of them did. 13Newspapers started to use the term “the Eda faction” and “the Sasaki faction” in early 1962 although Kokkai Binran, kept using the term “the Suzuki faction” until its 1963 edition. 14Standard error of gamma statistic is an asymptotic value and thus does not make sense for a small sample like this. 76 Table 6.3: Cross-Tabulation Analysis (1) The Suzuki Faction Pro-SR ln-BetVi/n Anti-SR total Pro-SR ln-Beth Anti-SR total Sole JSP Yes 0 1 8 DSP‘s Vaner 0 0 8 8 candidate % 0.0 11.1 88.9 100.0 presence % 0.0 0.0 100.0 100.0 in a 4or 5 No 7 2 25 34 in the Candidate 7 2 25 34 member % 20.6 5.9 73.5 100.0 home % 20.6 5.9 73.5 100.0 district total 7 3 33 43 district No Candidate 0 1 0 1 % 16.3 7.0 76.7 100.0 % 0.0 100.0 0.0 100.0 total 7 3 33 43 % 16.3 7.0 76.7 100.0 (2) The Wada Faction Pro-SR ln-Beth Anti-SR total Pro-SR ln-Beth Anti-SR total Sole JSP Yes 2 4 3 DSP's Vaner 1 3 0 4 candidate % 22.2 44.4 33.3 100.0 presence % 25.0 75.0 0.0 100.0 ina 4or5 No 8 13 3 24 in the Candidate 7 13 6 26 member % 33.3 54.2 12.51000 home % 26.9 50.0 23.1 100.0 district total 10 17 6 33 district No Candidate 2 1 0 3 % 30.3 51.5 18.2 100.0 % 66.7 33.3 0.0 100.0 total 10 17 6 33 % 30.3 51.5 18.2 100.0 (3) The Heiwa Faction Pro-SR ln-Beth Anti-SR total Pro-SR ln-Betw’n Anti-SR total Sole JSP Yes 0 0 3 DSP's Nnner 0 0 3 3 candidate % 0.0 0.0 100.0 100.0 presence % 0.0 0.0 100.0 100.0 ina 4or5 No 2 0 12 14 inthe Candidate 1 0 10 11 member % 14.3 0.0 85.7 100.0 home % 9.1 0.0 90.9 100.0 district total 2 0 15 17 district No Candidate 1 O 2 3 % 11.8 0.0 88.2 100.0 % 33.3 0.0 66.7 100.0 total 2 0 15 17 % 11.8 0.0 88.2 100.0 (4) The Leftist Group Pro.SR ln-Betwn Anti-SR total (No one from this group was a sole JSP DSP's Vaner 0 1 1 2 mndidateina4or5member district.) presence % 0.0 50.0 50.0 100.0 in the Candidate 2 2 1 5 home % 40.0 40.0 20.0 100.0 district No Candidate 1 0 0 1 % 100.0 0.0 0.0 100.0 total 3 3 2 8 % 37.5 37.5 25.0 100.0 77 TH 34 members of the Suzuki faction had a DSP candidate in their districts who did not win a seat. They had less pressure from the DSP than the 8 members with DSP incumbents; and not surprisingly, about 20% of them (7 members) supported the SR—ethis makes a clear contrast with the group with DSP incumbents in which everyone opposed the SR. There is also 1 member whose district did not have a DSP candidate. According to the hypothesis, he was most likely to have supported the SR, but actually he was in the “in-between” group. Yet, we cannot test hypothe- ses by the observation of one person. The gamma value is -0.88, which shows an exceptionally strong relationship. Hypothesis 2 is supported by the analysis of the members of the Suzuki faction. (2) The Wada Faction About a half of the 35 members of the Wada faction in 1961 were in the main- stream group of the Wada faction in 1963 and are classified as “in-betweens,” and there were 10 members who followed Eda or Katsumata and are considered as pro- SR. An interesting thing about this faction is that 6 members switched to the Suzuki faction and joined the Sasaki group within it. We can speculate that they had a special reason to do somprobably a strong opinion against the SR. The left-hand panel of Table 6.3 (2) shows that the “safe-seaters” were less supportive of the SR (22% favored and 33% opposed) than the “non-safe-seaters” (33% favored and 13% opposed). The gamma value is -0.37. Hypothesis 1 is again supported here. The right-hand panel of Table 6.3 (2) does not give a clear support for Hypoth- esis 2. The 4 members who were competing against a DSP incumbent in their home district—the ones whom I expect to have resisted the SR—were either pro-SR or in-between. Yet it is also notable that 2 of the 3 members whose districts did not have a DSP candidate were pro-SR. The gamma value, -0.22, does not show that the relationship is strong, but the direction of the relationship is consistent with 78 Hypothesis 2. (3) The Heiwa Faction Since there are only 17 members in the Heiwa faction, the results should be interpreted with caution and taken tentatively. However, the left-hand panel of Table 6.3 (3) shows a clear pattern that supports Hypothesis 1: all 3 members with “safe seats” were anti-SR and both of the 2 members who were pro-SR were “non-safe-seaters” (gamma 2 -1.00). The right-hand panel of Table 6.3 (3) shows a similar pattern. All 3 legislators who had a DSP incumbent in the same district opposed the SR; and both of the 2 members who supported the SR did not have a DSP winner in their districts. This result is consistent with Hypothesis 2 (gamma 2 -0.78). (4) The Leftist Group The Leftist Group in 1961 had only 8 members, and hence it is difficult to draw any systematic evidence from here. Also, Hypothesis 1 cannot be tested with this sample because no one in this group was elected from a four- or five-member district as a sole J SP candidate. However, the result from the right-hand panel of Table 6.3 (4) is consistent with Hypothesis 2 (gamma =2 -0.83). Overall, the cross-tabulation analysis of the members of four J SP factions sup- ports my two hypotheses fairly well. 6.3.3 Maximum Likelihood Estimation In this subsection, I utilize the maximum likelihood method to test my hy- potheses. The dependent variable is the members’ attitudes toward the SR, which I identified earlier. It is coded 1 if pro-SR, 2 if in-between, and 3 if anti-SR. Since it is an ordinal variable, the ordered logit model is used. It may be argued that the number of observations (136) is too small to make a valid inference using the 79 ordered logit model. The results from this analysis should therefore be interpreted with caution; and this is why I present the results from both the cross tabulation analysis and the ordered logit analysis. The two analyses would complement each other. Unlike the cross tabulation analysis presented above, all 136 legislators make up a single sample for this analysis. To control for the effects of their ideologies, I included a control variable that takes a value of 1 if the member belonged to one of the leftist factions in 1961 (Suzuki, Heiwa, Kuroda, Nomizo, and the Leftist Group) and 0 otherwise (Leftist).15 To test Hypothesis 1, a dummy variable was included that takes the value of 1 if the member was a sole JSP candidate in a four— or five- member district and 0 otherwise (Sole). For Hypothesis 2, I employed two ways to operationalize the presence of the DSP. The first method is to include two dummy variables in the model: one is whether the member’s district had a DSP candidate who won a seat (DSP Winner), and the other is whether there was a DSP candidate who lost (DSP Loser). If both variables take 0, there was no DSP candidates in the district. The second method to operationalize the presence of the DSP is to include a variable of DSP’s vote share (DSP Votes). If the two methods produce comparable results, it indicates robustness of the findings. Two more variables are included in the model as control variables. First, the JCP’s vote share is included (JCP Votes). Yet, as discussed above, there was not much variance in this variable and hence a significant effect may not be found. Second, the variable of the member’s age is added to the model (Age). Older members may be more emotionally attached to the party’s platform than younger members. There is almost no correlation between Age and Leftist. 15There was one member, Kyoko Asanuma, who did not belong to a faction. She is the widow of the former party leader Inejiro Asanuma who was assassinated during the 1960 election campaign. Since her husband was a rightist in the party, I tentatively coded her as non-leftist. She did not run for the parliament again. 80 Table 6.4: Ordered Logit Estimation of the JSP Members’ Attitudes toward the Structural Reform Model (1) Model (2) Coefficient p-value Coefficient p-value (SE) (SE) Leftist 2.756 0.000 *** 2.773 0.000 *** (0.415) (0.411) Sole 1.010 0.062 * 1.036 0.054 * (0.541) (0.538) DSP Winner 1.838 0.037 ** (0.880) DSP Loser 1.335 0.040 ** (0.650) DSP Vote 0.065 0.085 * (0.038) ICP Vote 0.020 0.788 0.005 0.942 (0.076) (0.068) Age 0.018 0.446 0.015 0.516 (0.024) (0.023) Thresholdl 2.927 2.047 (1.387) (1.257) Threshold2 4.134 3.241 (1.409) (1.276) N 136 136 Log(L) -106.31 -107.63 * p<0.10, ** p<0.05, *** p<0.01 81 Table 6.5: Predicted Probabilities Table 5 Predicted Probabilities Model (1) Legislator's Characteristics Leftist : Non-Leftist Leftist Sole : Yes Yes Yes No No No Yes Yes Yes No No No DSP :Winner Loser No Winner Loser No Winner Loser No Winner Loser No Pr (Pro-SR) 0.32 0.41 0.70 0.53 0.65 0.86 0.04 0.05 0.17 0.08 0.11 0.33 Pr (In-Betw'n) 0.26 0.27 0.17 0.24 0.21 0.09 0.07 0.09 0.20 0.14 0.18 0.27 Pr (Anti-SR) 0.42 0.31 0.13 0.23 0.14 0.05 0.90 0.86 0.62 0.78 0.71 0.40 Model (2) Legislator's Characteristics Leftist : Non-Leftist Leftist Sole : Yes Yes Yes No No No Yes Yes Yes No No No DSP : 20% 10% 0% 20% 10% 0% 20% 10% 0% 20% 10% 0% Pr (Pro-SR) 0.27 0.40 0.55 0.49 0.65 0.77 0.03 0.05 0.09 0.07 0.11 0.19 Pr (In-Betw'n) 0.26 0.27 0.23 0.25 0.21 0.14 0.05 0.08 0.14 0.11 0.17 0.24 Pr (Anti-SR) 0.47 0.33 0.21 0.25 0.15 0.08 0.92 0.87 0.77 0.82 0.72 0.57 Table 6.4 shows the result of the ordered logit analysis. The two models corre- spond to the two methods by which I operationalized the presence of the DSP. In both models, the variables of interest have the expected signs and are statistically significant. For easier interpretation, I used the “Clarify” program (King, Tomz and Wittenberg 2000; Tomz, Wittenberg and King 2003) to calculate the predicted probabilities while changing the combination of the independent variables and keep- ing JCP Votes and Age at their respective median values (1.93 and 51). The results are presented in Table 6.5. For each type of legislators, the predicted probabilities of falling into each of the three categories of the attitudes toward the SR (i.e., Pro-SR, In-Between, and Anti-SR) are calculated. Since the results from the two models are quite similar to each other, let us focus on the results from Model (1). The type of legislators who have the highest 82 predicted probability to be against the SR (90%) is, as my hypotheses state, those who were leftists (Leftist-:1), who were the sole candidates from four- or five-member districts (Solezl), and whose districts had a DSP candidate who won a seat (DSP Winnerzl). Indeed, there are 6 such members in the data, and they are all coded as anti-SR. The probability to be anti-SR declines to 62% if his or her district did not have a DSP candidate, to 78% if Sole=0, and to 42% if Leftist20. For the candidates who were not leftists, not sole candidates, and did not compete against a DSP candidate, the predicted probability to be anti-SR is just 5% and the probability to be pro-SR is as high as 86%. There are 10 such members in the data; 9 of them are indeed pro-SR, and the other 1 is in-between. Overall, the large differences in the predicted probabilities presented here show that the models have strong explanatory power and the characteristics of the JSP members examined here explain much of the differences in the members’ attitudes toward the SR. This result provides a strong support for my hypotheses. 6.4 Conclusion The empirical analyses in the previous section strongly support my argument that those legislators whose re—election prospect would have been jeopardized by the party’s transformation to a center-left party were the people who opposed the SR. The SNTV electoral system appears to have trapped some legislators in a unique dilemma in which an increase in the popularity of their own party decreases the probability of their re-election. It is not difficult to think that there were members who chose to pursue personal gain (re-election) over the party’s collective gain. If we assume that a political party is a unitary actor and behaves in order to maximize its votes and seats or to gain power, moving closer to the center of the political spectrum by adopting the SR should have been the optimal action for the 83 JSP. Yet, if we shift our focus to individual JSP politicians, it can be understood that the optimal behavior of a vote/seat/power seeking party was not necessarily consistent with every member’s interest. The JSP’s non-optimal behavior can be explained by its incumbent legislators’ purposive actions. It should be noted that my argument does not rule out the other explanations of the J SP’s ideological immobility outlined earlier since they are not mutually ex- clusive. Also, the findings from this analysis cannot be easily generalized since the SNTV system is used only in a few countries. However, I believe this analysis is a meaningful contribution for two reasons. First, although re-election incentive as a determinant of legislators’ behavior has been repeatedly studied and its importance is now widely accepted, the J SP politicians’ behaviors have not been analyzed from this perspective in the literature. My analysis has demonstrated the strong explana- tory power of re-election incentive in the context of Japan as it is in other societies. In that sense, my argument and findings are a rebuttal to some Japanologists who excessively emphasize the uniqueness of Japan and reject the comparability between Japan and other countries. Second, although the unique effects of the SNTV I discussed here will not be observed in the countries where other electoral systems are used, the findings here suggest the importance of paying attention to intra—party politics when we examine political parties’ behaviors in other settings as well. As I noted in Chapter 2, Laver (1998, 22) argues that going beyond the unitary actor assumption is “One of the biggest unclaimed prizes in the field.” I believe this study has made a step toward this direction. 84 Chapter 7: Conclusion In the preceding chapters, I explored the links between institutional rules, be- haviors of opposition parties, and the electoral fortunes of ruling parties. In this concluding chapter, I will summarize my findings and discuss their implications and the agendas for future research. Institutional rules, such as electoral systems and parliamentary rules, create the environment in which parties compete. It is well known that electoral systems affect party systems, as the literature reviewed in Chapter 2 shows. My discussion in this dissertation, however, was more general and paid attention to how institutions influence strategies and preferences of party leaders and legislators. Those strategies and preferences partly determine the behaviors of political parties. I argue that when parties merge, split, form alliances, dissolve alliances, join coalitions, quit coalitions, and change or maintain policy positions, those behaviors are often the end results of intra—party politics, in which various actors pursue their respective goals that are shaped by institutional rules. In Chapter 5, I argued that disproportional electoral systems and non-deliberative parliamentary rules make the environment of party competition more difficult for small parties to survive in. Those institutional features, therefore, make the oppo- sition less fragmented. In Chapter 6, I analyzed the ideological immobility of the Japan Socialist Party and argued that the electoral system gave the incentive to resist the policy change for some incumbent Socialist politicians. These analyses showed that institutional rules are important factors that influence the characteris- tics of opposition parties—~more specifically, degree of opposition fragmentation and ideological positions of opposition parties. Parties in power, in general, try to stay in office as long as possible. Among the many factors that determine how long they continue in office, I focused on the 85 role of opposition parties, which does not appear to have received much attention in the literature. I argued with supportive evidence in Chapters 3 and 4 that when the opposition is unified, the voters have a visible alternative party to choose at the ballot box, and thus dissatisfaction to the incumbent is more directly reflected to the vote loss of the governing parties. On the other hand, if the opposition is fragmented to many parties, the voters may not find any of them as a credible alternative, and thus the incumbent parties do not lose many votes even if their performance is not popular among citizens. Therefore, it is implied that fragmented opposition makes the tenure of govern- ments longer, although this proposition is not directly tested in this research. Since the length of time governments stay in power would make significant differences in political and economic performances, the degree of opposition fragmentation is an important characteristic of political systems. Also, whether the opposition is unified or fragmented would change the amount of the pressure the government feels from the opposition and thus responsiveness and accountability of the government. Hence, opposition parties are not just the groups of politicians staying out of power, but what they do and how they behave would make significant differences in the functioning of democracy. The role of opposition parties in political systems has not received much attention so far, but I believe this research has shown its importance and the need for more research on this topic. In closing, I would like to note two issues that were not addressed in this dissertation but are important agendas for future research. The first is the extension to presidential systems. This study has solely focused on parliamentary systems, and thus it is unknown whether the findings hold in presidential systems as well. Parties in presidential systems compete in both presidential and legislative elections. Hence, the behavior of parties and the pattern of party competition are affected by the electoral systems of both levels of elections and whether the two levels of elections 86 are held concurrently or not. Since many factors are involved, it would not be easy to analyze the mechanism of institutional effects on party competition; yet it is an interesting research topic nonetheless. The second issue is the behavior of governments to change the characteristics of opposition parties. If the behavior of opposition parties affects the fate of a government, the government may try to influence what oppositions do so that it can stay in power longer. As shown in Chapters 5 and 6, institutional rules affect the degree of opposition fragmentation and ideological locations of opposition parties. Hence governments may try to modify the institutional rules to make the opposition parties less serious in winning power. Changes of electoral systems are not very frequently observed, and thus it may be difficult to study them empirically. But it would be interesting to examine them as the governments’ attempts to weaken the opposition. Another aspect of political institutions governments may try to change is par- liamentary rules. As I argued in Chapter 5, if opposition parties are allowed to influence policy making, they may become less serious in pursuing power. By lis- tening to the opposition’s opinion in policy making, the government may be able to “tame” the opposition and extend its own tenure in power. Although official par- liamentary institutions, such as committee systems, seldom change, governments’ attempts to tame the opposition may be observed in consultation and negotiation between government and opposition over important policy issues. From the con- ventional view, such bipartisan (or multi-partisan) agreements may be seen as the parties’ efforts to make consensual decisions and ensure smooth implementation of important policies. Yet, from the viewpoint that emphasizes the strategies of par- ties, they could be interpreted as the government’s tactics to avoid responsibilities and deprive the opposition of the chances of anti-government campaigns. Throughout this dissertation, I emphasized the importance of paying attention 87 to the strategies of parties that stem from whether they are in power or in opposition and the need for examining intra—party politics as the determinants of party behav~ ior. I believe that the analyses presented in the preceding chapters have shown that these perspectives are useful and that they also opened up many promising areas for future research. 88 Bibliography Amorim Neto, Octavio and Gary W. Cox. 1997. “Electoral Institutions, Cleavage Structures, and the Number of Parties.” American Journal of Political Science 41(1):149—174. Andersen, Uwe and Wichard Woyke. 2000. Handwo'rterbuch des politischen Systems der Bundesrepublik Deutschland. 4th ed. Opladen: Leske + Budrich. Anderson, Christopher J. 1995. Blaming the Government: Citizens and the Economy in Five European Democracies. Armonk, NY: Sharpe. Anderson, Christopher J. 2000. “Economic Voting and Political Context: A Com- parative Perspective.” Electoral Studies 192151—70. Barker, Fiona and Elizabeth McLeay. 2000. “How Much Change?: An Analysis of the Initial Impact of Proportional Representation on the New Zealand Parlia- mentary Party System.” Party Politics 6(2):131—54. Bartolini, Stefano and Peter Mair. 1990. Identity, Competition, and Electoral Avail- ability. Cambridge: Cambridge University Press. Beck, Nathaniel and Jonathan N. Katz. 1995. “What to Do (and Not to Do) with Time-Series Cross-Section Data.” American Political Science Review 89(3):634— 47. Blais, Andre and R. K. Carty. 1991. “The Psychological Impact of Electoral Laws: Measuring Duverger’s Elusive Factor.” British Journal of Political Science 21(1):79—93. Blondel, Jean. 1968. “Party Systems and Patterns of Government in Western Democracies.” Canadian Journal of Political Science 1:180—203. Chappell, Henry W., Jr. and Linda Goncalves Veiga. 2000. “Economics and Elec- tions in Western Europe: 1960-1997.” Electoral Studies 19:183—97. Chhibber, Pradeep and Ken Kollman. 1998. “Party Aggregation and the Number of Parties in India and the United States.” American Political Science Review 92:329—342. Cox, Gary W. 1994. “Strategic Voting Equilibria under the Single Nontransferable Vote.” American Political Science Review 88:608—621. Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World’s Electorial Systems. London: Cambridge University Press. Dahl, Robert, ed. 1966. Political Oppositions in Western Democracies. New Haven: Yale University Press. 89 Day, Alan, Richard German and John Campbell, eds. 1996. Political Parties of the World. 4th ed. New York: Stockton. Duverger, Maurice. 1954. Political Parties: Their Organization and Activity in the Modern State. New York: John Wiley. Fidrmuc, Jan. 2000. “Economics of Voting in Post-Communist Countries.” Electoral Studies 19:197—217. Flanagan, Scott C. 1984. “Electoral change in Japan: A study of secular realign- ment.” In Electoral change in advanced industrial democracies, ed. Russell J. Dalton, Scott C. Flanagan and Paul Allen Beck. Princeton: Princeton Univer- sity Press. Franklin, Mark N. and Wolfgang Rudig. 1995. “On the Durability of Green Poli- tics: Evidence from the 1989 European Election Study.” Comparative Poltical Studies 28(2):409—39. Friedrich, Robert J. 1982. “In Defense of Multiplicative Terms In Multiple Regres- sion Equations.” American Journal of Political Science 26(4):797—833. Golder, Matt. 2004. “Codebook: Democratic Electoral Systems Around the World, 1946-2000.”. URL: http://homepages.nyu.edu/~mrg217/elections.html Gorvin, Ian, ed. 1989. Elections since 1945 : a worldwide reference compendium. Harlow: Longman. Harmel, Robert and Kenneth Janda. 1994. “An Integrated Theory of Party Goals and Party Change.” Journal of Theoretical Politics 6(3):259—87. Host, Viggo and Martin Paldam. 1990. “An International Element in the Vote? A Comparative Study of Seventeen OECD Countries, 1948-85.” Europrean Jour- nal of Political Research 18:221—39. Inoguchi, Takashi. 1983. Cendai nihon no seiji keizai no kouzu. Tokyo: Toyo keizai shinpo sha. International Labour Office. various issues. Bulletin of Labour Statistics. Geneva: The Office. Ishikawa, Masumi. 1984. Beta sengo seijishi. Tokyo: Iwanami Shoten. Jackman, Robert and Karin Volpert. 1996. “Conditions favoring parties of the extreme right in Western Europe.” British Journal of Political Science 26:501— 21. 90 Janda, Kenneth, Robert Harmel, Christine Edens and Patricia Goff. 1995. “Changes in Party Identity: Evidence from Party Manifestos.” Party Politics 1:171—96. Kabashima, Ikuo. 1988. Seiji sanka. Tokyo: Tokyo daigaku shuppan kai. King, Gary, Michael Tomz and Jason Wittenberg. 2000. “Making the Most of Statistical Analyses: Improving Interpretation and Presentation.” American Journal of Political Science 44(2):347—61. Kitschelt, Herbert P. 1988. “Left-Libertarian Parties: Explaining Innovation in Competitive Party Systems.” World Politics 40(2):194—234. Knigge, Pia. 1998. “The ecological correlates of right—wing extremism in western Europe.” European Journal of Political Research 342249—79. Kohno, Masaru. 1997. “Electoral Origins of Japanese Socialists’ Stagnation.” Corn- parative Political Studies 30(1):55—77. Laakso, Markku and Rein Taagepera. 1979. “Effective Number of Parties: A Mea- sure with Application to Western Europe.” Comparative Political Studies 12:3- 27. Laver, Michael. 1989. “Party Competition and Party System Change: The Interac- tion of Coalition Bargaining and Electoral Competition.” Journal of Theoretical Politics 1(3):301—24. Laver, Michael. 1998. “Models of Government. Formation.” Annual Review of P0- litical Science 1:1—25. Laver, Michael. 2003. “Governemnt Termination.” Annual Review of Political Sci- ence 6:23—40. Lewis-Beck, Michael S. 1988. Economics and Elections: The Major Western Democ- racies. Ann Arbor, MI: University of Michigan Press. Lewis-Beck, Michael S. and Mary Stegmaier. 2000. “Economic Determinants of Electoral Outcomes.” Annual Review of Political Science 32183—219. Lijphart, Arend. 1994. Electoral Systems and Party Systems: A Study of Twenty- Seven Democracies, 1945-1990. New York: Oxford University Press. Lijphart, Arend. 1999. Patterns of Democracy: Government Forms and Perfor- mance in Thirty-Sir. Countries. New Haven, Connecticut: Yale University Press. 91 Lipset, Seymour M. and Stein Rokkan. 1967a. “Cleavage Structures, Party Systems and Voter Alignments: an Introduction.” In Party Systems and Voter Align- ments: Cross-National Perspectives, ed. Seymour M. Lipset and Stein Rokkan. New York: Free Press. Lipset, Seymour M. and Stein Rokkan, eds. 1967b. Party Systems and Voter Align- ments: Cross-National Perspectives. New York: Free Press. Mackie, Thomas T. and Richard Rose. 1991. The international almanac of electoral history. 3rd ed. Washington, DC: Congressional Quarterly. Maeda, Ko and Misa Nishikawa. forthcoming. “Duration of Party Control in Par- liamentary and Presidential Governments: A Study of Sixty-Five Democracies, 1950-1998.” Comparative Political Studies . Mair, Peter. 1996. “Party Systems and Structure of Competition.” In Comparing Democracies: Elections and Voting in Global Perspective, ed. Lawrence LeDuc, Richard G. Niemi and Pippa Norris. Thousand Oaks, CA: Sage Publications. Mair, Peter. 1997. Party System Change: Approaches and Interpretations. New York: Oxford University Press. Mair, Peter and Richard S. Katz. 1997. “Party Organization, Party Democracy, and the Emergence of the Cartel Party.” In Party System Change: Approaches and Interpretations. New York: Oxford University Press. Mori, Hiroki. 2001. Nihon Shakaito no kenkyu: Rosen tenkan no seiji katei. Tokyo: Bokutakusha. Nadeau, Richard, Richard G. Niemi and Antoine Yoshinaka. 2002. “A Cross- National Analysis of Economic Voting: Taking Account of the Political Context Across Time and Nations.” Electoral Studies 212403—23. N ihon Seikei Shinbunsha. 1961. K okkai binran. Tokyo: Nihon Seikei Shinbunsha. Nihon Seikei Shinbunsha. 1963. K okkai binran. Tokyo: Nihon Seikei Shinbunsha. O’Handley, Kathryn, ed. 2003. Canadian Parliamentary Guide 2002. Farmington Hills, MI.: Gale. Ordeshook, Peter C. and Olga V. Shvetsova. 1994. “Ethnic Heterogeneity, District Magnitude, and the Number of Parties.” American Journal of Political Science 38(1):100—23. Pacek, Alexander C. 1994. “Macroeconomic Conditions and Electoral Politics In East Central Europe.” American Journal of Political Science 38:723—44. 92 Pa P Pacek, Alexander C. and Benjamin Radcliff. 1995. “The Political Economy of Com- petitive Elections in the Developing World.” American Journal of Political Science 39:745—59. Paldam, Martin. 1981. “A Preliminary Survey of the Theories and Findings on Vote and Popularity Functions.” Europrean Journal of Political Research 9:181~99. Paldam, Martin. 1986. “The Distribution of Election Results and the Two Explana- tions of the Cost of Ruling.” Europrean Journal of Political Economy 2(1):5—24. Paldam, Martin. 1991. “How Robust Is the Vote Function?: A Study of Seventeen Nations over Four Decades.” In Economics and Politics: The Calculus of Sup- port, ed. Helmut Norpoth, Michael S. Lewis-Beck and Jean-Dominique Lafay. Ann Arbor, MI: University of Michigan Press. Pedersen, Mogens N. 1983. “Changing Patterns of Electoral Volatility in European Party Systems; 1948-1977: Explorations in Explanation.” In Western European Party Systems: Continuity and Change. Beverly Hills, CA: Sage Publications. Pempel, T.J., ed. 1990. Uncommon Democracies: The One-Party Dominant Regimes. Ithaca: Cornell University Press. Pennings, Paul and Jan-Erik Lane, eds. 1998. Comparing Party System Change. London: Routledge. Powell, G. Bingham. 1982. Contemporary Democracies: Participation, Stability and Violence. Cambridge, MA: Harvard University Press. Powell, G. Bingham and Guy D. Whitten. 1993. “A Cross-National Analysis of Economic Voting: Taking Account of the Political Context.” American Journal of Political Science 37(2):391-—414. Powell, G. Bingham, Jr. 2000. Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven: Yale University Press. Rae, Douglas W. 1971. The Political Consequences of Electoral Laws. Revised ed. New Haven: Yale University Press. Ramseyer, J. Mark and Frances McCall Rosenbluth. 1993. Japan’s Political Mar- ketplace. Cambridge: Harvard University Press. Reed, Steven R. 1990. “Structure and Behaviour: Extending Duverger’s Law to the Japanese Case.” British Journal of Poltiical Science 20:335-356. Reed, Steven R. 2001. “Duverger’s Law is Working in Italy.” Comparative Political Studies 34(3):312—27. 93 Reed, Steven R. and John M. Bolland. 1999. “The Fragmentation Effect of SNTV in Japan.” In Elections in Japan, Korea, and Taiwan under the Single Non- Transferable Vote: The Comparative Study of on Embedded Institution, ed. Bernard Grofman, Sung-Chull Lee, Edwin A. Winckler and Brian Woodall. Ann Arbor, MI: The University of Michigan Press. Remmer, Karen L. 1991. “The Political Impact of Economic Crisis in Latin America in the 19805.” American Political Science Review 85:777—800. Remmer, Karen L. 1993. “The Political Economy of Elections in Latin America, 1981-1991.” American Political Science Review 87:393—407. Riker, William H. 1986. “Duverger’s Law Revisited.” In Electoral Laws and their Political Consequences, ed. Bernard Grofman and Arend Lijphart. New York: Agathon Press pp. 19—42. Rokkan, Stein. 1968. “The Growth and Structuring of Mass Politics in Smaller European Democracies.” Comparative Studies in Society and History 10:173— 210. Rose, Richard and Thomas T. Mackie. 1983. “Incumbency in Government: Asset or Liability?” In Western European Party Systems: Continuity and Change, ed. Hans Daalder and Peter Mair. Beverly Hills, CA: Sage Publications. Royed, Terry J ., Kevin M. Leyden and Stephen A. Borrelli. 2000. “Is ‘Clarity of Re- sponsibility’ Important for Economic Voting? Revisiting Powell and Whitten’s Hypothesis.” British Journal of Poltiical Science 30:669-98. Sartori, Giovanni. 1976. Parties and Party Systems: A Framework for Analysis. New York: Cambridge University Press. Shepsle, Kenneth A. 1992. “Congress is a “They,” not an “It”: Legislative intent as oxymoron.” International Review of Law and Economics 12(2):239—56. Shiota, Ushio. 1994. Eda Saburo: hayasugita kaikakusha. Tokyo: Bungei Shunju. Shugart, Matthew Soberg. 1992. “Electoral Reform in Systems of Proportional Representation.” European Journal of Political Research 21(3):207—24. Siaroff, Alan. 2000. Comparative European Party Systems. New York: Garland Publishing. Soga, Yuji. 1989. “Shakaito rosen ronso shoshi: Ketto kara ‘Shinsengen’ made no keifu.” Ekonomisuto (Rinji zokan) . October 23. 94 F- Mir-EMT; Stockwin, J. A. A. 1986. “The Japan Socialist Party: A Politics of Permanent Opposition.” In The Japanese Party System: From One-Party Rule to Coalition Government, ed. Ronald Hrebenar. Boulder, CO: Westview Press. Stockwin, J. A. A. 1992. “The Japan Socialist Party: Resurgence After Long Decline.” In The Japanese Party System, ed. Ronald Hrebenar. 2nd ed. Boul- der, CO: Westview Press. Strum, Kaare. 1990. Minority Government and Majorty Rule. Cambridge: Cam- bridge University Press. Strom, Kaare and Seymour Martin Lipset. 1984. “Macroeconomics and Macrop- olitics: The Electoral Performance of Democratic Governments.” Presented at the 1984 annuan meeting of the American Political Scicence Association, Washington, DC. Taagepera, Rein. 1997. “Effective Number of Parties for Incomplete Data.” Electoral Studies 16(2):145-51. T aagepera, Rein. 1999. “The Number of Parties as a Function of Heterogeneity and Electoral System.” Comparative Political Studies 32:531~48. Taagepera, Rein and Matthew S. Shugart. 1989. Seats and Votes: The Effects and Determinants of Electoral Systems. New Haven: Yale University Press. Taagepera, Rein and Matthew S. Shugart. 1993. “Predicting the Number of Parties: A Quantitative Model of Deverger’s Mechanical Effect.” American Political Science Review 87:455—64. Tani, Satomi. 1992. “The Japan Socialist Party before the Mid-19603: An Analysis of Its Stagnation.” In Creating Single-Party Democracy: Japan’s Postwar Political System, ed. Tetsuya Kataoka. Stanford, CA: Hoover Institution Press. Tomz, Michael, Jason Wittenberg and Gary King. 2003. “CLARIFY: Software for Interpreting and Presenting Statistical Results. Version 2.1.”. Stanford Univer- sity, University of Wisconsin, and Harvard University. January 5. Available at http://gking.harvard.edu/. Ware, Alan. 1996. Political Parties and Party Systems. Oxford: Oxford University Press. Whitten, Guy D. and Harvey D. Palmer. 1999. “Cross-National Analysis of Eco- nomic Voting.” Electoral Studies 18:49—67. Wildavsky, Aaron. 1959. “A Methodological Critique of Duverger’s Political Par- ties.” Journal of Politics 21:303—318. 95 Wilkin, Sam, Brandon Haller and Helmut Norpoth. 1997. “From Argentina to Zambia: a World-Wide Test of Economic Voting.” Electoral Studies 16:301—16. Woldendorp, Jaap, Hans Keman and Ian Budge. 2000. Party Government in 48 Democracies (1945-1998). Dordrecht, The Netherlands: Kluwer Academic Publishers. 96 IIIIIIIIIIIIIIIIIIIIIIIIIIIIIII ll(in)((1))I))I)()l))((i|